Glucose sensors: toward closed loop insulin delivery Chee W. Chia, MD, Christopher D. Saudek, MD * Division of Endocrinology and Metabolism, Department of Medicine, The Johns Hopkins University School of Medicine, 600 North Wolfe Street, Osler 575, Baltimore, MD 21287, USA Although the symptoms of severe hyperglycemia have been known for millennia, the laboratory measurement of blood glucose was a lengthy and labor-intensive process until 1915, when Lewis and Benedict [1] were credited with simplifying the analysis to a point that glycemia could be documented relatively quickly. This advance is considered a crucial step in the chain of events that led up to the isolation of insulin by Banting and Best in 1921 [2]. Glucose measurements remained laboratory-based until the 1940s, however, when self-assessment of glucosuria became feasible by measuring total ‘‘reducing substance’’ in the urine at home. It took sophisticated physical chemistry to fix glucose oxidase to a reagent strip with a development of color proportional to glucose in a drop of blood, initiating the era of self-monitoring of blood glucose (SMBG) with reflectance meters in the 1970s [3,4]. This technology was made faster and more accurate with electrochemical sensors that replaced reflectance meters in the 1990s. SMBG now requires small amounts of blood (0.3–10 lL), is rapid (5–15 seconds), and relatively accurate (less than 10% analytical error) [5]. The 10% error is still higher, however, than the American Diabetes Association (ADA) goal of less than 5% error [6]. The importance of SMBG in diabetes self-care, as the first step in blood glucose sensor development, can hardly be overemphasized. Along with hemoglobin A1c (HbA1c), it is a basic pillar in assessing diabetic control, and by allowing hour-to-hour measurement, it makes practical the sort of intensive self-care regimen that prevents This work was supported by National Institute of Health grant R01 DK55132 and General Clinical Research Center grant RR00051. Dr. Saudek is on the Medical Advisory Board for DexCom, Inc., the advisory panel for Cygnus, Inc., and has received research support from Medtronic MiniMed, Inc. * Corresponding author. E-mail address: [email protected](C.D. Saudek). 0889-8529/04/$ - see front matter Ó 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.ecl.2003.12.001 Endocrinol Metab Clin N Am 33 (2004) 175–195
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Chee W. Chia, MD, Christopher D. Saudek, MD*Division of Endocrinology and Metabolism, Department of Medicine, The Johns Hopkins
University School of Medicine, 600 North Wolfe Street, Osler 575, Baltimore, MD 21287, USA
Although the symptoms of severe hyperglycemia have been known formillennia, the laboratory measurement of blood glucose was a lengthy andlabor-intensive process until 1915, when Lewis and Benedict [1] werecredited with simplifying the analysis to a point that glycemia could bedocumented relatively quickly. This advance is considered a crucial step inthe chain of events that led up to the isolation of insulin by Banting and Bestin 1921 [2]. Glucose measurements remained laboratory-based until the1940s, however, when self-assessment of glucosuria became feasible bymeasuring total ‘‘reducing substance’’ in the urine at home. It tooksophisticated physical chemistry to fix glucose oxidase to a reagent strip witha development of color proportional to glucose in a drop of blood, initiatingthe era of self-monitoring of blood glucose (SMBG) with reflectance metersin the 1970s [3,4]. This technology was made faster and more accurate withelectrochemical sensors that replaced reflectance meters in the 1990s.
SMBG now requires small amounts of blood (0.3–10 lL), is rapid (5–15seconds), and relatively accurate (less than 10% analytical error) [5]. The10% error is still higher, however, than the American Diabetes Association(ADA) goal of less than 5% error [6]. The importance of SMBG in diabetesself-care, as the first step in blood glucose sensor development, can hardly beoveremphasized. Along with hemoglobin A1c (HbA1c), it is a basic pillar inassessing diabetic control, and by allowing hour-to-hour measurement, itmakes practical the sort of intensive self-care regimen that prevents
This work was supported by National Institute of Health grant R01 DK55132 and
General Clinical Research Center grant RR00051.
Dr. Saudek is on the Medical Advisory Board for DexCom, Inc., the advisory panel for
Cygnus, Inc., and has received research support from Medtronic MiniMed, Inc.
176 C.W. Chia, C.D. Saudek / Endocrinol Metab Clin N Am 33 (2004) 175–195
complications [7]. About 56% of people with diabetes are now performingSMBG on a regular basis, at least once a day [8].
