Process Analytical Technology for assessment and control ...pqri.org/wp-content/uploads/2015/08/pdf/hammond.pdf · The sample size calculation assumed the following: • Acquisition

Process Analytical Technology for

assessment and control of API

content uniformity

Available Technologies

Steve Hammond

Pfizer Global Manufacturing

Presentation Outline• Standard At-line application

• Batch blender system

• Continuous tablet core measurements

• Continuous Processing PAT systems

• Blend in a transfer chute

• Future RTR measurement strategy

• Potential for feed frame measurements

• Conclusions

1030 1040 1050 1060 1070 1080 1090 1100 1110 1120 1130 11401/0.0001

-3.1854

-2.6869

-2.1884

-1.6899

-1.1915

-0.6930

-0.1945

0.3040

0.8025

1.3010

1.7995

5mg: Loadings

Wavelength/nm

Inte

nsit

y X

10^4

10mg Calibration Regression Line

y = 0.9915x + 0.8313

R2 = 0.9915

50.00

60.00

70.00

80.00

90.00

100.00

110.00

120.00

130.00

140.00

50.00 70.00 90.00 110.00 130.00 150.00

%Intent by HPLC

%In

ten

t b

y N

IR

3

NIR for At-line CU Determination• Conventional lab-based NIR system

– Uses discriminating spectral signature of API

– Validated examples over range 0.25 – 75%

– Can be compressed blend or tabletsBlack Active

Red placebo

Aromatic C-H

NIR_Intact/MS/24Nov08 4

Capsule NIR Transmission SystemSource

Detector

NIR_Intact/MS/24Nov08 5

PLS Regression - API

912 nm 1156 nm

3 Factors selected

(94% of X variance

explained)

Very little

unrelated

spectral

variance within

the data set

1018 nm

1350 nm

API & Excipients

Dispensing Blend BlendSievingDispensing Blend BlendSieving Granulation

Mag

Stearate

Mag

Stearate

PAT1

PAT1

PAT2

PAT2

BlendTablettingCoating PAT3

PAT3

PAT4

PAT4

PAT2

PAT2

PAT2

PAT2

PAT:

1 = Identification

2 = Blend Uniformity

3 = Weight, Hardness, Disintegration, Potency, Content Uniformity

4 = Water Determination

Product “A” Real Time Release and PAT

NIR Mounted on a rotating blender

NIR in enclosure

Rotation point

-1

0

1

0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180

tPS

[1]

$Time (normalized)

Full_Set_SNV.M1:Step_1, PS-Full_Set_SNV_NEW_Jan

Predicted Scores [Comp. 1]

+3 Std.Dev

t[1] (Avg)

-3 Std.Dev

tPS[1] (Batch 810005500)

tPS[1] (Batch 810011300)

SIMCA-P+ 11.5 - 21/01/2008 12:44:22

-1.4

-1.2

-1.0

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180

t[1]

$Time (normalized)

Full_Set_SNV.M1:Step_1

Scores [Comp. 1] (Aligned)

+3 Std.Dev

t[1] (Avg)

-3 Std.Dev

SIMCA-P+ 11.5 - 18/12/2007 10:54:09

-3

-2

-1

0

1

-1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

t[1]

$Time (normalized)

Full_Set_SNV.M2:Step_2

Scores [Comp. 1] (Aligned)

+3 Std.Dev

t[1] (Avg)

-3 Std.Dev

SIMCA-P+ 11.5 - 18/12/2007 11:43:42

-0.2

-0.1

-0.0

0.1

0 10 20 30 40 50 60 70 80 90 100 110 120

t[1]

$Time (normalized)

Full_Set_SNV.M3:Step_3

Scores [Comp. 1] (Aligned)

+3 Std.Dev

t[1] (Avg)

-3 Std.Dev

SIMCA-P+ 11.5 - 18/12/2007 11:06:43

-5

-4

-3

-2

-1

0

1

0 10 20 30

t[1]

$Time (normalized)

Full_Set_SNV.M4:Step_4

Scores [Comp. 1] (Aligned)

+3 Std.Dev

t[1] (Avg)

