Achieving Six Sigma Quality in Medical Device ...

Achieving Six Sigma

MARCH/APRIL 2007 PHARMACEUTICAL ENGINEERING On-Line Exclusive 1www.ispe.org/PE_Online_Exclusive

©Cop

yrig

ht IS

PE 2

007

Exclusive On-Line Article

PHARMACEUTICAL ENGINEERING®

The Official Magazine of ISPE

March/April 2007, Vol. 27 No. 2

This articlehighlights how amanufacturer ofmedical devicesobtained SixSigma quality inproduction bythe use ofDesign ofExperiments(DoE) andStatisticalProcess Control(SPC) anddiscusses howthese tools canbe an importantstep toward theFuture DesiredState.

Achieving Six Sigma Quality inMedical Device Manufacturing by Useof Design of Experiments andStatistical Process Control

by Per Vase

% of organization Training Subjects Training Duration Roles

100 How to read a control chart and a capability 1 day Act on control chartsindex

10 How to perform a DoE, create a control 2 weeks Green Belts.chart, and select the right capability index Project participant in DoE and

SPC projects with supervision

1 How to manage projects using DoE and 4 weeks + project Black Belts.SPC Project Manager.

Supervisor.

Table A. Typical SixSigma training.

Introduction

A major healthcare company wanted tointroduce an ultrasonic welding tech-nique for making a critical componentfor one of their new medical devices. A

failure in a welding would have serious conse-quences for the customer. The Acceptable Qual-ity Level (AQL) was a sub-ppm error rate sincemillions of weldings have to be made each year.Such a low AQL can not be ensured by atraditional offline QC sampling inspection.Instead, a lean production layout was needed.All welded components should be monitoredfor welding quality in-line at production speed.Bad parts should be sorted out automaticallyby the welding equipment. To ensure on-targetquality and high yield, the monitoring of weld-ing quality should be used to control the pro-cess from Statistical Process Control (SPC)charts. Prior to the implementation of SPC,Design of Experiments (DoE) was used to cor-relate Critical To Quality (CTQ) attributes toparameters that can be measured quickly andnon-destructively on all samples to obtaintimely measurements. In addition, DoE hasbeen used to establish the correlation between

process result and process settings, the so calledtransfer function. By using the transfer func-tion, it is possible not only to monitor, but alsoadjust the process and control manufacturingto ensure final product quality. Finally, DoEhas been used to establish the Design Space.Data is quickly, conveniently, and visually dis-played using SPC charts on monitors as imme-diate operator information and stored in a da-tabase for trend analysis over a longer period oftime. By using the SPC system, Six Sigmaquality has been obtained.

A general description of the Six Sigma toolsused and methodology employed is presented,including how they can be of value for thepharmaceutical industry.

BackgroundThe FDA defines in their Guidance for Indus-try1 Process Analytical Technology (PAT) as:“The Agency considers PAT to be a system fordesigning, analyzing, and controlling manufac-turing through timely measurements (i.e., dur-ing processing) of critical quality and perfor-mance attributes of raw and in-process materi-als and processes with the goal of ensuring final

Achieving Six Sigma

2 PHARMACEUTICAL ENGINEERING On-Line Exclusive MARCH/APRIL 2007 www.ispe.org/PE_Online_Exclusive

©Copyright IS

PE 2007

product quality.” Tools for controlling manufacturing frommeasurements of CTQ parameters have been available formore than 80 years since W.A. Shewhart in 1924 introducedthe control chart concept in Bell Laboratories. Althoughfrequently used in some industries (e.g., the automotiveindustry), control charts have never obtained as widespreaduse as they deserve, and especially within the pharmaceuti-cal industry they are rarely used. There are several reasonsfor this. Three of the main reasons why control charts havenever previously made the breakthrough within the pharma-ceutical industry are:

1. no urgent need for change

2. lack of operational process understanding before imple-menting SPC

3. implementation attempt by statisticians instead of endusers

No Urgent Need for ChangeThe pharmaceutical industry has for many years been in aspecial environment with strong regulation and patent pro-tection. Production efficiency and yields have not, as in manyother industries, been the major competition parameter. Asa result of this, pharmaceutical manufacturing has a lowmanufacturing performance compared to other industries.2,3

A famous article in The Wall Street Journal expressed it thisway: “pharmaceutical manufacturing techniques lag far be-hind those of potato-chip and laundry-soap makers.”2 Inorder to avoid defective products reaching the market, heavyQuality Assurance (QA) and Quality Control (QC) strategieshave been established. A recent study by IBM3 shows that

pharmaceutical manufacturing typically has a process sigmalevel of 2.5 in productions, corresponding to a Cp of 0.83 or150000 ppm defects. In comparison, pharmaceutical releasehas a quality sigma level of 5 corresponding to a Cp of 1.67 or200 ppm defects. No other industry has this three orders ofmagnitude defect difference between produced quality andreleased quality. It is the result of an incredible effort in QAand QC, especially in end-product testing and sorting, lead-ing to Quality by Inspection. This is done to absolute perfec-tion and there is not more to gain following this route.However, there are two drawbacks to this working practice:

1. It drives the prices up, due to high Costs of Poor Quality(CoPQ).

2. It makes it impossible to improve the released qualityeven further.

As it is said in the FDA PAT Guidance, “The health of ourcitizens depends on the availability of safe, effective, andaffordable medicines.” The pharmaceutical industry has tofind a more efficient way of controlling manufacturing pro-cesses to make medicines affordable for a larger group ofcustomers. In addition, the quality needs to be improvedfurther; 200 ppm is not good enough for critical characteris-tics. The industry can not continue to increase the QC effortsby even larger sample sizes in end product testing; the limitis reached!

This general industry trend also can be seen in the latestISO sampling standard,4 which moves away from traditionalAQL sampling methods and recommends screening (continu-ous monitoring) and process control instead for critical char-acteristics. This issue also is highlighted in a recent publica-

Figure 1. Illustration of Capability index Cp and Cpk.

Achieving Six Sigma


©Cop

yrig

ht IS

PE 2

007

tion from the FDA5 inspired by recent Military Standards.6

There is now an urgent need to change that had not beenrecognized previously.

Lack of Operational Process UnderstandingBefore Implementing SPCMany SPC implementations have failed due to lack of opera-tional process understanding. In order to be able to establisha proper control strategy, processes need to be understood, toknow which parameter to measure and plot in a control chart,and when there are points out of control to know what to doabout it. As it is written in the FDA PAT Guidance forIndustry:1

“A process is generally considered well understood when:

1. all critical sources of variability are identified and ex-plained

2. variability is managed by the process”

If it is not known how to act on points out of control on thechart, the control chart only creates panic, not improvedprocessing. Of course there will be process understandingbased on learning by doing in all companies. However, this isvery person dependent and typically people act differently onprocess measurements. The understanding is not opera-tional, it is subjective. With subjective process understand-ing, control and adjustment often make things worse com-pared to not doing anything. This has resulted in the typical“don’t change anything after PQ strategy,” strong changecontrol, and the belief that if it worked in PQ, it will work atall times disregarding e.g., equipment wear, raw materialvariation, and climatic changes. Running three PQ batcheswith minimum variation between them, just after each other,heavily monitored by process experts and engineers (who willnot be there in normal production) does not solve this issue.

Fortunately, the tool is there to obtain process under-standing and test if processes are robust: DoE. Again, this isa more than 80 year old tool originally developed by R.A.Fisher in 1922. By systematically varying all factors of

interest in a DoE, it is possible with a minimum number ofexperiments to create operational process understandingthat can be shared within the whole organization. When thisprocess understanding is established in the organization, therisk is minimized for the customer and for the company. Riskis inversely proportional to understanding. It will be knownhow to control and adjust processes in order to managevariation in process conditions going away from the “don’tchange anything” strategy. Process understanding also willlead to a more lean regulatory approach. It is written in theguidance:1 “For processes that are well understood, opportu-nities exist to develop less restrictive regulatory approachesto manage change.”

Implementation Attempt by Statisticians Insteadof UsersMost pharmaceutical companies above a certain size have astatistical department that takes care of analyzing results ofclinical testing, input to product registration, and dimension-ing sampling plans for end product testing and release. Forthese companies, it has been obvious to try to use thesestatistical departments for implementing DoE and SPC inmanufacturing. However, this has often resulted in proce-dures that are too complex, reports that no one outside the

Sigma Level Yield % Cp before Sorting System CoPQ % of CoPQ % ofDowntime each Sales (8) Sales (9)

year (days)

1 30 0.33 255 >40 >70 Non competitive

2 69 0.67 112 30-40 >40 Non competitive

3 93 1.00 24 20-30 25-40 Average PharmaSigma = 5 after sorting (3)

4 99.4 1.33 2,27 15-20 15-25 Average Other Industries

5 99.98 1.67 0.085 10-15 5-15

6 100 2.00 0.0012 <10 <1 World Class Pharma (3)Automotive Industry

7 100 2.33 0.000069 ? ?

8 100 2.67 0 ? ? Semiconductor Industry

Table B. Relation between Sigma Level and Cost of Poor Quality. Sigma Level is the number of standard deviations between target valueand specification limits.

Figure 2. Cost of Poor Quality Iceberg. Cost of Poor Quality ismuch more than the direct costs.

Achieving Six Sigma


©Copyright IS

PE 2007

statistical department could understand, too slow responseto production needs, perfectly analyzed DoE’s with the wrongfactors tested, and SPC on wrong parameters. Other indus-tries overcame these challenges in the 1990s by implement-ing Six Sigma, originally developed by Motorola. A veryimportant part of Six Sigma is not to use statisticians toperform and implement DoE and SPC. Statisticians shall beused to train the organization in these methods so they can doit themselves. This ensures test of the right factors in DoEs,process experience used in the analysis phase, and it endswith control strategies that can be used on the shopfloor. Thisrequires an extensive training program where the wholeorganization is trained in applied statistics to different levelsas shown in Table A.

Previously, intensive training was needed to be able toperform DoE and SPC, but with today’s statistical softwaretools, it is possible to be operational after a few weeks oftraining especially with guidance from statistical experts.

Cost of Poor QualityCost is the driving force behind most decisions. In order to getmanagement attention to implement DoE and SPC, theimplementers need to be able to address the cost savings fromusing the tools. It is obvious to get inspiration from the workdone during implementation of Six Sigma. A Six Sigmaproject will typically minimize variation and drive sigmalevel and capability index Cp up. Six Sigma projects arealways cost/benefit driven. Models for the relation betweenCost of Poor Quality (CoPQ) and sigma level and/or Cp havebeen developed. Before proceeding, Cp and sigma level will bedefined. Figure 1 shows the formulas for and a schematic ofthe capability index Cp and Cpk. Sigma s represents thestandard deviation of the distribution of measured data. Cp isthe ratio between tolerance window and process width (6s).A Cp of 1 corresponds to the width of the tolerance window isequal to the width of the process (i.e., the process widthexactly fits the tolerance window). This does not allow anydrift of the process; it needs to be on target at all times.Having a Cp of 1 there is room for +/- 3s within the tolerancewindow, which is called sigma level 3. In order to allow fordrift, Cp needs to be higher than 1. When Cp=2, there is roomfor +/- 6s within tolerance window, called sigma level 6 or SixSigma Quality. Often, an analogy is made to driving a car intoa garage. The tolerance window is the width of the garage and6s corresponds to the width of the car. In order to ensure thatany driver will never hit the edges of the garage, the width ofit has to be twice the width of the car, corresponding to a Cp=2.

The Cp index alone is not enough to describe the process.It is possible to have a high Cp and a low yield if the process

Figure 3. Ishikawa cause and effect diagram.

Lean Six Sigma

Numberof Steps Cp=1,00 Cp=1,33 Cp=1,67 Cp=2,003 sigma 4 sigma 5 sigma 6 sigma

1 66807 6210 233 3

2 129151 12381 465 7

3 187330 18514 698 10

4 241622 24608 930 14

5 292287 30665 1163 17

6 339568 36684 1395 20

7 383689 42666 1627 24

8 424863 48611 1860 27

9 463287 54519 2092 31

10 499143 60390 2324 34

20 749142 117133 4642 68

50 968481 267617 11565 170

100 999007 463615 22997 340

1000 1000000 998029 207574 3392

Table C. Relation between ppm error rates, number of process steps, and Cp for each step.

Lean

Six Sigma

Achieving Six Sigma


©Cop

yrig

ht IS

PE 2

007

is far away from target as illustrated in the lower left cornerof Figure 1. To solve this issue, another supplementary index:minimum capability index Cpk is used. It is the distance fromthe mean value to the nearest specification limit divided by3s (half process width). If the process is on target, Cpk=Cp; ifthe process is not on target, Cpk<Cp.

Capability indices Cp and Cpk are excellent key perfor-mance indices to describe a process. However, they shall beused with care. They are calculated on the assumptions thatthe process is in statistical control and data are normaldistributed. This is often not the case and other types ofindices (e.g., Pp and Ppk) should be used as described in arecent ISO standard.7

The capability index is used as a parameter in the costmodels correlating Cp and CoPQ measured in percentage ofsales as shown in Table B. The CoPQ numbers for lowersigma levels might seem very high. This is due to CoPQcalculations that take into account all contributions to Costof Poor and not just the tip of the iceberg10 as shown in Figure2. In addition, the yield column in Table B is for one process.Typically, many processes or components need to work at thesame time to have a successful product so the yield of theproduct is lower than the yields of the individual processes asshown in Table C. By applying lean (reducing the number ofsteps) and Six Sigma (improve Quality of steps), low errorrates of 3 ppm can be obtained.

Design of Experiments (DoE)The first step in a DoE is to define the response variables (i.e.,what is to be measured on the runs in the experiment todistinguish between good and bad runs). Typically, external(seen from the customer) CTQ parameters will be measuredtogether with internal (measured quickly, non-destructively,and correlating with external) CTQ parameters. An impor-tant part of a DoE is to correlate internal to external CTQparameters. The next step is to identify the factors that areexpected to influence the responses and will be varied in theexperiment. This could be done in a process review.11 Typi-cally, the list of factors is too long to be able to make a precisemathematical function relating the factors to the responsesin one reasonably sized experiment. It is normal to start witha screening experiment where the many factors are varied intwo levels only and assuming no interactions to keep thenumber of runs low. The purpose of this experiment is not tomake the mathematical model, but only to find the factorsthat has the largest influence and requires a more detailedinvestigation. This step will typically cut the number offactors down to a level where they can be coped within a singlemodel experiment. This model experiment, called a responsesurface experiment, establishes the precise mathematicalrelationship between factors and responses, including non-linearities. An important output of the DoE is the DesignSpace: “The multidimensional combination and interactionof input variables (e.g., material attributes) and processparameters that have been demonstrated to provide assur-ance of quality.”12

Statistical Process Control (SPC)To ensure continuous optimized performance, processes needto be controlled during production to adjust for, e.g., rawmaterial differences, equipment wear, and environmentalchanges. From the DoE results, the external CTQ’s can bepredicted by measuring the internal CTQ. If the level ofinternal CTQ changes, the process can, based on the DoEresults, be adjusted to change level. However, two questionsremain to be answered:

1. How big shall the change in internal CTQ be before theprocess parameters are adjusted?

2. Which process parameter shall be used to adjust theprocess?

Two types of variation exist: random variation and system-atic variation. Random variation is characterized by beingunpredictable and having no assignable cause (or a sum ofmany small contributions, where it is practically impossibleto assign causes). If it is tried to adjust on random variation,it will only make the variation larger. If the internal CTQincreases due to random variation, it does not mean that thelevel has actually changed and if the level is adjusted downbased on this, variation is added. Systematic variation ischaracterized by having an assignable cause behind thechange and the process can be back on track by eitherremoving the assignable cause or compensate the assignablecause by adjusting process parameters. So in short, theoperator shall act on a change in CTQ if it is due to systematicvariation and leave the process as is if it’s due to randomvariation. The obvious question now is how will the operatorknow? The answer is straightforward: use SPC. Basically,SPC distinguishes between random and systematic varia-tion. When control charting, the measurements are typicallydivided in two subgroups and the mean and range (maxi-mum-minimum) are plotted versus time. Based on the varia-tion within a subgroup, control limits can be calculated forboth the mean and the range. When new mean values arewithin the control limits the process is only subjected torandom variation and process adjustment will only increasevariation. If it is outside the control limits, there is anassignable cause that either should be removed or compen-sated by adjustment. From the DoE, it will be known whichprocess parameter is optimal for adjustment.

DoE and SPC Case StudyTrain StaffIn order to ensure that the use of DoE and SPC resulted insomething that could be used on the shop floor, an intensivetraining program was initiated as shown in Table A. Allemployees were given a one day course in SPC, includingcapability indices and how to act on a control chart. Tenpercent of the employees were trained in using DoE and SPC,including training in the selected statistical software. Thetraining was a part of the general lean implementation at the

Achieving Six Sigma


©Copyright IS

PE 2007

company. Now the users of the processes can perform DoEand SPC with support of their statistics department, insteadof the statistics department doing it for them.

Select CTQ ParametersThe customers risk analysis was studied and CTQ attributeswere identified. From the risk analysis, the external (seenfrom the customer) CTQ attributes were the strength of thewelding and there was no loose, excess material from thewelding, so called flush.

Operationalize the CTQSince the inspection had to be made on all welded compo-nents, destructive testing of the welding strength was not apossibility. It was necessary to use other characteristic re-sponses to the process that could be measured quickly andnon-destructively (i.e., internal CTQ parameters). The heightreductions during welding and welding time were chosen.The height reduction is an indicator of the process result. Thewelding time is an indicator of the process itself. The welding

is done with fixed energy, i.e., the welder uses the time neededto deliver the set point energy.

Identify Potential Influence FactorsTogether with the process experts, potential influence factorswere identified in a process review and the result plotted inan Ishikawa diagram as shown in Figure 3. In the firstscreening DoE, many factors were investigated. This experi-ment had the double purpose of finding critical sources ofvariability and identifying factors for further analysis. Thisinitial study identified both the needs of modification of theheight measurement system and a small redesign of thecomponent. Because these needs were identified early in theproject (before Factory Acceptance Test), the modificationsdid not cause any delay in the project.

Establish RelationshipsBased on the conclusions from the screening experiment, themost important control variables were investigated in moredetail in a Response Surface Experiment. This was done after

Figure 4. Result of the response surface experiment.

Figure 5. Relation between external CTQ’s (separation force left, flush right) and internal CTQ (height difference).

Achieving Six Sigma


©Cop

yrig

ht IS

PE 2

007

the machine had been moved to the healthcare company.After this study, relationships between both control param-eters (Pressure, Energy, Trigger, Amplitude) and internalCTQ (height difference and welding time) as well as betweeninternal and external CTQ (force and flush) were found.Figure 4 shows the result of the response surface experiment,the transfer function between process parameters (Pressure,Energy, Trigger, Amplitude), and process results (heightdifference). These curves show which parameters are themost critical (i.e., have the largest slope). They also can beused to control the height difference and the welding strengthin future production. Since the transfer functions in this caseare non-linear, they also can be used to find the settings thatwill lead to the most constant height difference. Finally, itgives the Design Space for the process, i.e., the processparameter window that will ensure good weldings.

Another advantage of DoE is that it creates samples witha lot of variation in CTQ parameters, which are ideal forcorrelating internal CTQ (height difference and weldingtime) to external CTQ (force and flush) as shown in Figure 5.In this way, a specification limit for height difference can beestablished in a scientific way based on process understand-ing. It is seen that height difference is a good indicator of bothstrength and flush. The amount of flush is characterized ona scale from one to three. To the left of the blue curve on thelogistic fit is flush grade 1 (low flush) area. To the right of the

right blue curve is flush grade 3 (high flush) area. Since thecurves are almost vertical, height difference is a good indica-tor for flush.

OptimizeFrom the results of the Response Surface Experiment, theoptimal setting of the control variables for obtaining theoptimal internal CTQ’s were found. The optimization waseasy to implement since it was only a matter of changing setpoints for the control variables. With the optimized settings,Ppk increased from 0.7 to 2.0.

Control StrategyTo keep the optimized conditions for the external CTQ’s(welding strength and no flush) over time it was decided tomake statistical control charts on the internal CTQ’s (heightdifference and welding time).

The plot of mean values of height differences versus timeis seen in the upper chart on Figure 6. It is easy to read for theoperator, who shall monitor if new mean values are in the redzone. If this is the case, actions are needed. Also, the capabil-ity and performance index are shown in the table in the upperright corner. Since this process is on target, in statisticalcontrol, and follows a normal distribution, there is not muchdifference between Cp, Cpk, Pp, and Ppk. They are all above 2equal to Six Sigma quality. Besides being shown on-line at

Figure 6. Shopfloor SPC control chart.

Achieving Six Sigma


©Copyright IS

PE 2007

the shopfloor in control charts, data also is stored in adatabase connected to a statistical software package. Thisallows fast reviewing of historical data, which is being usedfor continuous improvements and troubleshooting.

When there are new mean values Out Of control (OOC), itis important to have operator instructions for what to do asshown in Figure 7. A typical reason for being out of control ismeasurement error, not that the process has actually changed.For this reason, a control chart also is made on the ultrasonic

welding time. So the first thing the operator does if there areOOC on height differences is to see if there are also OOC onwelding time. If this is not the case, the first action will be toclean the height measurement systems because the processindicator (welding time) shows no abnormal behavior. If thecleaning does not help or if both height difference and timeare OOC, welding parameters are adjusted or maintenance isperformed. The decision to adjust parameters or to do main-tenance is dependent on the position within Design Space.

As can be seen in Figure 4, there are several options forchoosing parameters to adjust height differences. Ideally, oneshould choose only one and fix the others to make it opera-tional on the shopfloor. When choosing the parameter, it canbe beneficial to look at how process parameters influence theCTQ relations shown in Figure 5. In Figure 8, it is seen thatthe lower the pressure, the better the height difference worksas a barrier for low forces due to a lower slope on thecorrelation curve. Therefore, it is not a good solution to usepressure to adjust height difference because it is best to haveit at a low value at all times. For this, process energy waschosen to adjust height difference keeping other parametersfixed.

ValidateDue to the process understanding obtained from the DoEexperiments, validation efforts were concentrated on valida-tion of the measurement systems and finding the final opti-mal setting for and correlation between internal and externalCTQ’s. The latter was done by running a final DoE as a partof Operational Qualification. In this study, it was demon-strated with statistical confidence that the on-line heightdifference measurements could be used to ensure sufficientFigure 8. CTQ relations for different pressures.

Figure 7. Operator instruction.

Achieving Six Sigma


©Cop

yrig

ht IS

PE 2

007

welding strength and low amount of flush. From the DoE, thefinal mathematical relationship between control variablesand height difference also was established (i.e., the finaltransfer function and Design Space). During PerformanceQualification, the validation efforts were concentrated ondemonstrating that the process was in statistical control andwith sufficient process capability. This was easily docu-mented with data from the SPC database.

Outcome and BenefitsThe company obtained a fully operational welding processproducing Six Sigma Quality with a Ppk>2. The starting pointwas a Ppk of 0.7. This corresponds to a reduction of CoPQ from25% to 1% of Sales. The process quality is documented duringproduction by in-line measuring of the height difference andwelding time for each welding. There is no need for lead timeincreasing offline QC controls. Due to the SPC system and theestablished transfer function, the company can now easilykeep the process on target from the shopfloor by adjusting thecontrol variables to compensate for changes in raw materials,climate, wear etc. Since the adjustment is based on processunderstanding, it can be done without change requests.

ConclusionImplementing PAT tools like DoE and SPC has a highpotential to increase quality and lower the costs of pharma-ceuticals and medical devices. Alot can be learned by lookingat how it was done in other industries. PAT has a lot ofsimilarities with Six Sigma and the pharmaceutical industryshould learn from the experience in implementing Six Sigma.Six Sigma cost savings models from variance reduction can beused to quantify PAT benefits. Another important learningfrom Six Sigma is that the statistical tools like DoE and SPCshall be used by the process users not by statistical experts.The case study shows an implementation example where SixSigma tools have been used within the pharmaceutical indus-try to improve quality from a Ppk of 0.7 to a Ppk higher than 2and lower Costs of Poor Quality from 25 to 1% of sales.

References1. http://www.fda.gov/cder/guidance/6419fnl.htm.2. Abboud, Leila and Scott Hensley, “New Prescription for

Drug Makers: Update the Plants,” The Wall Street Jour-nal, 3 September 2003.

3. http://www-1.ibm.com/services/us/imc/pdf/ge510-4034-metamorphosis-of-manufacturing.pdf.

4. ISO21247 (2005).5. Watts, D.C., and J.E. Clark, The Journal of PAT, Vol. 3,

Issue 6, December 2006, pp. 6-8.6. Department of Defense Test Method Standard, DoD Pre-

ferred Methods for Acceptance of Product, Mil-Std-1916,April 1996.

7. ISO21747 (2006).8. Harry, M.J., Quality Progress, May 1998, pp. 60-64.9. Clark, T.J., Success Through Quality, Quality Press 1999,

ISBN 0-87389-441-3, www.successthroughquality. com.10. DeFeo, Joseph A., Quality Progress, Vol. 34, No. 5, May

2001, pp. 29-37.11. Brindle, A., and Per Vase, “Process Review for PAT –

Selecting Cost Efficient PAT Projects,” PharmaceuticalEngineering, November/December 2006, Vol. 26, No. 6,pp. 70-78.

12. ICH, Q8 Pharmaceutical Development, Glossary, May2006.

About the AuthorPer Vase holds a PhD in materials scienceand an MSc in experimental physics. Herecently joined NNE’s Process AnalyticalTechnology team as a data analysis expert,but has more than 10 years of experiencefrom previous employments in various in-dustries. Vase has worked with Six Sigma inNew Businesses and has a proven track record

of bringing products rapidly from the Research Phase to theQuality Controlled Production Phase by the use of Six SigmaTools, especially DoE and SPC. Vase has a proven trackrecord in combining compliance efforts with process optimi-zation efforts within the manufacture of medical devices,ensuring both high quality and low costs. He can be contactedby email at: [email protected].

NNE - Process Consulting, Vandtaarnsvej 108-110, 2860Soeborg, Denmark.

Barrier Vial Technology


©Cop

yrig

ht IS

PE 2

007





This articledescribesBarrier VialTechnology(BVT) atechnology forthe filling ofliquids underasepticconditions.

Barrier Vial Technology: A GlobalApproach to the Aseptic Filling Process

by Diego López-Álvarez, Sergi Roura, and J. A. Garcia

Figure 1. A typicalcontainer in the BVTaseptic filling process.

Introduction

Microbial contamination is a concernand a constant struggle in researchlaboratories, as well as in sterilemedicine production plants. Al-

though the distinguished scientist AlexanderFleming discovered penicillin, one of the mostoutstanding discoveries of modern medicine,as a consequence of an accidental contamina-tion in a bacteria culture, it is vital to preventsuch contaminations in sterile formulas.

In order to minimize the risk of contamina-tion in sterile filling, the industry has imple-mented more and more rigorous proceduresand technologies. As a result, a leading com-pany specialized in the manufacture of plasmaderivatives applied its many years of experi-ence to developing an aseptic filling process:Barrier Vial Technology (BVT). This processhas continued to be refined over the last 20years.

BVT does not simply cover aseptic filling,but every single step of the aseptic process;

from the preparation and sterilization of con-tainers and closures to the laser etched identi-fication of vials after dosing. Both liquid andfreeze-dried products can be dosed with theBVT sterile filling process. The description ofthis process in this article refers to a liquidsterile filling plant.

The set of practices and procedures describedin this article demonstrate the unique approachused throughout the aseptic process which mini-mizes the risk of particulate and microbialcontamination every step of the way.

The most important safety measures to takeagainst particulate contamination include ahigh quality clean area and the use of physicalbarriers to protect sterile containers and stop-pers.

Among other relevant features in BVT, thevial is partially closed during handling withphysical barriers. The fact that the vial is pro-tected with physical barriers means that con-tamination risk is extremely minimized. How-ever, the vial allows the steam to enter for aproper sterilization.

Before describing the BVT aseptic fillingprocess in detail, the above mentioned physicalbarriers are described.

The Container: DescriptionA typical container handled in the BVT asepticfilling process comprises three elements: a vial,a capsule-stopper set, and a protector - Figure1.

The vial is a standard container made ofglass or plastic for pharmaceutical or medicinaluse.

The capsule - Figure 2, also is standard, butthe stopper is specially designed for asepticprocessing. The main features of this stopperare:



©Copyright IS

PE 2007

• A stepped outside contour allows the stopper to stay in twodifferent stable positions inside the neck of the vial.1. partially inserted to allow the container to be sterilized

- Figure 2a2. fully inserted to seal the vial after filling - Figure 2d

• A set of grooves in the body of the stopper (similar to thegrooves of stoppers for lyophilized products) allows thesterilization steam pass to the inside of the container.

• A flange shape provides a tight fit between the stopper andthe capsule.

The protector - Figure 1 and Figure 2a rests on the vial neckcovering the capsule-stopper set. This piece has the followingtwo functions:

1. to create a labyrinth-like path, between the vial and thecapsule-stopper set, that prevents particulates from en-tering into washed and sterilized containers

2. to prevent the capsule-stopper set from being fully in-serted due to improper handling before sterilization (onlywhen being handled manually)

The originality of the container is not based on the use ofstandard vials or redesigned closures, but on the handling ofthe vial: the capsule-stopper set and the protector are in placeon the vial, from the earliest steps of the aseptic processing,

creating a physical barrier against microbial contamination.

BVT: Aseptic Process DescriptionThe BVT flow diagram is illustrated in Figure 3.

Container and Closure PreparationThe specially designed, pre-washed, Gamma-radiated, andclean packed stoppers are automatically inserted inside thecapsule bodies. The product contact surfaces of the stoppersare rinsed with water for injection and blown with filtered air.

The vials are thoroughly rinsed and blow-cleaned insideand out at different stations to meet pharmaceutical stan-dards in conventional washing machinery.

The capsule-stopper sets are partially inserted into thevials and afterward protectors are simply placed over thevial.

Once this is done, the partial closure creates a labyrinth-like path, which reduces the probability that particulates willbe able to enter, but vials can still be sterilized. The containerwill continue to have a “labyrinth-like seal” until dosing.

Containers are arranged on trays, and the trays are loadedon wheeled racks.

Component SterilizationWheeled racks are conveyed into an autoclave where thecontainer (including the capsule-stopper set and protector)are sterilized by moist heat.

The labyrinth-like seal of the container permits the air tobe removed with a preliminary stage of vacuum pulses and

Figure 2. The container in BVT aseptic processing.



©Cop

yrig

ht IS

PE 2

007

then steam to enter and sterilize the vial. Just after thesterilization stage of the autoclave cycle, a drying stage takesplace to prevent condensation from forming inside the vials.

The intensive vial washing together with the long steril-ization process guarantee the reduction of endotoxins (by atleast 3 logs) of the containers as specified in cGMPs.1

After the sterilization cycle, the wheeled racks remainunder a laminar flow to cool the containers in the asepticprocessing area.

Aseptic FillingWheeled racks are brought near the filling room. The opera-tor places the containers of each tray onto the infeed rotarytable of the filling line.

Dosing takes place in a Grade A (Class 100) environmentequipped with a horizontal laminar flow. Inside this area, theprotector is discarded.As soon as the vial reaches the filling point, the capsule-stopper set is removed and the filling nozzle doses the phar-maceutical product. After dosing, the capsule-stopper set isinserted completely into the vial.

Therefore, the amount of time during which the vialremains open within the Grade A environment is reduced tothe time required to unstopper the vial, fill and restopper thevial (full insertion).

Unlike conventional filling lines, there is no need for extra

machinery to feed stoppers and capsules, because the vialreaches the filling point with the stopper-capsule set alreadymounted on the vial.

The absence of stopper feeder equipment in the filling areareduces the particulate count and obtains better particleresults during monitoring.

A video camera records the whole filling process.

Sealing and IdentificationStoppered vials are conveyed outside the filling room wherethe capsules are crimped under a laminar flow by means ofstandard crimping machinery.

The filling and crimping processes take place in differentrooms (physical separation and different pressure levelsavoiding pressure reversal),2 so that no particulates gener-ated during the crimping operation will reach the filling area.

After crimping, a laser system marks the batch code, thefilling time, and the vial number on the glass vial.

Laser marking is durable and cannot be eliminated with-out damaging the container. In addition to anti-counterfeit-ing benefits, laser marking also is helpful for traceability.

The filling process recording and the laser marking is veryuseful if a quality investigation is performed. Both the fillingtime, which is etched on every filled vial, and the videorecording of the whole filling operation allow complete track-ing of the filled units.

Figure 3. The BVT process



©Copyright IS

PE 2007

The Filling Suite and the Filling AreaThe filling suite is comprised of the following four rooms -Figure 4:

• the vial loading room• the filling room where the filling area is located• the vial finishing room• the service room behind the filling room

The operator supplies vials to the filling line from the vialloading room. The control and oversight of the filling lineare done from this room. This design features make it pos-sible to minimize the presence of the operator inside thefilling room.

The operator’s tasks inside the filling room are limitedto: preparing for the vial filling (set-up the non-viable par-

ticulates monitoring system and the sterilized filling equip-ment which includes tubing, filling nozzles, containers, etc.),troubleshooting, and environmental control for viable par-ticulates.

This equipment is designed to operate in an “at rest”occupancy state (at rest is when the equipment is installedand operating, but with no operating personnel present.”)3a

Because the container used in the BVT aseptic fillingprocess consists of a pre-assembled stopper (capsule-stopperset - Figure 1), the stopper feeding system in conventionalfilling rooms is eliminated meaning that:

• The operator does not have to enter the filling room to loadstoppers into the feeding system.

• There are fewer particles (any feeding device generatesparticles).

Figure 4. The filling room and the filling area.



©Cop

yrig

ht IS

PE 2

007

• The filling room is smaller.

The filling area (see blue area in Figure 4) is a tiny spaceprotected with a horizontal laminar flow where the followingsteps take place - Figure 5:

• The vial arrives with the protector and the capsule-stop-per set already mounted (Step 1).

• The protector is discarded (Step 2).• The vial moves to the filling position (Step 3).• The vial is unstoppered (Step 4).• The nozzle fills the vial (Step 5).• The vial is fully stoppered (Step 6).

Distinctive features of such an extremely small filling areaare:

• Size: the height and the length of the filling area has thesame dimensions as the HEPA filter of the laminar flow.

• Horizontal laminar flow: a horizontal air flow reducesthe risk of particulates entering into vials.

• Proximity of the vial to the HEPA filter (150 mm): thepotential for contamination of air flow that reaches thevial is reduced.

• Location of equipment within the filling area: move-able parts are placed downwind of the filling point and arecarefully designed to maintain the characteristics of lami-nar air flow.

• Vial handling: a device located outside the filling roompushes the vials into the filling area eliminating belts,chains, and similar conveying systems which are difficult

to clean. A sensor system detects the proper positioning ofthe vial under the filling nozzles.

• Restricted access: safety barriers are installed in thefilling area to protect the sterility of the process. If the lightbarrier detects a breach in the filling area, the fillingmachine automatically stops the process and all the vialsare immediately stoppered. If the process is re-started, themachine will run some cycles without filling, reducing thepossibility of contamination from intrusions into the fill-ing area.

Once the units are filled and stoppered, the stoppered vialsreach the vial discharge room where they are crimped (seeyellow area in Figure 4) and laser marked.

Any maintenance is performed from outside the fillingroom. Because the filling line is integrated into the wallpanel, the inside of the machine is accessible for maintenancefrom the service room in accordance with GMP equipmentdesign recommendations.3b

The design of the filling machine installed in the fillingroom is compatible with any filling system: piston pump,diaphragm pump, time pressure system, weight control,disposable filling, peristaltic pump, etc.

The selection of the most suitable system depends on:accuracy (the more expensive the product, the greater theaccuracy), volume adjustment (fixed volume versus variablevolume depending on product activity), amount of liquid to befilled (small versus big volumes), filling time, batch size, etc.

Validation and Production ExperiencesExtensive validation work has been performed to test theprotective qualities of physical barriers on sterile containersand stoppers in preventing contamination.

Two studies tested the effectiveness of sterile containers

Figure 5. Step-by-step filling process inside the filling area.



©Copyright IS

PE 2007

with a physical barrier used in the BVT aseptic filling process- Figure 1.

• exposure of sterile containers with a physical barrier todifferent microbial environments4

• an airborne microbial challenge of sterile containers witha physical barrier5

The first study consisted of a comparison between sterileopen vials and sterile containers with a physical barriercontaining sterile culture medium (aseptically filled) whenexposed to different environments, specifically grade A, gradeB, and “non-filtered air” for a period of seven days.

The results of this study demonstrate that:

• No single sterile container with a physical barrier wasfound to have microbial contamination after seven daysexposure to any environment.

• Every sterile open vial was found to have microbial con-tamination in the case of non-filtered air, and 1.4% of thesterile open vials were found to have microbial contamina-tion after seven days exposure to grade A and grade Benvironments.

The second study was an airborne microbial challenge ofsterile containers with a physical barrier containing sterileculture medium (aseptically filled). A microbial suspension ofbacillus (Bacillus atrophaeus) was aerosolized over the con-tainers (inside a sealed chamber) at a final concentration ofbetween 25 and 50 times the maximum microbial levelaccepted for a grade D area. After 60 minutes of exposure, thecontainers were fully stoppered, crimped, and incubated for14 days at 30 to 35°C.

The results of this study show that in both concentrationcases, not a single container with a physical barrier hadmicrobial contamination after having been exposed to envi-ronments between 25 and 50 times the limit allowed in agrade D area.

The results of these studies demonstrate that the “laby-rinth seal” created by the vial and the physical barrierincreases the safety against microbes of aseptically filledcontainers.

If “bacteria-carrying particles in room air are large andthat gravitational settling is the most important way they aredeposited,”6 it can be asserted that containers with a physicalbarrier contribute to minimizing the risk of contamination ofthe vials, because potential microbe-carrying particulatesshould not be able to overcome the “labyrinth seal” againstgravity.

Taking into account that personnel is the primary sourceof bacterial contamination in an aseptic cleanroom, the twokey factors that increase the confidence of sterility of filledunits are the “at rest” occupancy state of the filling room andthe physical barrier of the container.

The BVT aseptic filling process is used for the manufac-ture of injectable products derived from human plasma ap-proved by the FDA and European authorities.

BVT has been developed over the course of the last 20years, adding improvements and integrating the latest tech-nology in areas such as filling techniques, microbial control,or machine automation.

Media fill simulations have been done extensively follow-ing BVT procedures and practices at existing productionfacilities. More than 350,000 vials have been filled withmedia since 2002 using the Barrier Vial Technology and norevalidation has been necessary for any batch.

Advantages of BVTBVT offers four advantages over the conventional asepticfilling process:

1. Particulates and Microbial Safety• Vials are kept closed, though not hermetically, thanks to

the “labyrinth seal” existing between washing and filling.This means that the time the sterilized unit is exposed tothe environment is minimized.

• Capsule and stopper feeding equipment (sources of par-ticles) is not needed within the filling room.

• Horizontal laminar flow reduces potential risk of particu-late entry into the vial.

• Vials stay very near the laminar flow during the filling.• The equipment inside the filling area is placed downwind

of the filling point, and was carefully designed to avoid airdisturbances.

• There is no need for the operator to work inside the fillingroom, limiting his/her intervention to critical areas (onlyfor troubleshooting and environmental control).

• Maintenance is performed from outside the filling roomthanks to the design of the filling line embedded in thewall.

2. Environmental Control• The small filling area makes it easier to monitor both

viable and non-viable particulates, and it is easier toguarantee its integrity.

Figure 6. An example of an existing filling room (human albuminproduction in progress).



©Cop

yrig

ht IS

PE 2

007

3. Traceability of Filling Operations• Video recording together with the laser marking means

that the dosing process can be tracked.

4. Cost and Size of the Installation• Less machine intensive: depyrogenation tunnel versus

autoclave (the autoclave is needed anyway in the conven-tional approach).

• The overall size of the facility required for BVT is muchsmaller than conventional filling and easier to maintainand validate than conventional processes.

Disadvantages of BVTThe use of an autoclave for the sterilization/depyrogenationof containers (plus a previous intensive washing of the vial)limits the BVT aseptic process to batch production (as op-posed to continuous production).

The autoclave must be designed according to the expectedmaximum batch size.

It should be observed that in order to minimize the timeduring which the vial is open and exposed to the environment,the three operations: unstoppering – filling – stopperingmust be performed sequentially (see sections A-A, B-B, andC-C in Figure 4). This means that unstoppering the vial forfilling and stoppering the vial for sealing increase the cycletime. Therefore, the throughput of the machinery designedfor BVT aseptic filling is slower than conventional fillinglines.

These two disadvantages make the BVT aseptic processsuitable for small/medium batch sizes (from 20 vials up to11,000 vials).

Further DevelopmentsExperimental studies and tests are being performed to inte-grate the protector element in the capsule.

As described in a previous section, the containers used inthe BVT aseptic filling process are made up of three elements:a protector, capsule-stopper set, and vial. It also was men-tioned that the capsule was standard and that the vial wascrimped by means of standard machinery.

A plastic capsule which clips to the vial has been designed.The special design of this plastic capsule is long enough toplay the role of the protector performing the “labyrinth seal.”

The three main advantages of this protector-capsule are:

• reduction in the number of components being handledalong the BVT aseptic process

• substitution of the crimping machine with a simple pressto clip the capsule-stopper set in the vial and consequentlyeliminate the particles generated during crimping

• complete sealing of vial in front of a Class 100 horizontallaminar flow immediately after filling

The BVT aseptic filling process also is applicable to the sterilefilling of freeze dried products. After a vial is filled with afreeze-dried product, the protector-capsule, which includesthe stopper, partially stoppers the vial allowing the lyo-

philization process to occur. The shelves of the freeze dryerthen clip the protector-capsule, securely closing the vials.Therefore, the use of the protector-capsule for freeze driedproducts eliminates the need for crimping found in a conven-tional manufacturing process.

These most recent advantages meet the requirementsstated in the recent proposed revision (approval pending) toAnnex 1 of EC Guide to Good Manufacturing Practices: “Thecontainer closure system for aseptically filled vials is not fullyintegral until the aluminum cap has been crimped into place.Vials should be maintained in a Grade A environment untilthe cap has been crimped”7 (clause 93).

There is a BVT specifically adapted for the aseptic fillingprocess in extremely small batch sizes of personalized medi-cines. In these cases, the drug product losses are minimizedthanks to the full drainability of the filling system.

Furthermore, the system can be mounted and installed ina modular cleanroom delivered and pre-validated prior to thefactory acceptance test.

ConclusionBarrier Vial Technology (BVT) is an aseptic processing ap-proach for high value-added pharmaceutical products, suchas biotech medicines, plasma derivatives, and others whichare not stable enough to undergo final product sterilizationby heat. BVT is applicable both to liquid and freeze-driedproducts.

Although some similarities with conventional aseptic pro-cessing exist (vial washing process, integration of any fillingsystems - time pressure system, weight control, peristalticpump, etc.), BVT increases microbial safety of asepticallyprepared products and maximizes the exclusion of particu-lates from all phases of aseptic processing.

BVT offers advantages such as particulates and microbialsafety, environmental control, traceability, anti-counterfeit-ing, and lower facility costs when compared to conventionalprocedures.

BVT has been employed for the last 20 years in a plasmaderivatives factory approved by the FDA and Europeanauthorities for the production of medicines.

References1. U.S. Department of Health and Human Services (Food

and Drug Administration), “Guidance for Industry: SterileDrug Products Produced by Aseptic Processing - CurrentGood Manufacturing Practice,” p. 17, September 2004.

2. ISPE Baseline® Pharmaceutical Engineering Guide, Vol-ume 3 - Sterile Manufacturing Facilities, InternationalSociety for Pharmaceutical Engineering (ISPE), FirstEdition, January 1999, p. 23, www.ispe.org.

3a. 3b. Ad Hoc GMP Inspections Services Group (EuropeanCommission), “Manufacture of Sterile Medicinal Products- Medicinal Products for Human and Veterinary Use,” EUGuidelines to Good Manufacturing Practice, May 2003,Vol. 4, Annex 1, clause 3, clause 33.

4. Instituto Grifols, “Evaluation of the Contamination ofSterile Vials to Different Exposure Conditions,” IG_ITEC-



©Copyright IS

PE 2007

000244_ING, internal report, 2006.5. Instituto Grifols, “Demonstrating the Effectiveness of

Physical Barriers in Sterility Assurance of Vials Subjectedto Airborne Microbial Challenge Test,” IG_ITEC-000244_ING, internal report, 2006.

6. Whyte, W. “Sterility Assurance and Models for the Assess-ing Airborne Bacterial, Contamination White,” Journal ofParenteral Science and Technology, Vol. 40, No. 5, Septem-ber – October 1986, pp. 1996.

7. European Medicines Agency – Inspections, “GMP Annex1: Proposals for amendment to the environmental classi-fication table for particles and associated text, amend-ment to section 42 concerning acceptance criteria formedia simulations, amendment to section 52 concerningbio-burden monitoring, and additional guidance in section88 on the sealing of vials,” September 2005, clause 93, pp.4.

AcknowledgementsThe authors of this article would like to thank Dr. VictorGrifols, inventor of the Barrier Vial Technology, for hiscomments and suggestions during the writing of this article,as well as Instituto Grifols, for the information provided onthe validation of this technology.