The main limitations of SMBG from the patient’s view are inconvenienceand discomfort, but the main limitations in terms of information obtainedare that SMBG is intermittent and depends on the patient’s initiative. Atbest, SMBG provides 2, 4, or rarely over 10 data points in a day, but bloodglucose itself varies continuously and unpredictably, up to as much assixfold in the course of 24 hours. The values obtained by SMBG are alwaysonly snapshots of a continuous variable. The next major step in diabetesself-care will therefore be practical use of continuous glucose monitoring.
The idea of continuous glucose monitoring is not new. Clark and Lyons [9]and Updike and Hicks [10] first described the electrochemical enzymaticglucose sensors in the 1960s. The concept of artificial pancreas with con-tinuous glucose monitoring and automated insulin delivery was proposed byKadish in 1964 [11], and made a reality by two different research teams in1974, Albisser and colleagues [12,13] and Pfeiffer and colleagues [14,15].Pfeiffer and colleagues [14,15] developed the Biostator, a large, bedside devicewidely used in research and for clinical assessment of blood glucose patternsin individuals. Using a continuous withdrawal of venous blood, with an on-line glucose oxidase–containing membrane that provided real-time glucosevalues, the Biostator was limited by its sheer size and the need for constantattention of a technician. Thus, despite over 30 years of research in this area,the field of continuous glucose monitoring for clinical use did not prosperuntil the turn of this century.
This article assesses the uses and requirements of continuous glucosesensing and reviews the different glucose sensor technologies now clinicallyavailable and that may become available in the near future. The authors closewith an assessment of prospects for the fully automatic artificial pancreas.
Uses of continuous glucose monitoring
Continuous glucose monitoring could help avoid severe highs and lows(the alarm function), identify patterns in diabetic control, and, ultimately,control an insulin delivery system. The requirements for each role differ indetail, although in the end an accurate, precise, and robust sensor is the ideal.
Any sensor that can signal when blood glucose is too low or too highserves an obvious purpose. The set point to trigger high or low alarms couldbe programmed individually, and could include rate-of-decline or rate-of-rise to anticipate the highs and lows. Accuracy may be less of a requirementin this application than reliable avoidance of false alarms. Some yearsago, a wristwatch-like device called the Sleep Sentry (Teledyne Avionics,Charlottesville, Virginia) monitored changes in skin temperature and sweat-ing, but it had far too many false alarms to serve as a reliable indication ofhypoglycemia [16,17].
177C.W. Chia, C.D. Saudek / Endocrinol Metab Clin N Am 33 (2004) 175–195
Patterns of blood glucose control should be apparent from use ofa continuous glucose monitor. The main challenge in this use, particularlywhen applied to type 1 diabetes mellitus, will be data management: whatpatterns are identifiable when glycemia seems to vary almost at random? Thefinal application, driving an artificial pancreas, is the most difficult challenge.
Assessment of glucose sensors
Glucose sensors can be assessed quantitatively in terms of precision,accuracy, sensitivity, and stability. Among the important performancecharacteristics are calibration requirements, availability of results, longevity,and robustness. In vivo use is to be distinguished from in vitro, or factoryoperation. Accuracy assesses how close sensor results are to some ‘‘goldstandard,’’ and can be expressed as mean error (eg, mg/dL) or relative error(% error).Precision is a different measure, indicating how closely grouped thesensor results are around the mean result, and is expressed as variance,standard deviation, or range [18]. For example, accuracy may be excel-lent—on average the sensor readings are close (eg, within 5% of the goldstandard), but precision may be poor—the sensor readings vary widely (eg,20% above and below the gold standard). The results of sensor measurescommonly are displayed on the backdrop of Clarke’s Error Grid Analysis[19,20], although the authors and others [21–23], including Clarke, discourageusing this analysis alone.
An important issue in assessing a sensor is the applicable gold standard.Usually, the accepted method is to test against whole blood or plasmaglucose. When subcutaneous tissue fluid or extracted interstitial fluid istested, however, the results may differ systematically or in time kinetics fromblood glucose.
General approaches to continuous glucose monitoring
The general approaches to glucose-sensing technology are summarizedin Table 1. Over 100 universities and companies are working on the subjectto develop these technologies [24], none of which is yet either ready to drivean artificial pancreas or ideally suited to routine clinical use.
Sensors based on the enzyme glucose oxidase are the most reliablecompared, for example, to spectroscopic or viscosity-based sensors, becauseglucose oxidase is specific for glucose. Most of the sensors under de-velopment today are amperometric, stemming from the work of Clark andLyons [9], Updike and Hicks [10], and Gough and colleagues [25]. Thehydrogen peroxide–based enzyme electrode sensor [26–28] has a membranewith immobilized glucose oxidase coupled to a peroxide-sensitive catalyticanode. Hydrogen peroxide is generated when glucose and oxygen react withthe membrane. The hydrogen peroxide diffuses to the anode where it is
A GM1, Subcutaneous Continuous Glucose Monitoring System;
TGM stem.