-3 Std.Dev

SIMCA-P+ 11.5 - 18/12/2007 11:11:14

8

Control Charts

NIR interfaced with Press

10X Sampling Unit

NIR Process Analyzer

Report

MES-Systems

Process Analyzers• Measure condition of the process material in real time

• Collect more information about the batch

6.305.955.605.254.904.554.203.85

4000

3000

2000

1000

0

Potency, mg/g

Nu

mb

er

of

Un

its

85%75% 115% 125%

Mean 4.958

StDev 0.1058

N 17547

Normal

Histogram of Potency (mg/g) - Nominal 5 mg/g

Results from 18,000 tablets for a 5mg dose in a 200mg tablet

5.165.105.044.984.924.86

Median

Mean

5.0255.0205.0155.0105.0055.000

1st Q uartile 4.9670

Median 5.0089

3rd Q uartile 5.0554

Maximum 5.1910

5.0047 5.0234

4.9971 5.0198

0.0597 0.0730

A -Squared 0.45

P-V alue 0.268

Mean 5.0141

StDev 0.0657

V ariance 0.0043

Skewness 0.247446

Kurtosis -0.004509

N 192

Minimum 4.8372

A nderson-Darling Normality Test

95% C onfidence Interv al for Mean

95% C onfidence Interv al for Median

95% C onfidence Interv al for StDev

95% Confidence Intervals

Summary

Position of mean and median

106104102100989694

160

140

120

100

80

60

40

20

0

% Nominal

Fre

qu

en

cy

Mean 99.59

StDev 1.657

N 1150

Histogram % Nominal (52 Batches)

Distribution for 1140 capsules from 52 lots

Next Generation Manufacturing

Initiative in Pfizer

• An extension of Quality by Design to achieve

–Continuous manufacturing of a drug product

–Control Strategy - use Process Analytical Technology

(PAT) and Design Space to demonstrate control

–Process validation by Continuous Quality Verification

(CQV)

– Introduction of Real Time Release (RTR)

• Immediate release solid oral dosage form (capsule)

Schematic of Continuous Process

Encapsulation

PAT Measurement Point: Potency

PAT Measurement : Particle Size Distribution

Raw Material Feeders

Mean residence time 200 secs

Turbine speed 50rpm

Retained mass 2.5Kg

Throughput 50Kg/h

Primary mixing

Lubrication mixing

PAT Measurement Point: Potency

Process parameters determining

potency and CU• In a typical DP process:

Primary Blend Potency (CU) = weighing + mixing

• In a batch process weighing and mixing are a one time event.

• In a continuous process, weighing and mixing operation is repeated many

ten’s of thousands of times, the operation is time variant in nature.

• For a continuous process, the material flow can be considered a stream of

“mico batches”, each having a discrete potency and CU.

• Sampling of a continuous process should take into account the rate at which

“mico batches” are flowing from the system.

• Sampling should be time variant, and match material throughput

Continuous Dry Granulation

PAT Systems – Probe Interface

Proprietary PAT Probe Interfaces have been developed to minimize the impact of the dynamic powder flow and blend cohesive properties on the measurement capability

Final Blend Probes Interface

for NIR & FBRMFBRM Probe Interface after

Roller Compactor

Blend Uniformity Measurement by NIR

• Two measurement points:

– Initial blend

–Final blend

oMeasurement is in real time

• Measurement is under dynamic flow conditions

o A probe based NIR spectrometer is used

o A sampling interface is constructed

o Critical is an understanding of measurement

characteristics

o Contributing mass to a spectrum, scan speed.

Sample Size Calculation

The sample size that ePAT611 sees depends on

• Probe spot size

• Penetration Depth of light

• Powder Density

Probe

Powder Bed

Sample that probe sees

The sample size calculation assumed the following:• Acquisition time is 20ms, each spectrum is the average of 50 scans

• Probe spot diameter is 2mm,

• Depth of penetration is 0.5mm,

• Powder density is 0.7mg/mm3

Under static condition, the amount each scan sees can be estimated by:

M = p (d/2)2DH

where d is the probe spot diameter, D is the powder density, and H is the penetration depth

M = 1.1mg/scan

Scanning for 1 second and adding the spectra -The sample mass for each spectrum is approximately :

55mg

Calibration Model Building

• Calibration mixtures were prepared below and above

the nominal concentration

• Pre-blended material was fed through the continuous

process

• Spectra were taken approximately every second

• Spectra collected were divided into Calibration set and

Test set

• Multivariate Statistical techniques were used to develop

calibrations.