About the AuthorsDiego López -Álvarez, EMBA from ESADEBusiness School, MSc, industrial engineer, isresponsible for the Applied Engineering De-partment at Grifols Engineering and is in-volved in the development, design, and inno-vation of tailor-made machinery for the phar-maceutical companies of Grifols Holding andthird-parties. He has led, promoted, coordi-

nated, and installed a number of aseptic filling processinglines in European and American facilities, applying BarrierVial Technology (BVT) for high added value products (liquidsand freeze-dried): therapeutic proteins from human plasma(coagulation factors, albumin, etc.). Among his last assign-ments, he managed and planned the design, construction,and start up of several fully automatic lines for the filling ofterminal sterile products (parenteral solutions), blood typingreagent cards (gel products), as well as machinery for themolding of plastic bags for medical use. López has contrib-

uted to the improvement of a number of pharmaceuticalmanufacturing processes such as vacuum inspection in sealedglass containers and aseptic filling. In addition, he has doneconsultancy in the field of advanced machinery and lasermarking for important companies within the pharmaceuticalindustry. He can be contacted by e-mail at [email protected].

Sergi Roura Adell, Industrial Engineer, isManaging Director of Grifols Engineeringand holds a PPD from IESE Business School.He is currently running an engineering com-pany specialized in biotech process and ma-chinery development. Roura is responsiblefor the engineering technology of the GrifolsHolding. His areas of expertise include fin-

ished pharmaceutical manufacturing, biotech processes, andclean process utilities (Pure Steam, Water-For-Injection,USP Purified Water, CIP, sterile filling, etc). Roura has beeninvolved in the design and construction of manufacturingplants worldwide, mainly for biotech sterile products. Prior tothat, Roura worked as project engineer in ATC in Los Ange-les. He is a member of ISPE since 1994 and is the Presidentof the ISPE Spanish Steering Committee. He can be con-tacted by e-mail at [email protected].

Grifols Engineering, 2, Can Guasch, 08150 Parets delVallès, Spain.

Juan Antonio Garcia is Vice President ofManufacturing of Grifols Biologicals, Inc. Heholds a degree in pharmacy and an MBA inthe Escuela de Administracion de Empresas(Barcelona). As Head of the Validation De-partment of Instituto Grifols for more than10 years, he acquired a deep knowledge ofsterile filling processes, sterilization, clean-

ing, and control of cleanrooms, virus inactivation processes,as well as pharmaceutical process machinery. Since January2004, Garcia has held the role of Vice President of Productionin Grifols Biologicals, leading the manufacturing of the plasmaderivatives plant in Los Angeles. He can be contacted by e-mail at [email protected]

Grifols Biologicals, Inc., 5555 Valley Blvd., Los Angeles,California 90032.

Manufacturing Vision Changes


©Cop

yrig

ht IS

PE 2

007





This articlepresents thechangingmanufacturingenvironment andhow companiescan develop aninfrastructure tocontinue tomeet theirstrategicobjectives.

Pharmaceutical Manufacturing:Linking Vision and Decision-Making toAchieve a Roadmap Toward cGMPs forthe 21st Century

by Beatrijs Van Liedekerke and Ingrid Maes

Introduction

Despite the innovatory and advancedscience nature of many of its products,the pharmaceutical industry has beenmore used to incremental change in

manufacturing rather than quantum leap ad-vances. Now, however, there is the prospect ofmore rapid change in the industry. Changes inthe regulatory stance and compelling businessreasons are prompting companies to consider‘big leap’ rather than ‘small step’ changes. Butmany companies remain wary of drastic change.How can companies judge how best to preparefor the future manufacturing strategy and in-frastructure? How fast and how far should theymove? Many companies are seeking to imple-ment manufacturing change, but are doing so

in sub-optimal ways that do not maximize ben-efit for the company. This is because, often,changes in manufacturing practice and infra-structure are not being informed by a clearmanufacturing vision. Such a vision must ad-dress the regulatory, market, scientific, andtechnological forces that will shape pharma-ceutical manufacturing in the future. Changesin regulation and technology are already influ-encing how existing products are tested. Look-ing ahead, regulatory, scientific, and techno-logical developments have the potential to pro-duce significant change in the interaction ofmanufacturing and the market. This articleconsiders this changing context and looks athow companies can develop a manufacturingvision. It outlines four possible manufacturing

scenarios that companies mayfind themselves considering.The IT/manufacturing infra-structure that will be impor-tant for each scenario is pre-sented.

The ChangingManufacturing

ContextThe pharmaceutical manu-facturing sector has been in-herently conservative in itsapproach to manufacturingchange. Regulation is a keydriver for change. Histori-cally, though, the regulatoryframework, with its relianceon batch inspection, has de-

Figure 1. Moving towardthe manufacturingvision.



©Copyright IS

PE 2007

terred manufacturing innovation.Regulation has driven change, but inan ‘after the event’ fashion with com-pliance reliant on enforcement and in-spection. Now, recent initiatives of theUS Food and Drug Administration(FDA) herald an era where regulation

Case Study 1: Manufacturing Vision Development

BackgroundA pharmaceutical company has a product that will soon run out of patent andgeneric manufacturers are becoming strong competitors. Reducing manufac-turing costs has been defined by this pharmaceutical company as a keybusiness objective.

A Typical ResponseThe company decides to appoint a team of experts whose task is to reviewmanufacturing and propose optimization proposals. After a couple of months,this team presents the cost reduction initiatives to their management. A listof suggestions have been made, such as better planning to remove Work InProgress (WIP) and to lower inventory; optimization of manufacturing yieldsand costs by enlarging the batch size (higher filling levels in manufacturingequipment); in-line inspection instead of manual inspection; and installationof process analyzers to detect batch end-points, for example for drying andblending. The team shows that these measures will deliver a reduction inmanufacturing costs.

A ‘Manufacturing Vision’ ResponseAnother company takes a different approach. Instead of appointing a team tolook for optimizations and improvements, it first organizes a high level meetingwith representatives from a range of departments - R&D, manufacturing,sales and marketing, regulatory affairs. The aim of the meeting is toinvestigate what will be needed in five to 10 years time, taking account ofbusiness challenges, technological options, and regulatory opportunities.

The group has already looked at their current product portfolio and futureportfolio, based on their pipeline. It has investigated the consequences of thisnew portfolio on the current manufacturing infrastructure. It has consideredwhat the future manufacturing landscape will look like to be able to cope, notjust with the new product portfolio, but also with the future market andenvironmental requirements, business model requirements, regulatory changes,etc. A scenario planning exercise has supported the exploration of possibilitiesand future scenarios. This study results in the identification of a manufactur-ing vision, which describes the future required manufacturing landscape thatwill best fit with the most likely scenarios.

This vision makes it easier to identify the gaps between the current “as is”manufacturing situation and the future “to be” one. It also helps to indicatethe improvements and changes that the company can already start toimplement. A roadmap linking the “as is” and the future “to be” situationenables the company to focus on the improvement and optimization projectsthat help it move to the future situation. The company can avoid investmentswhich, taken in isolation, might have a sufficient Return On Investment (ROI)to implement, but when looked at in a fuller context, would not achieve a moresustainable advancement for the company. This broader perspective enablesthe company to move forward in the knowledge that it is not just investingin little islands of optimizations, but is linking them to a wider and biggerquantum leap forward.

can act as a more dynamic driver ofchange with both quality and regula-tory compliance ‘designed in’ to themanufacturing process. The FDA’s PATframework and its cGMPs for the 21st

Century initiative provide significantopportunities for improvement and in-

novation in pharmaceutical manufac-turing. The FDA talks about a ‘desiredstate’ of manufacturing with:

• product quality and performanceachieved and assured by design ofeffective and efficient manufactur-ing processes

• product specifications based onmechanistic understanding of howformulation and process factors im-pact product performance

• an ability to affect continuous im-provement and continuous “realtime” assurance of quality1

The final report of the FDA’s cGMPsfor the 21st Century Initiative2 high-lights the choices that pharmaceuticalcompanies face:

“At the end of the cGMP initia-tive, the pharmaceutical commu-nity has arrived at a cross-road;one path goes toward the desiredstate and the other maintainsthe current state. The path to-ward the desired state is unfa-miliar to many, while the currentstate provides the comfort of pre-dictability. The Agency hopes thepharmaceutical community willchoose to move toward the de-sired state.”

This new regulatory approach presentscompanies with the possibility of newmanufacturing visions. It also comesat a time when the risk reward contextfor pharmaceutical manufacturing ischanging. Companies are becomingmore exposed to powerful wider mar-ket forces. The pharmaceutical indus-try is at a key turning point in manyrespects. Historical ways of deliveringvalue will not be sustainable on theirown in the future. All the key planks ofvalue are in transformation – drugdevelopment pipelines are drying out,pricing is under pressure, and genericcompetition is more intense. Cost con-tainment is the name of the game bothfor the government customer bodiesthat play a lead role in the pharmaceu-tical market around the world and theprivate insurance customers in mar-



©Cop

yrig

ht IS

PE 2

007

kets such as the US. Double-digit salesand income growth has come to an endunder pressure from patent expira-tions, generic competition, and OverThe Counter (OTC) switches.

Alongside these trends, we are notso far from a future where it will bepossible to develop drugs that are tai-lored to the individual genetic andproteomic profile of the patient, mak-ing the therapy more effective and hav-ing less side-effects by optimizing dos-age and drug composition for each pa-tient. An investigation by the nationalacademy of science of the UK concluded:“personalized medicines; tailoring drugtreatments to a person’s genetic pro-file, also known as pharmacogenetics,have a promising future,”3 predictingthat “over the next 10 to 20 years, weexpect to see several pharmacogeneticproducts enter mainstreamhealthcare.”4 The report pointed outthat “industry will continue to favordrug candidates that avoid the effect ofgenetic variation, but where that is notpossible, the development of drugs withan associated diagnostic test is expectedto become routine in the next 10 to 20years.”5 In part, mainstream pharma-ceutical M&A companies have reflectedthis future with repeated acquisitionsof biotechnology companies. Thesemoves have been designed to boostdrug pipeline portfolios in the short tomedium term and build capacity for amore genetically-driven industry of thefuture in the medium to long term.

Such a future is very relevant to acompany’s manufacturing vision. As aconsequence, drugs will need to bemanufactured or produced in smallerbatches that are formulated on requestto match the profile of certain segmentsof patients or even a single patient.There will be fewer big blockbuster drugsand more personalized medicines. Toaccommodate these changing produc-tion needs, new flexible regulatory ap-proaches and batch control strategieshave to be developed. Moreover, sincethe treatment is formulated on requestand is intended for a patient who mayurgently need the medication, productdevelopment and manufacturing leadtime and release times will have to bedrastically reduced.

Developing aManufacturing Vision

Therefore, pharmaceutical manufac-turers face a complex and in some re-spects, contradictory set of demands.On the one hand, they have the oppor-tunity to make significant investmentsin automation and process technology,but on the other hand, they face costpressures, meaning that such invest-ments must deliver the maximum ben-efit. They face a future drug marketthat may be more personalized, posingkey dilemmas for whether the manu-facturing plant development should belarge scale or small scale.

Mergers and acquisition activity hasmade it easier for some companies toclose or modify existing outdatedplants. In our practical experience, wesee companies starting a lot of invest-ment projects both as part of post ac-quisition activity and elsewhere. Theyare called various names, such as im-provement projects or cost containmentprojects, but they have in common theaim of manufacturing modernization.However, they are rarely informed by areal look at the bigger picture of wherethe company wants its manufacturingto be in five to 10 years time (see CaseStudy 1). Classically, when companiesconsider investment in Process Ana-lytical Technology (PAT) for example,they often see it as replacing one formof testing with another form of testingwithout considering its full potential.No wonder Dr. Ajaz S. Hussain, who atthe time of being quoted was DeputyDirector at the Office of Pharmaceuti-cal Science CDER at the FDA, wasprompted to remind companies: “you’vegot to remember that PAT is not aboutjust throwing in-line sensors at a pro-duction line. It is more about under-standing the sources of product vari-ability during production and control-ling your processes in a flexible way toallow you always to produce a qualityproduct.”6

Investment tends to be on a limitedscale and fragmented, focusing per-haps on one production unit or process,but not making connections across themanufacturing software and infra-structure which, often, remains stand-ing alone or only present on isolated

production units. This often results insub-optimizations instead of an over-all optimization. In the future, the re-quirement will be for all the support-ing software and different applicationsto be interconnected. As Graham Cooke,Director Technology and External Sup-ply EMEA of Wyeth, has emphasized,companies need to avoid developingisolated islands of innovation: “’Islands’of PAT (need) to be tied together aspart of an overall strategy. Feed backand feed forward controls. (Companiesneed to) develop the ‘integrated plan’first and then create focus and divedeep into individual unit operationsbefore extending to other unit opera-tions.”7 In addition, whether it is PATor other innovation, the infrastructurewill need to be of high quality andreliability because the recourse to run-ning the production manually will notbe an option.

How can companies judge how bestto reshape their manufacturing strat-egy and infrastructure? In the contextof PAT, Cooke emphasises the need for‘wider company’ multi-disciplinarythinking: “…a number of success fac-tors have been identified for imple-mentation of PAT. These include theneed for multi-disciplinary projectteams, a clearly defined implementa-tion process, and a strong businessrationale.”8 Companies need to addressthe culture change implications of in-vestments such as PAT which includebreaking down silos within organiza-tions and also rethinking job roles.Far-sighted companies seeking to cap-ture the full competitive advantagepotential of PAT will, for instance, belooking at the links outside of manu-facturing into the consumer-facingfunctions of product development andmarketing. Skill-set requirements willchange significantly. Enterprise-widedata management, retrieval, and que-rying will be vital. Pharmaceutical sci-entific skills will need to extend intounderstanding the supportive databasestructure and be capable of managingknowledge retrieval systems in an effi-cient, usable, and timely manner.

In our view, the starting point has tobe the manufacturing vision and allparts of the business need to be in-



©Copyright IS

PE 2007

volved in looking ahead on a 10 to 15year time frame. The following caseillustration highlights the importanceof framing decisions in such a contextand contrasts that with the typicalapproaches that we, as authors, seemany pharmaceutical companies tak-ing.

The approach outlined in Case Study1 allows companies to prioritize spe-cific problems within the context oflong-term change. The range of specificconcerns could include a need to fix orimprove existing processes, speed upnew product development, reduce site

to site transfer risk and times, reducevalidation costs, or improve qualityreliability. Most companies are likelyto want to realize a blend of thesebenefits. Their immediate prioritieswill be determined by the current stateof play of their manufacturing and itsfit with their regulatory compliance,market and business goals. Most im-portantly, though, they need to com-bine this review of current wider con-cerns with the type of longer-term widerscenario planning outlined in the caseillustration above. Figure 1 outlinesthe steps companies might take to put

this process into practice.Figure 2 provides an overview of the

type of overall decision-making pro-cess that a company needs to under-take. The current manufacturing in-frastructure has to be assessed in thelight of the future manufacturing vi-sion (in line with the global company’sobjectives). What are the current bottle-necks and what are the improvementpossibilities? The resulting list of im-provement proposals have to be evalu-ated to judge just what they bring tothe company and whether they helpachieve the manufacturing vision andits objectives. Depending on whichmarket the company is in, the regula-tory constraints need to be superim-posed in order to make sure no sur-prises are encountered. Even for thosecountries that are actively drivingchanges (such as the FDA in the US), itis important to involve the regulatorsearly on in the process.

Four Change ScenariosThe outcome of this type of process willbe a view about what type of manufac-turing strategy and plant the companyneeds in a more medium to long termtimeframe, say five to 10 years time.The answer may be different from plantto plant and many companies are likelyto need to plan for a mix of scenarios.For example, a company may choose toimplement relatively modest improve-ment investment in a plant that ismanufacturing a product that is near-ing the end of its patent period (sce-nario one in Figure 3). Elsewhere itmay choose to plan for a rapid and fullscale move to PAT enabling full real-ization of the FDA’s vision of real timeproduct control and release, based oncontinuous manufacturing operations(scenario 2 in Figure 3).

Companies also will be mindful thata possible trend toward more personal-ized medicines will increase manufac-turing complexity, and in turn, posechallenges for Manufacturing Execu-tion Systems (MES) and quality sys-tems. A larger variety of products andvariation of the same products willrequire greater flexibility of produc-tion as well as closer integration alongthe whole pharmaceutical chain - R&D,Figure 3. Four change scenarios.

Figure 2. The pharmaceutical manufacturing change context.



©Cop

yrig

ht IS

PE 2

007

manufacturing, sales, and the end cus-tomer.

Scenarios three and four in Figure 3highlight how companies will face achoice between big plant with flexiblerecipe production versus small-scaledevelopment (pilot) plants which alsowill be production facilities with dedi-cated lines. For both models of produc-tion, industrial IT systems will play astrategic role, requiring tremendousflexibility, in the first model, to sup-port the flexibility of production thatwill be necessary, and in the secondsmaller scale model, to link productionwith continuous development andlearning from clinical trials. The regu-latory stance will be a key factor in thismix and at present, regulators are in-vestigating how to support this evolu-tion with the appropriate regulationsand guidelines.9

A key influence will be the demandside and we are likely to see a mix oflarge scale, very high throughput fa-cilities handling generic production,and micro-process centers concentrat-ing on higher end personalized medi-cines. Therefore, pharmaceutical com-panies need to investigate the invest-ment in planning for a potentially verydifferent manufacturing future as wellas responding to pressures on theircurrent manufacturing set-up.

Choosing BetweenScenarios –

Evolution or DrasticChange?

A critical issue for companies contem-plating scenarios such as outlined inFigure 3 will, of course, be how to makechoices between them. The identifica-tion of the right evaluation criteria(Key Performance Indicators (KPIs) forimprovement) is crucial for evaluatingthe options and for monitoring progressand achievement of the objectives. Eachcompany’s situation will be differentand judgements on the focus and paceof change will vary according to theROI analysis of the different optionsopen to them. For example, some com-panies may consider that certain plantsor processes do not merit investment,others will only need minor invest-ments and others require drastic

Case Study 2: Status Quo vs. Automation vs. Full PATImplementation in a Vaccine Plant

BackgroundA vaccine plant was seeking to achieve cost savings through modernizationof manufacturing infrastructure. Interviews with different stakeholders andanalysis of manufacturing data led to:

• the identification of areas for cost savings through the assessment ofpossible improvement scenarios

• an outline of operational and financial benefits for these various scenarios• assessment of the impact of different scenarios on the following KPIs:

- labor (people) - waste- manufacturing throughput time - inventory levels- quality

Improvement ScenariosThree improvement scenarios were identified. Each of these scenariosdescribe the various steps toward optimal PAT-enabled manufacturing,delivering the maximum benefits in terms of cost savings.

The scenarios are built up in such a way that maximum benefits are realizedwith minimal investments. They start with the quick wins followed by asequence of medium to longer term improvement investments. Each improve-ment investment goes hand in hand with benefits which are displayed as aneffect on the Key Performance Indicators (KPIs).

• Some of the scenarios can be executed in parallel; however, when activitiesare carried out in parallel, the necessary skilled resources need to be availablein order to deal with the complexity and the project management.

• A timeline was developed illustrating how much time it takes to implementthe improvements as well as the resources and skill set needed for eachof the improvement projects. The time to get regulatory approval shouldbe superimposed on the outlined project execution time lines.

• In parallel with the timeline, the sequence of investments needed to realizeimprovements was established.

ResultsThe result was a calculation of the optimal scenario (in this case, scenario 3)and its impact on the KPIs:

• Labor: 1/4 of operations people could be re-allocated and 1/3 of the QA/QC people could be freed up for other work.

• Manufacturing throughput time: throughput time decreased with 1/3freeing up capacity and allowing extra production with the same headcount.

• Quality: 13% of the cost of QA and QC are eliminated because ofimprovement in right first time.

• Waste reduction: 3.5%• Inventory: inventory could be reduced by 1/3 (representing about US

$14.3 million in this case).

ObservationsIn terms of PAT implementation, maximum benefits were achieved with abroad PAT definition. This means looking at the full opportunities offered byPAT, as outlined in the FDA PAT Guidance (e.g., real-time product release,manufacturing performance improvement, quality consistency improvement,and regulatory flexibility). This was preferable to a “limited PAT” approachbased only on the implementation of an on-line sensor. We found that thefeasibility of a broad PAT enabled manufacturing process could be demon-strated with much more certainty.



©Copyright IS

PE 2007

Figure 4. Impact of scenario implementation on various KPIs.

Figure 5. Manufacturing infrastructure scheme.

change.Even in the case of drastic change, it

is the authors’ experience in many real-life cases that a change, which at firstsight may appear quite drastic andassociated with big investments, canbe shaped into smaller pieces, solvingat the same time some technical is-sues. This allows a step-by-step invest-ment and implementation with eachstep having a ROI case, providing jus-tification of the investment. The com-

pany, although taking small steps, isdoing so in the context of a journeytoward a manufacturing infrastructurewhich meets the future business chal-lenges. This will enable companies tobe ready for the possible future busi-ness scenarios and to take advantageof adopting new technologies early. Thecritical elements are the selection ofthe improvement options, the identifi-cation of the right KPIs, the size andsequence of the steps, and last but not

least, the fit of the future manufactur-ing vision with the possible future busi-ness landscape. Case Study 2 illus-trates how this might work in action ina vaccine plant.

ManufacturingInfrastructure

Once they have chosen between differ-ent possible manufacturing visions andcompleted some scenario planning,companies will, of course, need to de-cide on the manufacturing and IT in-frastructure that will be required forthe chosen scenario. Decisions aboutthe future architecture will differ be-tween the various scenarios, and cru-cially between those with smaller sizeprocess equipment and larger scalemanufacturing. As an example, Figure5 outlines a manufacturing infrastruc-ture scheme corresponding to scenario2 of Figure 3. The PAT solution hasinterfaces to the process equipment,the process automation, and will takecare of data collection from the process,eventually from extra real-time mea-surements (PAT Analyser) as well asdata storage and retrieval. It consistsalso of an MVDA engine able to inter-pret quality data and translate thisinto control and correction actions. Thehigh level PAT solution will combinevarious unit operations and will takecare of the overall product release ofthe final product.

In general, the role of the qualitymanagement system will shift to themanufacturing floor and will be of morestrategic importance, as it is essentialfor real-time product release. Greaterintegration of multi-disciplinary teamswill be an important factor alongsidethe hardware and software. The qual-ity management system will consist ofa LIMS system and PAT systems (onunit and on line level). It will allowProduction Performance Analysis(PPA). In turn, for faster time-to-mar-ket, a closer link between developmentand manufacturing is required thatallows for continuous improvement.Figure 6 outlines the wider architec-ture that is needed. A central role willbe occupied by knowledge managementsystems and data portals, but also byadvanced data mining techniques. The



©Cop

yrig

ht IS

PE 2

007

changes are made in relative isolationwithout maximizing their potentialincremental contribution to longer termimprovement or, worse, moving thecompany further away from the manu-facturing it will need in the future.

We have shown how companies canuse a range of tools – scenario plan-ning, ROI analysis, KPIs – to constructsuch a roadmap to ensure changes arelinked together, thereby avoiding piece-meal and sub-optimal change. There isa need for companies to more consis-tently align investment in IT and manu-facturing with their vision of the manu-facturing that will be needed in thefuture. In doing so, companies will beable to ensure that investments don’tjust deliver specific gains, but also helpaccelerate the company’s progress to-ward longer term goals.

References1. Innovation and Continuous Improve-

ment in Pharmaceutical Manufac-turing: The PAT Team and Manu-facturing Science Working GroupReport: A Summary of Learning,Contributions, and Proposed NextSteps for Moving Toward the “De-sired State” of PharmaceuticalManufacturing in the 21st Century.http:/ /www.fda.gov/cder/gmp/gmp2004/manufSciWP.pdf.

2. Pharmaceutical cGMPs for the 21stCentury - A Risk-Based ApproachFinal Report. Department of Healthand Human Services, US Food andDrug Administration, September2004. http://www.fda.gov/cder/gmp/gmp2004/GMP_finalreport2004.htm#_Toc84065738.

3. The Royal Society, PersonalisedMedicines: Hopes and Realities,

role of knowledge management sys-tems and data portals will be essentialfor this change.

ConclusionA combination of regulatory, market,scientific, and technological forces islikely to mean that pharmaceuticalmanufacturing will undergo rapidchange in the next five to 10 years.Many companies are already investingin change projects, but they are oftenpiecemeal and not accompanied by aclear manufacturing vision. The ab-sence of such a vision also means thatcompanies sometimes feel caught be-tween ‘big leap’ and more incrementalchanges. In fact, incremental change isvital to achieve a longer term ‘big leap.’But, in the absence of a manufacturingvision, companies find themselves withno roadmap. The consequence is that

Figure 6. Tighter integration of development, manufacturing and knowledge to achieve continuous improvement.



©Copyright IS

PE 2007

press release, 21 September 2005,ref 18/05.

4. The Royal Society, PersonalisedMedicines: Hopes and Realities,September 2005, p 2.

5. ibid. p 1.6. Quality Manufacturing: A Block-

buster Opportunity for Pharmaceu-ticals, Economist Intelligence Unit,2005.

7. PAT Strategy: Islands of PAT to beTied Together, Graham Cooke, Eu-ropean Pharmaceutical Review, is-sue 5, 2006.

8. Ibid.9. Innovation Stagnation: Challenge

and Opportunity on the Critical Pathto New Medical Products, US De-partment of Health and HumanServices, Food and Drug Adminis-tration, March 2004.

About the AuthorsBeatrijs VanLiedekerke is Asso-ciate Director atPriceWaterhouseCoo-pers, China. VanLiedekerke is a phar-macist and has a PhDin the pharmaceutical

sciences. She has industry experiencein the food, biotech, and pharmaceuti-cal industries where she held posi-tions in QA, QC, manufacturing, R&D,

and customer relations. Over the lastfour years, she has been an adviser tocompanies on operational improve-ment, overall business optimization,and regulatory compliance. She canbe contacted by e-mail at: [email protected].

PriceWaterhouseCoopers, BeijingFortune Plz, 26/F Office Tower A 7,Dongsanhuan Zhong Rd. ChaoyangDist, Beijing 100020, China.

Ingrid Maes is SeniorConsultant in thePharma CompetenceCenter of SiemensA&D. She has an engi-neering background inthe chemical and agri-cultural industries

with a specialization in biotechnology.Through different positions at Siemens,Maes has a wide expertise in processanalytical tools, multivariate dataanalysis, automation, and control ofmanufacturing processes in the foodand feed, biotech and (petro) chemicalsectors, as well as in the pharmaceuti-cal industry. She can be contacted by e-mail at: ingrid.maes@ siemens.com.

Siemens AG HQ Pharma Compe-tence Centre Pharmaceutics - Consult-ant Advanced Technologies, NieuweWegl, B-2070 Antwerpen (Zwijndrecht),Belgium.

Process Description Application


©Cop

yrig

ht IS

PE 2

007





This articledescribes how adetailed ProcessDescription (PD)can help toalleviate manyof the pitfallsthat areencountered inthe automationof a lifesciencesprocess.

Using a Process Description to DefineAutomation Needs for a Life SciencesProject

by Steve Murray, Amit Shah, and Dawn Marruchella

Figure 1. Basic GAMP“V”.

The Need for a ProcessDescription

A detailed Process Description (PD) canhelp to alleviate many of the pitfallsthat are encountered in the automa-tion of a life sciences process, the ma-

jority of which are batch processes. Deliveringa properly automated process solution on sched-ule and within budget can be challenging onmost projects, even when requirements arewell defined and re-work is minimal.

A properly structured PD with appropriatedetails can help the automation team writeFunctional Design Specification (FDS) docu-ments that effectively achieve the followingcritical business goals:

• Gather together the knowledge and insightsof laboratory science, process design, qual-ity control, and pilot plant personnel into asingle location.

• Translate this expertise into automationrequirements that define how a productshould be made consistently and repeatablyin compliance with regulatory requirements.

A thoroughly reviewed and mutually acceptedFDS allows the automation team to develop ahighly modular specification and design struc-ture. This modular structure can lead to acomponent-based automation software appli-cation that allows consistent reuse betweenplants and processes, improves built-in qual-ity, and speeds software development, whileminimizing the addition of cost and/or resources.Furthermore, modular software reduces test-ing, commissioning, qualification, and mainte-nance requirements.

Understanding the complete process require-ments for a batch project allows the automationteam to take their knowledge of automationcontrol systems and the ANSI/ISA S88 stan-dard for batch control1 and apply it to the

process in the most ef-ficient manner. Theresulting automationdesign not only meetsthe requirements forthe current process,but also allows for fu-ture flexibility and fol-lows a common set ofstandards across theproject.

On the contrary, ifthe PD is poorly writ-ten without input fromall relevant groupsand lacks clear re-quirements, the auto-



©Copyright IS

PE 2007

mation team is forced to make assumptions, while writing theFDS documents. This typically leads to several iterations ofthe FDS documents with numerous communications backand forth or, worse yet, some details not being accounted foruntil much later in the project execution when changes aremore painful and costly. In determining commonality acrossthe project, missed details will cause greater problems. Ifsufficient time is not taken, the chance for success in meetingschedule, budget, and product quality demands and also inminimizing stress on personnel will be significantly de-creased. Starting slowly at the beginning of a project will payoff many times over by reducing rework and speeding imple-mentation. A deliberate start also will take advantage of theinherent interdependency among the many contributors.

This article is intended to serve as a guide for members ofprocess design teams needing to communicate process auto-mation requirements to their automation team. While thelevel of automation may be adjusted somewhat in the FDSportion of the project, there needs to be a documented hand-over of the process automation requirements and a definedprocess for communicating requirement changes that mayoccur after the PD is issued by the process design team to theautomation team.

There is a misconception that if a PD is created, it mustalways be a “living” document that must be managed for thelife of the system, using change control procedures. This isactually not the case - those familiar with GAMP GoodAutomated Manufacturing Practice Guide for Validation ofAutomated Systems 2 model will recognize Figure 1 as GAMP’sbasic framework for the specification and qualification of anautomated system. This model addresses the lifecycle docu-ments that need to be created, but it never details how tocommunicate the information required to generate thesedocuments. A PD document is a very efficient and many timesa necessary document to develop automation FDS. Depend-ing on the company’s needs, the PD may be maintained underchange control as a living document or obsoleted by the FDS.

What is a Process Description?Sometimes called a process narrative, a Process Description(PD) is a well organized, detailed account of what the process

does and how it works that accompanies Process Flow Dia-grams (PFDs) and Piping and Instrumentation Diagrams(P&IDs). One of the goals of the process description is tocreate a clear understanding between the process designteam and the automation team about the process controlrequirements for a particular process or process area withina facility. PDs can serve as the origin of requested processchanges if the document is maintained as an ongoing commu-nication tool, and can serve many other roles in an organiza-tion. PDs have been successfully used in other industries toserve this purpose, but the challenge in the Life Sciencesindustries is to determine the role that the document servesin the project lifecycle (if any).

Providing too much detail and/or making too many designassumptions in the PD can lead the automation FDS for thatprocess or process area down a path that may not produce thebest automation solution for the process. This frequentlyresults in inadvertent grouping of sequences considered bythe process design team to be similar, but that really have toomany differences to be implemented as the same softwareelement. It is best to let the process design team describe theprocess requirements and then let the automation teamdetermine the best way to automate these requirements. It isimportant to note that the authors of this article do not wishfor the term automation FDS to be confused with automationcontrol system requirements that would normally be pro-vided in a User Requirements Specification (URS).

A well written PD should contain elements that address,at a minimum, the following:

• objective• reference documents• overview of the process• general process flow information• general and specific equipment and instrumentation in-

formation• auxiliary systems requirements• sequence of operations• continuous control requirements• equipment scheduling requirements• detailed narrative of the process describing (where appli-

cable):- equipment states- equipment cleaning and sterilization requirements- sequencing requirements- normal operating conditions- abnormal operating conditions- input parameters- report parameters- critical parameters- alarms- operator interactions- process interlocks and permissives- time critical operations

An example of portions of a process description is included atthe end of this article.

Figure 2. An optimal method to develop automation functionaldesign specification.



©Cop

yrig

ht IS

PE 2

007

This information can all be provided in the User Require-ment Specification (URS) documents called out by the GAMPstandard, but this is generally an inefficient way to managethe process information. GAMP suggests, but does not dic-tate, that the content of the URS include an OperationalRequirements section. The GAMP guide directs that “processdescriptions or flowcharts may be included as appropriate.”For a small project, the URS may be used to communicate allof the process requirements since the document in such casesis usually small and developed by a few people. For a largeproject, the process requirements are more complex; andmaintaining and controlling the specifications as part of thesystem lifecycle can be an onerous effort. It can be inefficientto include process information since it is repeated in thefunctional design specification. As the functional designspecifications evolve through the project and through theyears of operation of the facility, there is a large amount ofduplicated effort in maintaining the same information inboth the FDS and the URS. The URS and the FDS togetherneed to provide the basis for the acceptance of the system -and should contain a minimal amount of repeated informa-tion. Therefore, it is the authors’ opinion that process require-ments should be written in a Process Description documentindependent of the URS.

Although the process design team, who is responsible fordeveloping the PD, may possess detailed knowledge of theS88 batch architecture, the PD should define little, if any, ofits structure. The PD, along with the other inputs shown inFigure 2, are most likely sufficient to produce a detailed FDS.Typically, the PD is written by the process group using, alongwith the information provided by the equipment vendors, theprocess specific applications as well as input from the QA andvalidation groups. It also is important to provide for theinvolvement of QA and validation during PD since they haveunique automation needs that need to be included as part ofthe requirements. For example, operator prompts that com-ply with 21 CFR Part 11 are clearly a validation requirementthat many times fail to be noted as a requirement early in theproject. Once the PD is written, the automation FDS isdeveloped using the URS, PD, and the P&IDs as shown inFigure 2.

As PDs and other inputs are considered, the appropriateS88 architecture will evolve as a result of the FDS develop-ment process. The benefit of permitting the S88 architectureto evolve in this manner is that the automation team is likelyto be more knowledgeable about the S88 model and itsapplication to projects. A major benefit of such an evolution-ary process is that software classes and templates can becreated, tested, validated, and then re-used with minimaladditional testing. Not only does this create a robust controlsolution, it creates a solution that will be easier to maintainin a validated state over time. Once completed, the FDSbecomes the primary communication mechanism among theautomation group members.

For additional information on the S88 standard and whatan S88 batch architecture might look like, the authors sug-gest visiting the ISA (www.isa.org) or WBF (formerly known

as the World Batch Forum) (www.wbf.org) Web sites wherethe various parts of the standard may be purchased and whitepapers on the standard can be downloaded, respectively.

Incomplete and/or ambiguous PDs often produce one oftwo scenarios: either they produce an ill-defined project thathas costly changes or schedule delays; or they lead to lengthydiscussions and/or incorrect assumptions during later phasesof the project. The process of waiting for answers to questions,while creating the FDS can cause project delays and/or re-work. Moreover, it can produce anxiety and frustration fromwithin the end-user organization. Not only are end-usersdistracted from other tasks, but they also begin to questionwhy the automation team cannot ask all the questions atonce, and may call into question the value of the automationteam.

Exception Handling Routines VitalMany PDs provide a list of equipment capabilities as per theequipment vendor documentation describing how the equip-ment operates. What’s missing is how the equipment will beused for a particular process application. Furthermore, ap-plication related information should enhance the operationalrequirements with any operating parameters and the legalranges that operators or batch recipe structures are permit-ted to modify.

Although it is necessary to include extensive documenta-tion of the normal equipment operation, equipment exceptionhandling and the conditions that constitute exceptions areequally important and must be extensively documented inthe PD as well.

Exception handling can require input from multiple groups,including safety, process, quality, and maintenance. In amanual plant, exception handling is covered by the experi-ence and judgment of the operations staff. However, when thefacility is highly automated all exception scenarios, events,and conditions are handled automatically. Thus, abnormalsituations must be identified and defined beforehand. Oftenthe best opportunity to explore and document abnormalsituations is during process Hazard and Operability (HAZOP)studies. In addition, the acceptance testing of automationconfiguration is generally focused more on normal operationsand ill-defined fault scenarios can make it to the qualificationprocess. Of course, correction of them at this point is muchmore costly and might concern the process owner that thingsmay not be properly thought out and fully tested.

Include Global Requirements and StandardsIn large automation projects, there are several commonrequirements and repetitive automation requirements thatspan multiple processing areas. These are sometimes re-ferred to as “global” requirements. For example, equipmentstatus tracking (e.g., clean, dirty, etc.), agitator operation,jacket temperature control, product transfer stations, etc.,are likely candidates for developing standardized solutions.Where possible, these requirements should be grouped in aGlobal (project-wide) PD and then referenced by other PDs.For example, once a standardized jacket temperature control



©Copyright IS

PE 2007

requirement has been defined in detail in the Global PD, thePD for fermentation would simply declare the start/hold/resume/stop of the jacket temperature control.

Another example is that almost every unit in a facilityneeds to perform a pressure test at some point in the process,whether it is prior to performing a sterilization or immedi-ately before performing a pressurized transfer. A section inthe Global PD can detail the requirements for performing apressure test, including indication of what the automationshould do in the event of a failed test. Area specific PDs wouldnow only need to make reference to the Global PD whenindicating the necessity for a pressure test that is the same asthe global definition. Even where there is a special scenario,only the deviations from the general requirements will needto be detailed in the area-specific document. Defining globalrequirements also promotes collaborative effort and a mindsetacross the process areas to standardize wherever possibleallowing for a cleaner, simpler automation solution.

Who Should Write a Process Description –and When?

A process description should be written by the process designteam with the assistance of the automation team. It may beadvisable to involve an automation team at this time if theyhave the background and skills to add significant value to theprocess - but most of the automation team’s effort will be atthe FDS phase. The automation team should provide detailsabout the type of information required from the processdesign team using good PD examples. This will avoid re-workfor the process design team and allow the project to stay on anaccelerated time schedule.

The details of the intended process are often not entirelyclear at the start of the FDS development phase. The goal isto minimize the need for the FDS author, a member of theautomation team, to request clarifications from the PD au-thor, a member of the process design team. This process is notonly time-consuming for everyone involved, but the commu-nication chain allows for miscommunications that can causescope and content disputes at the point of system acceptance.

Creating specifications about scenarios that are not requiredis undesirable. Change orders that result from misunder-standings are never well received and should be avoided.Figure 3 shows the desired document flow of the PD andwhere it fits into the project execution.

There are times when a Process Description may not benecessary for a life sciences project. If a project is smallenough to cover all of the required details in a URS, then itmay be enough documentation for the automation team touse for the FDS development. Furthermore, the automationteam, who would be responsible for authoring the FDS, maywork very closely within the same organization as the au-thors of the Automation and Process Requirements Specifica-tions, thus, eliminating the PD need. The PD is a communi-cation tool - if the communication can take place throughongoing interaction, then the PD would not provide as muchvalue. In contrast, for large projects, the process require-ments, automation requirements, and functional design speci-fications often do not come from closely integrated groups;therefore, there is usually a compelling reason for such acommunication tool.

Where Does a PD Fit into theSystem Life Cycle?

A URS will be written to contain many general requirementsof the system, but will probably not contain the detailedrequirements of how the process is to operate. The detailingof the process requirements to the automation team will thenrequire the supplemental PD to provide the basis for theirFDS efforts. It is important to note that this document doesn’tneed to be maintained past the point where it adds value.

Remember that a PD is an excellent communications tool.Many organizations do not see the need to maintain the PDas a living document under change control. This is especiallytrue if an FDS is prepared with a multi-disciplinary effortbetween Process Development, Automation, Validation, Qual-ity Control, and Operations. Figure 2 illustrates that it ispossible and practical to use P&IDs and associated processdescriptions as source material for developing an FDS, while

Figure 3. Collaborative and iterative process description development.



©Cop

yrig

ht IS

PE 2

007

still working within the recommended GAMP structure. Thisstructure provides a good basis on which to build an S88-based structure that promotes modular design.

The concept of using the PD as a means to an end ingenerating a comprehensive FDS can be a challenging under-taking. The automation team will need to stay vigilant inensuring that any evolutionary changes to the FDS are in linewith the intentions of the process design team. These evolu-tionary changes can result from a number of different origins,including inconsistent PDs across multiple areas; simpleerrors in the PDs themselves; or enhanced software modular-ity goals. While the PD may be approved, an analysis by theautomation team may reveal that the operation of equipmentlike filtration and Clean-In-Place (CIP) skids may be defineddifferently in different areas for no reason other than the factthat they were authored by different process design teammembers at different times. Similarly, the automation teammay seek to reduce the maintenance and training burden ofthe operations staff by standardizing the functionality ofmore common functions, like jacket temperature control. It isin this respect that the automation team can add great valueto the project.

The PD must set the expectations for the level of automa-tion. There are likely to be critical bottlenecks in the processthat need to be highly automated to attain maximum productthroughput. Generally, bottlenecks represent “non-negotiable”portions of the automation process. These should be articu-lated so the automation team has a clear picture of theproduction requirements to sustain the performance of theprocess.

While scope can certainly be reduced during the genera-tion of the FDS documents, it may be beneficial to specify upfront that certain automation is not required to save part ofthe budget for unknowns yet to come.

What Comes After the PD?Once the PD is handed off to the automation team so that theymay write the FDS, the true design of the automation for theprocess begins. For this reason, it is important for the processengineers to note that if there is a controls requirement,whether it is an alarm, a report parameter, an operatormessage, or an interlock, it is essential that the requirementis included in the PD. The PD, coupled with the P&IDs andURS documents are the basis for the automation design of theprocess, and as such should be the basis upon which thecompletion of the automation process is measured.

There is significant benefit in having automation-focusedindividuals involved at this point to interpret the PD andtransform it into a structured FDS that allows for traceabilitythrough a modular design and implementation of the appli-cation software. What makes this a desirable and workablesolution is that the FDS is best created by those most familiarwith the application of each process control requirementwithin the integrated software application solution. Thisgroup generally includes both the application software sup-plier and the automation engineers within the end-userorganization.

Prototyping activities and design guideline documents forthe project are among the first deliverables to be developed.The prototype and resulting guideline documents are animportant part of the project standards. Once approved, thedesign guideline documents and the comments from theprototyping effort are released for implementation of code. Achange to the requirements at this point in the life cycleresults in an increase in cost and more importantly, a delayin schedule. Depending upon the change, other process areasalso may be affected. For example, a change to the CIP Skidhas the potential to affect all of the units for which it cleans.

Upon completion of implementation, internal testing isdone prior to performing the software acceptance testing.When changes to requirements are made at this stage, notonly are FDS documents affected, but automation configura-tion also must be re-worked, test protocols rewritten and re-executed, or existing tests re-executed prior to code release.The increase in cost and schedule due to changes in automa-tion requirements continues to grow exponentially as theproject life cycle progresses.

Sample Excerpts from a Fermentor PDIn an effort to further explain the level of information thatshould be contained in a PD, some selected sample informa-tion from a Fermentor PD is included. These are intended toillustrate the level of detail appropriate to the PD and are notintended to be complete or depicted as the only method ofdeveloping the PD, hence, some of the sections only showoutline items and don’t include detailed information. Addi-tionally, some of the tables are not intended to be complete(for example, instrumentation and equipment tables), butare there to provide enough information to provide generalguidance.

Table of Contents1 Introduction

1.1 Objective1.2 Reference Documents and Drawings

2 Process Overview2.1 Process Flow Diagram

3 Equipment Information3.1 Equipment3.2 Instrumentation3.3 Auxiliary Systems

4 Sequence of Operations4.1 Fermentor Clean-in Place (CIP)4.2 Fermentor Pressure Test4.3 Fermentor Steam-in Place (SIP)4.4 Inoculation4.5 Fermentation

5 Phase Sequences5.1 Sample Bottle Sterilize-in-Place (SIP)5.2 Fermentor Set-Up5.3 Fermentor Pressure Test5.4 Fermentor SIP5.5 Fermentor CIP5.6 Fermentor Inoculation



©Copyright IS

PE 2007

5.7 Fermentor Transfer Out6 Continuous Control

6.1 Pressure Control6.2 Temperature Control6.3 Antifoam Control6.4 pH Control6.5 Dissolved Oxygen Control6.6 Agitator Control

1 Introduction

1.1 ObjectiveThe ABC Manufacturing Facility will be designed tohandle the fermentation and purification of the processfor types 1, 4, and 9 of the XYZ product. The purifiedproduct will be shipped to the HIJ facility for finishing.The scope of this project covers only the fermentationand purification of those products mentioned above.Standard operating procedures will be followed forcontrol of the process and equipment. The scope of thisdocument covers the fermentor portion of the process.

1.2 Reference Documents and Drawings

2 Process OverviewThe fermentor area consists of a series of operationsthat are either automated or manual. In either case, theoperators initiate the sequences, which include thefollowing:

• Manual Set-Up Fermentor• Initialize Fermentor• CIP Fermentor• Pressure Test Fermentor• SIP Fermentor• SIP Fermentor Sample Port• Inoculate Fermentor• Fermentation• Sample• Fermentor Transfer Out

There are four seed fermentors (45L, 150L, 1,000L, and2,750L) and one production fermentor (8,000L). Mediais transferred into the fermentors using bags for thetwo smallest seed fermentors and from the media tanksthrough the media filters for the two larger seed andproduction fermentors. Acid and Base tanks supply theproduction fermentor as does a Nutrient tank as shownin the flow diagram in the next section. The threesmallest seed fermentors each have sparge lines forcontrol of dissolved oxygen and the largest seed fermen-tor and production fermentor have both upper andlower sparge lines. Upon completion of fermentation,the production fermentor transfers its product to theharvest tank for further processing.

2.1 Process Flow Diagram

3. Equipment Information

3.1 EquipmentThe fermentation area includes the equipment shownin Table A:

3.1.1 Equipment StatesThe equipment listed below will always be in one of thefollowing states:

Dirty: equipment has completed processing operations,or has just been returned from being out of service or itstime in the clean or steamed state has expired.

Clean: equipment has successfully completed CIP op-erations.