179
C.W
.Chia,C.D
.Saudek
/Endocrin
olMeta
bClin
NAm
33(2004)175–195
operfusion Similar concept as
microdialysis except the
dialysate is in direct
contact with interstitial fluid
Disetronic ++
sdermal
verse
iontophoresis
Interstitial fluid is extracted
through the skin using
various techniques and
glucose is measured using
electrochemical sensor
GlucoWatch/
Cygnus
+
ophotonic SpectRx +
venous
ng-term Electrochemical sensor
with oxygen-based
enzyme electrode
inserted through the
subclavian vein and
positioned in the superior
vena cava
Medtronic
MiniMed
++++
ort-term VGMS/
Medtronic
MiniMed
++++
From least invasive (+) to most invasive (++++).
bbreviations: CGMS, Continuous Glucose Monitoring System; N/A, not available; SC
S, Telemetered Glucose Monitoring System; VGMS, Vascular Glucose Monitoring Sy
180 C.W. Chia, C.D. Saudek / Endocrinol Metab Clin N Am 33 (2004) 175–195
oxidized to produce an electrical current. The amount of current generatedis proportional to the glucose concentration. The hydrogen peroxide sensorshave several limitations, such as the dependence on oxygen concentration,interference from other chemical compounds, and inactivation of enzymeover time [29]. The most recent sensor developed by Updike and colleagues[30] may have overcome most of these limitations.
Another design is the oxygen-based enzyme electrode sensor [10,25]. Thisdesign has a membrane with immobilized glucose oxidase coupled to anelectrochemical oxygen sensor. The glucose and oxygen from the reagentfirst come into contact and react with the glucose oxidase. As a result, theamount of oxygen detected by the oxygen sensor is reduced. The signalcurrent generated by the oxygen sensor is subtracted from that of a referenceoxygen sensor without the glucose oxidase, and the difference is pro-portional to the glucose concentration. The design of the oxygen-basedenzyme sensor makes it less susceptible to oxygen concentration [29].
There are a number of comprehensive reviews of different types ofelectrochemical sensors [29,31,32].
Source of analyte measured
Although many fluid sources can be imagined (eg, tears, saliva, urine,cerebrospinal fluid), two locations are used commonly for measuringglucose: interstitial and intravascular. For continuous monitoring, theinterstitial approach is the most popular because it is less invasive and lessprone to triggering a clotting reaction. There are several ways to measureinterstitial glucose: (1) inserting the glucose sensor into the subcutaneoustissues, either short- or long-term; (2) microdialysis technique; and (3) open-flow microperfusion. The analyte most familiar to patients, clinicians, andresearchers, however, is blood or plasma, and the relationship betweenplasma and interstitial glucose is still not completely understood [33]. All thecurrent sensors that measure interstitial glucose are calibrated using bloodglucose measurement (obtained from capillary glucose meter) so the glucosereadings are meant to represent plasma glucose equivalents.
Exposure of the sensor to blood is relatively straightforward: either theblood is externally accessed, as in SMBG or the Biostator, or the sensor isplaced intravenously. The latter procedure can be potentially dangerous (ifan externalized line is in place long-term), and always prone to walling off bythe clotting mechanisms.
Clinically available subcutaneous glucose sensors
Regulatory approval and market availability of medical devices changesfrom month to month, but it is useful to consider which continuous glucosemonitors are available, which may become available soon, and which are
181C.W. Chia, C.D. Saudek / Endocrinol Metab Clin N Am 33 (2004) 175–195
less certain or unlikely to be available for some years. There are threecurrently available glucose sensors.
Continuous Glucose Monitoring System
The Continuous Glucose Monitoring System (CGMS; MedtronicMiniMed, Northridge, California) is the first commercially availablecontinuous glucose sensor and the most extensively studied. This first-generation sensor is a Holter-style device, with data collected in memory forup to 3 days and then downloaded for graphic display; the glucose data arenot available in real time.