• Two confirmation runs were used to evaluate the model

performance

NIR Calibration Model

Predicted Potency

Ac

tua

l P

ote

ncy

Sample Number

Actual Potency

NIR Predicted

Potency

Predicted vs. Actual Potencies for Calibration dataset

Po

ten

cy +/- 12% Nominal

Nominal

NIR Calibration Model

Predicted Potency

Ac

tua

l P

ote

ncy

Sample Number

Actual Potency

Predicted Potency

Predicted vs. Actual Potencies for Internal validation dataset

Po

ten

cy

+/- 12% Nominal

Model PerformanceNIR predicted Potencies for two confirmation runs

targeting at nominal potency

Nominal Potency

Nominal Potency

- 10%

Nominal Potency

+ 10%The Average

Capsule Potency by

HPLC = 99.3%LC

Standard platform for RTR

• RTR has significant benefits to Pfizer

• To supply chain management

• Reduced testing in a laboratory

• Increased process understanding

• Whilst also increasing the level of Quality

Assurance of the product

• Require an easy to implement, low cost,

standard platform for the application of PAT

Schematic of Feed Frame PAT

Installation

Right Paddle wheel

direction of rotation

Empty

Tablet DiesDirection of Tablet Die

Movement

Left Paddle Wheel

direction of rotation

Powder Inlet

Chute

Full

Tablet

Dies

Example Probe InstallationsManesty Unipress Diamond (USA) and Fette 2090 (Germany)

SentroProce-PAT 611

The design and engineering of the probe interface is still on-going,

but expected to be complete by the end of 2011

26

Technology AdvancementsQuantitative Method Development – Norvasc

Example

85

90

95

100

105

110

115

0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100 2200 2300 2400 2500

Num

Observed vs. Prediticted Line Plot, e-PAT 611 YPred[2](%_LC)

YVar(%_LC)

SIMCA-P+ 12 - 2010-05-02 19:02:36 (UTC-5)

Norvasc blends premixed at 85%, 93%, 100%, 107%, and 115% LC

Norvasc is 3.5% API and a good

test case for measurement capability+ /- 15% of nominal

nominal

Potential Benefits

• Increased process understanding of blending and

compression processes.

• Understand and monitor feed-frame function

• Ability to detect segregation during powder transfer from

IBC to the tablet press

• Applicable to both Batch and Continuous Processes

• Integration of PAT signal and tablet press weight control

signal into compression machine logic.

– Advanced Process Control

• Opportunity to implement as part of RTRt paradigm

RTR Approach for Continuous

Processes

• ICH Q8 (R1) - “Unit dose uniformity performed

in-process (e.g., using weight variation coupled

with near infrared (NIR) assay) can enable real

time release testing and provide an increased

level of quality assurance compared to the

traditional end-product testing using

compendial content uniformity standards”

Question?

• With the development of fast continuous

measurement systems in the feed frame, is a

CU measurement still a relevant statistic?

• If yes how do you deal with 10,000’s of

results?

• Is it a “discrete” measurement approach

• Or a statistic designed for a continuum of

data?

Summary

• Conventional At-line NIR transition measurements

now common – tablets and capsules

• On-line tablet core analysis is providing good

capability data

• PAT is important for Continuous Manufacturing

• NIR methods have been developed for real time

monitoring of Blend Uniformities in a Continuous

Dry Granulation process.

• NIR capability is close for feed-frame measurement

of CU, and may become the standard approach

• The predicted NIR results will be used for Process

Control; Continuous Quality Verification and Real

Time Release of the final product.

Acknowledgement of the whole army

Acknowledgements

• Many, many people have contributed to this

presentation:

• Pfizer World Wide R&D: Groton Pharm Sciences

• Pfizer Global Supply: Process Analytical Support

Group

• Pfizer Global Supply: Technology, Science and

Operation Group

• Pfizer Global Supply: Caguas Site, Illertissen, Freiburg

• Pfizer Global Supply: Product and Process

Development, Freiburg