Steamed: equipment has successfully completed SIPoperations.

Out of Service: equipment is currently out of serviceand must undergo cleaning/steaming upon return toservice for use.

3.1.2 Cleaning and Sterilization RequirementsRefer to the Common Process Description for Cleaningand Sterilization requirements for all vessels.

3.1.3 Equipment Scheduling Requirements

Equipment DescriptionSFE1001 45L Seed FermentorSFE1002 150L Seed FermentorSFE1003 1,000L Seed FermentorSFE1004 2,750L Seed FermentorPFE1000 8,000L Production FermentorAG1001 SFE1001 AgitatorAG1002 SFE1002 AgitatorAG1003 SFE1003 AgitatorAG1004 SFE1004 AgitatorAG1000 PFE1000 AgitatorVF1001 SFE1001 Vent FilterVF1002 SFE1002 Vent FilterVF1003 SFE1003 Vent FilterVF1004 SFE1004 Vent FilterVF1000 PFE1000 Vent FilterPU1001 SFE1001 Additive PumpPU1002 SFE1002 Additive PumpPU1003 SFE1003 Additive PumpPU1004 SFE1004 Additive PumpPU1000 PFE1000 Additive Pump

Table A.



©Cop

yrig

ht IS

PE 2

007

Figure 4. Recipe operationsequencing details.

3.2 InstrumentationTable B shows the instrumentation for the fermentors.

3.3 Auxiliary SystemsThe fermentors require auxiliary support as indicatedin Table C:

4. Sequence of Operations

4.1 Fermentor Clean-in-Place(CIP)

4.2 Fermentor Pressure Test

4.3 Fermentor Steam-in-Place(SIP)

4.3.1 Operating RequirementThis operation may runalone or as part of largerrecipe.

4.3.2 Sequential Flow DiagramSee Figure 4.

4.4 Inoculation

4.5 Fermentation

5. Phase Sequences

5.1 Sample Bottle Sterilize-in-Place (SIP)The operator is prompted to attach sample bottles tothe sample port. Bottle connection is SIP’d and valvesare returned to pre-SIP condition.

This sequence can not be run until a successful SIPof the Fermentor has been completed.

There are no time critical operations associated withthis sequence. However, once the bottle has been suc-cessfully SIP’d, the bottle must either be used or SIPrepeated before 24 hours has elapsed.

There are no critical alarms which need to be re-ported in association with this sequence. Successfulcompletion of the sequence will indicate a successfulSIP.

5.1.1 Input Parameters

5.1.2 Report Parameters

5.1.3 Operating Parameters

5.1.4 Abnormal Operations

Tag Name Description Operating RangeProduction Fermentor PFE1000

AI-1000-001A Dissolved Oxygen Probes 0 – 100%AI-1000-001BAI-1000-004A pH Probes 0.00 – 14.00AI-1000-004BFSH-1000-003 Foam Detector Switch On/OffPIC-1000-005 Vessel Pressure 0 – 50 psigTIC-1000-001 Vessel Temperature 0.0 – 200.0°CTI-1000-004 Jacket Inlet Temperature -35.0 – 250.0°CTI-1000-006 Jacket Outlet Temperature -35.0 – 250.0°CWI-1000-001 Load Cell Weight 0.0 – 10,000.0 kg

Seed Fermentor PFE1001Seed Fermentor PFE1002Seed Fermentor PFE1003Seed Fermentor PFE1004

Table B.

Support Item Operating ConditionsCarbon Dioxide 38 – 40 psigClean Air 35 – 40 psigClean Steam 35 – 40 psigChilled Glycol -50.0 – -35°C Instrument Air 100 – 110 psigNitrogen 30 – 35 psigOxygen 30 – 35 psigPlant Steam 65 – 70 psigHot WFI 90 – 95°C

Table C.

Table D.

Recipe Parameter Typical Value Allowable RangeTime to achieve SIP temperature 60 minutes 50 – 70 minutesInitial temperature 100°C 100 – 110°C

Parameter specified in recipe(may change between batches).

Table E.

Reported ParameterTime to achieve SIP temperature

Parameter to be recordedwhen temperature is achieved.

Table F.

Operating Parameter Typical Value Allowable RangeMaximum time to achieve 15 minutes 10 – 30 minutesSIP temperature

Parameter available foradjustment – not in recipe.

Table G.

Exception Handling Required ActionHOLD Functionality Close clean steam supply valves. Open trap

valves to vent pressure.RESUME Functionality Open clean steam supply valves. Restart

timer. If temperature at any trap fallsbelow 121°C, start timer from zero onceTIs are at temp.

ABORT Functionality Discontinue vessel jacket temperaturecontrol. Release vessel pressure. Close allsteam valves and appropriate processvalves as necessary.

Abnormal Operation Requirements



©Copyright IS

PE 2007

5.1.5 SequentialFlowDiagramSeeFigure 5.

5.2 Fermentor Set-Up

5.3 Fermentor Pressure Test

5.4 Fermentor SIP

5.5 Fermentor CIP

5.6 Fermentor Inoculation

5.7 Fermentor Transfer Out

6. Continuous Control

6.1 Pressure ControlThe pressure in the fermentor is monitored using apressure transmitter and maintained using abackpressure control valve. There are different modesof operation for Fermentation, Transfer, Sterilize-in-Place (SIP), Cool-Down, and Vent.

6.1.1 Interlocks and PermissivesThere are no interlocks or permissives associated withthe fermentation pressure control module.

6.2 Temperature Control6.3 Antifoam Control6.4 pH Control6.5 Dissolved Oxygen Control6.6 Agitator Control

SummaryToo little, incomplete, or incorrect detail in the PD frequentlyleads to developing an FDS that ends up requiring costly andtime-consuming re-work as new or conflicting requirementsare included or resolved. A properly structured PD withappropriate details can help an automation team write FDSdocuments that effectively translate client process informa-tion into automation requirements that define how a productis made consistently and repeatably, while complying withrelevant regulatory requirements.

Finding the balance is critical in ensuring a predictableautomation project. The PD should focus on detailing whatthe control requirements for the process are so that the FDScan define how the automation will accomplish meetingthese requirements.

Good communication between the authors of both docu-ments is imperative throughout the development and dura-tion of the project, with the automation team providingconsultative inputs to the PD and the process design teamdoing the same during the development of the FDS. Workingcollaboratively ensures mutual understanding of the processand the automation which will reduce the risk of costlyrework and unnecessary frustration by the members of boththe automation and process design teams as a whole. In theend, the project will be successful and provide a positiveexperience for all.

References1. Instrumentation, Systems, and Automation Society (ISA),

ANSI/ISA-S88.01-1995: “Batch Control Part 1: Modelsand Terminology,” 1995.

2. GAMP® 4, Good Automated Manufacturing Practice(GAMP®) Guide for Validation of Automated Systems,International Society for Pharmaceutical Engineering(ISPE), Fourth Edition, December 2001.

About the AuthorsSteve Murray is a Senior Principal Engi-neer in the Life Sciences Group at EmersonProcess Management in Austin, Texas.Murray has more than 14 years in processautomation experience, including 11 years ofexperience in the execution of projects forpharmaceutical/biotech manufacturing facili-ties. Murray has been the automation lead

on large projects and has been involved in all stages of designand deployment of the automation system ranging fromconsultation on software architecture to start-up support. He

Figure 5. Phase Operation sequencing details.

Table I.

Operating Parameter Typical Value Allowable RangeFermentation Pressure 4 psig 2 – 8 psigSIP Pressure 15.5 psig 12 – 18 psigCool-Down Pressure 5 psig 3 – 7 psig

Parameter required to be available for adjustment– may be written by recipe as well.

Operating Mode Required ActionFermentation Control vessel pressure at Fermentation Pressure (psig)Transfer Controls vessel pressure at recipe-specified value (psig)SIP Controls vessel pressure at SIP Pressure (psig)Cool-Down Controls vessel pressure at Cool-Down Pressure (psig)Vent Vessel vented – no pressure control

Table H.



©Cop

yrig

ht IS

PE 2

007

can be contacted by e-mail at: [email protected].

Amit Shah is a Principal Engineer in theLife Sciences Group at Emerson ProcessManagement in Austin, Texas. He has morethan 13 years of progressive experience inlife sciences and information technology in-dustries, including process development/manufacturing improvement, automation de-sign, IT consulting, and academic research in

team lead and individual contributor roles. Shah holds a BS/MS in chemical engineering and an MBA. He can be contactedby e-mail at: [email protected].

Dawn Marruchella is the Batch ProductManager for Emerson Process Managementin Austin, Texas. With more than 11 years ofcombined experience, Marruchella’s back-ground includes seven years of work execut-ing batch automation projects for the lifesciences industry. Marruchella has providedtechnical leadership for several automation

projects, beginning with front end engineering efforts andcontinuing throughout project execution and on-sitesupport. She can be contacted by e-mail at: [email protected].

Emerson, 12301 Research Blvd., Research Park Plaza,Bldg. III, Austin, Texas 78759.

MARCH/APRIL 2007 PHARMACEUTICAL ENGINEERING 1

PAT Implementation

©Copyright ISPE 2007

This articleoutlines aneight-step sixSigma toll-gateapproach toPATimplementation.

Eight Steps to PAT: Using the Designfor Lean Six Sigma Toll-Gate Processas Best Practice

by Bikash Chatterjee and Jeremy Green

Overview

The FDA’s recent guidance regarding Pro-cess Analytical Technology (PAT) offersthe pharmaceutical and biotech indus-

tries an unprecedented opportunity to leveragehard-won experience with scientific inquiryand innovation. However, the leap to PAT issignificant for even the most rigorous develop-ment program. Many aspects of Six Sigma,including its use of statistical tools and its

phase- or toll-gate approach to project manage-ment, can facilitate and accelerate a PAT ini-tiative. Rather than advocating company-wideSix Sigma adoption as a prerequisite to effec-tive PAT implementation, an eight-phase De-sign for Lean Six Sigma approach is recom-mended that can be used on a project-by-projectbasis.

Introduction:The Shift from Product Control to

Process ControlPrior to 2002, regulatory oversight focused pri-marily upon adherence to pre-defined proce-dures, record keeping, and an audit trail as ameans for ensuring product safety and efficacy.Due to the emphasis on oversight control, mostfirms would ‘lock-down’ their processes andcontrol methods once process validation wascomplete. Product quality was achieved throughoff-line inspection, rather than through identi-fying, understanding, controlling, and optimiz-ing critical process parameters. The FDA rein-forced this mindset by requiring regulatorypre-approval before any changes could be madeto the process. The FDA Modernization Act of1997 initiated a change in policy and thinkingthat culminated in the release in 2002 of theguidance document Pharmaceutical cGMPs forthe 21st Century - A Risk Based Approach. Tostreamline the regulatory approval process andenhance patient safety, this document proposeda shift to a science-based compliance model,integrating the disciplines of quality, safety,and risk management.

Since 2002, the FDA has released guidancedocuments on risk-based inspections, Part 11electronic records and signatures, quality sys-tems approach to pharmaceutical cGMPs, and

Figure 1. The Six SigmaDMAIC model.

Reprinted from




2 PHARMACEUTICAL ENGINEERING MARCH/APRIL 2007

PAT Implementation


Process Analytical Technology (PAT). The FDA was not theonly regulatory body to recognize this need and solicitedinput from its counterparts in Canada, Europe, and Japan,and from industry and academics worldwide. Several keyguidance documents from the International Conference onHarmonization (ICH), ICH Q8 (Pharmaceutical Develop-ment), ICH Q9 (Quality Risk Management), and the forth-coming ICH Q10 (Quality Management) have become the defacto standard for transforming organizations that aspire tothe highest degree of scientific rigor in the product develop-ment process. These documents comprise the basis for shift-ing manufacturing and regulatory philosophy from inspec-tion and oversight to managed risk: a scientific approachcapable of providing a higher level of product quality assur-ance.

Six Sigma and PATIn September 2004, the FDA issued its final guidance, “PAT:A Framework for Innovative Pharmaceutical Development,Manufacturing, and Quality Assurance.” The fact that theguidance extends beyond a pure hardware solution under-scores the Agency’s desire to shift product quality assessmentaway from a product-centric approach, based on inspectionand final testing, to one that is more process-centric and builtupon understanding the variables that affect overall productquality. Process Analytical Technology (PAT) represents theculmination of a true process-centric quality system. Unfor-tunately, the industry has had difficulty embracing the totalvision for PAT, partly because of its radical departure fromhistorical methods of process and product development, andpartly because of the lack of a definitive implementationmodel.

A criticism leveled at some PAT implementations is thatefforts have focused on the application of on-line analytical

technology (as a replacement for off-line laboratory testing),rather than on understanding control and reduction of varia-tion.1 In other words, the focus has been on the measurement,rather than the improvement of product quality. Statisticaltools for characterization and optimization of manufacturingprocesses have been quietly in use in industry for more than50 years although often confined to use by corporate statisti-cians. A renaissance in the more widespread use of industrialstatistics by non-statisticians came with the advent of theimprovement methodology of Lean Six Sigma in the mid-1980s.2 A search of the PAT literature reveals an emphasis onthe use of statistical tools, particularly multivariate meth-ods.3, 4, 5 However, the Lean Six Sigma approach to projectsprovides many advantages over isolated use of statisticaltools. Among those advantages are:

1. selection of limited scope improvement projects, accordingto verified bottom line cost savings and increased cus-tomer satisfaction, achievable in two to six months

2. use of cross-functional teams led by a Six Sigma Black Belt(not a degreed statistician although trained in the use ofstatistical tools)

3. sponsorship of projects by a corporate executive champion,whose role is to remove political, financial, and otherbarriers that stand in the way of the team’s success

4. Structuring the project in phases or stages, with eachphase having defined statistical and lean tools, and objec-tive criteria and metrics for the success of each phase. Themost common approach is known by the acronym DMAIC;which stands for Define, Measure, Analyze, Improve,Control.

Figure 2. Controlled release tablet process flow.


PAT Implementation


5. Toll-gate review meeting at the end of each phase. Duringthis meeting, the team presents progress to date to mem-bers of senior management. If the team pays the “toll” bymeeting the criteria and metrics for the phase, manage-ment raises the “gate” and allows the team to move to thenext phase.6

The improvement methodology of Six Sigma provides aneffective framework for the process characterization andoptimization required by PAT that is superior to the use ofstatistical tools in isolation. A company-wide conversion toSix Sigma and its requirements for significant cultural changeis not required. A less resource-intensive alternative is to usea Six Sigma project structure and associated statistical toolsto characterize, control, and reduce process variation toachieve more consistent product quality. It is the authors’opinion that a phased Six Sigma toll-gate approach to productand process improvement significantly enhances and accel-erates PAT implementation.

Based upon an actual PAT deployment initiative facili-tated by Pharmatech Associates, the methodology employedand challenges encountered in the course of a PAT implemen-tation project for a business unit of a major pharmaceuticalcompany will be presented

Business ProblemThe business unit had identified a number of improvementopportunities in its manufacturing process flow at one of itssolid dosage manufacturing plants. The product was a con-trolled-release tablet which utilized a high molecular weightpolymer to control the diffusion of the drug. The releaseprofiles of the drug had been inconsistent since its marketintroduction two years previously resulting in rejected lotsand a higher than desirable incidence of stage two dissolutiontesting. The inconsistent performance presented a potentialregulatory risk to the product unless the variable dissolutionperformance could be addressed. In addition, the productdemand was growing and the yield impact contributed to anerosion of plant capacity causing missed shipments and lostrevenue to the business unit. Based upon projections therewould be insufficient capacity to meet next year’s demand.

A series of characterization studies was initiated usingorthogonal experimental designs, evaluating API, granula-tion and compression parameters to identify the key param-eters which affect dissolution performance. The investigationidentified the root cause as periodic over-mixing of the lubri-cant during the final blending step, resulting in more hydro-phobic surface properties. The inconsistent mixing perfor-mance was ultimately attributed to varying raw materialproperties, in particular particle size distribution, due toalternate lubricant suppliers. This source of variation wasaddressed through the establishment of a particle size distri-bution specification. In addition, only suppliers that demon-strated a process capability greater than 1.0 against thespecification were qualified for the process.

Faced with the looming regulatory risk and capacity short-fall, management decided to initiate a PAT program to

determine if the inconsistent product performance issuecould be addressed through the use of in-line analyticalmeasurements and closed loop control. The decision wasmade to deploy a PAT team to implement improvements withthe objective of eliminating or significantly reducing processinstability.

Plant Process FlowThe process flow for manufacturing is shown in Figure 2. Themajor unit operations are compounding (API addition insolution), granulation, milling, blending, and tableting. Giventhe results of the root cause analysis exercise the projectfocused upon the final blending step for the PAT project.

PAT TeamTo deploy the project, the business unit established a PATproject team that consisted of experts from across the businessunit, an outside pharmaceutical consulting firm, and an auto-mation supplier. Although they were not a formal Lean SixSigma organization, the pharmaceutical company had experi-ence using many of the Lean Six Sigma statistical tools. ThePAT team decided to apply the Six Sigma structure to theproject from the outset because of the perceived advantage ofa toll-gate approach with its built-in checks and balances. TheSix Sigma approach allowed the team to clearly articulatesuccess metrics for the individual stages of the project, as wellas for the project as a whole, and align the project with currentbusiness objectives and strategy. The toll-gate approach, withits use of incremental success metrics, was instrumental ingarnering senior management support from across the organi-zation throughout the project. The team’s first task was todefine the deliverables and ensure consistency with currentbusiness and regulatory objectives. As described earlier, theplant was suffering from a capacity shortfall. The team metand summarized the situation as follows:

1. At the current manufacturing rate, the inconsistent tabletdissolution profile was costing the company $20 million on

Tablet Lot Percent Lubricant Comment

Lot A 1.58 Control Space Lot

Lot B 1.68 Control Space Lot

Lot C 1.77 Control Space Lot

Lot D 1.54 Control Space Lot

Lot E 1.60 Control Space Lot

Lot F 1.44 Control Space Lot

Lot G 1.46 Control Space Lot

Lot H 1.63 Control Space Lot

Failed Lot 1 4.62 Failed L1 Dissolution




Table A. Lot-to-Lot tablet lubricant content.


PAT Implementation


an annual basis, not including the cost of poor qualityassociated with handling rejected material.

2. Rejected lots needed to be reduced to no more than 5% inorder to recover the necessary manufacturing capacity forthe coming year.

3. Estimated cost to the business unit due to dissolutionfailures or Stage 2 testing requirements was approxi-mately 8% of the standing Work-In-Process (WIP) cost.

Based upon this assessment, the team determined it wasappropriate to proceed with the PAT project. It is importantto note that the objective in this case, from a businessperspective, was not to replace the quality overhead associ-ated with the tablet release, but rather to prevent the loss ofproduct due to poor dissolution. This greatly simplified theinitial regulatory strategy for the project, while leaving thedoor open for a future filing to replace product release testingwith an in-process control strategy.

Six Sigma or Design for Six Sigma?The team explored several models to evaluate the applicationof Six Sigma to the project. The classic Six Sigma DMAICmodel provides a good framework for objective scientificinquiry and is typically used to improve existing processes(and products). However, the team decided that Design ForLean Six Sigma (DFLSS), with its focus on the developmentof new products and processes, would be a more appropriateapproach for the PAT project. Subsequently, the team evalu-ated several of the current DFLSS Models as alternatives tothe classic DMAIC model. DFLSS models provide a struc-tured, phased approach to the design of a product, process, orservice with Six Sigma Quality (target of 3.4 defects permillion opportunities) and efficiency as key design criteria.Risk management is easily incorporated in the approach, as

Failure Modes and Effects Analysis is a standard DFLSS tool.Integration of DFLSS with PAT answers the criticism of somecurrent PAT implementations that focus too much on on-lineanalytical instrumentation rather than on the sources ofprocess and product variation. This DFLSS toll-gate ap-proach to PAT provides the additional advantage of a set ofmeasurable success criteria for completion of key milestoneswithin each phase of the process so the “gate” can be closed.Comparison of progress with such criteria provides objectiveevidence of incremental team success (that can be celebratedand communicated to the rest of the organization) and helpsprevent team self-delusion. The DFLSS models7 under con-sideration were Define, Measure, Analyze, Design, Verify(DMADV), Identify, Design, Optimize, Verify (IDOV), andDefine, Characterize, Optimize, Verify (DCOV) - Figure 3.

Six Sigma PATIn a review of the literature and with the recommendation ofthe pharmaceutical consulting firm, the team decided to usethe DCOV DFLSS model with its focus on process character-ization and optimization. The team expanded the DCOVroadmap into the following eight phases: Identify, Character-ize, Define, Optimize, Measure, Automate, Verify, and Vali-date. The modified DCOV project management approachallows the business to make the best possible decisions withthe available data and resources. The purpose behind eachstep of the eight-phase process is as follows:

1. Identify: clearly identify key elements of the project,including: regulatory strategy, regulatory commitment toKey Process Output Variables (KPOVs).

2. Characterize: what are the Key Process Input Variables(KPIVs) that have been characterized as they relate to theKPOVs?

3. Define: what is the defined design space for the process?

4. Optimize: what is the control space that defines theallowable KPIV levels in order to maintain the processwithin the design space?

5. Measure: what analytical solutions are possible surro-gates for the existing offline measurement systems?

6. Automate: what control solutions can be applied to lever-age?

7. Verify: prepare a proof-of-concept, process model.

8. Validate: complete the IQ, OQ, PQ, method validationand comparability study.

The PAT model adopted is shown in Figure 4.Within each of the phases, there are a set of deliverables

that must be completed to ensure all project requirements aremet. Each will be discussed as follows:Figure 3. Lean DFSS models.


PAT Implementation


IdentifyIn the identify phase, the PAT team is tasked with determin-ing the design criteria for moving forward with the PATstrategy. The key process parameters, such as API physicalcharacteristics, granulation process/control, and compres-sion force/tablet hardness had previously been determinednot to be the source of dissolution variation, leading the teamto focus on the blending step. On the process side, the teamdeveloped a flow chart to identify the Process Input Variablesand the Process Output Variables to design into the PATsolution. The input variables for the blending process identi-fied were as follows:

1. Granulation Particle Size Distribution (PSD)2. Mixing Time3. Intensifier Arm4. Lubricant PSD5. Lubricant concentration

The process utilized a 60 cu.ft. mixer. The Chemistry, Manu-facturing, and Controls (CMC) commitment during the origi-nal drug filing was to mix for five to 15 minutes, at a mixingspeed of 10 rpm with the intensifier arm on. The KPOVs filedin the NDA for this step were content uniformity and tabletdissolution at two, four, and eight hours. The specificationwas 10 to 20%, 21 to 60%, and 61 to 100% respectively forthese time points. The team determined it would use thetablet dissolution, API content uniformity, and concentrationof lubricant as benchmarks for evaluating the content unifor-mity of the lubricant and the mixing effectiveness during thatstage of the process. The regulatory strategy initially focusedon establishing a control range within the NDA commitment.Since PAT focuses on a feedback control architecture, theintent was to establish a scientifically rigorous comparabilitydata set using the optimized control range, then steer thetarget metrics for PAT automation to the same endpoints.

At this and subsequent phases, success metrics wereestablished. Progress and metrics were presented to manage-ment at a toll-gate review meeting with management’s char-ter to give the team approval to move to the next phase or totake additional action to resolve any open issues. For anyopen issues, the team would submit a formal corrective actionto get management’s approval to move to the next phase. Thisprocess continued through the subsequent seven phases.

CharacterizeA retrospective review of the process development data indi-cated that there was no evaluation of the impact of granula-tion PSD or lubricant PSD. Lubricant concentration wasevaluated, as was mixing time. Neither evaluation used anorthogonal experimental design; hence, the data could not beregressed. The development data evaluated 1% and 2% lubri-cant concentrations. Based upon this development work, thesignificant KPIV identified was mixing time with the KPOVsbeing tablet appearance and dissolution. Tablet appearancewas representative of the tablet compression process. A finalconcentration of 1.5% was chosen.

Given the lack of information from the original develop-ment work, a characterization study was initiated to evaluatethe impact of lubricant concentration, mixing time, andwhether the intensifier arm was used. The ICH Q8 guidancedescribes this evaluation as defining the knowledge space forthe process. An orthogonal experimental design, using ablocked design for the intensifier bar, was performed. Meangranulation size and PSD and lubricant PSD data weremeasured and kept constant for the study. The results indi-cated that lubricant concentration and mixing time KPIVswere both significant at all three dissolution points. Theintensifier bar did not have an effect. Key KPOVs measuredwere drug dissolution, drug content uniformity, and tabletappearance.

DefineICH Q8 discusses identifying the optimum design space forthe process. The design space is a subset of the overallknowledge space for the manufacturing process. A graphicalrepresentation of the relationship between the knowledge,design, and control space is shown in Figure 5. In evaluatingthe influence of key process inputs, the team focused upon atiered approach to reducing PAT risk. It was agreed theminimum acceptance criteria was to achieve drug contentuniformity. Once the control space was established, the

Figure 4. Design for Lean Six Sigma model applied to PAT.


PAT Implementation


Figure 5. Relationship of knowledge, design, and control spaces.

behavior of the lubricant would be evaluated. The objectivewas to find a control space in which drug content and lubri-cant uniformity could be assured.

The knowledge space defines the boundaries within whichthe process inputs or KPIVs can be varied. However, withinthe knowledge space, some parameters at their limits maynot produce acceptable product and some parameters mayhave no impact on the critical KPOVs (in this case study -content uniformity, dissolution, and lubricant content). Thedesign space then represents the widest range of each of theKPIVs within which acceptable product meeting all of theKPOV specifications can be manufactured under ideal andcontrolled conditions. The control space represents a furthertightening of the design space of the KPIVs in which accept-able product is assured of meeting specifications, allowing forprocess drift and measurement and sampling uncertainty.

The PAT team initiated a follow-on study, designed tocharacterize the design space. The lubricant concentrationwas fixed at 1.5%, and the intensifier bar was not used. Inorder to evaluate the impact of granulation PSD, the percent-age fines was evaluated. Two different suppliers of lubricantalso were evaluated. All satisfied the revised specification forthe lubricant. The DOE evaluated the following possibleinput parameters:

1. Mixing time: 7 to 12 minutes2. Granulation Percent Fines: 10 to 40%3. Lubricant Lots: 1 to 2

The study revealed that mixing time and granulation PSDwere significant KPIVs for drug dissolution at the two, fourand eight hour time points. The lubricant lots were notsignificant contributors to either drug content uniformity orlubricant content uniformity. All lots passed Level 1 dissolu-tion testing.

OptimizeThe next step in identifying the final processing space is toidentify the control space. The control space represents arange of critical parameters within which the process willyield an assured output within the KPOV specifications

allowing for sources of statistical uncertainty. It also repre-sents the basis for the control architecture to be adapted forthe PAT solution. Powder mixing theory states that thecomponents that impact blend uniformity are: granulation/blend physical characteristics, including particle size distri-bution, shape and moisture content, powder bulk density,and Van Der Waals forces. Of these, granulation particle sizeis the most significant factor. Given that granulation PSDwas identified as a significant contributor, a Six Sigmaexercise was initiated to understand the variability in thefinal granulation. Milling steps upstream of the blend stepwere evaluated and modifications to the milling set-up tocontrol the feed rate of granulation were made. A screeningstudy was repeated to determine if the new granulation PSDwas still a significant contributor to blend uniformity and itdid not come up as significant at the 95% confidence interval.Based upon this, the control strategy established a baselineof lubricant distribution to serve as the comparability criteriafor the PAT solution downstream. The team did not focus onthe drug content uniformity since the knowledge and designspace studies had moved the process away form the edge offailure, while characterizing the variability around the KPIVsthat would affect drug content uniformity.

MeasureThe challenge in developing an in-line metric for ensuringproper mixing of the lubricant was the lack of an off-line testcurrently being performed for lubricant content in tablets. Abaseline examination of tablets manufactured during thecontrol phase was performed using Mass Spectroscopy (MS)in order to understand the variability around the controlspace. Tablets which exhibited poor dissolution also wereevaluated from the original failed lots. The results of the MSdata are shown in Table A. The most striking observation isthat the tablets which exhibited poor dissolution had signifi-cantly higher levels of lubricant.

AutomateThe team had sufficient understanding of the behavior of thecurrent process, its KPIVs and KPOVs, to move to identifyingan automation solution. The PAT team included an externalautomation firm with a strong understanding of process,Design for Six Sigma, and GAMP 48 to complement theirexperience in custom automation. This is a significant consid-eration given the intimate relationship between the technicalsolution and quality and regulatory considerations for theproject. Having a solutions provider that possesses the sys-tems to integrate the requirements of Quality by Design(QbD) is a major advantage in developing the scientificargument that the in-line solution is an equivalent or supe-rior surrogate to the off-line analytical solution and in gener-ating the necessary documentation trail to support subse-quent validation.

The team approached the automation solution in phases.The first phase was designed to ensure there was a solidunderstanding of the existing process performance using off-line analytical tools. Tablet performance was currently mea-


PAT Implementation


Figure 6. Process capability curve for eight hour tablet dissolution time point.

sured using HPLC for content uniformity and potency assess-ment. Tablet dissolution coupled with UV spectroscopy deter-mined the tablet’s release profile. Currently no off-line as-sessment was performed on the lubricant. This had beenfound to be a key factor in achieving the desired tabletdissolution profile. The second phase was to establish acorrelation with the new surrogate analytical method. Thelast phase was to demonstrate that the hardware solutionand control algorithm resulted in tablets that satisfied theproduct’s release criteria. The team focused on establishing acorrelation between an off-line and in-line method for lubri-cant concentration. Tablets were analyzed using MS and FT-NIR to establish the correlation. The results are shown inTable B. Based on the results of the correlation study, theteam demonstrated that the in-line solution was viable andthey could proceed with developing an in-line solution.

VerifyThe verify step is used to establish a basic proof of conceptthat the principles of the solution are viable. In the previousphase, the team attempted to establish a correlation betweenthe offline and in-line measurement systems. Adherence tothe Six Sigma methodology had narrowed the control space tokeep the process sufficiently far from the edge of failure. Thefirst focus was establishing a measurement for the lubricantin the blending step, which could be used to dictate theblending time. It is important to note we are not concernedwith lubricant weighing errors; rather with the distributionof lubricant throughout the granulation. Lubricant integra-tion with the granulation particle has been shown to impact

the dissolution of some controlled-release tablets as mixingtime increased.

Since confidence was high that material mixed for seven to12 minutes resulted in tablets with acceptable dissolution,then the correlation with lubricant concentration could beone trigger used to prevent overmixing. A non-destructivetest was required to demonstrate comparability. A samplesize of 100 tablets was selected. With the selection of FT-NIRas the measurement tool, a blender was modified with a self-contained analytical probe and analyzer and equipped with awireless transmission system to deploy as a proof of conceptsystem. The final control space screening study was repeated.Measurements were taken from the in-line sensor and tab-lets were tested using MS for lubricant content. In addition,tablets were tested for content uniformity, potency, anddissolution at the three, four, and eight hour time points.One of the key metrics for process performance is a measureof process capability, Cpk. This metric for a process with anormal (or close to normal) distribution and a two-sidedspecification is described by the equation:

_ _USL - x x - LSL

Cpk = min __________ , __________3s 3s

Where, USL/LSL = Upper/Lower Spec LimitX = MeanS = Standard Deviation (sigma)

Cpk assumes that the process is in statistical control. Insimple terms, Cpk compares the spread of your process to thespread of your specification and how close the mean of that


PAT Implementation


Tablet MS% T-NIR% CommentLubricant Lubricant

Lot A 1.62 1.43 Control Space Lot

Lot B 1.76 1.68 Control Space Lot

Lot C 1.53 1.37 Control Space Lot

Lot D 1.67 1.59 Control Space Lot

Lot E 1.48 1.44 Control Space Lot

Lot F 1.51 1.49 Control Space Lot

Lot G 1.39 1.21 Control Space Lot

Lot H 1.71 1.47 Control Space Lot

Failed Lot 1 5.12 4.46 Failed L1 Dissolution




Table B. Analytical method comparison.

process is to the specification limits. A high number (>1.33)indicates that the probability of getting an out-of-spec prod-uct is very small. Cpk was calculated after determining thatthe process was in control through a control chart. All disso-lution time points had a process capability greater than 1.33(4 sigma process). The process capability chart for the eighthour time point is shown in Figure 6.

The results illustrate that the FT-NIR system was capableof controlling the process and delivering compliant product.Based upon these studies, a change control notice was initi-ated and the production equipment was modified.

ValidateThe final step in the process was to validate the equipmentand process. The Six Sigma process dictated the elements tobe completed as follows:

1. Generate the Final Development Report2. Baseline the Equipment3. Modify Operational SOP4. Modify Maintenance SOP5. Modify Calibration Program6. Software Validation-Part 11 Compliance7. IQ/OQ/PQ8. MS and FT-NIR Method Validation9. Regulatory Update

ConclusionThe statistical tools and toll-gate process of Six Sigma pro-vides a best practice process for characterizing, controlling,and reducing process variation that is necessary to success-fully deploy PAT. The partnership of Six Sigma and PAT wasintended to characterize and implement a control and mea-surement solution, which would minimize the likelihood of acontrolled-release tablet failing dissolution. The team used amodified DCOV model subdivided into eight phases designedto ensure that the basic requirements of the ICH Q8 require-

ment for QbD were satisfied. This framework ensured allaspects of the project were addressed in an efficient andmethodical manner with the scientific rigor necessary toimplement an in-line control architecture integrated through-out the process.

References1. Maes, I. and B. Van Liedekerke, “The Need for a Broader

Perspective if Process Analytical Technology Implemen-tation is to be Successful in the Pharmaceutical Sector,”Journal of Pharmaceutical Innovation, Volume1, Number1, September/October 2006, pp. 19-21.

2. Kourti, T., PhD, “Process Analytical Technology and Mul-tivariate Statistical Process Control: Wellness Index ofProduct and Process – Part 1,” PAT Journal, Volume 1,Issue 1, September/October 2004, pp. 13-19.

3. Kourti, T., PhD, “Process Analytical Technology and Mul-tivariate Statistical Process Control: Wellness Index ofProduct and Process – Part 2,” PAT Journal, Volume 2,Issue 1, January/February 2005, pp. 24-29.

4. Kourti, T., PhD, “Process Analytical Technology and Mul-tivariate Statistical Process Control: Wellness Index ofProduct and Process – Part 3,” PAT Journal, Volume 3,Issue 3, May/June 2006, pp. 18-24.

5. Harry, Mikel and R. Schroeder, Six Sigma: The Break-through Management Strategy Revolutionizing the World’sTop Corporation, Doubleday, 2000.

6. Pyzdek, T., The Six Sigma Handbook: A Complete Guidefor Green Belts, Black Belts and Managers at All Levels,Revised and Expanded, McGraw-Hill, 2003.

7. Brue, G., Design for Six Sigma, McGraw-Hill, 20038. GAMP 4: The Good Automated Manufacturing Practice

Guide for the Validation of Automated Systems, Interna-tional Society for Pharmaceutical Engineering (ISPE),Version 4, December 2001.

About the AuthorsBikash Chatterjee is President ofPharmatech Associates, Inc. He has beeninvolved in the bio-pharmaceutical, pharma-ceutical, medical device, and diagnostics in-dustry for more than 20 years. Most recently,he served as the vice president, Pharmaceu-tical Operations for Aradigm Corporationwhere he was responsible for establishing

their process development, engineering, validation, facili-ties, and manufacturing capabilities. Prior to joining Aradigm,Chatterjee was the plant manager responsible for manufac-turing Boehringer- Mannheim’s disposable coagulation test-ing system. Chatterjee has designed, built, and qualifiedmultiple facilities throughout his career, ranging from cen-tralized warehousing to large capacity BSL-2 manufacturingfacilities. His expertise includes site selection, project man-agement, and design and validation of facilities for both USand European regulatory requirements, and he has beeninvolved in the design and validation of biopharmaceuticalfacilities in the US, Europe, and Asia. From 1988 to 1992, heheld a number of senior manufacturing positions at various


PAT Implementation


pharmaceutical companies, including Syntex Corporation.Chatterjee is a certified ISO 9000 Lead Assessor, a Six SigmaMaster Black Belt, and has more than 15 years of experiencein the implementation of Lean Manufacturing programs inthe life sciences industry. He holds a BA in biochemistry anda BS in chemical engineering from the University of Califor-nia at San Diego. He can be reached by telephone at: +1-510-760-2456 or by e-mail at: [email protected].

Dr. Jeremy D. Green is a Senior Consultantfor Pharmatech Associates. He is a CertifiedLean Six Sigma Master Black Belt, AmericanSociety for Quality Certified Six Sigma BlackBelt, and Certified Quality Engineer with 20years of experience in quality and manufac-turing management. His background includeswork in the pharmaceutical process automa-

tion, food, semi-conductor, and electronics industries. Dr.Green received his PhD from Indiana State University and hisBA from the University of Illinois. He can be reached bytelephone at: +1-541-754-9164 or by e-mail at: [email protected].

Pharmatech Associates – SIPR, 1098 Foster City Blvd.,Suite 303, Foster City, California 94404.


Lean Data Analysis


Figure 1. PAT DataIntegration, Modeling,Improvement, andControl Process.

This articleintroduces a dataanalysis maturitymodel that mapsvarious tools andmethodologies aimedat predicting,analyzing, improving,or controlling thedrivers of productquality to the extentto which thesetechniques may helpreduce defects. Bymapping toolscurrently deployed ina particularmanufacturingfacility to thematurity model, it ispossible to define acost-effective roadmap for variousinitiatives aimed atimproving productquality throughincreased processunderstanding.Pragmatic dataanalysis andreporting approachesare introduced to aidprocessunderstanding formainstream usersand the deploymentof thatunderstanding inmanufacturing toincrease productperformance.

Lean Data Analysis: Simplifying theAnalysis and Presentation of Data forManufacturing Process Improvement

by Malcolm Moore

Introduction

To achieve increased process understand-ing via Six Sigma, Process AnalyticalTechnology (PAT), or other methodolo-gies requires adoption of at least three

types of technology:

1. measurement technology to gauge processand material inputs and intermediate prod-uct

2. data integration and cleansing technologyto bring together disparate sources of data –including process, material, intermediate,and final product data sources – in a timelyand effective manner

3. data analysis and reporting technology tobring understanding from integrated datacollected in the context of a problem orimprovement opportunity

Emphasis on measurement technology alonewill increase the extent to which process and

materials are measured, and will drive up costsand data volumes. The lack of effective dataintegration and data analysis methods for allconsumers of the data will limit the growth inprocess understanding and the ability of manu-facturing to exploit this understanding.

Figure 1 presents a high-level process modelof data integration and data analysis in manu-facturing. The components represented in bluedepict the IT function of integrating disparatedata sources, including databases, electronicand paper sources, then cleansing and trans-forming data to an analysis-ready state with adata model that is easily maintained and ex-tended as the number and type of data sourcesgrow.

The need for a data integration solution –and the level of sophistication required of it –will depend upon the extent to which inputs aremeasured. In newer production lines, it may becommon to measure hundreds of input vari-ables via NIR spectroscopy, and other inline,

Reprinted from





Lean Data Analysis


at-line, online, or offline methods, requiring a data integra-tion solution. However, older production facilities may focuson offline laboratory testing of intermediate and end-of-lineproducts and use measurement technologies for process in-puts on an as-needed basis.

The process steps represented in orange symbolize someways that business users might analyze their data. At leasttwo different approaches to modeling the relationships incleansed or analysis ready data are available and the termsefficiency and effectiveness modeling are introduced to distin-guish the two approaches. Efficiency modeling is used toclassify models that use multivariate relationships to predictmanufacturing problems. Such models do not necessarilyresult in model simplification or reduction of the number ofdimensions that need to be measured, nor do they greatlyincrease understanding of how the key inputs drive variationin product quality. Effectiveness modeling, on the otherhand, is used to describe approaches that identify the criticalfew inputs and define empirical transfer functions that de-scribe how these key inputs operate together to drive manu-facturing problems or issues, increasing our understandingof how those inputs affect variations in quality. These twomodeling approaches are described in more detail below andare illustrated by case studies.

This article focuses on pragmatic approaches to dataanalysis and reporting that work regardless of the extent to

which inputs are measured. It introduces ways of simplifyingdata analysis and reporting approaches associated with PAT,Six Sigma, and related methodologies and proposes a way todefine a road map for the adoption of manufacturing improve-ment technologies relative to the current level of measure-ment maturity. Mapping of a broad set of tools to a dataanalysis maturity model are presented along with examplesof various data analysis approaches, including a set of prag-matic analysis techniques that are simple to apply andunderstand at all levels of an organization.

PAT Data Analysis MethodsModeling ApproachesStatistical modeling approaches to PAT are classified in twoways: models for increasing the efficiency of manufacturing– reducing waste; and models for increasing effectiveness ofmanufacturing – enhancing process understanding and uti-lizing it to improve manufacturing performance.

Efficiency models consist of classification modeling tech-niques, such as discriminant analysis, cluster analysis, anddecision trees, along with predictive modeling techniquessuch as Partial Least Squares (PLS) and Principal Compo-nent Regression (PCR). These techniques exploit the multi-variate relationships among a large number of measuredinputs to predict product performance or batch failures aheadof time. Compared with effectiveness modeling methods,

Figure 2. Mapping of data analysis technology to process capability and dependence on extent and relevance of measured inputs.


Lean Data Analysis


efficiency models require a large number of measured inputs– in fact, the more the better – and tend to be used for “blackbox” batch classification or prediction of likely product perfor-mance. In other words, they make good predictions, but theydo not necessarily deliver fundamental changes in processunderstanding.

Effectiveness models consist of variable reduction or ex-ploratory data analysis methods, such as data mining, corre-lation analysis, process mapping, cause-and-effect analysis,Quality Function Deployment (QFD), Failure Mode EffectsAnalysis (FMEA) to identify the critical few inputs that areinvestigated in more detail via Design of Experiments (DOE),multiple regression, and generalizations of multiple regres-sion for non-normally distributed measures of product qual-ity. With care, these latter techniques develop empiricalmodels that approximate the causal relationships betweenthe critical few inputs and product quality.

Examples of some of these different modeling approachesare provided in the case studies below.

Maturity ModelFigure 2 may be useful when considering the best mix of dataanalysis methodologies to increase process understandingfor a particular manufacturing facility. It may help establisha baseline for your manufacturing facility with regard toproduct quality performance, define goals for a proposed PATinvestment, and help define a road map for getting to thoseperformance goals.

This maturity model maps data analysis methodologiesagainst sigma capability. Sigma is the measure of variabilityin the product quality measure, usually calculated by assum-ing the product quality measure is normally distributed. A 2sigma process is one where the mean ± two standard devia-tions coincide with the specification limits of the productquality measure. In this case, approximately 5% of batcheswould not meet the required quality specification (approxi-mately 2.5% in each tail of the distribution). Defects PerMillion Opportunities (PMO) is calculated after assuming ashift of 1.5 sigma in the mean of the product quality measure.Hence, a 2 sigma process encountering a 1.5 sigma shift in themean from target would result in 308,537 defects PMO. Thus,a high sigma capability value such as 5 or 6 is required toensure little or no defects after allowing for a shift in theprocess mean.

Most mature manufacturing facilities deploy a combina-tion of QA inspections, Statistical Quality Control (SQC) –control charts applied to product quality measures, and QAinvestigations in an attempt to trace the cause of batchexceptions. Such approaches generally achieve sigma capa-bility of up to 2.5. The introduction of SPC, where controlcharts are applied to intermediate product measurements,may get performance up to the region 3 sigma.

More sophisticated control methods, such as End-PointDetection (EPD) and Advanced Process Control (APC) can bedeployed to reduce variation in intermediate product andhelp reduce variation in final product to 3 sigma or there-abouts. Utilization of inline measurement tools in conjunc-tion with EPD to achieve a specified moisture content in

Figure 3. Key processes and inputs associated with excessive variation in 60 minute dissolution.

Actual

Count Too Low Good Too HighTotal %Col %Row %

Too Low 1 1 0 21.39 1.39 0.00 2.78100.00 1.54 0.0050.00 50.00 0.00

Predicted

Good 0 63 1 640.00 87.50 1.39 88.890.00 96.92 16.670.00 98.44 1.56

Too High 0 1 5 60.00 1.39 6.94 8.330.00 1.54 83.330.00 16.67 83.33

Total 1 65 6 721.39 90.28 8.33

Table A. Predicted by actual batch classification.


Lean Data Analysis


Figure 4. Visual exploration of relationships.a. Distributions of 60 minute dissolution and key inputs with

dissolution failures identified in dark green.b. Parallel of key inputs with the values of the processing

conditions identified for a failing lot.c. Parallel plot with settings of some key inputs worthy of further

investigation.

drying operations or specified blend uniformity in blendingoperations, or APC to have short loop controls to make“corrective” adjustments during unit process operations areappealing control strategies. However, to have real impact,these approaches require detailed understanding of the ex-tent to which changes to the mean, within-batch and be-tween-batch variability of factors, such as moisture content,affect the sigma capability of final product quality. Used inisolation, EPD and APC are unlikely to get final productcapability much beyond 3 sigma. Multivariate StatisticalProcess Control (MSPC) is an extension of SPC that exploitsthe correlations between measured inputs to give greaterresolution in detecting problems ahead of time and reduce theextent of false alarms. Like EPD and APC, MSPC is mosteffective when deployed in conjunction with effectiveness

modeling strategies. Early indications of process drift, de-tected by a control mechanism when investigated via effec-tiveness modeling methods, may result in increased processunderstanding that may be used to revise the control strategyto sustain the gain. Used in isolation, neither MSPC nor anyother control method is likely to get final product qualityappreciably beyond 3 sigma capability.