CGMS is an electrochemical sensor placed subcutaneously andconnected to a pager-sized monitor worn on the belt. Physically, it lookslike an external insulin pump. This sensor is placed subcutaneously witha needle insert and held in place by clear tape. Sensing is based on thehydrogen peroxide–based enzyme electrode technology described above[34]. The glucose sensor signal is measured in nanoamperes, acquired every10 seconds. An average of these signals is saved in the CGMS monitor every5 minutes. At least four calibrations by SMBG per day are recommended,each entered into the CGMS monitor. These SMBG readings are used tocalibrate the sensor readings retrospectively when the data are downloadedto a computer [35]. The clinical trials on the first-generation CGMS showedgood correlation between sensor readings and SMBG [36,37].
There have been a number of small trials to evaluate whether CGMS useimproves metabolic control, and to assess the incidence of ‘‘undiscovered’’hypoglycemia. Two studies showed an improvement in HbA1c with CGMSuse [38,39] although one showed no difference between CGMS versusfrequent SMBG alone [40]. A recent randomized multi-center study in-volving 128 patients with type 1 diabetes mellitus found improvement inHbA1c in both the control group and the CGMS group, but significantlyless hypoglycemia in the CGMS group [41].
Others have questioned whether the CGMS data are reproducible [42], orwhether the asymptomatic nocturnal hypoglycemia may be spurious [43].With the introduction of a redesigned sensor and an updated softwarepackage by Medtronic MiniMed in 2002, there was improvement incorrelation with mean absolute difference and improved performance [44].More studies need to be done using the current improved CGMS. Itsintended use is to provide glucose trend information and to supplementSMBG [45,46].
The limitations of the CGMS include the requirement for frequent (fourtimes daily) calibration, sensor warm-up period of an hour or more, and therare possibility of irritation and infection at the sensor insertion site.
Nevertheless, since its availability in 1999, CGMS has been usedrelatively widely by clinicians and researchers, and has been shown to beuseful in detecting glycemic patterns even in children with type 1 diabetes
182 C.W. Chia, C.D. Saudek / Endocrinol Metab Clin N Am 33 (2004) 175–195
mellitus [47,48], as well as asymptomatic hyper- and hypoglycemia in adultswith well-controlled type 2 diabetes mellitus [49]. It also has been used asa research tool in various studies looking at glycemic excursions underdifferent conditions ranging from the drug effect of pramlintide in type 1diabetes mellitus [50] to patients who underwent islet cell transplant [51]. Inmany ways, CGMS has ushered in the era of continuous glucose monitoringon a clinical basis.
GlucoWatch Biographer
The second continuous glucose monitor to become available in theUnited States was the GlucoWatch Biographer (Cygnus, Redwood City,California). Unlike the CGMS, the GlucoWatch is worn on the arm likea wristwatch, and provides a real-time glucose display. It extracts interstitialfluid through the skin using a process called reverse iontophoresis [52,53].The glucose concentration of the extracted fluid is measured by hydrogenperoxide–based enzyme electrode sensor in the disposable membrane on theback of the device. The new version of GlucoWatch G2 Biographer takes 2hours to equilibrate, and then it is calibrated using a single SMBG value.Thereafter, it provides a glucose reading every 10 minutes for up to 13 hours[54]. The glucose readings displayed in real time have a mean lag time ofabout 15 minutes [55].
Clinical trials in different patient populations (type 1 and type 2 diabetesmellitus) and different environments (controlled versus home) showed goodcorrelations between Biographer readings and capillary blood glucosereadings [55–58]. To optimize the trade-off between true positives and falsepositives for hypoglycemic alarms, the best setting of the alarm level wasfound to be 1.1 to 1.7 mmol/L (20–30 mg/dL) above the level of concern[59]. A recent study in a pediatric population showed statistically significantimprovement in HbA1c in the biographer group compared with the controlgroup where the biographer was worn four times a week for 3 months [60].
The clear advantages of the GlucoWatch system include the real-timedisplay of data and the need to calibrate only once. Its limitations includethe relatively short duration of function and the variable comfort of wearingthe device.
GlucoDay
The GlucoDay system (Menarini Diagnostics, Florence, Italy) is currentlyavailable in Europe, not in the United States. It is the first continuous sensorto be clinically available that uses a microdialysis technique to obtaininterstitial fluid.
In the early 1990s, several groups in Europe began using the microdialysisapproach for glucose monitoring [61–63]. Pfeiffer’s group [64,65] inGermany developed the so-called ‘‘Ulm Zucker Uhr System.’’ As sum-marized in an excellent review of microdialysis-based glucose sensing by
183C.W. Chia, C.D. Saudek / Endocrinol Metab Clin N Am 33 (2004) 175–195
Heinemann [66], a small catheter is inserted into the subcutaneous tissueand perfused with an isotonic solution without glucose. Glucose in theinterstitial fluid surrounding the catheter diffuses down the concentrationgradient across a semipermeable membrane into the catheter. This dialysateis collected outside the skin, and an amperometric glucose sensor measuresglucose concentration of the dialysate [66]. The limitations of this systeminclude problems with catheter insertion sites, signal drift caused by localtissue trauma and inflammation [67], and an inevitable lag time, dependenton the dialysate flow rate [66,68,69].
The GlucoDay system’s dialysate is pumped through the catheter at 10lL/min, and the glucose concentration in the effluent dialysate is measuredex vivo by an electrochemical sensor with immobilized glucose oxidaseenzymes. The dialysate and waste fluid are stored in two small plastic bagscarried by the patient [69]. The physical lag time for this system is under 3minutes because of the high dialysate flow rate, but this high flow rate maycause incomplete equilibrium of the dialysate with the interstitial fluid andthus incomplete recovery of the interstitial glucose [66]. The device iscalibrated at 2 hours after catheter insertion. It acquires glucose level everysecond with an average glucose value stored every 3 minutes, and can beused for up to 48 hours. The glucose values are available in real time alongwith hypo- and hyperglycemia alarm. They can be also be downloaded toa computer at the end of the monitoring period. A multicenter trialinvolving 70 subjects with diabetes showed promising results [66,69].
Systems approaching clinical availability
The exact status of clinical availability changes from month to month.The authors describe the clinical availability of the systems as of this writing,but specific devices may move in (and out) of clinical availability quickly.
Telemetered Glucose Monitoring System
Medtronic MiniMed’s Telemetered Glucose Monitoring System (TGMS)is a next generation of the CGMS. Its main new features are real-timeglucose values and hypo- and hyperglycemic alarms according to preset,adjustable parameters [70]. Otherwise, the TGMS will operate like theCGMS, with a sensor placed subcutaneously, lasting about 3 days, andrequiring SMBG calibration.
Subcutaneous Continuous Glucose Monitoring System 1
The Subcutaneous Continuous Glucose Monitoring System 1 (SCGM1)(Roche Diagnostics GmbH, Mannheim, Germany), like the GlucoDaysystem, uses the microdialysis technique coupled to an external ampero-metric glucose sensor. With a low flow rate (0.3 lL/min) of the perfusate, it
184 C.W. Chia, C.D. Saudek / Endocrinol Metab Clin N Am 33 (2004) 175–195
is said to last for up to 4 days with glucose value calculated every second andthe average stored every minute [66]. According to plans, the system will becalibrated with a SMBG glucose value no more than once a day. It willprovide trend information and hypo- and hyperglycemic alarms. The sensorunit holds the microdialysis catheter that is attached to the patient, and thewaste fluids from the dialysate. The glucose measurements obtained in thissensor unit are displayed, stored, and can be transmitted wirelessly to a datamanager [66].
Some limitations of the SCGM1 include a warm-up time of about 12hours [67], and because of the low perfusate pump rate, an inherent lag timeof 31 minutes (�2 minutes) [68]. A feasibility study of this device in 23ambulatory inpatients, however, showed promising results [71], anda multicenter clinical study is reportedly underway [67].
External continuous glucose monitoring systems under development
GluOnline
The GluOnline system is being developed by Disetronic Medical Systems(Burgdorf, Switzerland). It uses microdialysis to obtain interstitial fluid, butthe analysis is coupled to a unique nonenzymatic glucose sensor referred toas the viscometric-affinity sensor. The working principle is that the viscosityof a liquid containing dextran and concanavalin A (ConA) is highly glucose-dependent [72]. Free glucose modulates the viscosity of this fluid by com-peting with dextran to bind at the glycoligand binding site of ConA [73].The glucose-sensitive liquid is pumped through the dialysis catheter at a flowrate of 5 lL/h. Viscosity after equilibration with interstitial fluid iscompared with the viscosity of the reference fluid, and the difference inthe measured viscosity signals is transformed into glucose concentration.There is a technical delay of 5 to 10 minutes and the read-out of theimplanted sensor has a time shift of 10 to 15 minutes [72].
An open-flow microperfusion approach
Using yet another novel approach, Disetronic Medical System isdeveloping the ‘‘Open-flow Microperfusion’’ technique as part of theAdvanced Insulin Infusion with a Control Loop (ADICOL) project. This isa European Union research program sponsored by a consortium of partnersfrom six European nations representing universities, research institutes, andindustry. The goal is to provide the first subcutaneous glucose-sensing/subcutaneous insulin delivery (SC-SC) closed-loop insulin delivery system[74].