Efficiency models such as PLS and PCR work best whentens or hundreds of measured inputs are available for eachproduction batch. They exploit the correlations between theinputs to produce reliable classification or prediction modelsof product quality. These techniques need to be used inconjunction with effectiveness modeling methods if the goalis to increase sigma capability of product quality throughincreased process understanding. Efficiency models providean effective basis for prioritizing the input variables forinclusion in effectiveness modeling activities. Used in isola-tion, efficiency models are likely to get capability in the regionof 3 to 4 sigma.

The focus of effectiveness modeling is to identify thecritical few inputs and to develop empirical models of theeffects of these on product quality that approximate thecausal relationships between inputs and product quality.This new knowledge is then deployed to reduce variation infinal product quality and achieve performance requirementsthrough revised process and material specifications andcontrols. Effectiveness modeling approaches applied to theprocess development of new products is otherwise calledQuality by Design (QbD). For new products, Quality byDesign is a good way of achieving six sigma quality perfor-mance. This approach ensures a high level of process under-standing along with cost-effective control strategies in manu-facturing that are based on measurement and control of thecritical few relevant inputs. Measuring and controlling ev-erything that can be measured increases production costsand cycle times unnecessarily. Compared to process develop-ment of new products in R&D, effectiveness modeling ofmature manufacturing processes requires a little more caredue to the inherent correlations in measured inputs and thelimited range over which inputs are varied. With appropriateconsideration of these constraints, effectiveness modelingcan deliver increased process understanding and higherlevels of sigma capability, along with specifications andcontrols concerning the relevant few process and materialinputs.

Efficiency modeling along with advanced control strate-gies, such as EPD, APC, and MSPC, are classified as PATControl Methods. Effectiveness modeling and QbD are clas-sified as Six Sigma Methods. Lean Data Analysis or Prag-matic PAT as illustrated in the case studies is an appropriateblend of simplified Six Sigma Methods and PAT ControlMethods. The best blend for a particular manufacturingfacility depends on where the facility is positioned within thematrix.

Various factors need to be considered when determiningthe best mix of data analysis methodologies to enable in-creased process understanding for a particular manufactur-


Lean Data Analysis


Figure 5. Recursive partitioning decision tree.

V9 V15 V18 V21 % at stage 3-4

0.25 0.5 0.9 0.7 29.24

0.25 0.35 0.9 0.95 27.10

0.25 0.5 0.65 0.95 17.88

0.25 0.2 0.65 1.2 21.93

0.2 0.2 0.9 0.7 33.24

0.3 0.2 0.4 0.95 22.47

0.2 0.2 0.4 1.2 19.46

0.3 0.2 0.9 0.7 34.69

0.3 0.5 0.9 1.2 17.83

0.2 0.5 0.9 1.2 14.61

0.3 0.35 0.65 0.7 25.96

0.25 0.2 0.4 0.7 23.77

0.3 0.5 0.4 0.7 20.36

0.2 0.5 0.4 0.7 19.52

0.25 0.35 0.4 1.2 14.80

0.2 0.35 0.65 0.95 20.56

Table B. DOE worksheet.

ing facility, including product maturity, sigma capability,and the extent and relevance of measured and unmeasuredinputs. A mature manufacturing facility is unlikely to haveextensive inline, at-line, or online measurement tools inplace; therefore, greater emphasis on effectiveness modelingand Six Sigma approaches will be appropriate. For newermanufacturing processes with an extensive number of mea-sured inputs, there may be a greater mix of PAT ControlMethods although appropriate use of effectiveness modelingmethods also will be required to ensure fundamental under-standing of the process. After positioning a particular manu-facturing facility within the maturity matrix, it is possible tomap out short- and long-term goals of a quality improvementor PAT investment and define a high-level road map forachieving those goals.

Case StudiesThese are fictional case studies based on simulated data,copies of which are available on request from the author. Thescenarios around which the data have been simulated arefairly typical of the data sparse situation of mature manufac-turing and data rich position of some manufacturing facilitiesof new products. These simulated situations are not based onany particular case, but they do try to reflect the realities ofthe two situations and by so doing provide data analysisexamples that are easier to apply and understand for themainstream.

Case Study 1: Mature Manufacturing with FewMeasured InputsThis case study concerns a manufacturing facility that hasbeen producing an established product in the form of soliddoses at various concentrations for several years. Currentmeasurement systems are based on storing finished mate-rial, while offline QA tests are performed to assure thefinished product meets the performance specification.

The case study focuses on investigating the process for

tablets produced at a single concentration. The key perfor-mance metric is 60-minute mean dissolution, which must beno less than 70%. Historically, 16% of production batches failto meet the 60-minute dissolution requirement and QA inves-tigations into these lot failures rarely find an assignablecause.

In this data-sparse scenario, the manufacturing team wascommissioned to investigate the process and dramaticallyimprove sigma capability. The team adopted a variety ofeffectiveness modeling techniques, starting with processmapping, which was used to identify the key process stepsand to identify the set of inputs that were most relevant to theproblem and easy to collect information about retrospec-tively. The results of this process-mapping exercise are docu-mented in Figure 3; the set of inputs that might have animpact on 60-minute mean dissolution and are easily col-lected retrospectively are identified in black type. Inputsoccurring above a process step represent material properties;inputs occurring below a process step represent processparameters.

Data on the inputs identified in black type along withmean dissolution were collated for the last two years ofproduction batches, which resulted in a data set consisting of90 rows and 19 columns.

Exploratory data mining methods as indicated in Figure 4were deployed to help determine the inputs most stronglyassociated with dissolution failures. Part (a) of Figure 4shows simple histograms of each variable with the failingbatches identified in dark green. This shows a particularlystrong relationship between screen size in the milling stepand batch failure with a larger screen size resulting in agreater proportion of failures – presumably a larger screensize results in larger API particle size and these largerparticles take longer to dissolve. Spray rate in the coating


Lean Data Analysis


step also has a strong association with batch failures; inparticular, lower spray rates have no batch failures. Part (b)of Figure 4 shows the multivariate relationship of the 18variables in a single graph called a parallel coordinates plot.There are 90 lines on this graph – one line for each productionbatch – with the path of each line representing the processingconditions of each batch along with the resulting dissolutiontest result for that batch. By design, there is no scale on they-axis; instead, a plotting range for each variable is selectedto show the span of data values for that variable. The failingbatches are identified in red and the passing batches in blue.The values of each variable for one of the failing batches isillustrated in bold red, showing API with a small particle size,processed with a mill time of seven minutes, screen size offive, and so on. Part (c) of Figure 4 shows the parallelcoordinates plot with all failing batches identified in bold red;this version of the graph enables the identification of process-ing conditions associated with passing or failing batches.Some processing conditions associated with passing batchesare circled, e.g., high mill time, low and high blend time (withone exception), high blend speed, low and high force, low andhigh coating viscosity, high exhaust temperature, and lowspray rate appear to be more favorable processing conditions.Potential interactive effects of two or more inputs on dissolu-tion can be investigated on both graph types by coloring thepoints according to different rules. For example, to investi-gate the size of the interactive effect of blend time and blendspeed on mean dissolution, a cut point would be defined foreach input (giving high and low values of each input) and then

Figure 6. Multiple regression analysis summary.

color the points differently for the four combinations. Wewould then look to see if there is an appreciable change inmean dissolution across the four combinations of the twovariables. One potential draw-back of the parallel coordi-nates plot is that it is not as effective at exploring the effectsof categorical variables such as API Particle Size, ScreenSize, and Coating Supplier, due to the inability to display theproportion of failing/passing batches processed at each levelof a categorical variable. Nonetheless, it is a good visual datamining tool that helps identify key continuous variables forfurther investigation.

Another useful exploratory data mining method is recur-sive partitioning. This method repeatedly partitions dataaccording to a relationship between the input variables andan output variable, creating a tree of partitions. It finds thecritical input variables and a set of cuts or groupings of eachthat best predict the variation in batch failures. Variations ofthis technique are many and include: decision trees, CARTTM,CHAIDTM, C4.5, C5, and others.

Figure 5 shows the resulting decision tree using recursivepartitioning to explore the main drivers of batch failures. Theright-hand branch of the decision tree shows that 47 of the 90batches were processed using a screen size of four or three inconjunction with a spray rate less than 404. All 47 batchespassed the dissolution test. At the other extreme, the left-hand branch shows that 10 batches were processed using ascreen size of five and a mill time of less than 11. Eight ofthese batches failed the dissolution test.

These exploratory data mining methods have collectively


Lean Data Analysis


Figure 7. Simulation study to investigate process robustness.

identified a subset of inputs – Mill Time, Screen Size, BlendTime, Blend Speed, Force, Coating Viscosity, Exhaust Tem-perature, and Spray Rate – worthy of further investigation.The methods have several advantages over conventionalstatistical approaches, including:

1. ease interpretation and communication, enabling every-one to gain insight into the potential key relationships indata

2. inform the mainstream about the principles of statisticalthinking, particularly those of modeling variation in pro-cess outputs and identifying the key drivers of processvariation

The effects of this subset of input variables upon 60 minutemean dissolution were investigated in more detail usingmultiple regression in Figure 6. The graph at the top showsthat the model predicts actual values of dissolution reason-ably well, and the effects tests summary shows that all, butblend speed, force, and exhaust temperature significantlycontribute to variation in 60 minute mean dissolution at the5% level. Further mill time and screen size have an interac-tive effect and mill time has a quadratic effect on 60 minutemean dissolution as illustrated in the interaction profile. Theprediction profiler at the bottom of Figure 6 shows thedirection and strength of the effects of each factor on 60minute mean dissolution with the optimum settings of eachinput given in red. Since blend speed, force, and exhausttemperature do not significantly affect 60 minute meandissolution at the 5% level, any value of these three inputswithin the observed range are acceptable. The anomaly of onefailing batch with a high blend time in Figure 4(c) was due toa low mill time (10 minutes) and screen size of five.

To investigate process robustness against the proposed

new set points, the simulation illustrated in Figure 7 wasperformed. Using the multiple regression model as the trans-fer function between the key process inputs and 60 minutemean dissolution, 1000 simulations were performed withmean settings close to the best setting of the inputs withtolerances as indicated in Figure 7. The target and toleranceused for blend speed, force, and exhaust temperature was thesame as currently used in manufacturing since 60 minutemean dissolution was robust to this level of variation in thesethree inputs. The target and tolerance of mill time, screensize, blend time, coating viscosity, and spray rate wereadjusted per the knowledge gained via multiple regression toensure acceptable distribution of 60 minute mean dissolutionrelative to the lower specification limit of 70%. The simula-tion confirms expectations of consistent product performancewith a predicted Cpk of 1.4 (equivalent to a sigma level of 5.6).The proposed solution is wholly within the bounds of thecurrently validated process.

Case Study 2: New Production Facility withMany Measured InputsThis case study concerns a relatively new manufacturingfacility that has been producing commercial batches of aninhaler product for a couple of years. Extensive inline mea-surement systems were designed into the facility, resultingin a data-rich environment of 520 measured inputs. The first30 inputs are processing parameters of the milling, blending,and packaging steps; variables 31-100 are properties ofingredient 1; variables 101 to 170 are properties of ingredient2; and the remaining variables are properties of ingredient 3(active ingredient).

The key performance metric is a percentage of a given dosereaching stage 3-4 of a cascade impactor, which must bebetween 15% and 25%. Since the start of commercial produc-tion, 240 batches have been manufactured, approximately


Lean Data Analysis


Figure 8. Recursive partitioning decision tree identifies inputs most strongly associated with variation in % at stage 3-4.

14% of which have failed to meet the performance require-ment of the cascade impactor test. QA investigations intothese batch failures have been unable to identify any obviousassignable causes.

In this data-rich scenario, the manufacturing team com-missioned with investigating the process and dramatically

improving sigma capability adopted a variety of effectivenessand efficiency modeling techniques. Figure 8 illustrates theresults of recursive partitioning to help determine the inputsmost strongly associated with the percentage of a dose reach-ing stage 3-4.

The decision tree shows how the distribution of % at stage


Lean Data Analysis


Figure 9. Tree map of PLS model coefficients.

3-4 changes according to splits derived from the levels of twoprocess variables - V21 and V15. The left-hand branch of thedecision tree shows that when V21 >= 1.1, the distribution of% at stage 3-4 has a mean of 14.3 and standard deviation of4.6. The graph at the top of Figure 8 shows a greater propor-tion of rejected batches (red points) than passing batches(blue points) in this subgroup. The middle branch of thedecision tree defined by V15 >= 0.4 and V21 < 0.9 yields adistribution of % at stage 3-4 with a mean of 20.4 andstandard deviation of 2.1. Just one of the 36 batches pro-cessed this way results in a rejected lot.

The decision tree model was built by excluding 72 (30%) ofthe 240 batches, these excluded batches were used to validatethe decision tree. The partitions defined by the levels of V15and V21 in the decision tree explain 40% of the variation in% at stage 3-4 of the 72 batches that were excluded frombuilding the model. Nine of the 72 model validation batchesmet the criterion of V15 >= 0.4 and V21 < 0.9 with all nine ofthese batches passing the compliance test. Thus, V15 andV21 are verified as being strongly associated with some of theexcessive variation in % at stage 3-4 and are potential driversof resultant batch failures.

Recursive partitioning is a good visualization tool to helpall consumers of process data see and communicate under-standing about the dominant drivers of product variation;however, the method requires a large number of batches toreliably build decision trees with a greater number of branches(possibly utilizing other input variables to define the addi-tional branches). Nonetheless, it is a great tool for aidingunderstanding and communication about the potentiallydominant drivers of a problem.

To determine if there are additional input variables thatmay enable us to further reduce variation in % at stage 3-4,a PLS analysis was performed using cross validation. Thecoefficients from the resulting model are illustrated in Figure9. The area of each rectangle of the Tree Map is in proportionto the size of the PLS model coefficient for the correspondinginput variable. Blue rectangles stand for negative coeffi-cients and red for positive coefficients. Two dominant factors,in addition to V15 and V21, are identified as V9 and V18. Thesign of these four model coefficients tell us that increasing thevalues of V9 and V18, and reducing the values of V15 and V21will result in higher values of % at stage 3-4.

Table A compares the observed vs. predicted result of thebatch acceptance test, based on 72 batches that were ex-cluded from the model fitting. The PLS model predicts batchperformance of the 72 batches excluded from the model withthree misclassifications. However, as a prediction model ofbatch performance, the center branch of the decision tree inFigure 8 works just as well as a predictive model of batchfailures in helping to reduce future occurrences of batchfailures. With nine of the 72 model validation batches meet-ing the criterion of V15 >= 0.4 and V21 < 0.9 and with all ninebatches passing the compliance test, the recursive partition-ing decision tree appears to be a simpler and sufficientpredictor of batch failures.

To investigate in greater detail the effects of the input

factors V9, V15, V18, and V21 on % at stage 3-4, a D-optimalDOE with a full quadratic model was performed. The result-ing DOE worksheet is presented in Table B.

Summary results for the DOE analysis are presented inFigure 10, which shows significant linear effects of all fourfactors and a significant interaction between V15 and V21 atthe 5% level. The direction of the relationship between % atstage 3-4 and each of the four inputs is in agreement with thesign of the coefficients from the PLS model. Multiple opti-mum solutions that get the mean of % at stage 3-4 on targetexist, one of which is to operate close to V9=0.25, V15 = 0.4,V18 = 0.9, and V21 = 1.2. To explore the viability of thissolution, the regression model was used to simulate thepropagation of variation from the four inputs when set at theabove values with a tolerance as indicated in the bottom partof Figure 10, and random batch to batch variability defined bya standard deviation of 0.5 (more than twice the standarddeviation of the residuals in the fitted regression model). Thispredicts a distribution of % at stage 3-4 wholly within therequired range (Figure 10) with a predicted sigma qualitylevel of 4.8. In practice, before accepting this solution, itwould be necessary to validate the model and predictedbehavior with model validation batches performed at orwithin the proposed tolerance of the four process settings.

SummaryThe blend of three key technology enablers – measurement,data integration, and data analysis systems – required toimprove product quality through increased process under-standing, depends upon the circumstances of the particularmanufacturing facility.

Mature manufacturing facilities are unlikely to have ex-tensive inline, at-line, or online measurement systems inplace for tracking process inputs. Thus, the adoption ofeffectiveness modeling is a way to improve product qualitythrough increased process understanding. The focus is toidentify the critical few inputs and to develop empiricalmodels of the effects of these on product quality that approxi-mate the causal relationships between inputs and productquality. These models are then deployed to reduce variation


Lean Data Analysis


Figure 10. DOE summary analysis.

in final product quality and achieve performance require-ments through improved process and material specificationsand controls. A subset of some visual and statistical effective-ness modeling techniques in the context of mature manufac-turing was illustrated in Case Study 1.

Manufacturing facilities for newer products are morelikely to have extensive inline, at-line, or online measure-ment systems for tracking process inputs. The path to im-proved product quality through increased process under-standing is a combination of efficiency and effectivenessmodeling. Efficiency modeling methods are deployed to pre-dict product performance, define some temporary controls toreduce batch failures, while effectiveness studies are con-ducted. The efficiency models also help identify and prioritizethe inputs to be investigated in detail through effectivenessmodeling techniques. The combined use of efficiency andeffectiveness models may help reduce the number of processinputs that are routinely measured to the critical few if this

helps accelerate cycle time or reduce other risks. A subset ofsome efficiency and effectiveness techniques in the context ofa data-rich measurement environment was illustrated inCase Study 2.

Quality by Design is effectiveness modeling applied inprocess R&D, where it is possible to explore wider ranges ofprocess inputs. The goal is to design a robust process thatidentifies the critical few inputs and tolerances for each keyinput that must be maintained in manufacturing. From ameasurement systems viewpoint, the goal is to define the fewinputs that must be measured or controlled in manufacturingand to achieve this knowledge through a high level of processunderstanding.

Simplifying data analysis and reporting is critical if morepeople in process development and manufacturing are tointerpret and communicate around models that enhanceprocess understanding. This article has introduced visualmodeling methods that are easy to deploy for mainstream


Lean Data Analysis


users and help them apply the principles of statistical think-ing, particularly those of modeling variation in process out-puts and identifying the key drivers of process variation.

About the AuthorDr. Malcolm Moore has worked with cli-ents to integrate statistical methods andsoftware into R&D, quality improvement,defect reduction, cycle time reduction, andcorporate Six Sigma consulting activities fora variety of industries, including pharma-ceutical and semiconductors. Prior to joiningSAS UK, he worked at Light Pharma, BBN,

Astra Zeneca, and lectured in medical statistics at NewcastleUniversity. He is an expert in design of experiments, andreceived his PhD in design of non-linear experiments atLondon University. He can be contacted by e-mail at:[email protected].

SAS Institute, Whittington House, Henley Road,Medmenham, Marlow SL7 2EB, United Kingdom.


Real Time Process Determination


This articleintroduces astate-of-the-artmethod to trackconditions forwet processesthat providescontinuousoperatingguidancethrough acombination ofmeasured andcalculatedvalues.

Introduction to Real Time ProcessDetermination

by Kim Walter

Many processes in the pharmaceuti-cal industry require mixing of ac-tive pharmaceutical ingredientswith inactive powders to transform

the mixture into useful solid dosage material.Frequently, the process incorporates the use ofwater or organic solvents, the so-called wetprocesses such as drying of granules from amechanical mixer, spray granulation of theproduct in a fluid bed, pellet coating for tastemasking, pellet coating to modify drug releasecharacteristics, and powder laying of the activedrug on an inactive powder.

To produce to specification, wet processesgenerally require control of the humidity in theprocess chamber, although in some cases, con-trol of temperature or partial pressure is criti-cal. For organic solvents, instruments cannotmeasure the relative humidity inside the pro-cess vessel. Common practice is to use a trialand error procedure, changing process condi-tions until all parameters are within toler-ances. This procedure is both difficult and waste-ful, depending mostly on the insight and deci-sions of process developers and experiencedoperators.

Real time process determinationTM is a state-of-the-art method to track conditions for wetprocesses that provides continuous operatingguidance through a combination of measuredand calculated values. It may be used to controla continuous or a batch process through all ofits steps and transitions, regardless of varia-tions in ambient or process conditions. It givesthe skilled process developer easy-to-interpretinformation in the form of a chart that guidesthe decision process.

Process VariablesPharmaceutical production demands consis-tent results, which are very difficult to achievewith batch processes since each batch is slightlydifferent. To apply real time process determi-

nation, the target conditions must be defined -endpoint humidity for drying, solvent encapsu-lation and applied membrane characteristicsfor pellet coating, residual moisture in tabletpressing, etc. Theoretically, if the process vari-ables are consistently on target, the specifica-tions of each batch will be identical.

However, running a process in precisely thesame way time after time is impossible, andeven very small deviations can have a signifi-cant influence on the end result. Different re-sponse times for the process variables also maybe a factor when adjustments are made duringproduction - a change in the spray rate affectsthe process nearly instantly, while a change inthe inlet temperature has a much longer re-sponse time.

A preferred thermodynamic condition existsin the process chamber in order to achieveconsistent results. Using coating and spraygranulation as an example, variables are feedrate, inlet temperature, and spray rate of sol-vent. The thermodynamic condition can be givenas a particular combination of relative humid-ity and temperature - the target condition. Inthis case, only two process variables must becontrolled during the process instead of allthree. In some processes, one of these condi-tions, relative humidity or temperature, maybe more critical than the other. Therefore, thecritical variable becomes the primary targetcondition and the not-so-critical variable be-comes the secondary target condition, enablingthe critical target condition to be reached faster.If there is a deviation in the target temperaturein the process chamber - the process variablewith the longest response time - the spray ratecan be adjusted, which has the shortest re-sponse time, which will change the tempera-ture nearly instantly. The process gas flow ratecan be changed, which has a median responsetime, if we want the target temperature to reactover a short time interval.

Reprinted from







Usually, only the process variables are controlled, withoutknowing the thermodynamic target. However, if thermody-namic laws are applied to the equipment, the thermodynamiccondition in the process chamber can be determined, and byexperimenting with different conditions, the critical thermo-dynamic condition can be determined. With the critical thermo-dynamic condition specified, the scale-up and transfer fromequipment to equipment become easier. Choosing the ther-modynamic condition for controlling the process, instead ofusing only the single loop recipe control, will ensure morereproducibility of both batch and continuous processes. Thisis the basic objective of real time process determination.

The Thermodynamic ApproachThe thermodynamic approach is built on two fundamentallaws, conservation of mass and conservation of enthalpy. Thelaw of conservation of mass says that the change of massinside a closed system in time is equal to the flux of massentering the closed system minus the flux of mass exiting.The law of conservation of enthalpy says that the change ofenthalpy inside a closed system in time is equal to the flux ofenthalpy entering the closed system minus the flux of en-thalpy exiting.

Figure 1 shows the heat and mass balance of the thermo-dynamic system used on the process equipment. The controlsurface represents the equipment walls. Through the inletenters some mass flow of process gas, atomizing gas, sprayliquid (which may consist of several solvents), solvent vaporin the process gas, and solids suspended or dissolved in thespray liquid. Through the outlet flows the process and atom-izing gases, which will contain some solvent vapor. Thedifference in the mass flow of solvents from the inlet to theoutlet is what is added or removed from the product over time.The amount of solvent inside the equipment, which is notevaporated, is depicted as an area on the drawing.

The enthalpy balance consists of the enthalpy flowing outminus the enthalpy flowing into the equipment. The diver-gence in the enthalpy flow is the change in the enthalpy levelinside the equipment and the heat flow “Q” through theequipment wall. The term “W” depicted on the drawing is thework done on the system.

The last thermodynamic term we need to understand isadiabatic. A process is adiabatic when the heat change insidethe closed system happens without exchange of heat with thesurroundings. When the process is adiabatic, the enthalpy isconstant. So if the heat loss from the equipment is identified,

Figure 1. Thermodynamic model for real time process determination.




the exchange of the enthalpy inside the equipment walls canbe calculated.

Equations to Determine Physical ValuesFive physical values are important to the process:

1. The temperature, which can be measured.

2. The pressures, including ambient pressure, total pres-sure, and partial pressures of each component. The totalpressure is the sum of the partial pressures of eachindividual gas and the partial vapor pressure of the sol-vents. The partial solvent pressure (vapor pressure) is theamount of the particular solvent present in the gas. Thesaturated vapor pressure is the maximum pressure theparticular solvent can have at a given temperature.

3. The concentration of solvent, also called the specific hu-midity – the mass of the particular solvent dissolved permass unit of gas.

4. The dewpoint temperature, at which, for a given solventconcentration (specific humidity) and total pressure, thegas/solvent mixture is saturated.

5. Specific heat capacity - the amount of heat necessary toincrease a mass unit of the product, the particular solventas vapor or liquid, and the gas, one degree.

To connect the thermodynamic laws, the equations of the fivephysical values must be used. The relative humidity iscalculated as the ratio between the actual specific humidityand the saturated specific humidity in percent for a given gastemperature. The relative humidity also is the ratio betweenthe actual vapor pressure and the saturated vapor pressurein percent for a given gas temperature. The relative humidityfor the solvent {i} is expressed by the equation:

p(partial - pressure)solvent{i}jsolvent{i} = ________________________________

p(saturated - pressure)solvent{i}

For water, there are instruments that measure the electricresistance of the air, which depends on the concentration ofwater vapor. This measurement, in combination with the airtemperature, enables the calculation of the relative humidityfor water vapor. Since instruments cannot measure therelative humidity of an organic solvent; it must be calculated.

The relation between the relative and the specific humid-ity can be calculated:

Msolvent jsolvent p(saturated)solventxsolvent = ________________________________

Mgas[ptotal - jsolvent p(saturated)solvent]

where Msolvent is the molecular weight for the particularsolvent and Mgas is the molecular weight of the process gas.

The enthalpy, the heat content of a mass unit of gas, iscalculated as the specific heat capacity of the gas plus the sum

of the specific humidities of the solvents times their specificheat capacities, taking the entire sum times the temperatureplus the specific humidity of the solvents times their heats ofevaporation.

ó T i=n i=n

h = ô [cpgas (T) + S x{i}solvent cpsol (T)]dT + S x{i}solventrsolvent(Tref)õ Tref

i=1 i=1

The term x{i} is the mass ratio of the particular solvent {i}dissolved as vapor in the process gas. The total pressure of thegas plus the sum of the partial pressures of the solvents,which is the ambient pressure, is constant. When the solventsare dissolved in the gas, the gas volume will expand, loweringthe density of the gas. If the solvent is water, the change indensity of the gas is negligible since the amount of watervapor that can be dissolved before the mixture becomes

M = sum of mass (product, equipment, coat,solvent, and gas) inside the control volume [kg]

t = time [sec]min = mass flow into the control volume [kg/sec]mout = mass flow out of the control volume [kg/sec]u = inner energy [Joule/kg]h = enthalpy [Joule/kg]V = velocity [m/sec]s = entropy [Joule/kg]Q = heat [Joule/sec]W = work [Joule/sec]Gs = entropy production [Joule/Kelvin*sec]jsolvent{i} = relative humidity for the solvent {i}

[%{saturated}]p(partial)solvent{i} = partial pressure of solvent {i} [Pascal]p(saturated)solvent{i} = saturated pressure of solvent {i} [Pascal]ptotal = sum of the gases and solvents partial pressure

[Pascal]xsolvent = general mass ratio between particular solvent

and the gas [kg/kg]Msolvent = molecular weight of the solvent [kg/kmol]Mgas = molecular weight of the process gas [kg/kmol]h = enthalpy, heat content per mass unit of gas

[Joule/kg]T = temperature [Celsius] or [Kelvin]. If Kelvin is

used in the enthalpy equation, all values haveto be expressed in Kelvin

Tref = chosen reference temperature, normally 0°Ccpgas = specific heat capacity of the process gas

[joule/kg*8C] cpsol = specific heat capacity of the solvent vapor

[joule/kg*8C]s{i}solvent = mass flux of the solvent {i} entering the

control volume as liquid into the control volume[kg{solvent}/sec]

xproduct = specific humidity on product surface in acoating process

xambient = ambient specific humidity [kg{solvent}/kg{gas}]

mprocess-gas = flux of process gas mass flow rate entering orleaving the control volume [kg{gas}/sec]

rsolvent = heat of evaporation for the solvent at thereference temperature [Joule/kg{solvent}]

Table A. Nomenclature.

•

•

•

•

•

•

•




Figure 2. Automation control screen for typical multi-purpose process equipment.

saturated is small. When a mixture of solvents is presentduring the process and some of the solvents are volatile, thechange in gas density has to be taken into consideration. Theenthalpy gives the value of the gas and the vapor heatcontent, calculated from a chosen reference temperature, Tref,which normally is the triple point of water 0°C. The termrsolvent in the equation for the enthalpy is the evaporation heatfor the particular solvent {i} at the reference temperature.

All the values in these three equations are physical mate-rial properties and are a function of the temperature. Thevalues have been measured by many people over the lasthundred years, published in tables, and organized as aphysical-chemical database. Over time, many have convertedthe physical-chemical data table into mathematic formulaewith the use of different approximations. The equations thatapproximate the physical-chemical data seem at first glancecomplicated. However, with current computer capacity, thetask is possible.

With a further analysis of the three equations, it can beconcluded that if two of the four values are known, the two

other values can be calculated. However, if all four values areknown and two values are enough to determine the thermody-namic condition, it produces six different ways to solve theequations. This sounds strange, because if the values areknown, why calculate them? The answer is: there is moreinformation from the normal control system than is needed todetermine the thermodynamic condition. This allows us todetermine the unknown values in the thermodynamic models,such as heat loss, heat exchange, measuring errors, and so on.

Determining the Thermodynamic ConditionFor a given process, the inlet, product, and outlet tempera-tures from the control system can be obtained. The flow rateof process gas is known, as is the concentration of solvent inthe inlet gas (known from the inlet gas dewpoint tempera-ture). Again, using spray granulation or coating as an ex-ample, the amount of solvent added to the process is known.The ambient pressure is either measured or can be assumedto be normal atmospheric pressure, 1013 hPa.

The first calculated value is the enthalpy of the inlet gas.




The equipment is started empty, with a given inlet tempera-ture and process gas flow rate, and observed. In an adiabaticprocess, the product and outlet temperatures should rise tothe inlet temperature as the equipment warms up. This is notthe case; the system or the equipment is non-adiabatic for tworeasons: the heat loss through the equipment wall and theheat transfer from the process gas to the equipment, both ofwhich change the temperatures. The product and outlettemperatures will start out lower than the inlet temperatureand gradually increase as the system reaches a steady state.After some time has elapsed, the product and outlet tempera-tures will approach fixed values. At this point, the outlettemperature will normally be lower than the product tem-perature. Two important inherent features of the particularequipment being tested has been observed: the time responseand the effect of the heat loss through the equipment wall,both unique for this equipment. By repeating this procedurewith different process gas flow rates and inlet temperatures,the heat loss of the particular installation can be determined.If the same procedure is executed with different products andproduct loads, information is determined about the totalsystem’s heat loss and time response. The heat loss can thenbe calculated, so with both the equipment running empty andwith product being processed, the real inlet temperature canbe determined.

The next investigation should be the accuracy of theprocess gas measurement. Measuring the flow rate is difficultand frequently inaccurate. The best example to use in aninvestigation of the gas flow measurement is coating. In thecoating process, processing time is normally long enough forthe equipment to reach steady state. With the knowledgeabout the heat loss, the real inlet enthalpy is known. Incoating, a small amount of residual solvent is encapsulated inthe coat; therefore, the process is close to adiabatic. Assumingan adiabatic process, the enthalpy of the inlet and the en-thalpy on the surface of the product must be the same. Usingthe equation for the inlet condition with the modified inlettemperature, the specific humidity of the inlet gas, and therate of process gas, the inlet enthalpy is known. Measuringthe product temperature and the spray rate, the relative andspecific humidity on the product surface can be determined.The specific humidity on the product surface is the ambientspecific humidity and the added solvents from the spraydivided by the process gas mass flow rate:

i=n•S s{i}solvent

i=1xproduct = xambient + ____________

•mprocess-gas

Performing this procedure with different process gas flowrates and spray rates will reveal the deviation between themeasured process flow and the actual flow rate. With the twocorrections, the heat loss and the deviation between themeasured and calculated process gas rate, the relative hu-midity in the process chamber can now be calculated at anygiven time.

The customary control procedure in coating is to adjust the

spray rate according to the product temperature. The producttemperature is governed by the spray rate and the timeresponse due to the thermal heat exchange between theprocess gas and the equipment and product mass. The prod-uct temperature is measurable and real and the calculationof the specific humidity, based on the equation, will give theactual relative humidity on the product surface. Combiningthe “two” relative humidities, one based on constant enthalpyand one on spray rate/process gas mass flow, will provide theadiabatic ratio, or how far the process is from adiabatic. Whenthe steady state is reached, the adiabatic ratio equals one.

Using the Thermodynamic Calculationto Guide the Process

The thermodynamic calculations utilize the information fromthe control system. Determining the numerical values fromthe equations is complex and time consuming, so the obviouschoice is to use a computer program. When this is written anddata are input from the control system, the logical step is tobring the computer program and the control system togetheras one unit and calculate the thermodynamic conditions inreal time. With the real time calculation, the program alsocan calculate how much each of the process values has to bechanged to bring the process to the target conditions. Theprogram calculates all possible changes and the consequenceof each single change. There are a total of 12 changes andeight consequences to choose among, depending on whichfinal thermodynamic condition the process demands. Thefinal challenge is to display the possible choices in a compre-hensive way.

Instruments make available information visible, puttingthe operator in the best possible position for making anoptimal decision. Figure 2 shows a control screen for a typicalmulti-purpose process scheme. This familiar configurationuses a combination of flow diagram and equipment schematicto display measured physical conditions such as tempera-ture, pressure, and flow rates, as well as setpoints for processvariables.

To display calculated conditions in addition to measured,a real time process determination screen may be added to thecontrol panel - Figure 3. In the case of bottom spray coating,where the relative humidity in the process chamber has thehighest priority, the value can be calculated and displayed.Because the change in the relative humidity is a result ofchanges in three other process variables (inlet temperature,solution spray rate, and process gas flow rate) the deviationmeter shows the results from the calculation and displays theproper action to take. The operator can, in a single glance atthe meter, take in all three values which can be changed (thegold-colored lines), and how much each of the values has to bechanged to reach the desired condition.

Experience dictates that the combination of analog anddigital displays is the best way to notify the operator aboutcurrent and desired conditions. On the deviation meter, thethree set points are shown both graphically and digitally.Because bottom spray coating is a dynamic process with along response time for one of the observed values (the process




chamber temperature), the operator also has to be informedabout the time development of the process. This also is acombination of analog and digital information. The digitalinformation is given by the adiabatic ratio. This is calculatedas a ratio between the evaporation rate (based on a combina-tion of the spray and process gas rates) and the measuredtemperature difference between the inlet and process cham-ber temperature. This number provides information on thecurrent adiabatic ratio, but to make the dynamics of theadiabatic ratio visible, a new method had to be developed.

The Adiabatic Ratio InstrumentThe layout of the adiabatic ratio instrument has been influ-

enced by the opinion that few persons can comprehend valuesthat are not given in a linear form. Some observers even go asfar as to say that no person can comprehend magnitude. Weseem all better equipped to comprehend linear changes thanchanges in magnitude.

Relative humidity is based on the saturated vapor pres-sure as a function of temperature. The saturated vaporpressure increases with around the sixth power of tempera-ture; therefore, a small increase in temperature creates alarge change in relative humidity. The aim has been to finda way to display the relative humidity in linear form. Thesolution is a dynamic psychrometric chart or the dynamicspecific humidity diagram - Figure 3.

The Thermodynamic Equationsh, (h+½V2+n), where h is the heat content of thematerial entering and leaving the control volume.

The third equation, increase in entropy or the secondlaw of thermodynamics, expresses that the change ofentropy in time inside the control volume plus the flux ofentropy out minus the flux of entropy in, is larger than theheat conveyed over the boundary of the control volumedivided by the control volume’s absolute temperature.

•

d• •

Q_____ (Ms) + S (ms)out - S (ms)in > ____dt T

• •

•In order to balance the equation, the production of entropyinside the control volume must be added. The term Gs

expresses this production of entropy inside the controlvolume in joule/KelvinBsec. The M, m and Q are the sameas in the first and second equations. The s is the entropyfrom Gibbs equation. For the determination of the entropydifference between the entering and exiting flow, we lookat an example. For a simple compressible substance s =s(u,v), the entropy s is a function of the inner energy u andthe volume per unit of mass v. Differentiate the functionand we obtain:

æ ¶s ö æ ¶s öds = ç____÷ du + ç____÷ dv

è ¶u øv è ¶v øu

Using the thermodynamic definitions of temperature andpressure, we find:

1 Pds = ____ du+ ____ dv

T T

Therefore the difference between the entropy s{out} ands{in} is:

ó uout du ó vout Psout - sin =ô ____ +ô ____ dv

õ uin T õ vin T

The entropy s is expressed as joule/Kelvin.

Figure 1 represents the basic thermodynamic system fora control volume. The three equations are conservation ofmass, conservation of enthalpy, and increase in entropy.

The control surface is an imaginary boundary, chosenso that the fluxes crossing the boundary are known valuesor can be determined. The fluxes are the mass flow, åm,the sum of all gases, vapors, liquids, and solids flowing inand out of the control volume. The term W is the workapplied to the control volume. In this specific case of realtime process determination, the work applied is themovement of the process gas and the product inside thecontrol volume, in short: the pressure loss experienced bythe fan. The term Q is the heat passing over the boundaryof the control volume. In the case of real time processdetermination, the heat is the heat loss through theequipment wall.

The first equation, conservation of mass, expresses thechange of mass in time inside the control volume plus theflux of mass out minus the flux of mass in. This is equalto zero. The mass inside the control volume is representedby M, where M is the sum of the equipment wall, theproduct, the solid delivered to the product as coat orlayering material, and the solvent retained in the product,coat, or applied material (all the solvent that has notevaporated).

The second equation, conservation of enthalpy or thefirst law of thermodynamics, expresses that the change ofenthalpy in time inside the control volume plus the flux ofenthalpy out minus the flux of enthalpy in, is equal to thedelivered heat and work flux to the control volume. Theterm (u+½V2+n) is the energy. The u is the internalenergy of all the material inside the control volume exceptthe kinetic energy and potential energy. The numericalvalue of u is assumed to be zero at the temperature scalezero point. The term ½V2 is the kinetic energy of thematerial inside the control volume, where V is the velocity.The n is the chemical potential of the material inside thecontrol volume. The term (u+½V2+n) is expressed injoules. In the term representing the mass flow in and outof the control volume, the internal energy u is replaced by

•

•

•




Figure 3. Thermodynamics process screen.

Along an adiabatic line, the temperature and the specifichumidity are nearly linear. So if the values of the ambientcondition, inlet, process chamber, and outlet conditions arerepresented as the specific humidity and used as ordinate,the relations of these conditions are linear. Now the specifichumidity of each relative humidity value along the inletenthalpy line can be calculated. The 100 percent relativehumidity, or saturated condition, gives the maximum ordi-nate value. The abscissa value is time. From testing it isknown that the time response is from 20 to 45 minutes so thisis the length of the chart. The background of the chart is therelative humidity expressed as specific humidity.

The process conditions - ambient condition, inlet, processchamber, and outlet condition - are shown in relation to therelative humidity. As the inlet temperature increases, thesaturated value of the specific humidity increases also, so thedynamic psychrometric chart grows. When the inlet tempera-ture decreases, the saturated value of the specific humidityalso will decrease, so the chart shrinks. When the process isadiabatic, the specific humidity based on the spray/processgas flow rate and the specific humidity of the process chamber

temperature will be equal, so the two curves will overlap.When the process is non-adiabatic, the specific humidity ofthe combined spray/process gas flow rate and the specifichumidity of process chamber temperature will be two differ-ent values, so the two curves will be separated.

With one glance at the dynamic psychrometric chart, theoperator can evaluate if the process is adiabatic or non-adiabatic, and how far from adiabatic the process is in thecurrent situation. Thus, the operator sees information in alinear form indicating how much the spray rate can be in-creased by evaluating the distance from the ambient conditionto the curve showing the specific humidity of the spray/processgas flow rate with the specific humidity line representing thetarget relative humidity. People who have worked with thesystem find the dynamic psychrometric chart easy to under-stand and say that it makes the decision process fast and easy.

An Example of Real Time ProcessDetermination

Using the example of a bottom spray coating process, thethermodynamics process screen contains both actual and




target process variables in both graphical and digital form -Figure 3. The deviation meter at the right shows that the inlettemperature (gold line) should be increased 15% to achievethe desired thermodynamic condition. Alternatively, the so-lution spray rate could be decreased less than five percent orthe volume flow rate could be increased by less than fivepercent. Below the deviation meter are the calculated nu-merical target setpoints for inlet temperature, solution sprayrate, and volume flow rate.

The current process values are given at the left side of thescreen, below the graph. The inlet temperature is 70.2°C, andthe deviation meter shows that a change of the inlet tempera-ture to 71.4°C will give the desired relative humidity. Like-wise a change in solution spray rate from the actual 25.0g{solution}/min to the target 24.3 g{solution}/min will pro-duce the desired condition. A third possibility is to change thevolume flow rate from the current 202 m3/h to 207.6 m3/h.

The graph on the left side is the dynamic specific humiditydiagram (Walter diagram). All the process values are calcu-lated as the specific humidity along the adiabat that goesthrough the inlet condition. The violet line is the ambientcondition, less than 10 g{water}/kg{air}, which is a dewpointtemperature of 14°C. The green line is the addition of wateror solvent from the solution spray rate. Until 12 minutes ago,the green line was overlapping the violet line, which indicatesthat no spraying was occurring.

The black lines represent where the relative humidityvalues cross the constant enthalpy line or adiabat. The upperblack line is the saturated humidity. The left side of the graphshows the specific humidity of the saturated line, which isapproximately 16 g{water}/kg{air}. This indicates that 6g{water}/kg{air} adiabatic could be added to the ambient gasbefore the gas would become saturated. The saturated linegrows over the next four minutes to a specific humidity of 31g{water}/kg{air} as the inlet temperature increases. The satu-rated line decreases when the inlet temperature passes thesetpoint as the temperature controller begins to take over. Thetemperature under-shoots and reaches the final inlet tempera-ture in less than four minutes. The other constant relativehumidity lines parallel the saturated specific humidity line.

The red line represents the product temperature. The linehas the same initial value as the violet line, which is theambient condition. The inlet gas flow is cooled by the inletduct and equipment plenum and delivers heat to the product.This is a non-adiabatic situation, because there is heatexchange with the surroundings. Initially, the product tem-perature is between the fourth and the fifth relative humidityline. As the heat is delivered to the equipment and product,the product temperature line decreases to the level of theninth relative humidity line. If the relative humidity could bemeasured inside the process chamber, it would show a simi-lar decrease during the elapsed time, real physical behavior.

The blue line represents a modified outlet temperature.Since the specific humidity of the outlet temperature is thesame as the specific humidity of the product temperature, thetwo lines should overlap. So the outlet temperature line ismodified by calculating the relative humidity of the outlet

temperature based on the cooling with constant specifichumidity (the product specific humidity), providing informa-tion on the relative humidity in the outlet of the processchamber where the filter is located. In top spray granulation,the outlet temperature line indicates when the relative hu-midity is high at the filter, which can produce tacky productthat will begin to block the filter. In bottom spray coating, theblue line can cross the red line, which means that the producttemperature is lower than the outlet temperature. Thisoccurs when the solvent is not totally evaporated from thecoat during the free flight of the product, causing an increaseof the solvent content in the product. The coat will becometacky and the product load will start to lump together,bringing the coating process to a standstill.

The graph shows that the spray pump was initiated eightminutes ago, because the specific humidity value increased,as shown by the increase of the green line. The differencebetween the violet line and the green line is the addition inspecific humidity due to the spraying. It was observed thatthe red line (the product temperature) stops decreasing afterthe green line has reached the same specific humidity as thespecific humidity of the product temperature.

The numerical values at the current condition are belowthe graph. The target relative humidity is 22.5%{saturated},which will, together with the inlet temperature and theambient specific humidity, give a product temperature of42.3°C. We also can see the product temperature that willgive a saturated condition, 24.8°C. The relative humiditybased on the product temperature is 15.9%{saturated} andthe outlet relative humidity is 22.9%{saturated}. The relativehumidity created by the spraying is 24.1%{saturated}.

Finally, the lower left corner shows general process infor-mation, such as the time that has elapsed, the amount ofsolution delivered to the process and the current rate ofspraying and process gas volume flow rate.

The product transport pressure is measured as the pres-sure difference between the empty equipment at the givenvolume flow rate and the current pressure loss over theequipment. With this value, it is possible to calculate howmuch product is moving around in the equipment, using thefirst law of thermodynamics, the conservation of enthalpy.

References1. Ebey, G.C., “A Thermodynamic Model for Aqueous Film-

Coating,” Pharmaceutical Technology, (1987), April.2. Okhamafe, A. O., and York, P.,” Characterization of Mois-

ture Interaction in Some Aequeous-Based Tablet CoatingFormulations,” Journal of Pharmacy and Pharmacology,(1985), pp. 37, 385-390.

3. Hyland, M., and Naunapper, D., Process DevelopmentGroup, Goedecke, A.G., and Glatt, “Continuous Control ofProduct Moisture Content in Drying Process, Drug madein Germany,” (1988), No. 1, p. 31.

4. Wantano, S., Hiroko, T., et. al, “Modeling Drug Releasefrom Granule Coated with an Aequous-Based System ofAcrylate Methacrylate: Effect of Moisture Content on theKinetics of Drug Release,” Chemical Pharmaceutical Bul-




letin, (1994), 42(11) 2338-2341.5. Watano, S., Kiyomi, Y., et al., “Evaluation of Aequous

Enteric Coated Granules Prepared by Moisture ControlMethod in Tumbling Fluidized Bed Process,” ChemicalPharmaceutical Bulletin, (1994), 42(3).

6. Watano, S., Ikuko, W., et al., “Modeling and Simulation ofDrug Release from Granule Coated with Aequous-BasedSystem of Acrylic Copolymer,” Chemical PharmaceuticalBulletin, (1995), 43(5).

7. Walter, K., “Real Time Process Determination to AchieveReproducible Coating Results,” Proceedings Second Euro-pean Coating Symposium, (1997), 227-235.

8. Walter, K., “A New Method of Controlling the ProcessHumidity when Coating Particulars,” Proceedings FifthEuropean Coating Symposium, (2003), 82-87.

9. Larsen, C. C., Holm, P., et. al., “A New Process ControlStrategy for Aqueous Film Coating of Pellets in FluidizedBed,” European Journal of Pharmaceutical Sciences, (2003),20, 273-283.

Note: Real Time Process Determination is a trademark of Niro Inc.

About the AuthorKim Walter acquired a Bachelors degree inmachine-engineering in 1968. After servingin the Danish Royal Air Force, he earned aMasters degree in machine-engineering anda PhD from the Department of Fluid Me-chanics at the Technical University of Den-mark in Copenhagen, (former DTH). Through-out his career, he has combined inventive-

ness and computer programming, beginning with his firstinvention in noise control and a PhD project that developeda calculation to determine flow around blunt objects using alaser doppler anemometer. He has been with GEA Niro since1978, beginning as scientist/researcher at Niro Atomizer A/Sin Copenhagen, where he worked on kiln technology andpatented a new kiln design. In 1991, he became a seniortechnical advisor responsible for process equipment for thepharmaceutical industry, designing and patenting new equip-ment - precision coater, roto-processor, precision granulator,and tablet coater. He also developed a new control philoso-phy, real time process determination, commercially usedsince 2001. He is now a Senior Process Engineer at Niro Inc.,USA, where he is an expert in coating technique and equip-ment, developing sophisticated and proprietary process tech-nology and building up a technology center used by GEA Niroworldwide. He has numerous patents and published articles.He can be contacted by e-mail at: [email protected].

Niro Pharma Systems, 9165 Rumsey Rd., Columbia, Mary-land 21045.


Validation Strategies


This articleevaluates thetime (laboreffort) requiredto validatecomputer orsoftwaresystems as afunction of theappliedvalidationstrategy.

Cost and Benefit Analysis ofValidation Strategies

by Kent Lohrey

Introduction

Each additional day in the marketplaceof patent protected sales for a pharma-ceutical product or medical device canhave significant revenue impact for a

company, sometimes on the order of millions ortens of millions of dollars per day. While theduration of patent protection is clearly defined,the portion of that time period where a productis available for sale or generating revenue isvariable. One factor influencing how soon aproduct can reach the market place (or howlong the patent protected sales period lasts) isthe time spent in development, deployment,

and validation of any computer or softwaresystem required to produce the pharmaceuticalor medical product or the clinical trial suppliesrequired to get the product to the point where itcan be sold. Some production lines also requirenew technologies and computer systems onceproduct sales have started. Increasing productdemand can require additional production ca-pacity, which can drive changes to the manu-facturing systems. As a result, there is signifi-cant pressure on the delivery of new computeror software systems that support or provide thecapability to deliver these revenue-creatingproducts. Regardless of business pressures,

Table A. Protocol detailsby customer, project,protocol, and strategy,including system typeand total test items perprotocol.

Only WhenAlways Different

Customer Project Protocol Type of System (ALW) (OWD)

A 1 1 extrusion 1760

2 extrusion 916

3 compounding 932

B 1 1 building management/room monitoring 1761

C 1 1 process analytical technology 554

2 process analytical technology 2694



D 1 1 building management/room monitoring 1444

E 1 1 building management/room monitoring 2634

F 1 1 solution preparation/tablet coating 6304

2 1 solution preparation/tablet coating 5171


4 1 cream production 3521


6 1 building management/room monitoring 2533

7 1 purified water production and distribution 1176

G 1 1 Chromatography 1008

7 total 13 total 18 total Average: 1449 4284

Reprinted from







these systems also must be deployed in a manner satisfyingall applicable regulatory requirements, including softwarevalidation.

Companies have chosen to apply different validation strat-egies that are conducive to rapidly delivered and regulatorycompliant systems. This article evaluates the time (laboreffort) required to validate these computer or software sys-tems as a function of the applied validation strategy. This isaccomplished by comparing testing metrics collected fromthe execution of many protocols on several system validationprojects where different strategies were applied by differentcompanies. Analysis of this data quantifies the impact of thediffering strategies. Advantages and disadvantages of eachstrategy are discussed in the context of regulatory require-ments, and some conclusions are suggested to consider whensetting validation strategy for future projects.

All of these validated systems controlled FDA regulatedactivities for pharmaceutical or medical device manufactur-ers. Most of the systems were manufacturing control systems(e.g., extrusion, Process Analytical Technology (PAT), solu-tion preparation, and tablet coating). The remaining systemsincluded building management systems, room monitoringsystems, and purified water production and distribution. Allof these systems were delivered for items already in produc-tion except for the PAT system which was used to produceclinical trial supplies.

DataData from validation test execution has been collected for avariety of purposes, including evaluating project or taskefficiency and estimating future work. This data also can beused to compare validation strategies applied by differentcompanies, as each company has its own method to satisfy theregulatory requirements for system validation, while at-tempting to meet business needs.

The data includes the cost of testing time in units of hours.This eliminates influences on the data, and the correspond-ing conclusions, due to different rates or hourly chargesrelated to resources on different projects. Time is an accept-able unit for comparing different validation strategies astesting cost in dollars is directly proportional to testing time,meaning that an increase in time will create an increase incosts.

Test execution is quantified by calculating the averagenumber of hours used to execute each test item or testing timeper test item. This calculation requires dividing the totalhours used to execute a test protocol by the total number oftest items within the protocol. The test execution time in-cludes all of the following tasks: creating test conditions,observing results, assessing the results (pass or fail), docu-menting actual results, writing deviations (test discrepan-cies), and implementing the resolutions defined within devia-tions. Deviations include specification changes, protocol

Figure 1. Normal distribution of scaled data (scaled OWD mean = 1.00, scaled ALW mean = 1.99).




changes, system changes, and retesting or additional testingrequired.

The data used in this analysis are from 18 separate testprotocols executed as part of 13 control and informationsystem projects where hardware and software elements of asystem were validated. These projects were performed with atotal of seven different companies. Table A includes impor-tant project and protocol attributes for the data. Commonelements of all of these validation studies included:

• protocols developed to be consistent with Good Auto-mated Manufacturing Practice (GAMP4) principles

• cGMP documentation rules in effect• structured protocols derived from the validation company

(contractor) test template• clearly defined and consistent test instructions and accep-

tance criteria (for example, in some cases, the exact sameuser interface test instruction language was used fordifferent protocols with different customers)

• pass or fail assessment made for each test item• deviations required for specification, system (hardware or

software), or protocol changes that are needed to addressfailed test items

Significant differences were present in the validation strate-gies employed. These differences surrounded two fundamen-tal choices in the strategies used, when to document actualresults and when to use a risk-based approach to testing.

Two different actual results strategies were used on theseprotocols:

• actual results recorded at all times (five companies, 10protocols), referred to from this point on as ALW for theAlways recorded strategy

• actual results recorded only when the actual results weredifferent than expected (two companies, eight protocols),referred to from this point on as OWD for Only WhenDifferent strategy

Two different risk strategies were used on these protocols:

• testing 100 percent of design specification content (fivecompanies, 12 protocols) or

• applying a risk-based (less than 100 percent) test ap-proach (two companies, six projects, and six protocols)

The same set of companies, projects, and protocols was usedto analyze both major strategies. Individual companies con-sistently used the same actual results strategy for all of theirprojects. Some of the companies applied only one of the risk-based validation strategies, while others used both risk-based strategies depending on the specific system and project.Each strategy is addressed separately below.

All test execution time included in this analysis wasexpended by either contractor employees or employees fromthe customer companies. The total test execution time foreach protocol was obtained from a combination of time sheets

and test activity reports. All contractor time was documentedon time sheets reporting testing hours on a daily basis. Thehours reported on these contractor time sheets also weresubmitted to and approved by the customer companies throughapproval of a daily activity report. This approval step and acustomer’s financial incentive to only pay for work performedensured accuracy in this time sheet data. The daily activityreports, generated by the contractor, also documented whencustomer employees assisted with test execution activities asdefined above. The accuracy of this total test time componentis robust, but not as robust as the time sheet data becausethese reports were based on contractor observations, notdirect input from the customer employees. As a result, someinaccuracy is possible in the customer time contribution tothe total time. The extent of this possible inaccuracy isunknown, but mitigated by the following:

• Only eight of the 18 protocols included customer time. Onthese protocols, the customer time averaged less than onethird of the total time spent on a protocol.

• The contractor employees were responsible for coordinat-ing all test activities, regardless of who performed them.

• Most of the customer contribution was performed in com-bination with the contractor employees or performed inde-pendently, but in the same room as the contractor andwhen the contractor was present.

Validation Strategy:Actual Results Documentation

DataThe actual results data analysis produced a mean test execu-tion time per test item metric and a standard deviation for theprotocols within each strategy. The units for these values arehours/item. Dividing the mean and standard deviation fromeach population by the mean for the OWD population scaledthe data, changing the values from hours/item to a percent-age of the mean for the OWD population. For example, theOWD mean value scaled results in a value of 1.00 (the meandivided by itself). The mean for the ALW population is 199%of the OWD mean, reported as a value of 1.99 in Table B (ALWmean divided by OWD mean = 1.99). Comparison of thescaled test time per test item for the ALW and OWD popula-tions shows the ALW method requires double the executiontime per test item, compared to the OWD method.

The scaled mean and standard deviation results are rep-resented in Table B. Table B also includes the scaled mini-

Always (ALW) Only When Different (OWD)

Count of Protocols 10 8

Mean 1.99 1.00

Standard Deviation 1.22 0.34

Minimum 0.91 0.61

Maximum 4.60 1.59

Table B. Scaled data comparing testing time impact of actualresults strategies (values divided by OWD mean).




mum and maximum test execution time per test item valuesfound within each population.

The scaled test time per test item results are presented inFigure 1. This figure uses a normal distribution of the datashown in Table B to illustrate the difference between the twopopulations. Not only was the ALW mean much greater thanthe OWD mean, but the standard deviation also was muchgreater (scaled value of 1.22 compared to 0.34 or approxi-mately 3.6 times greater). This shows the execution time forthe ALW tests was far more variable than the OWD tests.Stated another way, the mean test time per test item in theOWD population was much more consistent. It is likely thatthe ALW population variability was driven primarily bydifferences in the types of actual results required within theALW population. For example, some ALW tests only requiredprinting and referencing a screen capture, which requiresmuch less time to document than re-writing the entire set ofexpected results in the protocol as was required for some ofthe other ALW tests.

Strategy ComparisonThe different actual results strategies have important simi-larities and differences. The differences create advantagesand disadvantages for each strategy as summarized in TableC.

Applicable Regulatory Requirements orGuidanceCurrent Good Manufacturing Practice (cGMP) requirementsare defined in 21CFR Part 820. These requirements include

at least two references to the results of validation activities.The regulation for production and process controls definesa requirement for results documentation during validationof automated processes as follows: “Automated processes.When computers or automated data processing systems areused as part of production or the quality system, the manu-facturer shall validate computer software for its intendeduse according to an established protocol. All software changesshall be validated before approval and issuance. Thesevalidation activities and results shall be documented.”1

Similarly, the cGMP regulation for process validationdefines the requirement for results documentation withinprocess validation as follows: “The validation activities andresults, including the date and signature of the individual(s)approving the validation and where appropriate the majorequipment validated, shall be documented.”2

These requirements call for documentation of validationresults. The requirements do not specify instructions forapplication of these requirements or if these results re-quirements apply specifically to actual test results. There-fore, individual companies must interpret the requirementsand decide how the requirements can, or should, be imple-mented. The number of companies and protocols within thedata analyzed in this article is only a small portion of thepharmaceutical and medical device industries. However,these companies do offer some insight into how regulatedcompanies have attempted to implement validation strate-gies to address this result requirement.

The companies choosing the ALW method often ex-plained this choice as being based on the presence of actual

Always (ALW) Only When Different (OWD)

Similarities • Testing is the same: test instructions, initial conditions, and expected results• Does not take varying priorities of tests into account• Relies on the integrity of the personnel involved

Differences • Every test requires the tester to document actual results • The tester is only required to document actual resultswhen they differ from expected results

Advantages • Evidence of actual results is provided for every test • Evidence is collected and documented when dictated by• Reviewers (including auditors) have more insight into test actual results (as part of a deviation)

results • Less time is spent recording actual results, requiring less• Availability of evidence may eliminate need for witness to time and cost (labor, either internal or vendor)

testing • Less documentation requires less time by reviewers• Can include some objective evidence (such as screen

prints or reports) for every test of a specific type

Disadvantages • Increases opportunities for human error (such as writing • No objective evidence is collected for successful oractual results incorrectly) passed tests (such as screen prints or reports) so no

• Some actual results can be generated by more than one evidence is available for reviewers or auditors on theseset of instructions and conditions and recording only actual testsresults provides no insight into this part of the test • May require additional resources for witnessed testing

• The method of collecting evidence can create issues (forexample some companies save screen prints as .jpg files -a format that can be easily edited - the use of electronicfiles for evidence may constitute use of electronic recordsto satisfy regulatory requirements, possibly invoking21CFR Part 11 for these actual results)

• More time is spent recording actual results, requiring moretime and cost (labor, either internal or vendor)

• More documentation requires more time by reviewers• Provides inconsistent levels of evidence when different

types of evidence are used (such as written vs. screenprints)

Table C. Comparison of different actual results strategies.




results, for all tests, generating a higher confidence duringinternal reviews. These documented results also were citedas a critical component when defending the documentation inany audit activities.

Those applying the OWD method decided documentingthat a test passes is the functional equivalent of writing downactual results when the results are the same as expectedresults. The test performer is documenting the actual resultswithout rewriting them in the protocol by indicating that atest passed. In these protocols a “pass” assessment wasdefined as the actual results matched the expected results.

ConclusionsWithin the limited population used in this analysis, compa-nies more frequently chose the ALW method, by a ratio of 5:2in this sampling. Validation strategy discussions with thecompanies using the ALW method revealed a very strongbelief that anything less than always documenting actualresults for each test was inviting regulatory failure. In manycases, validation personnel called on their experience withpast regulatory audits to explain the necessity for theirchosen strategy. These experiences hinged on a greater com-fort level that auditors had expressed with the actual resultdocumentation provided for every test. These companies didnot cite specific regulatory requirements as part of the ratio-nale for choosing the ALW strategy.

Even within the ALW strategy companies, there wasvarying confidence placed on different types of actual results.Objective forms of results evidence like screen captures orprints generated a much higher level of confidence than thewritten, subjective observations of a test performer. As notedin Table C, even verifiable objective evidence like screenprints have limitations in the insight or value they canprovide to a reviewer. The fact that a screen print cannotprovide specific definition of the actions taken to generate the

actual result prevents even this objective method of captur-ing results from providing a faultless illustration of allcritical aspects of the test. The chain of evidence used tosupport a pass or fail assessment on an individual test canonly be as strong as the weakest link. If only the test resultsare documented with evidence, written or otherwise, theunsupported or weak link in the chain of evidence is stillrelying on the integrity of the tester to have used the instruc-tions and initial conditions provided to generate this docu-mented result. If a company must rely on the tester’s integrityfor the instruction part of the test, then is it possible orreasonable to rely on the same integrity for the result?

This was a central part of the OWD strategy justificationfor individuals within the two companies not using the ALWmethod. Both of these companies believed it was completelyreasonable to rely on the tester for the accuracy of both theinstruction and result portion of a pass or fail assessment.Both also concluded the regulatory requirements did notnecessitate documenting actual results when the resultswere a match with the expected results defined in the proto-col. Additional results documentation and the associatedeffort did not provide a significant compliance advantage, intheir opinion. However, recording actual results was anobligation on these protocols and provided additional resultsevidence when the expected results defined in a protocol werenot observed. In this scenario, the actual results were docu-mented as a deviation to the protocol. For example, if a valvegraphic turned the wrong status color when a valve alarmoccurred, the deviation would document the behavior found,any required corrective action (e.g., specification or softwarechange), and any retesting required.

Like the ALW method, the OWD strategy also has disad-vantages. Screen prints, and other evidence like reports, canoffer objective evidence that can support a test assessment of“pass” in a visual way that can be very powerful. Not using

Table D. Comparison of different risk strategies.

100 percent Test Risk-based

Similarities • System design included GMP-critical and non-GMP features and functions

Differences • All specified design elements are fully tested • Some specified design elements are fully tested, othersare partially tested

Advantages • System is comprehensively tested creating less risk of • Reduces execution time and paperwork generated byeven minor issues going unnoticed testing, reducing time, and cost

• No justification needed for testing reductions (less • Risk assessment exercise focuses project and personneldocumentation) on highest priorities

• Less perceived risk by individual reviewers and approvers

Disadvantages • Increases execution time and paperwork generated by • System is not comprehensively tested, increasing thetesting which increases time and cost risk of some issues going unnoticed

• Treats all design features and tests as equal in importance • Justification needed for testing reductions (additionalor priority documentation)

• Can prevent inclusion of some useful features due to • Perceived risk by some reviewers or approvers as theyassociated testing costs approved the reduction in testing

• Requires additional design considerations to support riskstrategy

• May need additional design work to limit access to non-GMP functions through security (user level) requirements

• May need additional design work to create features thatare universal and can be reused on different systems(designing to satisfy multiple systems)




evidence like this for passing tests denies future reviewers ofevidence that could support at least the result portion of thetester’s assessment. Reduced evidence for reviewers andauditors is a disadvantage in the OWD method that couldoffset some of the time and cost savings offered by thismethod.

Both strategies present disadvantages or challenges thatmust be carefully considered before choosing a validationstrategy. The impact to schedule and cost is significant withthe ALW method taking double the test time per test item fortest execution. A lack of actual results evidence for passedtests in the OWD method, even for the most critical tests,could invite or influence a future regulatory review.

One possible solution to these challenges is to apply a risk-based approach to the need for actual results. A hybridstrategy could be adopted to require the ALW method for onlythe highest risk items. This limits cost and schedule impactsof the ALW method, while still providing evidence of actualresults for every high-risk test. The OWD method could beapplied for non-critical tests. For example, in a system wheretemperature is a critical process variable, temperature alarmsare likely to impact the quality, safety, or efficacy of theproduct. As a result, these critical alarms deserve a highdegree of scrutiny. At the same time, system usability fea-tures such as colors displaying device status do not have adirect impact on the product, requiring less emphasis on theverification of these system functions. Using this mixedstrategy requires the instructions and acceptance criteria toclearly define where the different actual results methodsapply and how they must be implemented. This hybrid andrisk-based approach to actual results recognizes that all testitems are not equal in terms of the functions or requirementsthey address while allowing companies to limit the time andcost impacts of validation testing.

Validation Strategy: Risk-Based TestingDataThe risk-based data analysis focused entirely on those projectsthat applied some level of reduced testing based on risksevaluated within the system. These projects fit into twodistinctly different types of risk-based approaches. In bothtypes, these systems were designed to support the applicationof a risk-based strategy. Reviewing the specific system designspecifications and counting the individual design elementsand conditions not tested quantified the amount of reducedtesting.

Five of the risk-based protocols applied a strategy ofreduced testing for the control system software by minimiz-ing testing of maintenance only functions. The design in-cluded some windows containing content only required formaintenance purposes. Access to these windows was re-stricted to prevent system operators from accessing thesefunctions. These windows also contained no GMP criticalinformation. All GMP critical information and process con-trol was included in other portions of the applications. Theserestrictions supported an approach to test a representativesample of the functions included in windows like a variable

frequency drive status window. Testing 100 percent of thesewindow features would have added on average approxi-mately 13 percent more test items to validation tests thatalready averaged nearly 6000 test items per protocol.

One of the projects in the data analysis applied a differentrisk-based strategy. In this project, a control system applica-tion was developed for use on a number of different systemsthat had many common components. This application wasdesigned generically in the windows that were used on eachof the systems (such as the security and alarm summarywindows). The application was used to control a suite of airextrusion systems. Some were single extruders and otherswere co-extruders (two extruders). The components of thesingle extruder were the same as the first or primary extruderin the co-extrusion systems. This allowed the primary extrud-ers and single extruders to be controlled through the exactsame set of user interface objects that were designed andprogrammed the same, except for linking to different fieldequipment. The design also disabled any window featuresthat did not apply on a specific system. Secondary extruderobjects were disabled when the application was installed ona single extruder system.

The first of these systems tested included almost all of thefeatures common to all systems (a co-extruder) and wasexecuted fully. The second installation of this applicationapplied a risk-based approach to testing by not repeatingtests of unchanged functions. For example, the access limitsdefined for a setpoint entry object in the user interface wereentirely a function of the user interface objects, not thesystem attached to the software. These features were notretested. Full testing was limited to features not previouslytested and those software components interacting with thespecific system devices. This included testing of analog in-puts like temperature, line speed, and pressure. Outputs likethose related to starting and stopping devices also weretested. The protocol for this second installation includedapproximately 900 tests. Testing 100 percent of the softwarefeatures, including those common features, would have re-quired more than 1600 additional tests, an increase of morethan 150 percent.

Strategy ComparisonThe different risk strategies have important similarities anddifferences. The differences create advantages and disadvan-tages for each strategy as summarized in Table D.

Applicable Regulatory Requirements orGuidanceCurrent Good Manufacturing Practice (cGMP) requirementsdefined in 21CFR Part 820.70 also apply to this risk-basedstrategy scenario. Part (i) for automated processes requirescompanies to “validate computer software for its intendeduse.”3 This regulation could be interpreted to demand 100percent testing of all specified software elements as addi-tional language in this section specifically directs validationof all software changes.

The door to the use of a risk-based strategy was opened in




2002 when the FDA announced a new initiative to enhancethe pharmaceutical GMP rules and regulations. “The firstgoal will be to enhance the focus of the Agency’s cGMPrequirements more squarely on potential risks to publichealth, by providing additional regulatory attention andagency resources on those aspects of manufacturing that posethe greatest potential risk.”4

The risk-based strategies used by the projects analyzedactually applied a minimal amount of risk when identifyingtesting that could be reduced. Risk-based decisions werelimited to areas that did not affect electronic records orelectronic signatures and software features that did notinfluence or alter product safety, quality, or efficacy. Thisavoided the aspects of manufacturing that pose a greatdegree of potential risk, minimizing the need for FDA scru-tiny of these risk-based decisions.

ConclusionsMore than half the projects included in this data analysis didnot apply a risk-based approach. These companies and projectsshared a common element: individuals were more comfort-able with the 100 percent testing method. Validation strategyand test protocol approvers perceived risk-based validationas a potential compliance risk. This, in turn, was seen as apersonal risk if they were associated with approving a poten-tial compliance risk and a future regulatory activity ques-tioned that choice.

Both of the risk-based approaches applied by the organi-zations in this sample delivered noticeable time and costresults for the companies. In the case of the maintenancefunction approach, the applications were able to includehelpful monitoring and troubleshooting capabilities withminimal additional testing. These additional features willaid the company during operation and maintenance activi-ties for years to come. Data from these projects demonstratehow risk-based testing allows an increase in applicationfeatures for the same or less testing time than a 100 percenttested application. These projects could have pursued aneven greater savings, beyond the average 13 percent, byapplying the risk-based approach to testing of other non-critical features within the applications that were not iso-lated from critical functions as well as the maintenancefunctions.

The second risk-based approach, multiple installations ofthe same application, cut the number of test items by morethan 50 percent on the second installation. The software wasconsidered a custom configured application (GAMP category5) for this company, which typically requires validation of thecomplete system.5 Through careful design choices and use ofcommon user interface objects, the company reduced valida-tion testing time and assumed very little compliance risk inthe process.

The risk-based method focused these project teams andtesting resources on the highest priority aspects of the spe-cific systems which impacted quality, safety, or efficacy of theproduct. These critical features were fully tested. In theseorganizations, the project decision-makers were encouraged

and supported in the risk-based work.Those companies applying a risk-based approach were

able to validate their systems, while avoiding significanttesting time which would have been required by using the 100percent testing method. This time savings was achieved alsowhile maintaining a strong position for any future compli-ance reviews through full testing of all critical functions.These projects proved a risk-based approach could provideregulatory compliance and reduce testing time (costs) simul-taneously.

DiscussionThe different validation strategies discussed above each haveadvantages and disadvantages. While some strategies mayappear to be more commonly accepted, the more commonlyused testing strategies – to always document actual resultsand to apply no risk-based reductions – drive validationtesting time and costs up as indicated by this data analysis.Full or limited use of the other emerging strategies cangenerate schedule and cost benefits that merit considerationby companies needing to design, validate, and deploy systemswithin their own budgetary environment and regulatoryhistory.

Considering these strategy options and their tangibleimpact on time and cost is likely to either generate moreconfidence in the current methods applied by a company orprovide ideas for changes to future validation strategies.Regardless of the validation strategy chosen, clearly definingand documenting the strategy applied will provide the basisfor validation decisions and support the defense of the ap-plied strategies in any future regulatory compliance evalua-tions or audits.

References1. 21 CFR Part 820.70 (i).2. 21 CFR Part 820.75 (a).3. 21 CFR Part 820.70 (i).4. FDA Press Release, “FDA Unveils New Initiative To

Enhance Pharmaceutical Good Manufacturing Practices,”August 21, 2002

5. GAMP® 4, Good Automated Manufacturing Practice(GAMP®) Guide for Validation of Automated Systems,International Society for Pharmaceutical Engineering(ISPE), Fourth Edition, December 2001, Appendix M4 -Guideline for Categories of Software and Hardware, p. 4.www.ispe.org.

About the AuthorKent Lohrey attended Princeton Univer-sity, earning a Bachelor of Science and Engi-neering (BSE) in mechanical and aerospaceengineering (with honors). While atPrinceton, he was a cadet in the Air ForceReserve Officer Training Corps and receiveda commission in the United States Air Force.Following graduation, Lohrey reported to

the Space and Missile Center, Los Angeles Air Force Base.




During his time in the Air Force, Lohrey held a variety ofpositions, including the liquid propulsion engineer for theTitanII and TitanIV launch vehicles where he supported thelaunch of the Cassini spacecraft mission to Saturn. After theAir Force, he joined Accenture as a process consultant. Whileat Accenture, he led the design and validation of controlleddocument management systems for pharmaceutical compa-nies to comply with 21 CFR Part 11 requirements for elec-tronic records and electronic signatures. Lohrey joined thevalidation department of Total Systems Design (TSD), Inc., acontrol system integrator, in 2002 and he is currently theValidation Program Manager. In this role, Lohrey sets vali-dation strategy for the company, manages all validationprojects, writes and executes validation protocols, and searchesfor opportunities to improve validation methods or tools. Hismost significant accomplishment at TSD is successfully de-veloping and deploying a process and tools to automate thecreation of test plan content from design specifications. Kentis a member of ISPE and was elected to membership inSigma Xi, the Scientific Research Society, for his researchwork as a Princeton undergraduate. He can be contacted bytelephone at: +1-610-857-1666 or by e-mail at: [email protected].


Industry Interview


This onceaspiringprofessionalbasketball playertalks about hisunconventionalcareer pathinto thepharmaceuticalindustry andForestLaboratories; hisexperiencerunningcompanyoperations andfacilities indifferent partsof the globe; histhoughts onQuality byDesign; thetherapeuticareas to watch;and the majorindustrychallengesahead.

PHARMACEUTICAL ENGINEERING InterviewsRichard S. Overton, Vice President ofOperations and Facilities, ForestLaboratories

Born in 1947 and raised in the NaugatuckValley of Connecticut, Rick Overton came toLong Island in 1977 from Miles (now Bayer)Laboratories to run operations for Forest Labo-ratories in Nassau County. He holds businessmanagement degrees from the University ofNew Haven and Excelsior College, but wasthrust into the area of building design, facili-ties and equipment acquisition, constructionand maintenance when a void occured in thecompany’s engineering staff. Since that time,he has been instrumental in the developmentof Forest’s facility expansion philosophy as thecompany emerged from its $3 million roots inInwood, New York to become a $3+billion mul-tinational corporation today. He currently over-sees Forest’s Supply Chain Team as well astheir Long Island, New York operations. He isone of the 26 charter members of ISPE andserves as Vice Chairman on the Board of Direc-tors of the Farmingdale State College Founda-tion in addition to numerous non-profit organi-zations.

Q What is your educational background?

A Oddly, I may be one of the only chartermembers of ISPE who was not an engi-

neer. My education is in business with an ASfrom the University of New Haven and a BSfrom Excelsior College. In 1980, I was fullyoverwhelmed in running the construction andmaintenance of Forest Laboratories’ facilities.Our fledgling company didn’t have any engi-neers on staff at the time. Most probably be-cause I had worked as a janitor and handymanduring summer school breaks in my youth, ourCEO, Howard Solomon, considered me as mostqualified to take on the task. When I wasapproached about joining an organization de-voted to the advancement of pharmaceuticalengineering, I jumped at the opportunity tocommunicate with others who, I later found,were suffering through the same uncertaintiesof creating an adequate environment to meetFDA guidance as was I. Since that time, I’vebecome reasonably proficient in the pharma-ceutical building trades thanks to my associa-tion with ISPE and its terrific membership andprofessional staff.

Q What lead you into a career in pharmaceu-tical manufacturing? What experiences

and training best prepared you for your currentposition?

A Maybe I can answer both questions atonce. I never anticipated that I would

have a career in pharmaceuticals. I fully ex-pected to be a professional basketball player ormusician when I left high school, but oncereality set in during my early college days, Ifound myself a dropout, working for Pratt andWhitney Aircraft doing everything from mail

Long-time ISPEsupporters, Rickand CaraOverton.

by Gloria Hall, Editor, Pharmaceutical Engineering

Reprinted from





Industry Interview


so that it would be difficult to tell thedifference between, for instance, walk-ing into our Irish manufacturing facili-ties or those at our Cincinnati, Ohiosite in the US. They both will be clean,neat, and compliant and although weuse local building materials and meetlocal building codes to take advantageof cost savings, they both will appearsimilar to the untrained eye. We usethe ISPE/FDA Baseline Guides reli-giously and they have been an enor-mous help in maintaining consistency.

Our operations have shifted focusover the last 10 years from a series ofsubsidiary silos producing indepen-dently, to a globally integrated supplychain. This came about by a mid-1980sdecision to move our primary manufac-turing for oral solid dosage forms by themid-1990s to Ireland to leverage a taxadvantage there as well as to have aEuropean base of operations that couldmore easily integrate with our partner,Lundbeck AG of Denmark for the pro-duction of our first blockbuster product,Celexa, an SSRI to fight depression. Wechose to ship bulk from Ireland to ourCommack, New York and Cincinnati,Ohio packaging facilities that in turnwould ship finished goods to our St.Louis, Missouri distribution center toserve the US wholesale market. Com-munication here is through a networkof collaborative site managers who worktogether on our supply chain teamthrough regularly scheduled conferencecalls and meetings. We’ve found thatour Informatics (IT) group has been akey component in keeping our lines ofcommunication fluid.

Q What are some of the key metricsused in your organization to

gauge operations and facilities perfor-mance or success?

A Profit and compliance, not neces-sarily in that order. Our mission

as a supply chain is “to collaborativelyensure the supply of quality productsat the right time, place, price, andquantity to fulfill the needs of our cus-tomers.” With our eye on this biggerpicture, each site and subsidiary istasked with developing their own mea-surements for throughput, inventorycontrol, and operating efficiency al-

room duty to stationery supplies, totruck driver, to timekeeper, to pay-master, to inventory analyst, to pro-duction planner, plus a few other oddjobs along the way. I realized at thatpoint that to advance further, goingback to school nights to obtain a degreewas a must. That led me to anotherreality. I needed to make more money.It was at this point that a friend toldme of a job opening at the Dome Labo-ratories, a division of Miles (later Bayer)Laboratories, in production planning.They were about to move to a newfacility in West Haven, Connecticutfrom Manhattan and were looking fornew faces. This was in 1969. It wasthere, in the development of that site,that I cut my teeth in pharmaceuticalproduction from OTCs to oral soliddosage forms to liquids, creams andointments, to sterile products and be-yond, became a supervisor and eventu-ally became a technical advisor andmoved into sales, both inside and out.

At about that time, around 1977, Igot an offer to join Forest Laboratoriesas their operations manager as theywere moving from Elizabeth, New Jer-sey to a new manufacturing site inInwood, New York near Kennedy Air-port. Starting at the ground floor in-side of a new company better fit myintroverted personality than acceptingrejection as a detail man in Miles’ Bos-ton sales territory so I accepted theposition. As a startup, I was forced tolearn every aspect of the business fromthe ground up. Not only was I runningoperations, but manufacturing andpacking orders to collecting money andloading trucks. That is when I drew theshort straw and took over the mainte-nance of the facilities. Over the years,I moved to Puerto Rico to build and runthose businesses as plant manager,then returned to New York to helpupgrade and construct a couple of mil-lion square feet of pharmaceutical floorspace, including operations in the USand overseas, primarily in the UK andIreland.

Q What are the primary responsi-bilities of your current position?

A Today, I am the Vice President ofOperations and Facilities and

have direct responsibilities for run-ning our Long Island operations andoversight for Forest’s global supplychain.

Q You’ve been with the companysince the late 1970s, watching it

grow from a small laboratory servicefirm that helped larger pharmaceuti-cal companies create new drugs to amajor pharmaceutical giant that de-velops, manufactures, and sells name-brand, generic, and over-the-counterproducts. What do you think were yourmajor contributions to the company’sgrowth?

A I like to think of myself as aliaison between our corporate en-

vironment and our production andmaintenance sub-cultures. You mightsay that I attempt to translate “finan-cial speak” to “production speak.”

Q What core philosophies andstrategies (the company’s and

your own) guide your leadership styleas VP of Operations and Facilities?

A Forest senior management hasalways had a strong entrepre-

neurial spirit and although we havehired employees over the years thatare more technically gifted in specificareas than we were, these folks stillmaintain the ability to think for them-selves in creative ways for the better-ment of the corporation. I don’t thinkyou’ll find a Forest executive who won’tcredit above all else to the strength ofour employees as the strength of For-est and its rapid growth.

Q Are there any differences betweenoperations and facilities in Eu-

rope and the US? If so, what are theyand how do you approach/handle thesedifferences?

A First, let me talk about facilities.Our Irish solid dosage manufac-

turing operation is both a GMP and EUcompliant facility, while our US basedoperations are mostly built around FDAguidelines. That said, we see more har-monization between agencies than everbefore and have tried, through my of-fice, to communicate a consistent ap-proach to building design at each site


Industry Interview


though our new President and ChiefOperating Officer, Dr. Larry Olanoff,has been instrumental in globalizingcertain KPIs.

Q How is Forest increasing efficien-cies and product quality? What

kinds of technologies are being em-ployed to accomplish this? Is the com-pany embracing the Quality by Designapproach? If so, how exactly?

A One of our goals is to seamlesslymeld our production and quality

groups. Their joint goals and objectivesare being coordinated locally and glo-bally. For example, we historically holda Quality Congress as well as a SupplyChain Team Annual Meeting to bringtogether all our key directors. In thepast, they were held at separate timesof the year and there was always thechance that a single message was notdelivered by management to bothteams. In January, we completed ajoint SCT/QA global meeting in whichthere were independent breakouts forgoal setting, but a single shared visionand a final coordinated meeting of theminds on actions for the coming yearand beyond without borders. Similarwork is being done within our ForestResearch Institute for new products aswell. Quality by Design is being em-braced, but is still a work in progress asyou might expect. Our R&D new prod-uct development staff are spearhead-ing this, MEP, and PAT initiatives,like NIR, so that our next pipelineproducts come to fruition using thelatest technology that has been provenrobust from product inception. Upgrad-ing our training programs for key man-ager development and to expose moreemployees to continuous improvementconcepts and tools are underway andhave been for some time, but in thisarea as well as in cGMP and SOPtraining, our concentration is on train-ing for effectiveness and coordinatingfirst time right as a function of not only

quality, but personal performance re-views. In the area of measurement, ourIT tools like SAP are being reexaminedfor reports that are more meaningful tothe average user as well as the sea-soned specialist.

Q What are some of the concerns orissues you have today in your

operations?

A As I said earlier, training, staffdevelopment, and succession

planning are always paramount foremployees. Risk management as a busi-ness and quality issue is a challengeespecially when the company is in adynamic change to globalization.

Q What therapeutic areas do yousee making the most news head-

lines in the next five years? How do yousee Forest being included in those head-lines?

A The biotech industry is comingup with new chemical entities

daily so what I think is a big deal todaymay become old news tomorrow, buttwo areas that come to mind immedi-ately are diabetes and the antibioticmarkets. Diabetes delivery systems forthe variety of therapies already avail-able may make the first news and mostcompanies are poised to jump in. How-ever, we are more likely, through therecent acquisition of the research com-pany Cerexa in California, to forgeahead with new antibiotics as we seethis as a constantly growing and chang-ing market both now and in the future.Of course, our primary corporate focushas and continues to be in the area ofCNS, including pain and cardiovascu-lar medications. Forest has a reputa-tion as a nimble company, capable ofrapidly responding to a changing envi-ronment, like traditional pharmaceu-tical API licenses moving to biopharma-ceuticals so who knows what or wherethe next opportunity will be? Our cor-

“Our operations have shifted focus over the last 10 yearsfrom a series of subsidiary silos producing independently,

to a globally integrated supply chain.”

poration was built on our ability tolicense and develop drugs for the USmarket efficiently in effective partner-ships with primary research compa-nies who are less capable than Forestin understanding how to do the “D” inR&D and weave through the maze ofthe FDA’s clinical and NDA process.Thus, I anticipate that we will con-tinue to fill our pipeline through thesepartnerships in whatever therapeuticarea that presents us with an opportu-nity to succeed.

Q What technological and opera-tional breakthroughs do you an-

ticipate within the next five years?What do you see as some of the emerg-ing technologies in the pharmaceuticalmanufacturing industry?

A NIR as well as other PAT initia-tives will start to pay off. More

companies are likely to use MES as theindustry moves away from a paper base.Certainly, rapid release technology ison the rise in drug development andseparation technology for more puri-fied APIs so impurities profiles areimproved from the get go are here. QbDwill enhance this effort overall as itwill point out inefficiencies and inef-fectiveness early in development.

Q What do you see as the key at-tributes and qualities in facility

design? What do you envision the phar-maceutical facility of the future to looklike, say in 20 years?

A Designers are starting to under-stand the principles of KISS

(Keep It Simple Stupid!). Better use ofmaterials of construction for seamlessinteriors, cleaner details for penetra-tions, simpler MEP systems that aremore self contained and most impor-tantly, miniaturization of equipmentenvironments and enclosures to condi-tion product away from personnel androom structures so that the cost of


Industry Interview


operating a facility can be put backunder control. We are no longer in aworld where a facility can run ineffi-ciently or environmentally out of con-trol no matter where it is located. Wecan’t assume that monster creationsthat are monuments to a designer orengineer’s overkill can just be writtenoff as an overhead that gets built intothe cost of goods. Newer facilities mustbe green, efficient, and effective, aswell as be a pleasant place with whichto attract employees.

Q What is your involvement withISPE? When did you first en-

counter ISPE?

A Today, I have been less activethan in the past primarily be-

cause of the scope of my daily duties. Inthe past, I was active on committees,especially for the annual meetings andfound the community spirit of the mem-bership has helped me form long stand-ing relationships throughout the phar-maceutical world.

Q What kinds of activities do youenjoy in your spare time?

A I enjoy boating in the summerand skeet shooting in the winter

and am active on the board of directorsof a number of non-profits and atFarmingdale State College inFarmingdale, New York where I amVice President of the Farmingdale Col-lege Foundation and Chairman of theFinance Committee.

Q In what ways do you believe aglobal organization such as ISPE

can assist regulators, pharmaceuticalcompanies, and individuals in the in-ternational arena?

A ISPE for me has been not only aneducational resource, but more

importantly a clearing house for prob-lem solving and meeting industry mov-ers and shakers. In addition, I’ve re-ceived a wealth of information aboutequipment and other vendor relatedsupport that I might otherwise haveapproached with skepticism or not in-vestigated at all. Most of all, I’ve had alot of FUN in doing so. This is an activegroup of friends who happen to get

together for a single purpose. As aglobal structure, ISPE has helped andwill continue to help expedite the har-monization of pharmaceutical engi-neering practices which may end upbeing its greatest legacy.

Q What have been the most signifi-cant changes in the industry in

the past two decades and how havethese changes affected your personalviews on the industry’s progress?

A The politicizing and ensuingmedia circus that has developed

over drug development and sales hashad its pros and cons. I’ve often thoughtthat if it weren’t for those devils in themedia, we’d have been able to go ourmerry way and probably have beenstuck in the 1980s technology (andprofits) forever. But the world is flatwhen it comes to our industry todayand we must realize that it is overallfor the better. We have more competi-tion creating quantum leaps in drugdevelopment and improved processesand equipment and organizations likeISPE that help us all keep up to dateand moving forward for the good ofpatients who have the misfortune toneed our concoctions, but the good for-tune to have us around to supply them.

Q What is your definition of innova-tion? How does innovation apply

to what you do at Forest?

A I may be a bit of an old schoolhard___, but innovation is noth-

ing more than common sense as ap-plied to the principles of continuousimprovement. We’ve lived by thatthinking at Forest and have done rea-sonably well over the past few yearsand I believe that it will serve us wellinto the future.

Q What do you think the majorchallenges will be for this indus-

try in the future?

A The never ending search for newchemical entities to develop and

deliver for the public good at reason-able prices, while maintaining ad-equate profit to do the research to con-tinue the search might be a concernthat we all should think about.


Data Acquisition of Water Instrumentation


The articlepresents highfrequency dataacquisition incontinuouspharmaceuticalprocesses andillustrates,through the useof EWPS, theneed for processunderstanding,control, andimprovement.

The Use of Exponentially WeightedProcess Statistics (EWPS) andStatistical Process Control (SPC) inHigh Frequency Data Acquisition ofPharmaceutical Water SystemsInstrumentation

by Nissan Cohen

Introduction

Statistical methods for the measure-ment of process variability are welldocumented. The greatest proponentof statistical process control was Ed-

ward Deming. Although Deming’s ideas wereoften dismissed by the manufacturing sectorsof the United States, the Japanese manufac-turing industries readily adopted Deming’sideas. Japanese quality in manufacturing sur-passed American industries’ expertise in themid 1970s and heralded the beginning of Japa-nese product dominance in the North Ameri-can market. Japanese “run-of-the-mill” prod-ucts were often superior to carefully craftedAmerican-made products.

Measuring changes in the variability of the

products produced can identify (and sometimesmodify) the characteristics of the process thatcreates it. Some of the basic tenets of ProcessAnalytical Technology (PAT) are based on thisvery idea. If the process is measured and under-stood with multi-variants, then process qualitywill increase and product deviations will beminimized or non-existent.

A definition of traditional batch orientedpharmaceutical manufacturing can be describedas Discrete Product (DP) processes or manufac-turing processes that produce discrete “prod-ucts.” Discrete product processes produce lim-ited data. DP statistics require that each mea-surement be independent. This data may besufficient for the discrete product process, butproblems arise when DP statistics are applied

to continuous flow produc-tion systems.

Modern pharmaceuticalwater systems are continu-ous operations with recircu-lating flow. The use of on-line instrumentation en-hances the monitoring,management, and compli-ance of the water systemsoften precluding the needfor laboratory off-line test-ing. Continuous processeswith on-line instruments

Figure 1. The rate of theweighing function.

Reprinted from







amass a series of timed individual measurements producedsequentially. Sequential measurements are usually depen-dent on the preceding one, particularly when a high samplingrate is used. Real-time critical processes require advancewarning of an impending upset and typically employ on-linemonitoring with alarm generation capabilities. Statisticalmethods are used to detect subtle process changes.

DP statistics measure the characteristics of a small numberof randomly selected samples from the production line andestimate, within known confidence levels, the characteristicsof the entire batch. This is known as statistical inference.

Continuous processes often monitor process parametersover long periods of time. There is no need to estimatepopulation statistics, as with DP statistics, simply to mea-sure them.

In DP statistics, control chart limits are generally set to±3s based on the expectation that 99.73% of the samples willfall within these boundaries if the variable is “in control.”This leaves only a .27% chance of an excursion beyond theselimits. In a continuous process, samples are taken continu-ously at higher sampling rates than those assumed in DPstatistics. The result is much larger sample sizes and apossibility of correspondingly higher frequency of excursions.For example, if frequency measurements are made once eachsecond on a continuous variable, ±3s limits would result in ana probability of an excursion approximately every six min-utes as only 99.73% of the readings would comply.

Ultimately, the goal of quantifying and understandingprocess variability remains unchanged. The methodology bywhich these statistical techniques are applied must be prop-erly selected to derive the desired results in discrete andcontinuous processes.

The Use of StatisticsThe first and perhaps most useful feature of statistics is theability to investigate and numerically quantify variations ina measured parameter. If size and variation patterns areestablished when the process is running well, natural limitsmay be applied to detect and identify small shifts in thestandard pattern. Statistical methods can be used to separate

variations due to definable causes rather than random chance.Statistics can be used to isolate and investigate assignablecauses. When one parameter appears to be related to an-other, correlation techniques may be used to quantify thedependency. Often seemingly unrelated parameters can cor-relate to root causes of deviation.