The open-flow microperfusion technique is based on a double lumencatheter with macroscopic (0.5 mm diameter) perforations, inserted into thesubcutaneous adipose tissue. The perfusate enters the double lumen catheter
185C.W. Chia, C.D. Saudek / Endocrinol Metab Clin N Am 33 (2004) 175–195
through the inner cannula and then into the annular space between the innercannula and the outer perforated catheter. With the perforated outercatheter, the perfusate is able to equilibrate with the interstitial fluid whilein the annular space. The outer lumen is connected to a peristaltic pump,which pumps the effluent perfusate to a collecting system [75,76]. Currently,the technique acquires a glucose reading every 30 minutes [76,77]. Resultsfrom six patients with type 1 diabetes mellitus showed good performance, atleast on the Clarke Error Grid [77].
SpectRx: a transdermal, biophotonic approach
The SpectRx’s biophotonic technology developed by SpectRx, Inc.(Norcross, Georgia) uses a laser device to create micropores (less than 100lm in diameter) in the stratum corneum of the skin. Interstitial fluid iscollected through thesemicroporeswithmild suction, and glucose ismeasuredin a biosensor patch containing an enzyme-based glucose assay. The patch isdisposable, lasting up to 3 or more days. This system requires at least onecalibration a day. Preliminary findings showed better results on the first daythan subsequent days, and more trials are currently underway [78,79].
Other techniques
Sonophoresis or phonophoresis techniques are in early development. Theprinciple is that applying ultrasound to the skin can increase its permeabilityand transdermal transport [80]. The use of ultrasound pretreatment fornoninvasive monitoring has also been investigated [81]. Whether this willever function practically as a glucose sensor remains to be seen.
Another technique for collection of fluid transdermally is the suctionblister technique to extract dermal interstitial fluid for glucose analysis[82,83].
Noninvasive optical glucose sensors
Each of the systems described above requires obtaining an interstitial orblood fluid to analyze. A truly noninvasive glucose sensor, however, wouldnot break the skin barrier at all. Optical glucose sensors fall into thiscategory, and sensing glucose by transmitting light into or through the skinhas a long, rocky history. There are two basic approaches to optical glucosesensing: absorbance of light and optical rotation of light. Each presentsconsiderable technical challenges because glucose provides an extremelysmall signal relative to water and solutes other than glucose in tissue andblood. Several areas of the electromagnetic spectrum (wavelengths) havebeen used, including infrared and near-infrared spectroscopy. The specifictechniques include Raman spectroscopy, fluorescence spectroscopy, andpolarimetry, each beyond the scope of this article. Interested readers arereferred to several excellent reviews [24,84].
186 C.W. Chia, C.D. Saudek / Endocrinol Metab Clin N Am 33 (2004) 175–195
At present, several groups continue to develop optical-based sensors, butin general they have involved large, bedside equipment rather than portablesensors. If they are to become practical, the optical sensors will have to beminiaturized and show that they function with various skin thicknesses andcolors, states of tissue hydration, and levels of potentially interferingsubstances, whether endogenous (eg, urea) or exogenous (eg, drugs).
Implanted glucose sensors
Implanted glucose sensors would have the obvious advantage of requiringno transdermal fluid collection, and no risk for infection or irritation oncethe device is implanted. Because they involve an initial surgical procedure,they must function for much longer periods than is the case if they can bechanged every few days, and because they cannot be removed as easily in theevent of failure, they must be more robust.
The long-term intravenous glucose sensor currently under developmentby Medtronic MiniMed is based on the oxygen-based enzymatic glucosesensor by Gough and colleagues [85]. The sensor is percutaneouslyimplanted by way of subclavian vein catheterization, with the sensor tipradiographically positioned into the superior vena cava or right atrium (ie,in areas of high blood flow). An external electronic module is used to recordsensor output every minute. Small clinical trials in human are underway.Preliminary findings based on the use of 15 sensors in 10 people with type 1diabetes mellitus over a total of 11.8 patient-years showed encouragingresults. In a preliminary report, no clotting or venous trauma was notedwith average sensor duration of 287 days (range 89–431 days) [86]. Althoughit shows great technical promise, the device has not moved into largerclinical trials as quickly as expected.