Averaging is a common method to depict and track thecentral tendency of the process. The purpose of averaging orthe subgrouping of datums is to reduce the spread in ameasurement and track its central tendency. The simplesttechnique is to average each n datums in succession, asshown below. Averaging is commonly used in DP statistics.

Datums 7 8 5 5 6 4 7 9 8 8 5 6 7 8 9 5 4 4 5 6 7 6 3 6 8 9 8 6 7 8Average 6.2 7.2 7.0 4.8 6.0 7.6

This method has the distinct disadvantage of producing anoutput only once every n readings. If the duration of theaveraging period is long, the delay in updates may be unac-ceptable. A better approach is to use a running average:

Data 7 8 5 5 6 4 7 9 8 8 5 6 7 8 9 5 4 4 6 5 7 6 3 6 8 9 8 6 7 8Running Average 6.2Data 7 8 5 5 6 4 7 9 8 8 5 6 7 8 9 5 4 4 6 5 7 6 3 6 8 9 8 6 7 8Running Average 5.6Data 7 8 5 5 6 4 7 9 8 8 5 6 7 8 9 5 4 4 6 5 7 6 3 6 8 9 8 6 7 8Running Average 5.4Data 7 8 5 5 6 4 7 9 8 8 5 6 7 8 9 5 4 4 6 5 7 6 3 6 8 9 8 6 7 8Running Average 6.2

This method produces an output every reading. A practicallimitation with this approach is that each reading in thehighlighted running subgroup must be retained in memoryand these archiving requirements can become substantial ifthe subgroup is quite large. Each time a new reading is addedto the subgroup, the oldest reading must be dropped. Anothershortcoming of both simple and running averages is the lackof any time reference in the calculation of the average.Readings are averaged in the same manner whether they aretaken at one second or three day intervals.

In a continuous process, variables average is over a certaintime period rather than a certain number of readings. A one-hour average, for instance, may contain three, six, or 3600readings dependent on the frequency of measurement andcould vary from one hour to the next. In relation to monitoringcontinuous processes, it is the dynamics of the process thatdictate the necessary smoothing, not the number of readings.When applying averaging to continuous process measure-ments, the time periods must be taken into account.

Exponentially Weighted ProcessStatistics (EWPS)

A technique to solve both problems of time and traditionalrunning averages is to produce a running average based on atime-weighted sum of all previous readings. In continuousprocesses, the influence of “now” is much greater than areading in the “past.” “Now” has a direct influence on theprocess. “Past” 30 seconds ago, one minute ago, five minutesago, one hour ago, and one day ago have a decreasing influ-ence on “now.” The purpose of Exponentially Weighted Pro-Figure 2. Noise: EWP vs. running average.




cess Statistics (EWPS) is to exponentially weight every mea-surement and reading. The most recent reading has thegreatest influence. As the readings age, their influence wanes.This resolves the issue of equal value for running averagesand discrete products. Since each new reading does notrequire a previous one to be dropped, the weighted sum maybe retained in a single register, regardless of the period overwhich the average is run:

Data ...7 8 5 5 6 4 7 9 8 8 5 6 7 8 9 5 4 4 6 5 7 6 3 6 8 9 8 6 7 8Running Average 6.2Data ...7 8 5 5 6 4 7 9 8 8 5 6 7 8 9 5 4 4 6 5 7 6 3 6 8 9 8 6 7 8Running Average 7.2Data ...7 8 5 5 6 4 7 9 8 8 5 6 7 8 9 5 4 4 6 5 7 6 3 6 8 9 8 6 7 8Running Average 7.0Data ...7 8 5 5 6 4 7 9 8 8 5 6 7 8 9 5 4 4 6 5 7 6 3 6 8 9 8 6 7 8Running Average 5.8

In this case, each reading must be weighted. The currentreading is weighted with a weight of “1” and each priorreading with an exponentially decreasing weight, dependingupon the time difference and the selected time constant. Ifeach input is designated as xk and each output as yk.

xk + e-dtk-1/t xk-1 + e

-dtk-2/t xk-2 + e-dtk-3/t xk-3 + …yk = ___________________________________________

sum of the weights

The weight given each prior reading is of the form e -dt/t: wheredt is the time prior to the current reading and t is the timeconstant. The rate at which this weighting function drops offis a function of the selected time constant t, as shown inFigure 1.

yk = e-dt/t yk-1 + (1 - e-dt/t)xk

This equation is equivalent to the last, but it requires only thecurrent input (xk) and the last output (yk-1). Thus, only oneresult has to be retained in memory regardless of the timeconstant selected.

This function is available for usage in any system toproduce exponentially-weighted running averages over anytime period. The syntax of the function is:

ewp(argument, t, I)where:

argument = any valid transfer equation argument: a chan-nel name, a raw parameter, or a mathematical expressioncontaining a channel name or parametert = the time constant, in minutesI = The initial value of the function at start-up. (Usually, 0)

As shown in Figure 1, higher frequency of readings has adiminishing influence over time. However, less frequentreadings have a great influence over time. The classic ex-ample of this phenomenon, in pharmaceutical water systems,is the viable microbial test. Microbial testing is an off-linelaboratory test and frequency is commonly once a day or everyfew days. Based on the results of the test, the pharmaceuticalwater system is operated on a single datum for the interimuntil the next test. Thus, the infrequent nature of the testing

has an extraordinary influence on the continued operation ofthe process.

In DP statistics, it is typical to think and describe in termsof samples and averaging instead of signals and filtering.Although similar in function, the two sets of terms implysubstantially different characteristics about the source of thereadings. “Samples” is correctly applied to a series of inde-pendent measurements if the change in each successivemeasurement has no relationship to changes in the priormeasurement. The example of the microbial testing is of thisilk. Each measurement is independent and has no influenceon the successive measurement.

However, in relation to continuous processes, a time seriesof measurements is constrained by the dynamics of theprocess itself. If we take the temperature of 5,000 gallonpurified water tank at intervals of 1 reading/second, wewould not expect the measurement to change or deviate muchfrom reading to reading. Thus, due to the inability of thatwater volume to change temperature quickly, each succes-sive reading is highly dependent on the previous one and theorder of reading in the time series becomes essential to thesignificance of the measurement.

The point of this discussion is to emphasize that continu-ous process measurements can no longer be viewed as inde-pendent samples in the DP sense. The time series of readingsmust be viewed as a signal, reflecting the dynamics of theprocess and the effect of the sampling rate. “Averaging Nsamples” becomes “filtering over a time period.” Randomvariations in the readings can be referred to as “noise.” EWPfiltering and conventional running averages both averagedata, but they employ different methods. While conventionalrunning average weights all the readings over the averagingperiod equally, the EWP filter assigns weights that decreaseexponentially with time. EWPS is more responsive to changein the data as the response is twice as fast given the same timeincrement. This is depicted in Figure 2. The response of theEWPS is faster at the initial start-up. The confluence of bothdata traces occurs only when the running averages trace hasaveraged enough data points to conjoin the EWPS response.In fast changing or dynamic environments, the running

Figure 3. EWPS control chart – Control Chart A.




averages will always lag behind the true changing nature ofthe process and the EWPS response.

Natural Control ChartsIn conventional or DP process statistics, the model for a processvariable assumes that the variable is constant and smallvariations are due solely to random influences. In fact, thepurpose of conventional control charts is to detect when this isNOT the case and alert the operator to this condition. Thisapplication of statistical methods is of great utility and rou-tinely applied, but it does not address a wide range of processmeasurements which do not fall into these conditions.

Typical measurements not suited to conventional controlcharts are process parameters not controlled by the process,but are part of the process. Examples are raw water qualitysupplied by well water, municipal supplied water affected byseasonal changes, and conductivity. In these cases, a mea-sured parameter may vary in what appears to be randomfashion, be affected by another measured parameter, orexhibit cyclic changes with no defined pattern. Typically, thegoal of measurement is to quantify the effect the measuredparameter has on process quality.

Even tightly controlled process variables exhibit system-atic changes in the mean value. For instance, the reaction ofa control loop to disturbances in the process is a function ofhow well the loop is designed, tuned, and maintained. Subtleinteractions of two or more controlled variables could upsetthe consistency of the delivered quality of the water.

No instrument, in a continuous process, always gives theexact same value. The values will deviate slightly fromreading to reading even if the actual value is the same,dependent on the sensitivity of the instrument. Subtle changesin the output of the signal can vary slightly in the real processworld. Sometimes, this slight deviation is compensated bythe use of truncated values.

Using data management software and data acquisitionsystems, one can construct unlimited variety of process controlcharts. Examples: real-time update and historical charts ofsingle parameter data, means, standard deviations, +/- 3scalculations with upper and lower control limits, Booleaneffects, multiple disciplinary charts with multiple data traces,cause and effect charts with multiple data traces, correlations,etc. A few charts are presented as examples. However, it is the

process engineer with specific needs and the ambition to deviseanalysis methods to solve real-world situations and problems.

To illustrate the value of EWPS and SPC, the followingcontrol charts were generated. The first example is Tempera-ture of an ambient pharmaceutical water system controlledat 70°F as denoted in the raw values of X. The control chartof X and the display readings are shown in Figure 3.

In Figure 3, the transfer equation involves simple scalingand offset of the raw measured parameter “a.” We have notyet determined what the control limits should be. Controlchart limits for individual measurements are calculated interms of mean and multiples of the standard deviation of theX measurements. The next control charts add the mean andthe standard deviation of X._

Mean is denoted by “X” and produces an on-line average ofX using EWP in the transfer equation. By setting the timeconstant at T1 long enough, the average is produced overenough readings to represent the population average or the“grand mean” of X as depicted in Figure 4.

_Channel Name: XTransfer equation: [°F] = EWP([x], 60, 0)

With the 60 minute time interval, the total values displayedat a 1/second frequency are 3600 readings. This chart islabeled as M:X or Mean of X

Channel Name: sTransfer equation: SQRT(EWP(([X] – [M:X])^2,T,I))

The second displayed channel measures the Standard Devia-_tion of X relative to the X. Using the same 60 minute constant,t1, the population statistic s is derived and shown in Figure4.

The final step in the control charts is to calculate and_establish the control limits for both X and s. Control limitsare set to the mean of the signal +/- n standard deviations ofthe signal. Not all the data must be displayed to calculate thecontrol limits. Displayed in Figure 5 is the readout of selecteddata points over 17 hours of readings. The mean of thetemperature is 69.754°F and the standard deviation of the

Figure 4. Deriving the population statistic s – Control Chart B.

Figure 5. Calculating and establishing control limits for X and s –Control Chart C.




Figure 6. Control Chart D. Figure 7. Control Chart E.

mean (X Bar) is 1.262° F. (The trace at the bottom of the chartis Standard Deviation of X, not X Bar)

As seen in Figure 5, the mean of X is 69.754°F and thestandard deviation of the mean of X is 1.262°F. If the controllimits for the parameter are to be set at +/- 3s, calculationsare 69.754 +/- (3*1.262) or 75.53°F as the Upper Control Limitand 65.97°F as the Lower Control Limit - Figure 7. However,this 6s spread may or may not be tolerated by the process.

Thus, 6s control limits may be too wide for your process andtighter controls may be needed. This is the reason to lookintently at the standard deviation and set tight controls onthe Standard Deviation data trace.

The Standard Deviation of X is the bottom trace in Figure6. The mean (X-Bar) of the SD(X) is 4.842 and the standarddeviation of the SD(X) trace is 0.907°F (noted as ss).

Upper Control Limits (UCL) and Lower Control Limits

Figure 8. Averaged particle concentration.




Figure 9. Particle concentration; addition of the EWPS function to the running average data.

(LCL) are calculated based on a +/- 3s or a 6s spread. Thereare UCLs and LCLs for both the mean value of X and theSD[X] (Standard Deviation). These are calculated as UCL =µs + nss and LCL = µs - nss. In Figure 7, all means, standarddeviations, and control limits are displayed.

The control charts depicted illustrate a single value oftemperature with a continual trace output. Pharmaceuticalwater systems have many on-line instruments for the man-agement of the water system not mandated by the USP. Flow,pressure, differential pressure, temperature, and liquid par-ticle counting are not mandated USP measurements, but areimportant measurements for the stability and operation ofthe water system. The use of EWPS has a direct correlationon multivariate readings with synergistic influence on thewater system.

Example: the start of a pump can initiate a “water ham-mer” effect, a sudden increase in water pressure affectingvarious downstream components. The increase in TOC due tosloughage of the filters during a “water hammer” effect is welldocumented in high purity literature. The increased pressurecauses the dislodging of trapped TOC in the filter. The suddenincrease in the TOC may cause a spike well above traditional

operating levels of less than 100 ppb of TOC. If this spikeshould register above the 500 ppb (0.5 ppm/L) level, is thisconsidered an action for the immediate shutdown of the waterfacility or the segregation of the water? The answer is no. Theuse of EWPS ensures a faster response to immediate changesin the water system without the violation of the protocols andshows the value of the data in short-term, medium term, andlong-term operations. If the initial readings were based on a0.1 minute interval when the spike occurred, then the use ofone minute intervals will smooth the value of the spike andrender the system within limits. This is not to negate thespike, but the spike can be explained with the data due to theincrease in pressure and the start of the pump, thus avertingan investigation in the sudden TOC increase. However, thenext TOC readings should show a decrease in TOC with arapid return to the sub-100 ppb level. Without the extensivemonitoring and the use of EWPS, the original TOC spikewould have never been explained, but relegated to a status ofan unexplained anomaly after extensive investigation withno resolve.

Known assignable causes to increases in conductivity: ionexchange exhaustion, R/O membrane breakthrough, small




molecular weight species of organics partially oxidized toionic components, atmospheric contamination, etc. With aconductivity measurement alone, how can the various causesof increased conductivity be investigated? The use of datastreaming from all of the various water train components canhelp, but the use of EWPS and SPC charts will notify theoperator immediately of a transition above the standarddeviation and its limits, depicting whether the spike was ananomaly or is instituting immediate trending toward controllimits.

Traditional running averages and averaging programswill not respond quickly to the changing values, especially iflong interval averaging is used, as the spike will be smoothedby the next averaging value. The longer the time interval, theless chance of finding spikes in the system. Note the followinggraphs of a running average of liquid particle counts wherethe intervals for sampling vary from 0.1 minutes, 1.0 min-utes, and 10 minutes - Figure 8. Each data set in runningaverage starts at the 10 minute interval increment as acertain number of values is needed for averaging. The in-crease in sampling interval misses the spikes and smoothesdata.

The EWPS response is immediate and faster than runningaverages for the 0.1 and 1 minute intervals - Figure 9. Whenthe interval is increased to 10 minutes the response of eachsmoothing technique is almost identical with EWPS havingan immediate response over the first 10 minute readinginterval and running averages initiated only at the 10 minutemark.

Findings and ConclusionsAlthough this information seems trite, having this data andgraphical expression can help monitor and manage the phar-maceutical water system to very tight tolerances with little orno downtime over long periods of time.

The graphical examples shown have been for the tempera-ture and liquid particle counters of a pharmaceutical watersystem only. Many other on-line sensors, instruments, anddevices are installed in a pharmaceutical water system.Signals vary from analytical instrumentation with multi-stream data to singular devices with a single output. Theadvantage gained is the complete integration of all possibleinstrumentation into one network, data acquisition, and dataarchival/retrieval system. Thus, flow, temperature, conduc-tivity, differential pressure, pressure, ozone concentration,TDS, TOC, chlorine, etc. can all be monitored with accurateprecision and trending capability.

The purpose of measuring the standard deviation, in real-time, in relation to the actual channeled data from theinstrumentation is to enable and understand the reality ofthe data at any given moment. The use of 6s does not preventoutliers and questions of instrumentation integrity, espe-cially, if data averaging is used. The use of a single standarddeviation with tight control limits will inherently alert theuser to any deviation which is outside the norm. If an increasein temperature is gradual due to averaging of the data, the

system could be out of control long before an alarm is alerted.This statement can be seen in the control charts of the mean.Long averages will cancel out short-term excursions that donot exceed a control parameter. However, the use of tightcontrol limits on the standard deviation trace will depict anyreading that exceeds the normal standard deviation. If thestandard deviation is .907° F degrees, then anything thatexceeds 1.5s of that number can be attributed to a faultyreading or a true trending. If the 6s values of the temperaturerange from 65-75°F with a mean of 70, a two degree shift ispermissible and within the control limits of the process. Theissue is how to identify a trend long before a control orspecification limit is achieved. The use of EWPS on both theraw data and the standard deviation will alert the user toprocess issues long before an alarm is tripped.

EWPS can be used on any episodic or analog data stream,for any instrument and any parameter. The exclusive use of6s for any given process parameter may have too wide atolerance and may not be indicative of the actual dynamicprocess parameters.

Continuous processes need accurate data acquisition andmanagement to help determine the dynamic makeup andchanges. Although Discrete Products management and itsdata have been traditional in pharmaceutical manufactur-ing, the need for continuous data acquisition, on-line trend-ing information, and data analysis as an overall feature ofpharmaceutical production processes cannot be overlooked.Continuous processes with dynamic and changing param-eters need to be monitored and managed closely to preventproduct deviations and quality upsets. Fast and responsivestatistical tools like Exponentially Weighted Process Statis-tics will enable the engineers to assess the issues at hand,take necessary steps, and ensure the process continuity.

References1. Lucas, J.M., Saccucci, M.S., “Exponentially Weighted

Moving Average Control Schemes: Properties and En-hancements,” Technometrics , (1990), Vol. 32, pp.1-12.

2. Woodall, W.H., Maragah, H. D., “Exponentially WeightedMoving Average Control Schemes: Properties and En-hancements: Discussion,” Technometrics, February 1990,Vol. 32, No. 1, pp. 17-18.

3. Berthouex, P.M., Wenje, L., and Darjatmoko, A., “A Sta-tistics-Based System for Treatment Plant Operation,”Earth and Environmental Science, November 1989, Vol-ume 13, Numbers 2-3, 1573-2959 (Online), pp. 247-269.

4. Cox, D.R., “Prediction by Exponentially Weighted MovingAverages and Related Methods,” Journal of the RoyalStatistical Society, Ser. B, (1961), Vol. 23, pp. 414-22.

5. Crowder, S.V., Hamilton, M.D., “An EWMA for Monitor-ing a Process Standard Deviation,” Journal of QualityTechnology, (1992), Vol. 24, pp.12-21.




6. “EWPS Exponentially Weighted Process Statistics,”DataTrax, A Division of Eaton Corporation, Internal Docu-ment, (1990), Graphs courtesy of DataTrax.

About the AuthorNissan Cohen has more than 30 years ofexperience in instrumentation and mission-critical monitoring with emphasis in semi-conductor manufacturing, pharmaceuticalprocess and production, ultrapure water,drinking water, waste water, chemical sys-tems, nuclear, hydroelectric, and fossil fuelpower generation, and environmental issues

of containment and remediation. Cohen has written morethan 30 technical and peer reviewed articles for variouspublications including Pharmaceutical Engineering, Phar-maceutical Technology, Ultrapure Water, Semiconductor In-ternational, Contamination, A2C2, and The Journal of theInstitute for Environmental Sciences. A recognized world-wide expert in TOC and water systems, Cohen is a memberof ISPE, the Institute for Environmental Sciences and Tech-nology (IEST), Technical Editor for the Journal of the Insti-tute for Environmental Sciences and Technology, member ofthe ASTM International E-55 Committee, member of thePharmaceutical Engineering Committee, Steering Commit-tee member of ISPE Communities of Practice for CriticalUtilities, Process Analytical Technology, and HVAC, formerChairman of the ISPE Internet Subcommittee, and PastChairman of the ISPE Membership Services Committee.

Start-Up Business Development, 267 S. Jefferson Ave.,Louisville, Colorado 80027.


Control System Architecture


Figure 1. S88 modelsand methodologyoverview.

This articlereports on aunified MCSarchitectureusingcommerciallyavailable MESand PCS. Itexplains thesteps in movingbeyondpaperlessfunctionality toa unified systemhelping manageinformation,processes, andpeople.

Unified Manufacturing Control System(MCS) Architecture for Pharmaceuticaland Biotech Manufacturing

by Ronald E. Menéndez and Darrell Tanner

Introduction

In recent years, Manufacturing ExecutionSystems (MES) and Process Control Sys-tems (PCS) have gained wide acceptance inthe pharmaceutical and biotech industries,

due to the adoption of industry standards andtechnology advancements. PCS for bulk thera-peutic and biotherapeutic manufacturingachieved uniformity in the past decade thanksto the establishment of the ANSI/ISA-88 mod-els for batch control. During the same period, abroader range of industries used MES and

ANSI/ISA-95 standards to improve their manu-facturing operations.

While MES and PCS found their place in theindustry, they were typically viewed as sepa-rate solutions within a manufacturing facility.This approach often led to a disparity of sys-tems and organizations responsible for devel-opment and maintenance. The resulting sys-tems were usually hindered by a lack ofinteroperability and dependence on custom in-terfaces for connectivity.

As companies pursue MES and process au-

Reprinted from







Evolution of Batch Control

A key requirement for effective batch control iscollecting useful data and information—and knowingwhat to do with it. To standardize the use of batchcontrol technology in the process industries, theInstrumentation, Systems, and Automation Society(ISA) established the S88 standard. The ISA guide-lines identified a common set of procedures that canbe used to describe and define batch manufacturingsystems in accordance with U.S. Food and DrugAdministration (FDA) current Good ManufacturingPractices (cGMP).

S88 Defines Manufacturing MethodologyThe S88 methodology breaks down each manufactur-ing module into a pyramid of smaller and smallerprocess steps, known (in descending order) as “proce-dures,” “unit procedures,” “operations,” and “phases.”

The models and terminology incorporated in the S88standard emphasize good practices for the design andoperation of batch manufacturing plants. They can beused to improve control of continuous or discreteprocesses, and applied regardless of the degree ofautomation. The standard includes both physical andprocedural models that are written once and thenemployed as templates.

Physical models define the equipment used in theprocess, such as units, equipment, and control mod-ules, whereas procedural models, which include proce-dure, operation, and phase modules, define the controlenabling the physical models to perform given tasks -Figure 1.

Proper implementation of S88 batch automationreduces the time required to reach full production levelsfor new products. It also helps vendors supply appro-priate tools for implementing batch control, and allowsusers to better identify their needs.

Standard Improves Process Design PhilosophyS88 isn’t just a standard for software, equipment, orprocedures; it’s a way of thinking, a design philoso-phy. Understanding S88 will help you better designyour processes and manufacture your products. Le-veraging the knowledge and experience contained inthe standard will enable you and your customers tobetter identify your needs, make recipe developmenteasier, and help reduce the time to reach full produc-tion levels with a new system or for each new product.Following the concepts explained in S88, you canimprove the reliability of your operations and reducethe automation lifecycle cost of your batch processes,including lowering the initial cost of automating youroperations.2

tomation initiatives, they are often challenged by varyingbudgets, schedules, and project methods. That is becauseautomation is traditionally viewed as an engineering disci-pline, whereas MES is regarded as an IT function. However,in a recent project at a brownfield biotherapeutic manufac-turing facility, a new aggregate approach referred to as theManufacturing Control System (MCS) was put forth as asolution to provide a single environment for manufacturingoperations and process automation meeting all requirementsof a paperless facility.

This article reports on this system integration effort andpresents a unified MCS architecture using commerciallyavailable MES and PCS. It further explains the steps inmoving beyond paperless functionality to a unified systemthat helps manage information, processes, and people.

Advancements in Automation TechnologyIn the 1990s, the advent of open systems in process automa-tion changed the way manufacturers operated their plants.Proprietary computer networks and control applications fromautomation vendors gave way to PC-based hardware usingcommercially available operating systems. Ethernet commu-nications employing standard wiring, switches, and routerssuperseded proprietary communication protocols.

Modern control systems utilizing Web-based Human-Machine Interfaces (HMIs) provide a single, facility-wideview of operations. These systems, designed to integratebusiness processes with a common HMI across the plant, alsoprovide seamless, third party integration through open Webstandards. This trend toward third party integration enabledthe advancement of batch control technology benefiting auto-mation end-users throughout the process industries.

Batch Management Increases FlexibilityBatch management software integrated in most PCS avail-able on the market today provides a robust solution fordesigning, modeling, and automating batch processes. Itenables flexible recipe building and management using ob-ject-oriented recipe structures aligned with the S88 models.On-line tools allow users to manage multiple batches fromthe same window, and navigate between displays based onbatch execution activities.

S88 batch management applications for automated recipemanagement and unit procedural control reduce latenciesand improve repeatability. This, in turn, improves produc-tion efficiency. S88 batch automation ensures procedures areexecuted in accordance with approved specifications andstandard work processes. Using these applications, manufac-turers have achieved faster response to production ordersand schedule changes, flexible processing to support newproduct introduction, and increased throughput to meet ex-panding production demands.1

Development ofManufacturing Operations Technology

In a variety of industries, MES has proven to be effective inmanaging all steps of the production lifecycle; from materials




gration, and production dispatching and execution in single,unified environment.

Benefits of a Unified MCSThe Manufacturing Control System (MCS) is the integrationof MES and PCS technology to provide a single solution forproduction management, process automation, and reporting.This unified MCS design utilizes the strengths of MES formaterial management and plant floor applications, and atthe same time, incorporates the latest advancements in PCStechnology - particularly in the areas of automated recipemanagement and unit procedure control. Together, the twosolutions are employed in a way that is most beneficial tooperational objectives.

Tight integration of MES and process automation allowspharmaceutical and biotech manufacturers to move beyond“paper-on-glass” functionality and leverage all of the robustcapabilities the two systems have to offer. These include:electronic work instruction execution and workflows, mate-rial reporting, asset management, laboratory data logging,production dispatching, and Electronic Batch Record (EBR)management.

Open Communications Interface to ERPImplementation of MCS requires an open, standards-basedprogramming interface allowing communication betweenMES and ERP solutions and business logic. Such integrationenables users to access production-related information fromthe MES and business applications in real time. This connec-tivity, made possible by S95 Parts 1 and 2 defining ERP/MEScommunications standards, is a precursor to MES/PCS uni-fication - and a new level of plantwide integration.

Within the integrated manufacturing architecture, MESserves as an interface to corporate-level 3 and 4 systems,electronic document management systems, laboratory infor-mation systems, Material Resource Planning (MRP) sys-

Figure 2. S95 production operations management model.

receipt to product shipment. The technology and S95 stan-dards assist production personnel in managing executiondecisions and information during the processes of planning/scheduling down to production execution.

Typical MES provide specification management tools al-lowing users to define the materials, equipment, and proce-dures required for production. In many cases, the systems canbe expanded to handle multiple production sites - enablingproduct development departments to quickly deploy newproducts or update existing product formulations.

Characteristically, MES benefits manufacturers by pro-viding a scalable, Web-based architecture that is easy todeploy and maintain. MES can form the central system forsynchronization of business systems with manufacturingand process control - Figure 2. Integration with other manu-facturing systems can be achieved using Web services andindustry standard technologies such as XML and OPC.

Paperless Records Reduce ErrorsKey to the adoption of MES technology was its promise ofeliminating paper-based batch recordkeeping. With the FDA’sre-examination of 21 CFR Part 11 and their issuance ofGuidance for Industry Part 11, Electronic Records; ElectronicSignatures – Scope and Application,5 there is a greater under-standing of the compliance requirements for paperless sys-tems in the regulated industries.

MES makes it easier for pharmaceutical and biologicsproducers to meet regulatory compliance by managing andrecording activities associated with personnel, manufactur-ing resources, and the process itself. In addition, the MESsolution is a direct means to reduced human error duringdata entry. Users can reduce paperwork, improve overallresource management, and produce fully compliant, paperlessproduction records.

MES provides a “paper-on-glass” replacement for tradi-tional paper formulations, typically referred to as “tickets,”by offering prompted data collection, electronic work instruc-tions, and e-signature-based review processes.

Challenges Facing MES SolutionDespite the merits of MES, the technology alone cannotadvance the state of biologics and pharmaceutical manufac-turing. This is because traditional “paper-on-glass” systemsdo not collect, organize, and manage all production informa-tion - particularly manufacturing and process data generatedby the PCS.

Although MES applications have matured around inte-grated material management and paperless plant-floor op-erations, which provide significant production efficienciesand cost savings, often personnel find themselves manuallymanaging vast amounts of information. Users are required torefine production data so operations and quality decisionscan be made in a timely manner.

Combining today’s MES with batch control provides abeneficial architecture for tackling activities such as mate-rial tracking/genealogy, barcode scanning, bills of materialand work instructions, asset management, lab systems inte-




tems, and other Enterprise Resource Planning (ERP) appli-cations.

The MCS provides a platform for handling both inboundtransactions (i.e., process orders and lab results) and out-bound transactions (i.e., inventory updated and lab requests)- Figure 4.

Typical MES/PCS TransactionsThe benefits of the unified MCS approach are demonstratedthrough MES/PCS transactions, such as production execu-tion, resource management, material tracking, and elec-tronic work instruction management. Unit procedural con-trol and phase execution with an MCS is more efficient thanin a traditional environment with separate system domains.Transactions between different systems and personnel areseamless; operators see a unified interface with a commonHMI environment, instructions, and displays.

Production Definition, Dispatching, andExecutionWith the unified MCS architecture, orders from MRP comedown to the plant floor through the MES - Figure 5. The MESautomatically dispatches recipes based on required equip-ment statuses and availability, and executes them in theprocess control system. This innovative approach eliminatesthe traditional requirement for operators to manually checkequipment status, assign equipment, load recipes, and ini-tiate batch execution. Rather, the MES handles these activi-ties as the operator fulfills the order at the PCS layer.

Consider a typical MCS batch processing application:after dispatching a unit procedure, the MES binds to theprocess unit for execution and starts the sequences. The PCSthen executes phases within operations at the equipmentlevel, performs automated tasks, and requests informationfrom the MES.

Figure 3. S95 control hierarchy levels.

Evolution of MES

In the 1990s, with adoption of the ANSI/S95-95(S95) standard, manufacturing companies began imple-menting MES technology to ensure their productionoperations were capable of delivering on theirenterprise’s supply chain commitments.

MES holds the potential to significantly improvemanufacturing excellence and compliance to regula-tions. However, realizing this promise requires tightintegration of information and work activity across allthe real-time levels of the S95 model. Integrated recipeauthoring and execution delivers the MES promiseacross both bulk production and finishing, while reduc-ing the time and risk required to deploy electronicrecipes.

Understanding the S95 Control HierarchyS95 Part 1 defines the interfaces between businesslogistics systems and manufacturing operations sys-tems. Part 2 doesn’t add any new concepts to theintegration model, but it contains additional detailsand examples to help explain and illustrate the Part 1objects. Part 3 defines models for the disparatecollection of activities that must occur in manufactur-ing operations for effective and efficient manufactur-ing. The goal is to provide manufacturing companieswith a common language to describe requirements tovendors and let companies compare alternate archi-tectures and solutions.3

Upcoming S95 Parts 4 and 5 will address objectmodels and attributes for Manufacturing OperationsManagement, as well as business to manufacturingtransactions enabling information collection, retrieval,transfer, and storage in support of enterprise/controlsystem integration.

“Shop Floor to Top Floor” IntegrationS95-compliant MES systems fill the complicated gapsbetween the “top floor and the shop floor,” linkingbusiness systems and the core automation, controls,and HMI/SCADA, and pure manual data collectionsystems existing in the manufacturing environment -Figure 3.

Although the S95 standard includes a modelsimilar to S88 that defines terms and transactions,the scope of S95 goes on to define activities andmodels at all levels of the production process. Userswho purchase systems from different suppliers fordifferent levels of the organization can have confi-dence that they will understand how they communi-cate along with finding greater ease with the integra-tion process if both are compliant with the S95standard.4




Management of Resources and their StatusesAt a typical pharmaceutical or biotech plant, operators aretasked with managing production resources and reporting ontheir statuses. The operator must verify the status of speci-fied equipment in a paper log or database before a batch canbe started or progressed.

The MCS solution automates this procedure since theprogrammed phase in the PCS controls specific equipment.The phase is designed to automatically request equipmentand assets from the MES based upon their required status.PCS requests for information are handled by a transactionexecuted to the MES via an OPC service. The MES automati-cally allocates resources and performs arbitration shouldconflicts arise. This allows the automation process to con-tinue without interruption.

For example, The PCS might issue requests such as, “Thistank is needed - is it sterile?” The MES will respond, “Yes, youcan acquire this resource because its status is correct for yourrequirements.” Once the operation is completed, the PCSphase will release the tank back to the MES with a messagesaying, “This equipment is being returned with a status of‘dirty.’” Such transactions are carried out automatically,without operator intervention.

Material Management and TrackingWhen it comes to material tracking and reporting, the PCSphase again interfaces directly with the MES, which in turn,interfaces with Manufacturing Resource Planning (MRP) asrequired for inventory updates. During execution of a par-ticular phase, the system might say, “Material ‘A’ for the

batch is needed.” The MES then reports,“The material’s quality is acceptable andthe expiration date has not been exceeded.Here is the quantity that should be added.”It then provides results regarding barcode scanning and performs system dataverification at the point of use (i.e., whenthe material is introduced into the batch).

When tracking material consumption,the PCS can send a transaction notifyingthe MES that it is time to automatically ormanually consume a particular additiveor ingredient. As the automated stepsexecute, a procedure pops up on theoperator’s screen with prompts for com-pleting the task.

Under normal circumstances using dis-parate MES and PCS systems, the opera-tor has to pull up a ticket or paper-on-glass in the MES environment to checkthe status of materials, and verify infor-mation indicating that he is adding theprescribed material. Then, he must ac-knowledge the material addition is com-plete and instruct the PCS to continueexecution.

In the case of manually consumed ma-terials, standard material add pages prompt the operator toscan the required material and then automatically executethe quality checks prior to prompting the operator to deliverthe material. For automatically added materials, the qualitychecks and consumption reporting are done without operatorintervention unless required.

Management of Electronic Work InstructionsThe MCS strategy also revolutionizes the handling of elec-tronic instructions and workflows and eliminates paper pro-cedures. Unlike a standalone MES, the integrated systemautomatically presents instructions or workflows (i.e., SOPs)on the HMI screen whenever and wherever they are needed.Operators are no longer burdened with coordinating MESactivities, while staying abreast of PCS execution. This en-ables a new level of plant production efficiency.

During a phase execution, for instance, the system calls upstandard faceplates on the process control graphic thatprompts the operator whenever his attention is required. Theoperator is presented with an “action list” displaying phaseswith their instruction, a button to display the detailed in-struction, and upon acknowledgement of the action, the typeof signature required. Operator instructions can be signed offdirectly from the HMI page - Figure 6.

Likewise, in the middle of a phase, required manualactions can appear on the MES page as a workflow thatincludes a variety of MES activities the operator must follow.Once the tasks are completed, the technician acknowledgesthe work with an electronic signature and the PCS resumesautomated control.

Figure 4. Typical system transactions.




Figure 5. Unified MCS architecture.

DiscussionFor pharmaceutical and biotech operations, the unified MCSnot only delivers new automation capabilities, but also presentsnew ways to manage manufacturing complexity and improveoperational efficiency. Industry analysts estimate that as muchas 20% of a firm’s costs of operations are associated withmanufacturing, which means even modest operational im-provements can have a significant financial impact.6

Greater People CollaborationFor a plant’s technology personnel, the MCS merges dispar-ate MES and automation departments into an integratedproduction team that works hand-in-hand to optimize manu-facturing operations. Under the new architecture, compo-nents such as work instructions, bill of materials, and assetdefinitions are supported in the MES, but requested from

phases executed in the PCS. As a result, the two departmentsinteract to ensure components are correctly configured andmanaged. This closer collaboration enabled a reduction ofsupport staff between the two departments of over one-thirdand the restructured groups operate as a single organizationas opposed to separate teams of engineers and IT specialists.

In addition, manufacturing personnel, quality depart-ments, and engineering staff now utilize a single, unifiedsystem with a common environment for accessing productiondata, viewing process displays, and making critical opera-tional decisions.

Faster Review and Product Release ProcessesThe MCS solution eliminates the need to manage paper batchrecords. The system provides electronic records of each batchof products produced, as well as the means to collect, store,




and analyze data more efficiently. Documents within regu-lated environments can be created, reviewed, approved, andissued electronically in a collaborative manner with fullchange control by the respective departments. This approachsimplifies the GMP-related document management process.

Manufacturers implementing MCS can streamline theeffort required for regulatory compliance and expedite therelease of manufactured product. The integrated system,with EBR and process automation information (i.e., alarms,unit control data, batch events, process history, etc.), enableseasier compliance and verification. The plant’s quality groupcan access robust, consolidated data assisting the reviewprocess prior to product release.

Previous paper-based systems required numerous weeksto collect paper records, review, reconcile discrepancies, andapprove for release. Subsequent designs of disconnectedMES and PCS architectures reduced the product releaseprocess to a couple of weeks, but the new MCS design isestimated to reduce this process to a few hours.

Better Information Equals Improved PerformanceThe integrated MCS system approach brings together infor-mation from key areas such as process control, MES, andlaboratory systems. This facilitates the discovery of newopportunities to improve operational performance and drivedown costs.

Pharmaceutical and biotech facilities gain a world-classsolution providing a single source of centralized manufactur-ing data. No longer must PCS data be duplicated for the MES

Figure 6. Unit control graphic.

environment, and then migrated into the ERP system. Insteadof distributing asset and process information between threedifferent systems, users attain a “Single-Source of Truth.”

Equally important, an interoperable MCS design withWeb/HTML-based applications and open, industry standardcommunications protocols provides a secure and predictablepath for future technology investments. Potential directionscan include RFID, biometric security, and wireless hand-heldmobile devices, to name a few.

ConclusionIn the life sciences industry, the constant challenges formanufacturing are efficient, streamlined operations withfewer errors, greater consistency, and unfailing compliancewith FDA regulations. Manufacturers seek shorter productcycle times, faster product changeover, and better mainte-nance scheduling - all adding up to improved operationalperformance.

To meet these challenges, a seamless MCS architectureproviding common electronic batch records and productionreporting for automation and production management withreliable traceability (i.e., materials, equipment, and person-nel) can be employed as presented in this article.

Traditionally, many manufacturing facilities have had adisconnected view of their automation and IT solutions. Asthe lines between systems blur, a paradigm shift is likely tomove the industry toward the next-generation MCS, mergingcurrent MES and PCS with integrated batch control technol-ogy. As this natural evolution progresses, companies will




actualize the benefits of managing plant-level and corporate-level systems as part of a single system. And, unified enter-prise architectures, as discussed in this article, will likelyemerge as the standard manufacturing solutions for newfacilities by the end of this decade.

AcronymsEBR Electronic Batch RecordsERP Enterprise Resource PlanningIT Information TechnologyHMI Human Machine InterfaceMES Manufacturing Execution SystemMRP Material Resource PlanningPCS Process Control SystemPC Personal ComputerOLE Object Linking and EmbeddingOPC OLE for Process ControlsRFID Radio Frequency IdentificationSCADA Software Control and Data AcquisitionUP Unit ProcedureXML Extensible Markup Language

References1. Menendez, R. E., “A New Suite of Automated Bioreactors

at Genentech: A Case-Study Employing Standards, Guide-lines, and Software: From Requirements to Validation,”Pharmaceutical Engineering, Nov/Dec 2004, Vol. 24, No. 6,pp. 8-16.

2. Parshall, J. and Lamb, L., Applying S88 Batch Controlfrom a User’s Perspective, ISA, 2000.

3. Enterprise Integration Takes the Fast Track, InTech, Sep-tember 2001.

4. Mintchell, G., “Drive Integration with Standards,” Auto-mation World, September 2004.

5. U.S Food and Drug Administration, Guidance for IndustryPart 11, Electronic Records; Electronic Signatures – Scopeand Application, August 2003.

6. Implementing Manufacturing Execution Systems in thePharmaceutical and Biotech Industry, Center for BusinessIntelligence, August 2006.

About the AuthorsRonald E. Menéndez has more than 16years of automation experience within thebiopharmaceutical, pharmaceutical, chemi-cals, and consumer goods industries, and isexperienced in a variety of control systemplatforms and batch management software.He is a Senior Automation Engineer in theCorporate Engineering Department at

Genentech’s headquarters in South San Francisco, Califor-nia, where he currently serves as the lead design engineer ofthe control system and MES for Upstream operations at anew bulk manufacturing facility in Vacaville, CA. Previously,he worked for the Sequencia Corporation, where he contrib-uted to the development of batch management software inthe early 1990s; performed system integration and plantstart-ups in Latin America, Southeast Asia, and Europe; andprovided S88 consulting to control system vendors,biopharmaceutical and pharmaceutical manufacturing com-panies. He holds a BS in chemical engineering from theUniversity of California, Santa Barbara. He can be contactedby e-mail at [email protected].

Genentech, 1000 New Horizons Way, MS# 181/Rm. #81002, Vacaville, California 95688.

Darrell Tanner has 17 years of experiencein the process automation industry design-ing, developing, and delivering solutions tothe life sciences, consumer goods, and chemi-cal industries. He is currently the HoneywellTechnical Lead for Genentech’s new bulkmanufacturing facility in Vacaville, Califor-nia. He has a BS in chemical engineering

from Arizona State University. He can be contacted by e-mailat [email protected].

Honeywell, 3079 Premiere Pkwy., Duluth, Georgia 30097.


Regulatory Requirements


This articlecompares therequirements inthe 21 CFR Part11, EU GMPAnnex 11, andPart 11’s EUcounterpart PI011-2.

Pharmaceutical Standards forComputerized Systems

by Siri H. Segalstad

Introduction

GxP is the term used to refer to pharma-ceutical and associated life scienceregulations, including those coveringmanufacturing (Good Manufacturing

Practice - GMP), non-clinical testing (GoodLaboratory Practice - GLP), and clinical testing(Good Clinical Practice - GCP).

IT systems may have a direct or indirectimpact on the product quality in various ways.Such IT systems are generally called GxP sys-tems. These systems include, but are not lim-ited to, systems handling materials used in theproduction, production planning and controlsystems, laboratory systems, and systems fornon-clinical and clinical testing. Documentmanagement systems also may be regarded asGxP systems if they handle the quality systemand Standard Operating Systems (SOPs), or ifthey handle production documents and valida-tion documents.

The GxPs themselves do not say very muchabout IT systems so appendices and/or addedstandards have been created to make peopleunderstand how the authorities expect IT sys-tems to be handled. In an old version of the EUGCP, the only requirement for computer sys-tems was “Computer systems shall be vali-dated and error free.” The next version hadremoved the requirement “error free…”

For medical devices,1,2 IT systems are usedin two different ways: one is when the ITsystem is used during development, produc-tion, and control of the medical device, just likeit is for any other pharmaceutical product. Inthis case, the system should be handled thesame way as GxP IT systems. The other case iswhen the system is, or is part of, the medicaldevice itself. Some of these medical devices areimplanted in the body. A pacemaker is anexample of that. Others are used for in vitro(outside of the body) testing, e.g., testing for

allergies, where the test instrument may beregarded as a medical device.

A pharmaceutical company selling to theEU market must comply with the EU regula-tions, and a company selling to the US marketmust comply with the US regulations, regard-less of where the company is developing ormanufacturing its goods.

The GxP Requirements forIT Systems

European GxP Requirements for ITSystemsGMP: The EU GMP added Annex 113 to explainwhat the regulators considered was needed forcomputerized systems. Annex 11 originated asPIC/S4 GMP Annex 5 in 1991 and was lateradopted by the EU GMP as Annex 11 and alsois now Annex 11 in the PIC/S GMP.5

Annex 11 has 19 clauses covering what theinspectors expect, but it is not very useful as atool to tell how to get there. Adding the fact thatthe standard is now very old, it is about time torevise it. As of February 2007, it is still thecurrent requirements for IT systems.

Pharmaceutical Inspection CooperationScheme (PIC/S)4 which is the organization forcooperation between regulators and inspectors,also thought it was about time to revise it,especially after the FDA had created the 21CFR Part 11 with its relatively detailed re-quirements.

The PIC/S finalized their interpretation docu-ment 20 August 2003 under the name PI 011-1 Good Practices for Computerized Systems inRegulated “GxP” environments. The currentversion is PI 011-26 from 1 July 2004. Wheninspectors get together and create a documentlike this, it is worth paying careful attention toit.

GLP has long used the OECD Monograph1167 for the pharmaceutical industry, which is

Reprinted from







Table A. Overview of Requirements in Part 11, Annex 11, and PI 011. The table includes the reference to paragraph or chapter instandard. The text is only an excerpt. For the full text and context please read the standard. (Table is continued on page 98.)

Topic US 21 CFR Part 11 EU GMP Annex 11 PIC/S PI 011-2

Validity of Standard 11.1 This part applies to records in Not mentioned, but it is an annex to GMP. 3.2 It is not intended to be a barrier toelectronic form that are created, modified, technical innovation or the pursuit ofmaintained, archived, retrieved, or excellence. The advice in this Guidance istransmitted, under any records requirements not mandatory for industry. However,set forth in Agency regulations…. industry should consider these

recommendations as appropriate.

Closed / Open Systems 11.10 and 11.30 defined N/A 19.7 The words are not used, but thedifference is made, and handling is aboutthe same as in Part 11.

Validation 11.10(a) Validation of systems to ensure 2 The extent of validation necessary will Included in all, but five of its 24 clauses oraccuracy, reliability, consistent intended depend on a number of factors, including the chapters. These include severalperformance, and the ability to discern use to which the system is to be put, requirements, many with reference toinvalid or altered records. whether the validation is to be prospective or GAMP 4 for how to do it.

retrospective, and whether or not novelelements are incorporated. Validation shouldbe considered as part of the complete lifecycle of a computer system.

Copies of Records 11.10(c) The ability to generate accurate 12 For quality auditing purposes, it should be 21.10 The ability exists to generate accurateand complete copies of records in both possible to obtain clear printed copies of and complete copies of records in bothhuman readable and electronic form suitable electronically stored data. human readable and electronic form.for inspection, review, and copying by theAgency.

System Access 11.10 (d) Limiting system access to 8 Data should only be entered or amended by 21.10 Access to records is limited toauthorized individuals. persons authorized to do so. authorized individuals.

ID and Password 11.300(b) Ensuring that identification code 8 ….Suitable methods of deterring 19.2Issuance and password issuances are periodically unauthorized entry of data include the use of • The management and assignment of

checked, recalled, or revised (e.g., to cover keys, pass cards, personal codes, and privilegessuch events as password aging).(c) Following restricted access to computer terminals. • Levels of access for usersloss management procedures to electronically There should be a defined procedure for the 19.3 …basic requirements are satisfied:reauthorize lost, stolen, missing, or otherwise issue, cancellation, and alteration of • Access rights for all operators arepotentially compromised tokens, cards, and authorization to enter and amend data, clearly defined and controlled, includingother devices that bear or generate including the changing of personal passwords. physical and logical access.identification code or password information, Consideration should be given to systems • Basic rules exist and are documentedand to issue temporary or permanent allowing for recording of attempts to access to ensure security related to personalreplacements using suitable, rigorous controls. by unauthorized persons. passwords or pass cards and related(d) Use of transaction safeguards to prevent system/data security requirements areunauthorized use of passwords and/or not reduced or negated.identification codes, and to detect and report • Procedures are in place to ensure thatin an immediate and urgent manner any identification code and passwordattempts at their unauthorized use to the issuance are periodically checked,system security unit, and as appropriate, to recalled, or revised.organizational management. Loss management procedures exist to

electronically invalidate lost, stolen, orpotentially compromised passwords. Thesystem should be capable of enforcingregular changes of passwords.• Procedures identify prohibited

passwords.• An audit log of breaches of password

security should be kept and measuresshould be in place to address breachesof password security.

• The system should enforce revoking ofaccess after a specified number ofunsuccessful logon attempts.

better detailed than the old GMP Annex 11. The new PIC/Sdocument also covers GLP.

GCP adopted the GMP Annex 11, and the new PIC/Sdocument also covers GCP.

The three major documents for IT system compliance inthe pharmaceutical EU and US all have a number 11 in theirdocument number. This is probably a coincidence; at least theauthor is not aware of any reason why they all include thisnumber.

21 CFR Part 1121 CFR Part 11 Electronic Records; Electronic Signatures8

was created after the industry requested guidelines for com-pliant use of electronic signatures in the early 1990s. The firstdraft was made public and received a lot of comments, someof which were contradictory. The FDA issued the final versionon 20 March 1997 and had expectations that industry wouldbe able to be in compliance four months later, by 20 August1997. The industry was surprised to see that very little of the




regulation actually covered the electronic signatures, whilethe majority of the text covered electronic records. However,it does make sense: you can only trust the electronic signatureif the electronic records can be trusted.

21 CFR Part 11 is nicknamed “Part 11.” People who haveworked with IT systems in the pharmaceutical industry inthe roles of key personnel, validation experts, or qualityassurance, are already familiar with Part 11. Numerouspublications have been written about the requirements andhow to comply with them. The exact interpretations of therequirements vary. The FDA also added to the confusion byfirst issuing a number of draft guidance documents makingthe interpretation of the regulation stricter, more and moreprescriptive, and then suddenly withdrawing them all whileissuing one new draft guidance document where risk-basedapproach is a key factor.9 In this, it is up to the discretion ofeach organization to assess the importance of their computersystems and handle the high-risk and lower-risk systemsappropriately. The risk assessment itself must be docu-mented. It focuses on risk assessment, but does not go intodetail on how to handle the risks – risk management.

Part 11 has two main sections: one on electronic records,and one on electronic signatures. There is no requirement forusing either, but if IT systems are used, they need to be incompliance with the electronic records requirements. It isoptional to use electronic signatures, but if E-signatures areused, the E-signatures also must comply with the require-ments.

The GxPs are called “predicate rules” to the 21 CFR Part11. These cover a lot of details not mentioned in 21 CFR Part11. The document is not to be read instead of the GxPs, but inaddition to them. One example is the section on electronicsignatures that does not say anything about where or whensignatures are needed. The user must find this in the appli-cable GxP. Then, if the computer system in question shall beused for E-signatures, those Part 11 requirements apply.Another example is that Part 11 does not state that thereshall be given a reason for data changes in the system. Thisis required by the predicate rules.

EU GMP Annex 11EU GMP Annex 11, or just “Annex 11,” is basically a list ofthings that must be in place for computerized systems usedin the pharmaceutical industry. It is not a prescriptive de-scription of what to do and it includes little detailed guidance.It can be used as a checklist for an organization to see if theycomply, but only if the organization understands the intentbehind the words in the document.

PIC/S Guidance - PI 011-2This document provides an EU perspective on electronicrecords and signatures, as well as the wider aspect of the useof computerized systems in the GxP environment, includingadvice on what to take into consideration when implementingand validating a system.

PI 011 has references to ISO,10 IEEE,11 ISPE,12 and GAMP,13

in addition to Part 118 and Annex 11.3 It encourages the use

of the existing standards, instead of repeating the sameadvice. But also it warns that no standard should ever befollowed without understanding how one’s own organizationworks. It is refreshing to see that we are encouraged to think.

It is surprising that the references to the ISO standards10

include 1995 versions and not the 2000 versions, which wereavailable four years before the PI 011-2 document was final-ized.

It is quite obvious that Part 11 has been scrutinized whencreating this document. A lot has been done to make it abetter and more useful document than Part 11. It coversmuch of the same areas as Part 11, but some details arespelled out in a much clearer way.

GAMPGood Automated Manufacturing Practice (GAMP)13 wasstarted as an industry initiative to explain to the pharmaceu-tical industry exactly what computer systems validation was,in order to fulfill the regulatory requirements. These require-ments were very hard to translate to “what do we actually doto validate our computer system.”

The first version came in 1995, and now, more than 10years later it is a widely-used industry guidance document,and is referred to by the FDA and the European agencies. Thecurrent version is GAMP 4.14 The GAMP organization is aCommunity of Practice within ISPE,12 where many peoplearound the world take part in preparing various good prac-tices guidance documents.

ISPE has issued several guidance documents as answersto the new regulatory initiatives, e.g., for Part 11 Risk-BasedApproach to Validation,15 Laboratory Systems,16 Testing ofGxP Critical Systems,17 IT Infrastructure,18 Calibration Man-agement,19 and Global Information Systems.20 They also havepublished Position Papers on topics, including Building Man-agement Systems.21 Each of the guides has detailed sugges-tions for how to practically deal with the topics in question.

ISO 9000-SeriesThe ISO 9000-series standards are not pharmaceutical regu-lations, but are still useful for pharmaceutical companies.

The ISO 9000 currently exists in a six year old version asISO 9000:2000.22 This standard describes requirements for aquality management system in a quality managed organiza-tion.

ISO 9001:200023 describes how to develop, manufacture,and test products, and how to deal with customers andsuppliers. ISO 9001 is a certifiable standard. This means thata company, who chooses to comply with the standard, can getan accredited standards organization to assess their compli-ance, and issue a certificate to prove that they are followingthis.

ISO 9001 is easy to use for a company producing tangibleitems. Software is an intangible product, and ISO 9001 is notequally good for software. A guideline to ISO 9001 wascreated to tell how to deal with software. This was called ISO9000-3, and has existed in several issues. In its newestedition it is called ISO 90003:2004.24 The standard explains




how a software vendor shall plan, program, test, and supportsoftware. When a software company wants to be ISO 9001certified, it is normally done “under the TickIT Scheme,” i.e.,the TickIT25 Guide and ISO 90003 are followed.

A pharmaceutical company will quite often perform avendor audit on its major software suppliers, and the ISO90003 standard tells what to expect from the vendor. Publi-cations also have been written on this issue.26, 27 The TickIT

Guide is very useful as additional reading of what to expect.In principle there is no difference between what is normallydone in a pharmaceutical company and what a softwarevendor shall do: plan what to do, do what you plan, anddocument what you have done.

GAMP 4 also covers vendor audits in an appendix, but inthis author’s opinion that appendix is a bit thin. ISO 90003gives a more thorough basis for understanding what needs to

Topic US 21 CFR Part 11 EU GMP Annex 11 PIC/S PI 011-2

Electronic Signature 11.100 (a) Each electronic signature shall be N/A 20.4 It is expected that appropriate controlsUniqueness unique to one individual and shall not be will exist such as the maintenance of a

reused by or reassigned to anyone else. register of authorized users, identification11.300 (a) Maintaining the uniqueness of codes, scope of authorized actions, ineach combined identification code and support of GxP electronic records.password, such that no two individuals have 21.5 An appropriate form of Electronicthe same combination of identification code signature or authentication/identificationand password. • Should be applied where external access

can be made to a computerized GxPsystem

• The system electronically generates GxPregulatory records or key decisions andactions are able to be undertakenthrough an electronic interface andelectronic signatures.

Electronic Signature 11.200 (1) Employ at least two distinct N/A 21.8Components identification components such as an • A unique combination of user ID and

identification code and password. password called for by the computerizedsystem and linked to the user’sauthorized account for the use of aspecific application.

• Permitted task functionality for thatuser

Audit Trail 11.10.(e) (e) Use of secure, computer- 10 Consideration should be given to building 21.10 Secure, computer-generated, time-generated, time-stamped audit trails to into the system the creation of a complete stamped audit trails to independently recordindependently record the date and time of record of all entries and amendments (an GxP related actions following access to theoperator entries and actions that create, “audit trail”). system are used.modify, or delete electronic records.

Data Changes 11.10(e) Record changes shall not obscure 10 Any alteration to an entry of critical data 20.1 All original data records and masterspreviously recorded information. should be authorized and recorded with the and any subsequent alterations, additions,

reason for the change. deletions, or modifications are to be retainedaccurately and comprehensively within theretrievable audit trail.

Training 11.10(e) (i) Determination that persons who 1 Persons in responsible positions should 15.3 Records of operator trainingdevelop, maintain, or use electronic record/ have the appropriate training for the (introduction and on-going training).electronic signature systems have the management and use of systems within their 20.1 Firms will need clearly documentededucation, training, and experience to field of responsibility which utilizes policies, standard operating procedures,perform their assigned tasks. computers. This should include ensuring that validation reports, and training records

appropriate expertise is available and used to covering such system controls.provide advice on aspects of design, 21.0, 22.6 Other training coveredvalidation, installation, and operation ofcomputerized system.

Genuine E-signature 11.200(3) Be administered and executed to N/A Definitions: (a) It is uniquely linked to theensure that attempted use of an individual’s signatory. (b) It is capable of identifying theelectronic signature by anyone other than its signatory. (c) It is created using means thatgenuine owner requires collaboration of two the signatory can maintain under hisor more individuals. control. (d) It is linked to the data to which

it relates in such a manner that any changeof the data is detectable.

Risk Assessment Included in the draft guidance document N/A 15.2 For all critical systems, a holisticrisk-based approach is necessary. Thisshould consider the risks from the entirepharmaceutical application.

Risk Management N/A N/A Reference to DISC PD 3002 Guide to BS7799 Risk Assessment and RiskManagement (ISBN 0 580 29551 6).

Table A. Overview of Requirements in Part 11, Annex 11, and PI 011. (Continued from page 96.)




be included in the checklist. PDA Technical report 3228 also isuseful when planning a vendor audit.

Comparison Between the StandardsThere are no contradictions in the requirements in the threestandards 21 CFR Part 11, EU GMP Annex 11, and PI 011-2, but there is a large difference in how much they actually tellthe reader.

Annex 11 is short and concise, but very difficult to useunless one actually understands what to do.

Part 11 has been interpreted differently over the years.Some of this has a root cause in warning letters from inspec-tions; some in what an inspector has publicly said, whichimmediately has been understood as the interpretation of therequirement; some have been various interpretations wherethe pendulum has moved from one extreme to the other.

The FDA Guidance for Industry on the Scope and Applica-tion of Part 1113 has clarified the FDA position and theirinterpretation of Part 11.

The FDA is currently re-examining Part 11 and is ex-pected to issue a revised regulation for comment.

PI 011-2 is a lot bigger and includes discussions of severalof the items. Many of these discussions have references tofurther reading in other standards, and GAMP 4 is one ofthem. The application of risk-based approach is heavilyemphasized. There also is a distinction in typeface to tellwhether the text is explanatory (normal) or what the inspec-tor expects to see (italic). Even people in US companies willdo well making themselves familiar with this document, as itgives a lot more detail to what Part 11 only states as expec-tations.

DifferencesThe differences between the standards and regulations aregenerally in the words and level of details in and not in therequirements or interpretation.

Standards generally cater to one specific type of organiza-tion: The ISO 9000-series is meant for manufacturers, re-gardless of whether they make safety pins or cars; the GxPregulations are meant for the pharmaceutical development,testing, and production; Part 11 and Annex 11 has the end-user as their primary focus, but Annex 11, while addressingthe integrity of electronic records does not cover electronicsignatures. PI 011-2 also has a section on some of the vendor’sresponsibilities for creating the software in a quality environ-ment, and suggests use of a few standards for that purpose aswell as inspector’s expectations. GAMP has chapters for thesystem developers, i.e., the vendors for how to develop thesystem in a quality environment; chapters for the QA auditorso that they will know what to look for when conducting avendor audit; and chapters for the end users so they know howto validate their system.

GAMP 4 is the most detailed document. This also is theonly one that explains in detail how to fulfill requirements,instead of just stating the requirements.

Probably the most significant difference is the level ofdetail in validation descriptions. While Part 11 mentions the

word “validation” once in §11.10(a), detailed requirementsand expectations of the validation effort are described through-out the PI 011 document. Annex 11 includes a short descrip-tion, and GAMP 4’s 200+ pages are dedicated to all practicalaspects of validation.

Risk assessment is included in the FDA Guidance forIndustry on the Scope and Application of Part 11, and also isin PI 011-2. GAMP 4 includes a separate guide for risk-basedapproach to validation of computerized systems.

SimilaritiesWhen looking at the three documents Part 11, Annex 11, andPI 011-2, we can see that there are few requirement differ-ences. The words may be different, and the level of detail iscertainly different, but the content is basically the same.None of the documents have requirements that can not beread into each of the other documents, perhaps with excep-tion of electronic signatures, which is not mentioned in Annex11.

The risk-based approach has been adopted both by the USFDA and the European authorities, and is now the currentway of thinking.

Table A gives a selected overview of the various standards’coverage of some of the requirements.

ConclusionCompanies have done a lot of work during the past seven toeight years to make sure that their systems are Part 11compliant. Systems that are Part 11 compliant also are likelyto be compliant with many aspects of the EU requirementsset forth in Annex 11 and interpreted in PI 011-2.

PI 011-2 has a lot of suggestions and details that FDAregulated companies can benefit from, and they should beencouraged to examine it, even if the organization does nothave to comply with EU regulations.

GAMP 4 and its associated Good Practice Guides all covervarious aspects of validation. These various guides reflect EUand US regulatory requirements and expectations, and givegood practical advice on what needs to be done. The GAMPguides are useful tools; however, none of the guides shouldever be used directly with copy-and-paste. All organizationswork differently and you must make sure you still assess yourorganization and your way of working. In other words: think-ing is still encouraged.

If you are still unsure of how to handle the computerizedsystems, the author can recommend one of the good classesthat ISPE/GAMP are giving around the globe during any oneyear.

References1. US Medical Device Regulations: US 21 CFR 820 Quality

Systems Regulation, Medical Devices. US 21 CFR 821Medical Device Tracking Requirements.

2. EU Medical Device Regulations: Council Directive 93/42/EEC of 14 June 1993 concerning Medical Devices.

3. EU GMP Annex 11: EU Annex 11 to the EU Guidelines ofGood Manufacturing Practice for Medicinal Products.




4. PIC and PIC/S: Pharmaceutical Inspectors ConventionPIC/Pharmaceutical Inspection Cooperation Scheme PIC/S http://www.picscheme.org.

5. PIC/S GMP: Document PE 009-5, 01 August 2006 http://www.picscheme.org.

6. PI 011-2 Good Practices for Computerised Systems inRegulated “GXP” Environments 01 June 2004 http://www.picscheme.org.

7. Monograph 116: GLP Consensus Document ‘The Applica-tion of the Principles of GLP to Computerised Systems,’1995, OECD/ OCDE/GD (95) 115 (Environment Mono-graph No.116).

8. 21 CFR Part 11: US FDA Register 21 CFR Part 11 Elec-tronic Records and Electronic Signatures www.fda.gov.

9. Guidance for Industry, Part 11, Electronic Records; Elec-tronic Signatures – Scope and Application, US Dept. ofHealth and Human Services and all FDA Centers/Offices,September 2003. www.fda.gov.

10. ISO Standards: www.iso.org. ISO standards can be pur-chased from the national standards organizations, oftenin the native language in addition to English.

11. IEEE organization: www.ieee.org.12. ISPE: www.ispe.org.13. GAMP: www.ispe.org/gamp.14. GAMP® 4, Good Automated Manufacturing Practice

(GAMP®) Guide for Validation of Automated Systems,International Society for Pharmaceutical Engineering(ISPE), Fourth Edition, December 2001, www.ispe.org.

15. GAMP® Good Practice Guide: A Risk-Based Approach toCompliant Electronic Records and Signatures, Interna-tional Society for Pharmaceutical Engineering (ISPE),First Edition, April 2005, www.ispe.org.

16. GAMP® Good Practice Guide: Validation of LaboratoryComputerized Systems, International Society for Phar-maceutical Engineering (ISPE), First Edition, April 2005,www.ispe.org.

17. GAMP® Good Practice Guide: Testing of GxP Systems,International Society for Pharmaceutical Engineering(ISPE), First Edition, December 2005, www.ispe.org.

18. GAMP® Good Practice Guide: IT Infrastructure Controland Compliance, International Society for Pharmaceuti-cal Engineering (ISPE), First Edition, August 2005,www.ispe.org.

19. GAMP® Good Practice Guide: Calibration Management,International Society for Pharmaceutical Engineering(ISPE), First Edition, December 2001, www.ispe.org.

20. GAMP® Good Practice Guide: Global Information SystemsControl and Compliance, International Society for Phar-maceutical Engineering (ISPE), First Edition, October2005, www.ispe.org.

21. ISPE GAMP Forum – BMS SIG, “Position Paper: Use ofBuilding Management Systems and Environmental Moni-toring Systems in Regulated Environments,” Pharma-ceutical Engineering, September/October 2005, Vol. 25,No. 5, pp. 28-78).

22. ISO 9000:2000 Quality Management Systems – Funda-mentals and Vocabulary.

23. ISO 9001:2000 Quality Management Systems – Require-ments.

24. ISO 90003:2004 Guidelines for the Application ofISO9001:2000 to Computer Software.

25. The TickIT Guide Using ISO 9001:2000 for SoftwareQuality Management System Construction, Certifica-tion, and Continual Improvement, Issue 5.0, ISBN 0-580-36743-9, DISC TickIT Office, January 2001,www.tickit.org.

26. Segalstad, Siri, “Vendor Audits for Computer Systems:An ISO 9000-3 Approach,” Laboratory Automation andInformation Management, 32, (1996), pp. 23-31.

27. Segalstad, Siri, “Supplier Auditing and Software,” Euro-pean Pharmaceutical Review, Vol. 1 Issue 3, (1996), pp.37-44.

28. PDA Technical Report No. 32, “Auditing of SuppliersProviding Computer Products and Services for RegulatedPharmaceutical Operations,” PDA 1999, www.pda.org.

AcknowledgementI would like to express my gratitude to Siôn Wyn who madecorrections and improvements to this article.

About the AuthorSiri H. Segalstad has worked with qualityassurance/validation of IT systems since late1980s, and started her own consulting com-pany in 1995. During the past 11 years, shehas worked with a variety of pharmaceuticalcompanies with a variety of validation tasks.She is also involved in an EU project wherethe goal is to create a complete curriculum for

a Master’s degree in IT validation. Segalstad is in charge offour classes for basic and advanced Quality ManagementSystems, Validation, and LIMS. She has written a number ofpublications on LIMS and validation issues, and has givenpresentations and classes in more than 20 countries fromTaiwan in the East to California in the West. She can becontacted by e-mail at: [email protected].

Segalstad Consulting AS, PO Box 15 Kjelsas, N-0411 Oslo,Norway.


FDA Q&A on Barrier Isolation


The followingquestions andanswers wereprovided bypanelists RickFriedman andBob Sausvilleduring the ISPE15th AnnualBarrier IsolationConference,held in June2006. Theseresponses donot necessarilyrepresent thepositions orpolicies of theFDA. They aresimply thepanelists’interpretationsof these cGMPmatters and arebased on theircollectiveexperienceswith isolatortechnology.

The FDA Answers Your Questions onBarrier Isolation Technology

RABS

1Q Many companies are evaluating theRABS vs. barrier isolator decision for

use in their new facilities. Recognizing thatthere are product-specific or economic reasonsto choose one technology over the other, andwhen all other factors are equal, the issue ofregulatory acceptance becomes a major factorin the decision. Given the state of both tech-nologies and the operational characteristics ofeach, does the FDA have an opinion regardingwhich technology should be chosen? Is there apreference for barrier isolators? Will that pref-erence ever be codified?

1A The FDA would not tell a manufac-turer that they must use a specific

single technology to assure adherence to asep-tic processing requirements. The FDA doesindicate its general preference for isolators andprovides corresponding regulatory incentivesfor them. A sound RABS concept also can pro-vide added protection versus traditional pro-cessing approaches.

2Q Recent opinion suggests that a RABSline should use a VHP decontamina-

tion system if it is to be classed as an “AdvancedAseptic” Installation. In the 2005, ISPE RABSdefinition, this was not mentioned. What is theFDA position on an enterprise using a RABSinstallation in a Class 100 cleanroom (ISO5)without a VHP system? Will this be an accept-able installation and for how long - five years?10 years?

2A While the FDA cannot forecast thefuture, a well designed and controlled

RABS (automated VHP or a robust manualapplication of sporicidal agent) should exceedregulatory expectations at this time. Policymodifications generally take a long time toevolve. Also, the FDA’s typical approach is to beessentially technology neutral, as we’re a per-formance-based agency. You must adopt a de-sign that achieves reproducible and compliantsterility performance in accord with GMP –how you get there is your decision.

3Q What specific types of validation reliefcan users of advanced aseptic process-

ing expect?

3A First an isolator filling line is permit-ted to be in Class 100,000 surround-

ings, rather than the Class 10,000 used formost traditional lines. So that is a lower aircleanliness classification qualification hurdle.

During a media fill, advanced aseptic pro-cessing lines do not require the same run size ascompared to more conventional or manuallyintensive processes. In the latter case, fillingrequirements need to be close or equal to nor-mal batch size.

For isolators, significantly lower numbers ofvials in a piggyback or staggered approach cangenerally provide an adequate assessment ofmedia filling time to provide ongoing simula-tions of the campaign. While isolator line simu-lations generally simulate a lower proportion ofthe batch size versus traditional lines, initialsimulations should have more vials to estab-lish that baseline.

Media fill of all aseptic processing linesshould be done semi-annually. When doing sofor isolators, more flexible approaches to pro-cess simulation program and study design canbe considered with respect to shifts. For ex-ample, instead of running a media fill with eachseparate shift on its own, one could, whenappropriate, propose a study approach thatincludes overlapping of shifts. The rationalewould be that the environmental affects ofchallenging shifts doesn’t have much signifi-cance with use of an isolator. The people on theshifts still matter, of course. But the environ-mental effects of a shift change are normallynot as meaningful anymore (unless there areother operational activities attendant to theshift change that might potentially impact theexposed sterile product in the isolator). In con-trast, while RABS applications can represent apositive step forward, the shift-related issues ofsuch non-isolated operations are still of signifi-cance.

Reprinted from







sufficient? Are fill needles monitored?Must gloves be sampled by RODACs orswabs?

10A Basic things that should bemonitored do not change –

yes for particles, continuous air sam-pling, remote active air monitoring of-ten in at least two to three positions ina larger isolator. The FDA does not seesettle plates that much in isolators.RODAC plates should be taken at theend of the campaign (e.g., 1X/week)and we want to see the gloves moni-tored. Monitoring the filling needleswould depend on risk and level ofmanual activity in the area; we do notgenerally see this practice in isolators,but do see it rather frequently in tradi-tional filling operations. Stopper hop-pers, however, are often monitored.

11Q Preference between activeair sampling vs. settling

plates (passive).

11A Active air sampling is theFDA preference.

Glove Integrity

12Q For glove leak integritytesting, it seems we now

have data to support the fact that awell-executed visual inspection withtrained personnel is at least as effi-cient as an automated physical integ-rity test. If you accept that premise, dowe still need to do the automated test-ing for micro holes? If yes, what is therationale?

12A We do not necessarily ac-cept the premise, and we

will look for evidence that a visualinspection is well executed, using de-tailed procedures (with provisions forsupervision, as appropriate). We willexpect data from a physical integritytest. Guidance is clear that the combi-nation of visual and mechanical isneeded to provide the right level ofassurance, in tandem with microbialmonitoring of gloves.

13Q If there is a leak detectedin a glove, must all product

be rejected in the filling or simply thevials which are exposed (i.e., not stop-pered)? Can the bulk be filled after thedecon methods completed and re-ster-

4Q What are the issues, if any,the FDA sees with “open” in-

terventions in RABS processing?

4A RABS units are, by definition,not generally meant to be open

during production, a situation that addsundesirable variables that underminethe advanced processing model (suchas impact on gloves surface by person-nel, airflow dynamics).

5Q Is there a need for a material-of-construction study with dis-

infectants used for RABS inside?

5A Yes, any decontamination ordisinfection qualification

studies should be adequate to addressdifferent types of surfaces in the RABS.Any special surfaces or materials ofvariable finish need to be addressed.There is literature out there already tohelp you choose on which materialsyou want to focus your attention.

6Q Would you approve a RABS inClass ISO 8 if: 1. the doors

cannot be open during operation (re-corded during production)? 2. The vali-dation has shown good decontamina-tion and cleaning process? 3. All manualoperations are made through gloves?

6A No, the FDA is willing to go toISO 8 for isolators and well-

designed Blow-Fill-Seal operations, butwe do not believe this background en-vironment would likely be appropriatefor RABS. Firms are always welcometo bring such details of their designconcept to the FDA’s Field Office orOffice of Compliance in their relevantCenter (for purposes of discussion andfeedback).

Isolators

7Q Since the FDA is not taskedwith operator safety, regula-

tion of toxic processing issues is not intheir scope. However, due to the sig-nificant impact that safe handling re-quirements have on engineering con-trols, there is the potential for cross-over impact with aseptic processingisolators. (The most obvious issue ispositive vs. negative pressure). Addi-tionally, there is a lot of discussion onsafe handling for sterile liquids vs.powders, where some companies al-

lowing RABS or even open fill opera-tions for very hazardous products thatrequire containment in powder form.Does the FDA have a position on con-tainment of toxic products?

7A GMPs do deal with cross con-tamination issues. Although

the Aseptic Processing guidance is in-tended to address positive pressureisolators, many of the concepts alsoapply to negative pressure isolators.Generally speaking, one should be care-ful in deciding whether it is appropri-ate to process a potent compound (es-pecially those that are allergenic orparticularly difficult to clean) on thesame line as non-potent compounds.The FDA is working on detailed guid-ance on betalactam products and alsois considering guidance beyond suchproducts. The industry and the FDAare both very interested in developingguides on potent compounds – it wouldmake sense to do so. One of the con-cepts is that degree of facility separa-tion depends on your own scientificallyrigorous risk assessment.

8Q How should instantaneousnegative pressure excursions

in an isolator be evaluated? (Example:negative pressure caused by half-suitmanipulation).

8A This situation is not commonand should result in a major

investigation.

Environmental Monitoring

9Q To what extent can risk analy-sis and practical evaluation of

risk in process reduce the need formicrobiological monitoring of Isolators?

9A To a significant extent, en-vironmental monitoring will

always be needed, but less than ontraditional lines.

10Q For a compounding isola-tor (cytotoxic products –

negatively pressured), what methodsof monitoring are used during the pro-cess? Same question for aseptic fillingisolator and transfer isolators whereclassifications are Class 100 laminarflow and Class 100 turbulent respec-tively? Continuous air sampling meth-ods? Are settling plates adequate and




ilization of equipment done? Must thebulk be re-filtered before filling?

13A Release is a layered issue,and to answer, one must

get granular on what happened duringthe event. A firm may or may not reject,and then we would be glad to review acase with the individual questioner.

14Q In past ISPE Barrier Con-ferences, the practice of

aseptic changing of isolator gloves wasconsidered by some to be acceptable ifthe process was included during theexecution of media fill validations andrevalidations. Have any recent inspec-tions or audits produced observationsor concerns contrary to this acceptance?

14A We do not see aseptic chang-ing of isolator gloves used

– it was tentatively proposed during aPre-Approval Inspection (PAI) by oneuser, and then voluntarily withdrawn.The particular firm had major prob-lems demonstrating the practice dur-ing PAI. If you had a major problemwith a glove, would you want to con-tinue on? Quality by design can se-verely mitigate the impact of such anevent.

15Q Can you share how a dam-aged glove that cannot be

aseptically changed be taken out ofservice? Suggestions?

15A If you don’t stop the fill,taking a very remote glove

out of service has been attempted bysome (no details offered on how). Itwould be critical for a firm to present avery convincing risk assessment to jus-tify such a practice, based on scientificrationale.

16Q What is the Agency’s posi-tion on glove leaks in non-

filling areas – accumulators, post stop-pering etc. (assuming a minor/pinholeleak)?

16A This situation may repre-sent a lower risk if it occurs

a significant distance from the sterileexposed product and its components,especially if the glove is not used. Youneed to fully explain the issue, explainwhy the glove did not affect product orpotentially render the product non-ster-

ile. It is possible that the glove wasinvolved in a peripheral way duringthe fill, and if the campaign is notended early, it might be justifiable in asituation of extremely low risk to con-tinue to the end of a campaign. Thiswould be when container closure, orproduct are not exposed to any in-creased contamination risk due to be-ing far from the area of the problem.Then you should look at all the dataand make a well-supported, risk-baseddecision.

Decontamination

17Q Has the FDA seen any newtechnology that looks prom-

ising for the decontamination/steril-ization of barrier isolators? Or is hy-drogen peroxide vapor the best choiceat this juncture in the technologies’development?

17A Whatever works for you.There are other options, but

we have not seen anything dramati-cally new recently.

18Q Is humidity and tempera-ture mapping required for

validating a VHP sterilized isolatorwhere the product contact filling needleand stopper bowl are not VHP steril-ized?

18A Temperature mapping inand out of the isolator is

still important, controlling room tem-perature as well as knowing the dewpoint is important as well. Firm maynot need rigorous mapping of humiditylevels.

19Q What are the FDA’s expec-tations regarding treat-

ment of stopper bowls in place withVPHP?

19A There are lots of scenarios(sterilize out of place and

aseptically install, sterilize in place). Afirm should show the >6-log reductionand the FDA will evaluate it with sup-porting data.

20Q Does the Agency expect thecrimping station to be VHP

sterilized in an isolator system thatVHP sterilizes the filling and stopper-ing station? (The crimping station is

separated by a mousehole to the fillingand stoppering section).

20A No to crimping station, butthe FDA recommends to

routinely perform manual sporicidaldisinfection, not necessarily VHP. Sepa-ration between crimping station andmain isolator’s egress mousehole is im-portant.

21Q How often do you recom-mend revalidating barriers

with BIs?

21A “Knowledge is power” –while the FDA does not

specify periodicity, it is important tohave the data. Certainly, perform a BIchallenge if there has been a change inthe isolator to warrant concern.

22Q Are penetration studiesnecessary during continu-

ing (on-going) validation studies? Areinitial validation penetration studiessufficient?

22A To answer this question,we need clarity on whether

this is penetration through media bagsor into product vials.

Regarding media bags or plates,contract management needs attention.Your supplier agreement needs to tellyou about penetration of chemical agentto the load. It may be prudent to dorepeat testing every few years to peri-odically verify. Also you’re only as goodas the ability of your vendor to avoidunanticipated change (i.e., your agree-ments should include provisions forcontractors to inform you proactively).

23Q Do you require VHP con-centration (ppm) as a pa-

rameter for an acceptable decontami-nation cycle? How do you recommendthe limits be set during validation stud-ies or from process data? Do you wantto see a minimum level only or a maxi-mum level as well?

23A The FDA wants to see chemi-cal concentration measure-

ment during validation, and theystrongly recommend you measure con-centration during routine decontami-nation. This is analogous to other ster-ilization processes to assure that youhave enough concentration to provide




cremental improvement that can beprovided by a good RABS design ap-proach, warrants use of advanced trans-fer technologies when possible.

27Q What trends do you see tomanage risk for decontami-

nating/sterilizing incoming material(e.g., syringe tubs) coming in to barrierisolators? How do you view effective-ness of different methods (manual wipedown, VHP, UV, and electron beam)?

27A VHP has been a commondecontamination method

for syringe tubs. UV is considered mar-ginal. Bioburden on syringe tub is im-portant, and is affected by how it hasbeen handled, where the exterior bagopened, whether and how long it isexposed to the cleanroom environment– all these factors affect the predictedbioburden. Ebeam is mentioned, and ismore often used in Europe to sterilizethe outside of the tubs, but the FDAdoes not have much experience withebeam in this application.

General

28Q Is the aseptic processingguidance being rewritten?

What is the expected issuance date?Will there be any changes in the gloveintegrity test section?

28A The FDA aseptic process-ing guidance was issued in

September 2004, and has been verywell received as it is flexible in itsdiscussion of design and encouragesmodern approaches! The Agency canget any related issues out on the FDA’sQ&A Web site (formerly Human DrugcGMP Notes). Further explication ofregulatory relief for media fills in isola-tors may be one area.

29Q What is your recommenda-tion of room classification

for Sterility Isolator Testing?

29A This is your choice. Usu-ally, we see controlled un-

classified background in a clean, or-derly room. A firm would likely excludeblaming the background environmentshould a positive result occur (as thethought is that the background envi-ronment is not of consequence).

surface decontamination. The genera-tor tells you what initially exits thegenerator, but not what is inside theisolator. You should measure in at leastone isolator location. To demonstratethe need, one company had a hole inthe delivery tubing and did not detectthe situation (i.e., reduced hydrogenperoxide was making it into the isola-tor) for a month. It was eventuallydiscovered during re-qualification. Per-oxide level and limits should be estab-lished during validation. The NIR mea-surement systems provide pretty goodaccuracy, and even if they might be tosome extent “precisely inaccurate,” thedata still reveals meaningful fluctua-tions or abnormally low concentrationlevels.

Qualification

24Q What is the Agency’s posi-tion on the use of vendor

testing/Factory Acceptance testingrather than traditional qualificationapproaches?

24A Be flexible, one can’t useFAT for OQ/PQ, but can be

leveraged for IQ.

Transfer Systems

25Q If a transfer isolator is un-docked, powered down, and

moved to a clean unload area to loadsterilized materials that are in con-tainers, does the transfer isolator re-quire battery power to keep the HEPAsystem on?

25A Possibly, but we really needmore information to an-

swer. Does turning the HEPA systemoff and on hurt the situation? If vialsare sealed and chamber stays positive,then that’s major risk mitigation so itcould be feasible depending on the ap-plication.

26Q Is the introduction of ma-terials (non-product con-

tact such as sterilized parts bags) intoa RABS of the same interest as into anisolator?

26A Yes it is, material intro-duction into isolators would

seem to be of similar significance asmaterial introduction into RABS. Ad-vanced approaches, including the in-

30Q For blow fill seal machines,is a Grade ‘C’ (Class 10,000)

environment a mandatory require-ment?

30A Aseptic guidance stateswith proper design Class

100,000 surrounding classification canbe justified. Older designs need Class10,000.

About the PanelistsRichard L. Friedman, M.S., is Direc-tor, Division of Manufacturing andProduct Quality (HFD-320), Center forDrug Evaluation and Research/Officeof Compliance. He can be reached bytelephone at 301-827-9036 or by e-mailat [email protected].

Robert Sausville is Director, Divi-sion of Case Management, OCBQ,CBER, FDA. He can be reached bytelephone at 301-827-6201 or by e-mailat [email protected].

Global Regulatory News


©Cop

yrig

ht IS

PE 2

007

InternationalThe Global Harmonization Task Force(GHTF)1 has issued draft guidelinesfor comments on:

• Regulatory Auditing of QualityManagement Systems of MedicalDevice Manufacturers –Part 3: Regulatory Audit Reports

• Role of Standards in the Assess-ment of Medical Devices (revised)

Australia/ New ZealandAdded to the Therapeutic GoodsAdministration (TGA) Web site2 inDecember 2006/January 2007 was:

• EU guidelines adopted now includeGuideline on Similar Biological Me-dicinal Products containing Biotech-nology-Derived Proteins as ActiveSubstance: Quality Issues (EMEA/CHMP/BWP/49348/2005). Thisguideline addresses the require-ments regarding manufacturingprocesses, the comparability exer-cise for quality, considering thechoice of reference product, analyti-cal methods, physicochemical char-acterization, biological activity, pu-rity, and specifications of the simi-lar biological medicinal product.

• TGA has published3 a new versionof its guidelines relating to GoodManufacturing Practice (GMP)clearance for therapeutic goodsmanufactured outside of Australia.The guidelines state that the TGAwill no longer automatically acceptGMP certification from Pharmaceu-tical Inspection Cooperation Scheme(PIC/S) member countries, unlessthey are also Mutual RecognitionAgreement (MRA) partners withAustralia.

In January 2007, the Australia NewZealand Therapeutic Goods Au-thority (ANZTPA)4 published on itsWeb site a ‘Questions and Answers’page on GMP procedures under themerged authority.

CanadaHealth Canada5 has issued guidance

on information to be provided to manu-facturers for the reprocessing and ster-ilization of reusable medical devices.This draft guidance document is in-tended to assist manufacturers in un-derstanding and complying with theregulatory requirements of section21(1)(i) of the Medical Devices Regula-tions as they pertain to the directionsfor use for reusable medical devices.

EuropeThe European Council and Parlia-ment6 have approved Reach, the pro-posed chemicals regulation on the reg-istration, evaluation, and authoriza-tion of chemicals. The regulation willenter into force progressively from June2007, and it is estimated that the reg-istration process will take 11 years tocomplete.

Reported on the Web site for theEuropean Medicines Agency(EMEA)7 in December 2006 and Janu-ary 2007 were:

• The European Medicines Agency haslaunched a new public database de-signed to facilitate access to infor-mation about medicines availablein the European Union. The data-base can be accessed atwww.eudrapharm.eu.

• Updated Scientific Data Require-ments for the Plasma Master File(PMF) (Effective as of 1 June 2007)

• Updated ICH Topic Q3A (R2) - Notefor Guidance on Impurities Testing:Impurities in New Drug Substances

The Committee for Medicinal Prod-ucts for Human Use (CHMP)monthly report8 from the DecemberPlenary meeting held 11 to14 Decem-ber 2006.

Documents prepared by theBiologics Working Party shown belowwere adopted at the December meet-ing:

• Overview of comments on guidelineon the Scientific DATA require-ments for a Plasma Master File(EMEA/CHMP/427732/2006).

• Draft Guideline on EnvironmentalRisk Assessments for MedicinalProducts Consisting of or Contain-ing Genetically Modified Organisms(GMOs) (EMEA/CHMP/BWP/473191/2006).

• Overview of comments received onDraft Guideline on EnvironmentalRisk Assessments for MedicinalProducts Consisting of or Contain-ing Genetically Modified Organisms(GMOs) (EMEA/CHMP/BWP/480303/2006).

• Concept paper on revision of theGuideline on plasma-derived me-dicinal products (EMEA/CHMP/BWP/495530/2006).

The Heads of Agencies9 Web site hasbeen updated with reports from theCMD(h) meetings held 13 to15 Novem-ber 2006 and 11 to 12 December 2006.

An updated Question and Answerdocument on Mutual Recognition Pro-cedures following the January 2007EU enlargement was made availableon the Web site at the beginning ofJanuary.

The European Commission DGEnterprise10 announces a revision toGMP Annex 3 “Manufacture ofRadiopharmaceuticals.” This draft re-vision for public consultation is pro-posed in the light of new GMP require-ments for active substances used asstarting materials. It specifies applica-tion of Part II for the manufacture ofradiopharmaceuticals. Comments arerequested by 30 March 2007.

The European Directorate forthe Quality of Medicines (EDQM)11

is now the European Directoratefor the Quality of Medicines andHealthcare (EDQM andHealthcare). Its Web site5 (updatedDecember 2006) advertises the avail-ability of European PharmacopoeiaSupplement 5.8, style guide, and struc-ture nomenclature guide.

IrelandIMB have published on its Web site12

in December 2006, a summary of theresponses to their public consultationinto dual labelling of parallel imported





Global Regulatory News


©Copyright IS

PE 2007

products. Some individual issues areaddressed in the form of a Q&A ses-sion. At the same time, they announcedthat downloads from its InformationDay 2006 – Manufacturing sessionsare now available from the same Website. Subjects include implications ofnew manufacturing legislation, sterilemanufacture, reference and retentionsamples, engineering inspections, de-viations, process quality review, andEMEA reflection document, Storage ofMedicinal Products, Market Compli-ance, and Upcoming Regulatory Com-pliance Inspections.

SwedenA new Swedish rule6 on activities re-lated to the handling of blood and bloodcomponents entered into force on 30October 2006. The new provision imple-ments all current European Unionregulations on blood and blood compo-nents that are used as raw materials inthe manufacture of medicinal prod-ucts. It describes in detail how thegathering, control, and manufacturingof blood and blood components shouldbe performed.

References1. http://www.ghtf.org/2. TGA - http://www.tga.gov.au/me-

dia/index.htm3. RAJ Pharma ,Vol. 17, No. 12, De-

cember 2006.4. ANZTPA - http://www.anztpa.org/

index.htm5. TPD - http://www.hc-sc.gc.ca/dhp-

mps/prodpharma/index_e.html6. RAJ Pharma, Vol. 18, No. 1, Janu-

ary 2007.7. EMEA - http://www.emea.eu.int/

whatsnewp.htm8. EMEA - http://www.emea.eu.int/

PressOffice/presshome.htm9. HOA - http://heads.medagencies.

org/10. EC - http://ec.europa.eu/enterprise/

pharmaceuticals/index_en.htm11. EDQM - http://www.pheur.org/12. IMB - http://www.imb.ie/

This information was provided by Pe-ter Hagger, Pharmaceutical ResearchAssociates (UK).


New Products and Literature


Water Treatment

Christ Water Technology Group hasavailable their newest innovationcalled the Hydrokat. This catalytic ex-haust-gas converter protects electro-chemical processes from possible dam-age resulting from the explosion of gasmixtures. Its heart is a catalytic com-bustion element with automatic oxy-gen supply and integrated tempera-ture regulation, making it capable ofconverting hydrogen-oxygen mixtureswith any relative concentrations of thetwo gases.

Christ Water Technology Group,www.christwater.com.

Process Monitoring Catalog

Millipore Corp. has available its newproduct catalog highlighting tools andservices to meet your process monitor-ing needs. The 92-page color catalogue(CA1002EN00) showcases Millipore’seffective and reliable products that testfor liquid and airborne contaminants,including systems, media, methods,validation protocols, and rapid detec-tion tools for time sensitive applica-tions.

Millipore Corp., www.millipore.com.

Glove-Port

Extract Technology recently launchedPharmaPort(PatentPending), a contamina-tion-free glove-port interface for thecompany’s renowned isolator rangecurrently in widespread use through-out the pharmaceutical and asepticprocessing industries. This new designwill help to eradicate contaminationhang-up around the operator accessglove/gauntlet and the glazing/windowpanel, while improving operator safetyby way of mechanically clamping theglove to the port.

Extract Technology Ltd., www. ex-tract-technology.com.

Dust Collector

A full line of HemiPleatTM retrofit car-tridges from Farr Air Pollution Controlmay be used to upgrade performanceor solve problems of existing dust col-lector systems. The key to enhancedperformance is a patent-pending pleat-ing technology that opens up the pleatsuniformly for more effective cleaningand better airflow. This filtration up-grade can greatly extend service lifeand reduce pressure drop compared tocompetitive dust collector cartridge fil-ters.

Farr Air Pollution Control, www.farrapc.com.

Tubing

Now available from NewAge Indus-tries are two types of polyethylene tub-ing: linear low density formula and astyle co-extruded with Ethylene VinylAcetate (EVA). While both are madefrom non-toxic ingredients conformingto FDA standards, they offer differentperformance characteristics. Uses in-clude air lines, chemical and fluid trans-fer, food and beverage processing anddistribution, pharmaceutical process-ing, pneumatics and instrumentation,potable water, deionized water, wirejacketing, laboratory applications, anddecorative coverings.

NewAge Industries, www.newageindustries.com.

Temperature Sensor

Weed Instrument has a new, novel-design sensor for temperature mea-surement in sanitary applications. Thesensor (Weed Instrument Model#3142B) is especially beneficial wherewipers or mixers can interfere with aninserted probe or thermowell. The sen-sor, attached inside a welding spud viaa CIP sanitary clamp, is easily removedwithout need to disturb the mountinghardware, thus simplifying cleaning,replacing, and calibrating operations.