Short-term intravenous glucose sensor
Long-term intravenous sensors raise difficult issues of fibrin deposition,stability of placement, and so forth, but essentially the same technology couldbe used in short-term applications such as intensive care unit glucosemonitoring. At the 2003 ADA 63rd Scientific Sessions, Medtronic MiniMedpresented a late-breaking abstract (unpublished findings, 2003) on the resultsof a feasibility study looking at intravenous glucose monitoring in critically illpatients. This is particularly attractive use of sensing because it is used in analready highly controlled, medically sophisticated environment; intravenous(even triple-lumen) lines are often already in place; and blood glucose swingsmay be particularly dangerous. Medtronic MiniMed’s Vascular GlucoseMonitoring System (VGMS) consists of an intravenous glucose sensor
187C.W. Chia, C.D. Saudek / Endocrinol Metab Clin N Am 33 (2004) 175–195
placed into the sheath port of a multilumen central venous catheter, withsensor tip position at the superior vena cava/right atrial junction. The sensoris calibrated using a standard glucose meter, and glucose data are collectedevery minute. The preliminary results from eight patients are promising, andthe application could prove extremely valuable in intensive care units.
Subcutaneous implantation is easier and less risky than long-termintravenous implantation, but the issue with long-term subcutaneousimplantation has always been that devices (such as pacemakers or implantedinsulin pumps) are quickly walled off by a thick capsule of scar tissue. Howdoes a sensor device maintain contact with the rapidly changing glucoseconcentration of interstitial fluid if it is walled off? DexCom (San Diego,California) is developing a device, based on technology developed byUpdike and colleagues [30], that not only resists tissue encapsulation butpromotes vascularization of the subcutaneously placed device.
The prototype developed by Updike and colleagues [10,27,30,87] hasa hydrogen peroxide–based enzyme electrode sensor [27]. It maintains long-term activity when implanted subcutaneously because of an angiogenic layerthat encourages formation of capillaries, rather than fibrous tissue, adjacentto the sensor on the tissue surface of the membrane. There is also a special‘‘bioprotective’’ membrane layer that prevents macrophages from contact-ing the enzyme-active membrane. The prototype reportedly can measureglucose at levels ranging from 40 mg/dL to 700 mg/dL (2.2–38.9 mmol/L),has a recalibration interval of 20 days, and has a maximum lifetime of over160 days [30]. Results are telemetered externally.
Preliminary data from a DexCom clinical trial involving 15 patients withtype 1 diabetes mellitus, presented as a late-breaking abstract at the 2003ADA 63rd Scientific Sessions (unpublished), showed promising results, withpatients able to better manage their blood glucose with fewer hypoglycemicand hyperglycemic excursions. A simply implanted, accurate, and reliablelong-term subcutaneous sensor would be a valuable addition to sensordevelopment.
Closing the loop: the artificial endocrine pancreas
Since insulin pump and sensor work began, the notion of a fullyautomated artificial pancreas has been the final goal. The basic elements ofthe closed loop artificial pancreas are not hard to imagine: a sensor, a deliverysystem, and programming to link the two. The delivery systems, externaland implanted, have developed far more successfully over the past decades,but sensor technology is now moving forward quickly.
As with most advances in medicine, the closed-loop insulin delivery basedon sensor-derived glucose values is not likely to come as a single
188 C.W. Chia, C.D. Saudek / Endocrinol Metab Clin N Am 33 (2004) 175–195
breakthrough, but incrementally. This discussion notes the status of varioussteps in ‘‘closing the loop.’’
Controlling external insulin pumps
In 1976, Pickup and colleagues [88] were among the pioneers indeveloping continuous insulin infusion (CSII) (ie, use of the external insulinpump). By 2001, an estimated 200,000 people worldwide and over 130,000people in the United States [89] were using CSII. Although all these pumpsnormally are controlled on an entirely open-loop basis (with the patientdeciding proper dosing based on SMBG), a meta-analysis of randomizedcontrolled trials showed small but statistically significant lower mean bloodglucose level, lower HbA1c, and lower insulin dose in the CSII groupscompared with intensive insulin injections [90]. Health care professionalshave taught and written a variety of sliding scales, but it has alwaysdepended on the patient to ‘‘close the loop’’ (ie, to judge what dose of insulinis needed when).
In 2003, Medtronic MiniMed, in partnership with BD (Becton, Dickinson& Co., Franklin Lakes, New Jersey), gained regulatory approval in theUnited States to market an external pump that begins to automate theprocess of dose selection. The system establishes a communication betweena blood glucose meter (Paradigm Link Blood Glucose Monitor) and theParadigm 512 external pump. With this telemetered link, the pump recordsthe blood glucose value. If the health care professional has programmed intothe pump the estimated insulin requirement for a given blood glucose anda given amount of carbohydrate to be consumed, and if the patient accuratelyestimates and enters the meal size, the pump then displays a suggested insulindose, which the patient can accept or modify. The programming can considerfactors such as a recent dose that is still having an effect.
This new device can be considered the first practical step in closing theloop because it involves communication of a blood glucose value andtranslation into a recommended insulin dose. Clearly, it is far from a fullyautomated system. Next steps could include telemetry between a continuousglucose sensor (such as the Medtronic MiniMed TGMS or the CygnusGlucoWatch) and an insulin pump.
Implantable pumps
The development of open-loop implantable insulin pumps (IIP) hasstretched over the past 25 years, but intraperitoneal insulin delivery by IIPis now largely successful. Blackshear and colleagues [91,92] pioneered theinvestigational use of IIP for the treatment of diabetes in humans in the1970s. Currently, the only available device is the Medtronic MiniMedMMT-2007C, which is available in selected centers in France [93]; trials arestill underway in the United States. Over the years, studies in types 1 and 2diabetes mellitus patients have shown consistently that HbA1c is modestly
189C.W. Chia, C.D. Saudek / Endocrinol Metab Clin N Am 33 (2004) 175–195
reduced, and glycemic fluctuations and hypoglycemic events are markedlyreduced [94–97]. Interested readers are referred to several comprehensivereviews on IIPs [98,99].
A recent short-term closed-loop trial of 10 patients with type 1 diabetesmellitus is reported in an abstract [86]. IIPs were linked to an intravenousglucose sensor for fully automated 48-hour closed-loop insulin delivery. Theresults were promising with the time recorded, with glucose levels less than70 mg/dL reduced from 18% to 6%, and glucose values greater than 240mg/dL levels reduced from 17% to 2%. The report is preliminary, however,and more research is needed to demonstrate the efficacy and practicality ofthis system.
Other approaches to ‘‘closing the loop’’
If both the potential sensors and delivery systems can be classified asexternal or implanted, then there are four conceivable approaches to‘‘closing the loop’’: (1) external sensor/external pump, (2) external sensor/implanted pump, (3) implanted sensor/external pump, and (4) implantedsensor/implanted pump.
All studies with closed loop systems are preliminary and in early stages.In one feasibility study, CGMS was modified and used in real time tocontrol blood glucose with intravenous insulin infusion using a closed-loopalgorithm in five critically ill patients. The results showed that the closed-loop system performed comparably to manual control [100].
Another, the ADICOL project, uses the external/external approach, withsubcutaneous sensing and subcutaneous, closed-loop insulin delivery system[74]. Using a model predictive control (MCP) algorithm with simulatedsubcutaneous glucose values (intravenous glucose values+30 minutes delay)in six people with type 1 diabetes mellitus, mean blood glucose levels were123 � 17 mg/dL in the fasting state and 116 � 32 mg/dL in the postprandialstate, compared with 179 � 70 mg/dL in the observation period (withoutclosed-loop control) [101]. In another study, the intravenous-subcutaneousroute was compared with the subcutaneous–subcutaneous route using theMCP algorithm with six subjects in each group. There was no significantdifference in blood glucose values between the two routes at any time duringblood glucose control with the systems [102]. Importantly, however, thesubcutaneous glucose values used in both studies were only simulated,meaning they were derived by intravenous glucose values +30 minutesdelay, greatly limiting their interpretation.
Summary
Sensor-driven, closed-loop insulin delivery has been a long-term goal ofmany researchers. Delivery systems (insulin pumps), both external andimplanted, are used widely and successfully, but the sensor component has
190 C.W. Chia, C.D. Saudek / Endocrinol Metab Clin N Am 33 (2004) 175–195
been more difficult to bring to practical reality. Over the past few years,however, rapid progress has been made in glucose sensing. CGMS was thefirst continuous glucose monitor to be approved for use by the US Food andDrug Administration, and the GlucoWatch was the second. Each hasgenerated considerable clinical and research interest, and undoubtedly therewill be more continuous glucose sensors on the market over the next fewyears. Indeed, we seem to be entering an era in which continuous glucosemonitoring will be a standard part of diabetes care, revolutionizing clinicalmanagement.
Many questions remain, and research will be stimulated by the easyavailability of continuous monitoring. The goal of the fully automated,closed-loop, IIP remains years away. But when it is attained, implantinga long-term, reliable, low-maintenance, closed-loop device could rightly becategorized as a technological ‘‘cure’’ for diabetes mellitus.
References
[1] Lewis RC, Benedict SR. A method for the estimation of sugar in small quantities of
blood. J Biol Chem 1915;20:61–72.
[2] Bliss M. The discovery of insulin. Chicago: The University of Chicago Press; 1982.