Weed Instrument Co., www.weedinstrument.com.

Reprinted from







Containment Valve System

L. B. Bohle LLC has a new contain-ment valve suitable for operation inareas where Operator Exposure Lev-els are limited and high containmentoperations are essential. The compactnew valve is designed in two sections,active and passive, and is air operatedusing only a single actuator. Centeringbolts ensure proper alignment whilevacuum sealing and a connection tothe dust collection system guaranteescomplete containment and increasedoperator safety.

L. B. Bohle LLC, www.lbbohle.com.

Peristaltic Pump

Precision dispensing of shear sensitivepharmaceuticals, biopharmaceuticals,and cell products by Flexicon peristal-tic pumps, available from FlexiconAmerica Inc., assures laboratories andmanufacturers of gentle handling,closed fluid path sterility, precisiondispensing accuracy and flexibility, and

multiple plate/flask formats and per-form cell maintenance, colony selec-tion, and RNAi studies.

Thermo Fisher Scientific Inc., www.thermofisher.com.

Peristaltic Pumps

Watson-Marlow Bredel will showcaseits biopharmaceutical processing solu-tions, which include the 520, 620, and720 Series of peristaltic pumps atINTERPHEX2007 from 24 to 26 April2007 at the Jacob K. Javits ConventionCenter in New York, New York, USA.The 520/620/720 family of peristalticpumps are designed for the accuratemetering, dosing, and transferring ofsensitive fluids in sanitary environ-ments, making them ideal for accuratefiltration, fermentation, dispensing,coating, and seamless integration intoreusable or disposable of bioprocessapplications.

Watson-Marlow Bredel, www.watson-marlow.com.

Air SamplerCoriolis® µ by Bertin Technologies isan air sampler that captures biologicalparticles such as bacteria, fungi, andpollen. Its patented cyclonic technol-ogy delivers a liquid sample compat-ible with all analyses, includingimmuno-assay, PCR, flow cytometry,and microbiology. Coriolis® µ is a solu-tion to collect a large panel of micro-organisms (particle size >0.5µm) witha collection time or air flow rate up to300L/min.

Bertin Technologies, www.bertin.fr.

Plant Asset ManagementHoneywell has an enhanced version ofits Field Device Manager, a key compo-nent of the company’s plant asset man-agement portfolio. Field Device Man-ager R301 is the first system to supportboth the latest Electronic Device De-scription Language and Field Device

speed in meeting production economies.The closed fluid path that character-izes the peristaltic pumping processassures that sterile product nevercomes in direct contact with any mov-ing parts before being dispensed.

Flexicon America Inc., www.flexiconamerica.com.

Software forAsset Information

Emerson Process Management hasavailable Version 2.5 of its AMS™ Suite:Asset Portal™ software, which expandsasset management capabilities. AMSAsset Portal Version 2.5 allows users tocustomize enterprise-wide asset infor-mation, including filtering and report-ing alerts, polling on demand, and view-ing graphical asset health reports. AMSAsset Portal is a Web-based tool thatenables maintenance management per-sonnel to obtain timely information toquickly identify critical equipment thatis not performing and predict unex-pected failures or off-spec product intime to take corrective action.

Emerson Process Management,www.emersonprocess.com.

Automation Platformfor Cell Culture

Thermo Fisher Scientific Inc. has avail-able its new Thermo Scientific CellGrowth and Discovery (CGD)WorkCellTM, a fully enclosed, environ-mentally controlled automation solu-tion designed for high capacity cellgrowth, supply, and in-line imageanalysis. Combining state-of–the-artsoftware with sophisticated robotics,the CGD WorkCell is a turn-key sys-tem which can simultaneously handle




Tool with Device Type Manager tech-nologies without the use of conversiontools or add-on devices. This providescustomers the broadest choice of in-strumentation and valves to integratewith their distributed control systemsand asset management systems to im-prove plant uptime and reduce mainte-nance costs.

Honeywell International, www.honeywell.com.

Visual Supervisors

Eurotherm has recently launched thenew Eycon series of visual supervisors.These visual supervisors provide multi-function control, recording, and visual-ization. Eurotherm has a long historyof experience in the control, data ac-quisition, and process automation mar-ket and now the Eycon Series bringsthat expertise into a single processunit.

Eurotherm, www.eurotherm.com.

Software for Submission ofMedical Device Reports

Sparta Systems, Inc. has released itsBeta version of the TrackWise eMDRSubmission ManagerTM software. Theadd-on software enables medical de-vice companies to electronically sub-mit Medical Device Reports (MDRs) tothe US FDA. TrackWise eMDR Sub-mission ManagerTM enables device com-panies to effectively manage complaintsand investigations, assess potentialadverse events for safety risk prob-lems, and comply with health author-ity reporting requirements, includingthose of the US FDA.

Sparta Systems Inc., www.sparta-systems.com.

Vessel Outlet Valves

Fluid Transfer’s Sanitary Flush-Bot-tom “Fluid Flow” Valves are uniquelydesigned to be welded flush with thebottom of vessels, thus eliminating anydead space between the valve and thevessel, so all product can be completelyprocessed. These valves come in sizesfrom 1-1/2" to 4" and can be automatedwith either pneumatic or electric ac-tuators. The maximum working pres-sure is 450 PSIG at temperatures to300º F.

Lee Industries, Inc., www.leeind.com.

Temperature Sensors

TURCK introduces a line of highly re-liable and easy to program tempera-ture sensors. The TS400 and TS500temperature sensors incorporate amultitude of design features that makethe sensors suitable for nearly all fac-tory and process automation applica-tions. TS400 and TS500 temperaturesensors are platinum resistance tem-

perature detectors, a technology com-monly referred to as Pt-100.

TURCK, www.turck.com.

Pressure/Flow Transmitter

A FOUNDATION Fieldbus version ofthe Yokogawa EJX910A multivariablepressure/flow transmitter, along witha flow configuration tool using FDT/DTM technology based software, is nowavailable from Yokogawa Europe. Thepatented DPharp single-crystal siliconresonant sensor technology ensureshigh accuracy, provides superior over-pressure protection, minimizes the ef-fect of temperature and static pressurechanges, and provides a unique multi-sensing capability.

Yokogawa Electric Corp., www.yokogawa.com.

To submit materialfor publication in

Pharmaceutical Engineering'sNew Products and Literature

department, e-mailpress releases with photos to

[email protected] forconsideration.


Industry and People


To submit materialfor publication in

Pharmaceutical Engineering'sIndustry and Peopledepartment, e-mail

press releases with photos [email protected] for

consideration.

Turner EmployeesEarn LEED Certification,

Chance to Win Car

Joseph Schilens, a project superinten-dent in Turner Construction’s Cleve-land office, was one of 250 Turner em-ployees to become Leadership in En-ergy and Environmental Design(LEED) certified in 2006. All Turneremployees who were LEED accreditedas of 31 December 2006 were enteredinto a lottery to receive the use of a2007 Toyota Prius hybrid as their com-pany-provided vehicle for three years.Joseph Schilens was selected as thewinner of the 2007 Toyota Prius.

Turner Construction, www.turnerconstruction.com.

Korsch and ThomasEngineering Announce

PartnershipKorsch AG and Thomas EngineeringInc. (TEI) have formed a strategic glo-bal partnership that will enable theglobal distribution of leading edge tech-nology for coating pans, tablet presses,and press tooling, while permittingtheir multinational customers to stan-dardize on critical process equipmentacross global sites. TEI will work jointlywith Korsch America to promote, sell,and support Korsch equipment in NorthAmerica. Thomas Engineering will nolonger sell new Manesty Tablet Presses;

however, all existing Manesty custom-ers will continue to be fully supportedwith regard to technical service andspare parts.

Korsch AG, www.korschamerica.com.

Thomas Engineering Inc., www.thomaseng.com.

Nicomac Partners withICOS Impianti Spa

Nicomac Inc. has partnered with ICOSImpianti Spa for marketing and servic-ing the entire ICOS line of products forthe pharmaceutical industry. In addi-tion to well established products suchas autoclaves, dry heat sterilizers, andstopper processors, Nicomac will pro-mote and provide after sales servicesin the US, Canada, and Puerto Rico forICOS products.

Nicomac Inc., www.nicomac.com.Icos Impianti Spa, www.icosimpianti.

com.

Rockwell AutomationAcquires ProsCon Holdings

Rockwell Automation Inc. has acquiredProsCon Holdings Ltd., a privately heldengineering firm offering proven andtechnically unique design solutions tothe process industry. Areas of exper-tise include process technology, con-trol systems, and information technol-ogy. ProsCon also provides modularsolutions as an innovative and cost-effective approach delivering fasterimplementation of new facilities, aswell as retrofits for existing plants.

Rockwell Automation Inc., www.rockwellautomation.com.

ProsCon Holdings Ltd., www.proscon.ie.

Sartorius Combines itsBiotech Division with

StedimSartorius AG has signed a bindingagreement with Stedim Biosystems SAand its founders thereby acquiring thecontrol of Stedim. Sartorius will com-bine its Biotechnology Division withStedim to create a globally leadingtechnology provider for the biopharma-

ceutical industry. The combined com-pany, to be named “Sartorius StedimBiotech SA” will be listed on the Parisstock exchange. The founders and ma-jority shareholders of Stedim supportthe transaction and will stay investedin the combined company.

Sartorius AG, www.sartorius.com.Stedim Biosystems S.A., www.

stedim.com.

Werum Software andSystems Receives Award

Werum Software and Systems is therecipient of the 2007 Frost and SullivanCompany of the Year Award in theManufacturing Execution Systems cat-egory for the pharmaceutical andbiopharmaceutical industries. Eachyear, Frost and Sullivan presents thisaward to the company that has demon-strated excellence in business develop-ment, competitive strategy, consistentgrowth, and leadership. The award alsorecognizes Werum’s continuing inno-vation and its commitment to customersatisfaction in the pharmaceutical in-dustry.

Werum Software and Systems, www.werum-america.com.

Reprinted from





ISPE Update


Shin (second from left) with Campbell students at the 2006ISPE Student Leadership Forum.

ISPE Student Chapter Profile: Campbell Universityby Daniel Shin, PhD, Faculty Advisor, ISPE Student Chapter, Campbell University,and Martin Rock, President, ISPE Carolina-South Atlantic (CASA) Chapter

Editor’s Note: One of ISPE’s goals is todevelop students into competent phar-maceutical professionals and encour-age them to pursue careers in theindustry. ISPE Student Chapters playan important role in this endeavor. Thisarticle is the first in a series profilingISPE Student Chapters and the people,education, research, and activities oftomorrow’s pharmaceutical profession-als.

Getting Started

In December 1995, Jane Brown, thenfirst chair of the Student Affairs Com-

mittee of the ISPE CASA Chapter, con-tacted Mark Yates, PhD, then a facultymember at Campbell University inNorth Carolina, USA. Brown wantedto know whether Yates was interestedin starting a Student Chapter atCampbell and serving as faculty advi-sor.

Yates was involved in initiating theBachelor of Science program in phar-maceutical sciences within the Schoolof Pharmacy at Campbell. But, a yearlater, he accepted the challenge of start-ing the ISPE Student Chapter atCampbell University. The first Stu-dent Chapter kickoff meeting atCampbell University was held in thespring of 1996 with seven students.

Brown currently serves as Chair-man of the International Board of Di-rectors of ISPE and is still actively

involved with the CASA ChapterStudent Development Committee.Yates, currently employed withWyeth Pharmaceuticals, is also stillinvolved with the CASA ChapterStudent Development Committee.

Growing the ChapterThe ISPE Student Chapter atCampbell was the first Student Chap-ter to send students to an ISPE An-nual Meeting (New Orleans – 1998).So far, four Campbell students havewon the local student poster compe-tition at the CASA Chapter level, allof whom participated in the interna-tional poster competition at ISPEAnnual Meetings.

To date, two international posterchampions were members of theCampbell Student Chapter. EricBlaesing won the undergraduate posteraward in 2003 and Wendy Haines, PhD,while a graduate student at UNC –Chapel Hill, won the graduate levelposter award in 2001. Haines nowserves as the Chair of the CASA Stu-dent Affairs Committee.

After Yates left Campbell Univer-sity in 2000, the activities of the Stu-dent Chapter slowed down due to thelack of a faculty advisor. Daniel Shin,PhD, joined the Department of Phar-maceutical Sciences at Campbell Uni-versity in the fall of 2001, and a yearlater, agreed to serve as a faculty advi-

sor to the StudentChapter.Shin made a plan to

recruit new studentmembers and orga-nize the student lead-ership. He has con-tacted some profes-sors within and out-side of the Depart-ment to get a few min-utes of their class timeto introduce ISPE.

Yates, who becamethe Student Chapter’s

industry advisor, and many other mem-bers in the ISPE CASA Chapter lead-ership, helped recruit featured speak-ers for the Student Chapter’s monthlymeetings. Many local leaders gave pre-sentations at those meetings.

The local CASA members were veryhelpful and enthusiastic about comingto Campbell and speaking to the stu-dents. Since Campbell is located in arural area, these guest speakers sacri-ficed their precious time and gas moneyto drive out to the campus in the eveningto help the students understand whatpharmaceutical scientists and profes-sionals do in the real world.

There was no explosive growth ofmembership at Campbell; rather it wasa gradual steady increase. The ISPEstudent membership at Campbell hasgrown from seven students in 1996 to60 active students in 2006. Most ofthem are majors in pharmaceutical sci-ences, but there are some pharmacystudents, clinical research, and otherscience majors. Last year, the StudentChapter at Campbell University wasthe recipient of the first ISPE StudentChapter of the Year Award.

Let’s MeetNetworking opportunities are one ofthe great benefits of being an ISPE

Campbell University students get a firm grip onthe dos and don’ts of networking from BoCrouse-Feuerhelm, Past Chairman of the ISPEStudent Development Committee.

Reprinted from




Concludes on page 2.


ISPE Update


ISPE Student Chapter Profile: Campbell UniversityContinued from page 1.

Student Member. There is always awarm and friendly atmosphere at vari-ous ISPE functions and meetings, in-cluding conferences, seminars, tech-nology shows, career fairs, and studentleadership forum meetings. Shin en-couraged students to attend these func-tions since they were sure to learnmore about the industry and careeropportunities through exposure atthese meetings.

Virtually all of the students whoattended ISPE events said that theywere excellent because they were greatopportunities to see a part of the realworld and interact with industry pro-fessionals. The students are interestedin attending as many of these activi-ties as possible even though they areunder severe time constraints pursu-ing their academic goals.

Most of the meetings are open to thestudents at reduced prices or even freeof charge. The financial support pro-vided by the ISPE CASA Chapter tothe Student Chapters is also helpful tothe students. Yearly monetary sup-port, reduced annual fees, and freeadmission to conferences, seminars,technology shows, etc., all help to en-courage student participation.

Another benefit to students is thatISPE Members, both on a local andnational level, are proactive in helpingthe students prepare and enhance their

resumes and their interview-ing skills. Seminar sessions areheld on internships, getting ajob, poster competitions, etc.Many of the graduates fromthese programs have landedjobs through ISPE at variousindustrial companies, includ-ing Biogen Idec, Novo Nordisk,Talecris, GSK, Eli Lilly, Wyeth,D&Z, CRB, Monsanto,Magellan/Cardinal, EISAI,AAI Pharma, and others.

The ISPE DifferenceThere are many other organi-zations available to studentsin the pharmaceutical field.

But, ISPE is the only organization thatsupports students with this depth andbreadth of industry interaction. Stu-dent initiatives are truly a wise invest-ment by ISPE and by the professionsince the students represent our fu-ture.

Many students who have benefitedfrom ISPE membership as students,and have graduated from their univer-sities, now serve in various voluntarycapacities for ISPE. These studentsare testimony to the on-going mutualbenefits of the student programs forboth students and ISPE.

The spirit of volunteerism and dedi-cation among the ISPE industry mem-bers is priceless and contagious. Theatmosphere created by this infectiousspirit of goodwill among the membersis very inviting to the students andmakes everyone feel welcome. The stu-dents learn from the devotion and ser-vice of the industry members, and manyof these students go on to reciprocatewhen they are in industry positions.

Now, there is a very positive momen-tum at work at Campbell University.The Campbell students who joined ISPEclearly see the opportunities providedby ISPE membership. Each studentmember then becomes an advocate forthe ISPE organization. Through wordof mouth promotion among the stu-dents themselves, many other students

A student gets information about ISPE and itsStudent Chapter during a High School Career Fairheld at Campbell University.

have joined ISPE. As result of this posi-tive momentum, it takes much less ef-fort now to sustain the student chapter.

For more information on the researchStudent Members are involved in at

Campbell, visit www.ispe.org.

April Paris ConferenceHighlights Nano andMicro Technology, andMore

ISPE will hold its Paris Conference 16to 19 April at the Hilton Hotel with

industry leaders presenting eight semi-nars on:

• Revision of the GAMP® Good Prac-tice Guide: Validation of ProcessControl Systems

• Nano and Micro Technology forPharmaceutical Products and Pro-cesses

• Process Analytical Technology(PAT)

• Pharmaceutical Water, Regulation,and Innovation

• Facility of the Year Exposé

• Biosafety

• Design Space

• Project Management – FacingToday’s Challenges

For more informationand to register,

please go to www.ispe.org/ParisConference.


ISPE Update


Washington Conference Highlights Facilities Summit and FDA Collaboration

Technical Documents and 16th Barrier Isolation ForumTake Spotlight

ISPE will feature four exciting highlights in addition tomany other opportunities for pharmaceutical manufac-turing professionals at the ISPE Washington Conference

to be held 4 to 7 June 2007 at the Crystal Gateway Marriottin Arlington, Virginia, USA.

Facilities SummitAs the international expert on pharmaceutical facilities,ISPE will offer presentations and innovative case studiesfrom leaders in the field. Panel discussions will includeFacility of the Year 2007 Category Winners with virtualfacility tours.

This multi-day, multi-track program, to take place 4 to 5June, will include content in three key areas, Project Deliv-ery, Regulatory, and Manufacturing Technology/Operations.Spotlighting case studies and state-of-the-art facilities, in-teractive discussions will focus on practical solutions tofacility design (new or renovated), construction, buildinggreen, and qualification for operational excellence.

The Summit also will feature the most up-to-date informa-tion on the Facility of the Year Awards, bringing the engi-neers and architects who designed those state of the artfacilities, together to share their insights with participants.

FDA-ISPE CollaborationISPE and the US FDA are co-sponsoring a series of first-everinteractive seminars designed to allow delegates to impacttheir own futures by participating in the development of howICH Q8, Q9, and ultimately Q10 guidelines will be imple-mented.

These sessions will focus on Product Quality LifecycleImplementation (PQLI), and will be the foundational meet-ings for an open, continuous dialogue between industrystakeholders and regulators, ultimately helping to generatea pragmatic approach to implementing Q8, Q9, and Q10regarding Quality by Design and Quality Systems.

Prominent US regulatory representatives will co-host thisground-breaking event, which also will comprise six break-out sessions for working groups to comment on and captureindustry input.

New ISPE Technical DocumentsThe “Ready for Release,” ISPE Good Prac-tice Guide: Commissioning and Qualifi-cation of Pharmaceutical Water andSteam Systems seminar will be held 4 to5 June.

This two-day seminar will explore thesuccesses and failures of design, installa-tion, qualification, continued operation,

and maintenance of qualified water and steam systemsthrough a series of regulatory updates and case studies.Participants will learn about the relationship between qual-ity impact of the utilities and the business risk associatedwith their operation. Impact classification and release high-lights of the ISPE Good Practice Guide: Commissioning andQualification of Pharmaceutical Water and Steam Systemsalso will be discussed.

The “Ready for Release,” Bulk Pharmaceutical Chemicals(BPC) Baseline® Guide, Review by Developers seminar willbe held 4 to 5 June. The BPC Guide, originally published inJune 1996, has been undergoing a complete revision. Thissession will allow an opportunity to interact with the authorsand review the completed Guide revision with the executivecommittee members. Participants will hear from experts inthe field to gain a real “hands-on” experience in the applica-tion of the Guide. With assistance from subject matter ex-perts from the API Community of Practice, key changes andhot topics of the revised Guide will be presented and dis-cussed.

Gold Standard in Barrier IsolationISPE will host the 16th Annual Barrier Isolation TechnologyForum – the longest running Barrier Isolation TechnologyForum in the world. ISPE’s Barrier Isolation TechnologyForum is the standard by which all others are measured, andcontinuously builds upon the foundation of knowledge andbest practices set in place during previous years, providing avital opportunity to gain updates and examine new casestudies.

This program will feature the latest developments inBarrier Isolation Technology. It will include backgroundinformation, technology updates, a series of new case studies,agency presentations, and industry comments. Offering aglobal perspective with speakers from Europe and the US,this seminar will present state-of-the-art advancements foruse in developing and manufacturing pharmaceuticals utiliz-ing Barrier Isolator Technology.

Participants will gain insight into updated technologiesapplicable to advanced aseptic processing using Barrier Iso-lation; learn “what not to do” from those who’ve done it;participate in peer discussion groups that will answer yourown questions on Barrier Isolation Technology issues; andidentify regulatory agency perspectives that will streamlineyour regulatory submission/approval process.

For more information about, or to register for, theWashington Conference, please visit www.ispe.org/

washingtonconference.


ISPE Update


April 20075 Central Canada Chapter, Student Poster Competition, Montreal, Quebec, Canada5 - 6 Japan Affiliate, Annual Meeting, Tower Hal, Tokyo, Japan10 San Diego Chapter, New Member Breakfast, San Diego, California, USA11 New Jersey Chapter, Dual Track Sessions, Session Topics: Error Reduction—the Human Factors and

Cleaning Validation, Holiday Inn, Somerset, New Jersey, USA12 Greater Los Angeles Chapter, Watson Laboratories Tour/Training, Corona, California, USA12 South Central Chapter, Education, Exhibits, and PPG Industries Plant Tour, Houston, Texas, USA13 Central Canada Chapter, Networking Event, Toronto, Ontario, Canada13 Czech Republic/Slovakia Affiliate, Risk Analysis Workshop13 Delaware Valley Chapter, Habitat for Humanity, Pennsylvania, USA14 Puerto Rico Chapter, Student Leadership Forum, Puerto Rico16 Delaware Valley Chapter, Meeting and Student Poster Judging, held in conjunction with the ISPE

Philadelphia Training Series, Philadelphia, Pennsylvania, USA16-19 Philadelphia Classroom Training, Hilton Philadelphia, Philadelphia, Pennsylvania, USA16-19 Paris Conference, Hilton Hotel, Paris, France17 Boston Area Chapter, Seminar on Contract Manufacturing, USA17 San Francisco/Bay Area Chapter, Commuter Conference, Primary Systems Panel, USA18 Nordic Affiliate, Event: How to Minimize Cleaning Validation, Copenhagen, Denmark19 Ireland Affiliate, Training Seminar (half-day) and Gala Dinner, Dublin, Ireland19 Puerto Rico Chapter, Biotechnology Track, Puerto Rico24 - 25 Poland Affiliate, Conference on Innovations in the Pharmaceutical Industry, Starogard Gdanski, Poland24 - 26 INTERPHEX New York, Javits Convention Center, New York, New York, USA26 New Jersey Chapter, Student Poster Competition, Hoboken, New Jersey, USA26 San Diego Chapter, Dinner Meeting, La Jolla, California, USA30 DACH Affiliate, Process Analytical Technology COP Meeting, Frankfurt, Germany

May 20072 Midwest Chapter, Extended Education and Vendor Day with Bayer Plant Tour, Sheraton Overland Park Hotel,

Overland Park, Kansas, USA3 Central Canada Chapter, Education and Networking, Quebec City, Quebec, Canada5 Puerto Rico Chapter, Annual Golf Tournament, Puerto Rico7 Carolina-South Atlantic Chapter, Bausch & Lomb Facility Tour, Fuquay Varina, North Carolina, USA7 Delaware Valley Chapter, Golf Outing, Philmont Country Club, Philadelphia, Pennsylvania, USA7 - 8 ISPE Singapore Training Course - Cleaning Validation, Singapore7 - 10 Barcelona Classroom Training, Renaissance Barcelona Hotel, Barcelona, Spain9 New Jersey Chapter, Golf Outing, Farmstead Country Club, Lafayette, New Jersey, USA10 Italy Affiliate, Colleretto Giacosa, Biotechnology Manufacturing, Turin, Italy10 Nordic Affiliate, Clean Utility - Purified Water and WFI, Stockholm, Sweden10 - 11 DACH Affiliate, Workshop, Theme: "Neubau Feststofffabrik/Prozessanlagen/Diagnostika/Medizinprodukte,"

Graz, Austria11 Ireland Affiliate, Abbott Plant Tour and Golf Outing, Sligo, Ireland15 Boston Area Chapter, Seminar on Product Contact Materials, USA15 Central Canada Chapter, Toronto Breakfast Seminar, Toronto, Ontario, Canada17 Central Canada Chapter, Montreal Breakfast Seminar, Montreal, Quebec, Canada17 New Jersey Chapter, Cardinal Health Tour, Somerset, New Jersey, USA17 San Francisco/Bay Area Chapter, Dinner Meeting, California, USA21 - 25 ISPE and Society of Bioprocessing Professionals (SBP), 5th Annual Bioprocessing Institute, Hyatt Regency

at Penn's Landing, Philadelphia, Pennsylvania, USA23 Argentina Affiliate, Masterly Conferences, cGMP for the 21st Century FDA Initiatives, Buenos Aires,

Argentina24 Belgium Affiliate, Technical Meeting on Disposables Technology, Brussels, Belgium24 Puerto Rico Chapter, Project Management Program, Puerto Rico24 San Diego Chapter, Dinner Meeting, La Jolla, California, USA

Dates and Topics are subject to change

Mark Your Calendar with these ISPE Events


ISPE Update


ISPE, SBP Host 5th Annual Bioprocessing Institute21 to 25 May

ISPE and the Society for BioprocessingProfessionals (SBP) will host the 5th

Annual Bioprocessing Institute, ahighly-regarded conference that willfeature both introductory and advancedcourses, with the opportunity to learnfrom and share insights with innova-tive biotechnology professionals. Theconference will be held 21 to 25 May2007 at the Hyatt Regency at Penn’sLanding, Philadelphia, Pennsylvania,USA.

This conference will include ninecourses with 18 workshops that willprovide delegates with an immediatereal life learning experience, includingexhibits and plant tours.

Courses include:

• Bioreactor Systems: Develop-ment, Design, and OperationThis course will help participantsgain an understanding of the devel-opment, selection, design, valida-tion, operation, and regulatory com-pliance of large scale bioreactor sys-tems used for production of thera-peutic proteins and related prod-ucts.

• Bioprocess Development andScale-up for Fermentation andCell CultureIn this advanced course, workshopteams will get involved in the deci-sion-making process that’s requiredto move fermentation and cell cul-ture toward manufacturing for acommercial bioproduct.

• Separation Technologies forBioprocessingThis course will cover the separa-tion technologies for downstreambioprocessing, including basic prin-ciples; cleaning methods; and prob-lem-solving workshops focusing onthe design of a protein purificationprocess.

• Bioprocess Development forDownstream PurificationThis advanced workshop will focus

on the points and tools to considerwhen developing a downstream pu-rification process. Participants canimprove their understanding of thebasic biochemical process technol-ogy typically used in purificationprocesses for proteins from recom-binant hosts.

• Scale-up of Bioprocessing Sys-tems for PurificationIn this advanced workshop, partici-pants will acquire an understand-ing of approaches and tools used inthe scale-up of chromatography andtangential flow filtration systemsincluding pumping, piping, supportequipment such as bubble traps,and instrumentation.

• Application and Design ofBioprocessing EquipmentParticipants will gain an under-standing of the principles and com-ponent design details of thebioprocess equipment necessary forlate stage phase III clinical trialsfor an FDA-approved drug.

• Cleaning Technologies forBioprocessing SystemsThis course will cover cleaning tech-nologies including Clean-in-Place(CIP). Participants will review theprinciples and practices of the ap-plication of CIP technology tobioprocess systems including con-siderations for compliance and vali-dation.

• Designing Facilities for Success-ful Bioprocessing OperationsThis course will focus on the con-cepts of designing multi-purpose,multi-cellular cGMP biologics pro-cessing facilities as they are becom-ing more complex and challengingas new cell lines, global harmoniza-tion, and processing technologies areadvancing.

• Bioprocessing OverviewA series of non-workshop sessionsfocusing on:- Bioreactor Systems: Develop-

ment, Design, and Operations

- Separation Technologies forBioprocessing Systems

- Application and Design ofBioprocessing Equipment

- Cleaning Technologies forBioprocessing Systems

For more details and on-lineregistration, visit: http://

www.bioprocessingprofessionals.org/Institute_InstituteCourses.htm

Philadelphia Training16 to 19 April

ISPE Philadelphia Training will beheld 16 to 19 April 2007, Hilton Phila-

delphia City Avenue, Philadelphia,Pennsylvania, USA. Courses taughtby leading professionals in their fieldsinclude:

• HVAC for Pharmaceutical Facilities

• Cleaning Validation Principles

• Auditing for GMP

• Drug Manufacturing Facility Design

• Complying with Part 11 – RiskManagement

• GMP Fundamentals for the Phar-maceutical Industry

• Basic Principles of Commissioningand Qualification

• Clinical Trial Materials

All courses will provide valuable infor-mation to take back and use in yourjob.

For more details on the courses andinstructors, visit www.ispe.org/

philadelphiatraining.


ISPE Update


INTERPHEX2007: What to Expect and What's Newexhibitors within the pharmaceuti-cal manufacturing industry

• the announcement and celebrationof outstanding achievements in fa-cility design and construction at thethird annual Facility of the YearAwards

• networking with peers from acrossthe country and around the world

• free daily keynotes by industrythought leaders and innovators

• technology pavilions focused on au-tomation and packaging

• exclusive ISPE Member discountson event registration fee – ISPEMembers receive 20 percent off con-ference registration fees

• complimentary passes to the Ex-hibit Hall

• exclusive ISPE Member Loungewith complimentary continentalbreakfast, beverages, Internet ac-cess, and small meeting rooms

NEW Life Sciences Job FairProduced in partnership with ISPE andAAPSThe Life Sciences Job Fair will provideopportunities to meet representativesfrom a broad range of companies whohave employment and career opportu-nities, including exhibitors and othersuppliers, pharmaceutical and biotechcompanies, industry recruiters, andmore.

ISPE Member Lounge andLife Sciences Job Fair Hours:

Tuesday, 24 April, 9 to 5Wednesday, 25 April 9 to 5

Thursday 26 April 9 to 3

FOYA Finalists: Find Out atINTERPHEX

This year, Facility of the Year Awards(FOYA) Category Winners will benamed at INTERPHEX2007.

Each Category Winner will be avail-able to discuss their facility operationas well as offer virtual tours of theirnew facility during exhibit hours.

Award winners for each categoryand the overall Facility of the YearAward winner will receive high profileattention and media coverage fromISPE, INTERPHEX, and Pharmaceu-tical Engineering magazine including:

• all Category Winners receive crys-tal awards

• Facility of the Year Awards compe-tition winner receives the presti-gious crystal and marble award

• worldwide distribution of press re-leases to global media outlets de-tailing facilities of the CategoryWinners and Facility of the YearAwards winner

• recognition via announcements dur-ing keynote sessions and special dis-plays at INTERPHEX and ISPE’s2007 Annual Meeting

Keep pace with the rapidly evolv-ing world of pharmaceutical andbiotech manufacturing at

INTERPHEX2007, where you can ac-cess the newest products from morethan 1,000 exhibiting companies, learnabout advanced solutions and innova-tive business practices, get expert per-spectives from industry movers andshakers, and network with thousandsof industry peers from across the coun-try and around the world.INTERPHEX2007 will be held 24 to 26April at the Jacob K. Javits ConventionCenter in New York, New York, USA.

Show Hours:Tuesday, 24 April, 10 to 5

Wednesday, 25 April 10 to 5Thursday 26 April 10 to 3

Conference Hours:Tuesday, 24 April, 9 to 4

Wednesday, 25 April 9 to 4Thursday 26 April 9 to 3

Sponsored by ISPE, this event is thelargest and most distinguished indus-try event taking place worldwide. Seewhat you have in store:

• a comprehensive conference pro-gram offering views and opinions ofindustry leaders, as well as strate-gic and technical applications ontopics focusing on IT, Biopharma-ceutical, Facilities, Manufacturing,Outsourcing, Supply Chain and Se-curity, Contamination Control, andPilot Plants

• new and innovative products andservice trends from more than 1,000

ISPE Member Lounge SponsorsPremier Tec-Ultra Clean Magnetic Mixer • Veolia Water Technologies

• Commissioning Agents •AWS Bio-Pharma Technologies • Burkert Fluid Control Systems

• Compliant Logistics •


ISPE Update


INTERPHEX2007: What to Expect and What's NewContinued.

• invitation to attend INTERPHEXLeadership Dinner with ISPE,INTERPHEX, and PharmaceuticalProcessing magazine leadership inNew York

• media advisory sessions featuringCategory Winners and Facility ofthe Year Awards competition win-ner during INTERPHEX and ISPE’sAnnual Meeting

• coverage in two Special Editions ofPharmaceutical Engineering maga-zine

• cover story in Pharmaceutical Engi-neering magazine and feature ar-ticle in Pharmaceutical Processingmagazine

Special Room Blockat INTERPHEX2007for ISPE Members

At INTERPHEX2007, you can connectwith other ISPE Members before andafter the show! ISPE has a specialroom block for ISPE Members attend-ing INTERPHEX2007 at the SheratonNew York Hotel and Towers (811 7thAvenue).

The ISPE Member rate is $259single/double. You can make your res-ervations through www.interphex.comand click to “travel/hotel.”

To receive your discounted room rate:

1. Click on the travel desk homepage

2. Select “ISPE” from the “I am a”menu

3. Use pass code ISPE2007

ISPE Containment COPFirst General Meeting

at INTERPHEXISPE’s Containment Communityof Practice (COP) is organizingits first general meeting on 24April from 5:00 to 6:30 pm to beheld in conjunction with theINTERPHEX Show at the JavitsConvention Center. The programwill include ice breaking activi-ties, a short presentation, net-working opportunities, and in-formation about upcoming COPevents. Light appetizers and bev-erages will be served. The eventis open to all Containment COPmembers and will be free for allISPE Members. Non-ISPE Mem-bers will have to pay a nominal$30 registration fee.

Here’s a brief overview of all that’s newand exciting at this year’s show:

Two Visionary Keynotes

What’s Next: Where Devicesand Medicine Go From Here...Tuesday, 24 April at 11:30 am

Bill Cook, Chair and CEO of the CookGroup Inc., will survey the challengesand potential presented by one of thelife science industry’s fastest-growingsegments— the convergence of drugand device technologies. With hisunique perspective of growing deviceand biotech companies, Mr. Cook ex-plores the partnership strategies, in-novative approaches to R&D, invest-ment, and regulatory issues needed tocreate winning combination productsthat will transform healthcare.

Innovation, Integration, andExploration: The Future of LifeScience IndustriesWednesday, 25 April at 11:30 am

G. Steven Burrill, CEO of Burrill andCo, has championed the growth and

prosperity of the biotechnology indus-try and life sciences industries for 40years. One of the industry’s originalarchitects, he continues to be an indus-try leading expert and visionary. Build-ing on this history, he will presentstatistics, predictions and visions forthe future. Mr. Burrill will explore howscientific advances, technological con-vergence and expanding global mar-kets will continue to transform the lifesciences industries and open up newfrontiers for personalization andcommericialization.

NEW Co-Located Event!Admission is FREE with your

INTERPHEX badge.

PharmaMedDevice 2007Conference and ExhibitionPharmaMedDeviceTM is the first eventto fully illuminate the convergence ofthe medical device, pharmaceutical,and biologic industries. This transfor-mational event addresses the needs ofthe emerging combination productmarket and the exciting innovationstaking place in drug delivery technol-ogy and healthcare today. PharmaMed-Device provides a dynamic platformfor education, partnering, sourcing, anddiscovery of innovative new products –and provides a unique opportunity forcross-sector collaboration across theseindustries.

ISPE and the Journal of Pharma-ceutical Innovation (JPI) are officialSponsors of PharmaMedDevice 2007.

BIO TO BUSINESS –Bridging People, Opportunities,

and TechnologiesIt’s a dynamic time for the biotechnol-ogy industry, and discoveries are oc-curring at a rapid pace. If you’re asmall- to medium-sized biotech com-pany interested in scale-up production,INTERPHEX is a valuable resourcefor partnerships, networking, educa-tion, products, technologies, and ser-vices in both small and large moleculeprocessing and production.

Concludes on page 8.


ISPE Update


NEW INTERPHEX/BioExecutiveInternational RoundtablePresented by INTERPHEX andBioExecutive International magazineWednesday, April 25, 8:00 am to 11:30am – Invitation only

In the ever changing business environ-ment, there is a growing necessity tostrategically align, partner, and workwith other entities to ensure the suc-cess of your future business. Key bio-pharmaceutical and pharmaceuticalindustry executives will discuss, de-bate, and provide insight on how theircompanies see the future in light of thecurrent business environment.

NEW Bio/Techfunding ForumFunding for Early Stage LifeSciences CompaniesPresented by Life Sciences Greenhouseof Central PAThursday, 26 April, 10:00 am to 11:30am

What kinds of funding are available tostartup companies, when is each typeappropriate, and how does one get con-nected to sources? What does each typeof investor look for? What should astartup look for in an investor (i.e.,what makes a good partner)? A panel ofindustry insiders, including represen-tatives from federal funding sources,angel investors, venture capitalists,entrepreneurs and industry technol-ogy councils will explore the issues andprovide practical answers.

NEW EducationalOpportunities

With nearly 100 cutting-edge sessionsled by respected industry experts, theINTERPHEX2007 Conference is un-surpassed for up-to-the-minute, inten-sive education geared toward problemsolving and productivity enhancement.New additions to this year’s conferenceprogram include:

NEW Management ConferenceTrackThese results-oriented sessions aredesigned to help you develop the criti-cal management skills to stimulateand sustain innovation, maximize pro-ductivity, reduce waste and achievemanufacturing excellence. Whetheryou’re looking to execute lean manu-facturing techniques, achieve six sigmaquality or create a culture of continu-ous improvement, you’ll find the toolsand strategies you need.

A STATE-OF-THE-ARTBiotechnology TrackSessions covering many of the mostchallenging issues in commercial scalebioprocessing, including disposablesand single use technologies, prefilledsyringes, cell rupture unit operation,direct oxygen injection in aerobic fer-menters, and more serve the educa-tional needs of the growing biotechnol-ogy audience at INTERPHEX.

NEW Supply Chain andSecurity TrackThis new track provides a strong foun-dation in the fundamentals of RFIDtechnology and its application to prod-uct authentification and supply chainsecurity. Topics include 21st Centurysupply chain models, cold chain, sup-ply chain risk, network critical infra-structure for RFID, removing risk fromthe pharmaceutical supply chain, trans-action life cycle management and sup-ply chain compliance, real world itemlevel tagging and many others.

NEW RFID Master ClassPresented by the International RFIDBusiness Association in collaborationwith the RFID Technical InstituteMonday, 23 April, 8:00 am to 4:00 pm

Leveraging RFID, Sensor, andWireless Technologies forManufacturing ProcessImprovement and SupplyChain Efficiency

This concentrated, thought provokingclass, created specifically forINTERPHEX2007 attendees, providesa solid foundation in the latest devel-opments in RFID technology and theirpractical application to the pharma-ceutical manufacturing and the supplychain.

NEW ISA CAP Review CourseThree days, 23 to 25 April, 8:00 am to4:00 pm / CEU credits: 2.1

This course reviews the Certified Au-tomation Professional Job AnalysisDomains, Tasks, Knowledge and SkillAreas, and Technical Topic areas de-veloped as the basis for the CAP ex-amination. It is designed specificallyfor experienced automation profes-sionals preparing to take the exam.An explanation of the examinationprocess, and sample test-taking ques-tions are provided. Separate registra-tion required.

NEW Package DesignShowcaseSponsored by Package Design magazine

See award winning package de-signs for pharmaceutical, cosmeticand medical device products that havebeen selected by respected industryorganizations, and explore new de-sign possibilities in the Package De-sign conference track. Sessions illus-trate opportunities for pharmaceuti-cal, cosmetic and medical device com-panies to strengthen their brandsthrough effective package design andinnovation.

INTERPHEX2007: What to Expect and What's NewContinued from page 7.


ISPE Update


Singapore Conference 10 to 12 June: Driving RegionalPharmaceutical Manufacturing Excellence

Held annually since 2000, the ISPESingapore Conference offers a

world-class educational and network-ing opportunity for pharmaceuticalmanufacturing professionals in Asia.

With a panel of 25 internationaland regional speakers, this is the onlyconference in Asia that addresses thelatest regulatory, technological, andpractical issues facing multinationaland regional pharmaceutical manu-facturers in API, secondary, and biotechmanufacturing.

Key features of the 2007 ISPESingapore Conference are:

• Six industry tracks and one SpecialFocus Track on Manufacturing Ex-cellence tailored to address specificpharmaceutical manufacturing issues

• Two pre-Conference workshops pro-viding in-depth knowledge of se-lected processes and procedures

• Pharma Nite – an opportunity tonetwork and build relationshipswith the international and regionalspeakers and delegates to the Con-ference

• Educational plant tours – a uniqueopportunity to visit some of themost advanced pharmaceutical

manufacturing plants in Asia. Reg-istered delegates can choose to visitany one of the following pharma-ceutical manufacturing plants inSingapore on a space-available ba-sis – Pfizer; Novartis; Schering-Plough; Merck, Sharp and Dohmeand GlaxoSmithKline

• Facility of the Year Award CategoryWinners – this Award recognizesstate-of-the-art pharmaceuticalmanufacturing projects that use newand innovative technologies to de-liver a quality product while reduc-ing the manufacturing cost. Thewinner will be selected and an-nounced at the ISPE Annual Meet-ing in November 2007. See the de-tails of the best pharmaceuticalmanufacturing facilities fromaround the world

• Tradeshow – an opportunity to see awide range of equipment, products,and services and to source for newsuppliers

• Student Poster Competition andCareer Fair – an excellent opportu-nity for students to present theirposter and win a grand prize of anopportunity to present their postersat the ISPE Annual Meeting 2007

For more information on the Conference Programvisit http://www.ispesingaporeconference.com


Classified Advertising


DID YOU KNOW...?More than 56% of

Subscribers ForwardPharmaceutical Engineering to

One or More PeopleSource: 2005 Publications Survey

For advertising opportunities, callISPE Director of Sales, Dave Hall at

+1-813-960-2105.

Architects, Engineers – Constructors

CH2M Hill, PO Box 22508, Denver, CO80222, www.ch2mhill.com. See our adin this issue.

CRB Consulting Engineers, 7410 N.W.Tiffany Springs Pkwy., Suite 100, KansasCity, MO 64153. (816) 880-9800. See ourad in this issue.

IPS – Integrated Project Services, 2001Joshua Rd., Lafayette Hill, PA 19444.(610) 828-4090. See our ad in this issue.

Parsons, 150 Federal St., Boston, MA02110. (617)-946-9400. See our ad inthis issue.

Stantec Consulting, 201 Old Country Rd.,Suite 301, Melville, NY 11747. (631)424-8600. See our ad in this issue.

Cleanroom Products/Services

AES Clean Technology, 422 Stump Rd.,Montgomeryville, PA 18936. (215) 393-6810. See our ad in this issue.

Employment Opportunities

Fabricating Distributors Wanted forAdvantaPure’s Sanitary Tubing, Hose,and Assemblies. Contact [email protected] for AvailableTerritories.

Employment Search Firms

Jim Crumpley & Associates, 1200 E.Woodhurst Dr., Bldg. B-400, Springfield,MO 65804. (417) 882-7555. See our ad inthis issue.

Filtration Products

Millipore Corp., 290 Concord Rd.,Billerica, MA 01822. (800) MILLIPORE.See our ad in this issue.

Siemens Water Technologies, 125Rattlesnake Hill Rd., Andover, MA01810. (978) 470-1179. See our ad in thisissue.

Hoses/Tubing

AdvantaPure, 145 James Way,Southampton, PA 18966. (215) 526-2151.See our ad in this issue.

Label Removal Equipment

Hurst Corp., Box 737, Devon, PA 19333.(610) 687-2404. See our ad in this issue.

Passivation andContract Cleaning Services

Active Chemical Corp., 4520 Old LincolnHwy., Oakford, PA 19053. (215) 676-1111. See our ad in this issue.

Astro Pak Corp., 3187 Redhill Ave., Suite105, Costa Mesa, CA 92626. (800) 743-5444. See our ad in this issue.

Passivation andContract Cleaning Services (cont.)

Cal-Chem Corp., 2102 Merced Ave., SouthEl Monte, CA 91733. (800) 444-6786.See our ad in this issue.

Oakley Specialized Services, Inc., 50Hampton St., Metuchen, NJ 08840. (732)549-8757. See our ad in this issue.

Pumps

Watson-Marlow Bredel, 220 BallardvaleSt., Wilmington, MA 01887. (978) 658-6168. See our ad in this issue.

Sterile Products Manufacturing

Tanks/Vessels

Lee Industries, PO Box 688 Philipsburg,PA 16866. (814) 342-0470. See our ad inthis issue.

Used Machinery

Validation Services

ProPharma Group, 10975 Benson Dr.,Suite 330, Overland Park, KS 66210;5235 Westview Dr., Suite 100, Frederick,MD 21703. (888) 242-0559. See our ad inthis issue.

Water Treatment

Christ Pharma & Life Science AG,Haupstrasse 192, 4147 Aesch,Switzerland. +41 617558111. See our adin this issue.

Reprinted from


