68

DIABETES TECHNOLOGY & THERAPEUTICSVolume 9, Number 1, 2007© Mary Ann Liebert, Inc.DOI: 10.1089/dia.2006.0059

Noninvasive Continuous Glucose Monitoring UsingPhotoacoustic Technology—Results from

the First 62 Subjects

RAM WEISS, M.D., Ph.D.,1 YEVGENY YEGORCHIKOV, M.D.,2ALEXANDER SHUSTERMAN, M.D.,2 and ITAMAR RAZ, M.D.1

ABSTRACT

Background: Continuous glucose monitoring is a useful tool for the achievement of improvedglucose control in patients with diabetes. We tested the ability of a novel noninvasive continu-ous glucose monitor to track glucose excursions.

Methods: The novel noninvasive Aprise™ sensor (Glucon Inc., Boulder, CO) utilizes the pho-toacoustic properties of the blood to infer the prevailing glucose levels. The performance of thissensor was tested in 62 subjects with diabetes who underwent an oral glucose tolerance, mixedmeal, or glucose infusion test.

Results: The analysis included 979 pairs of reference values and sensor-derived glucose de-terminations. The mean of the relative absolute difference (RAD) of the sample was 19.9% witha median RAD of 13.2%. When results were plotted using the Clarke error grid, 66.5%, 28.1%,1%, 4.4%, and 0% were within the A, B, C, D, and E ranges, respectively. The mean RAD for theoral glucose tolerance tests and meal tolerance tests was superior to the mean RAD of the glu-cose infusion studies (17% vs. 19% vs. 22%, respectively).

Conclusions: Continuous noninvasive glucose monitoring using the Aprise sensor providespromising results. Such device may increase patient compliance and improve diabetes manage-ment.

INTRODUCTION

PREVENTION OF LONG-TERM micro- andmacrovascular complications of diabetes

involves intensive insulin treatment regimens(as shown in the Diabetes Control and Com-plications Trial).1,2 The fine-tuning of such in-sulin regimens depends on frequent plasmaglucose level determinations and the under-standing of the timing and frequency of glu-cose measurements at specific periods. Mostpatients with diabetes use conventional glu-

cometers and measure their glucose levels priorto meals and at bedtime. Since glucometer mea-surements involve pricking of the finger orforearm, the number of daily measurements islimited by patient compliance and by the priceof the disposable glucometer sticks.3 Prepran-dial glucose monitoring is beneficial yet missesthe postprandial glucose excursions that arecritically important for the planning of insulinboluses per meal.4 Moreover, solitary glucosemeasurements do not provide information re-garding nocturnal glycemic excursions—a time

1The Diabetes Unit, Hadassah Hebrew University Hospital, Jerusalem, Israel.2Glucon Inc., Boulder, Colorado.

period prone for hypoglycemia and hyper-glycemia.5

Because of the above-mentioned factors, con-tinuous glucose monitoring has recently beengaining increasing popularity.6 Besides pro-viding real-time glucose data and enabling im-proved glycemic control, such systems hold thepromise of a future combination with subcuta-neous insulin infusion pumps, thus creating the“artificial pancreas.”7 The continuous glucosemonitoring devices presently on the market arestill considered invasive, involving a subcuta-neous catheter or probe; thus they still face thelimitations of acceptance by patients who arereluctant to maintain an invasive device con-tinuously attached. Moreover, the performanceof such devices may be limited by local reac-tions at the insertion site. The novel noninva-sive Aprise™ sensor (Glucon Inc., Boulder, CO)utilizes the photoacoustic properties of theblood and tissues to infer the prevailing glu-cose levels. The ultrasound waves are gener-ated by illuminating the tissue with laserpulses. Analysis of the acoustic signals can mapthe depth profile of the absorbance of light inthe tissue. The sensor is attached to the skinabove a blood vessel and does not involve anydisruption of the tissue. In order to test the po-tential of this technology for continuous glu-cose sensing in the clinical setting, we con-ducted a series of studies using oral andintravenous routes of glucose administration inpatients with diabetes.

Preliminary results of this study have beenpresented at the American Diabetes Associa-tion’s 66th Scientific Sessions, in Washington,DC, June 9–13, 2006.8

SUBJECTS AND METHODS

Sixty-two subjects with type I (n � 23) andtype II (n � 39) diabetes were recruitedthrough posted advertisements. Inclusion cri-teria were known diabetes, age between 18 and60 years, and hemoglobin A1c within the rangeof 6–12%. Pregnant or lactating women andsubjects with uncontrolled hypertension (sys-tolic blood pressure greater than 180 mm Hg)were excluded from the study. The study wasapproved by the Human Investigation Com-

mittee of the Hadassah–Hebrew UniversitySchool of Medicine.

Participants arrived fasting at 8:00 a.m. andwere instructed to withhold their morning in-sulin dose. Thirty subjects underwent a stan-dard oral glucose tolerance test (OGTT), and 17underwent a standard meal test. An intra-venous catheter was inserted in the antecubitalvein and kept patent using normal saline.When glucose levels were greater than 250mg/dL, the patients received their standardmorning insulin regimen. Fifteen subjects un-derwent a glucose infusion study using 20%dextrose in order to reach hyperglycemic lev-els at a faster tempo. Glucose was infused at aninfusion rate of 15 mg/kg/min until blood glu-cose was greater than 250 mg/dL and was thendiscontinued, and the patients received theirstandard morning insulin regimen. In all stud-ies, an initial period of up to 90 min, duringwhich a change of at least 80 mg/dL of plasmaglucose was achieved, was used for calibrationof the Aspire glucose sensor. Reference glucosevalues taken during the calibration period werenot included in the analysis.

In all studies, glucose was monitored every10 min using a YSI 2700 glucose analyzer (Yel-low Springs Instruments, Yellow Springs, OH).

The photoacoustic technology is a novelmethodology for measurement of optical prop-erties in a heterogeneous, diffusive medium(e.g., glucose in blood).9 The photoacoustic ef-

NONINVASIVE CONTINUOUS GLUCOSE MONITORING 69

FIG. 1. Image of a vein, using Glucon’s photoacousticsensor. The capability of Glucon’s device to image thevein and localize the measurement is depicted. The wave-length used for imaging is highly absorbed by hemoglo-bin, and the boundaries of the blood vessel are clearly ob-servable.

WEISS ET AL.70

0

50

100

150

200

250

300

350

400

0 30

Calibration

mg

/dL

60 90 120min

150 180 210

predictionglucose

0

50

100

150

200

250

300

350

400

0 30

mg

/dL

60 90 120min

150 180 210

predictionglucose

0

50

100

150

200

250

300

350

400

0 30

mg

/dL

60 90 120min

150 180 210

predictionglucose

240 270 300 330

FIG. 2. Representative OGTT (toppanel), meal (middle panel), and glu-cose infusion (bottom panel) studies.A calibration period of up to 90 minwas chosen in order to achieve achange in plasma glucose of greaterthan 80 mg/dL. The reference glucosemeasurements are in gray dashes, andthe sensor output is in gray spheres.

fect involves sound or pressure waves createdby the absorption of light. These sound wavescan be generated by illuminating tissue withshort, low-energy laser pulses at selected wave-lengths absorbable by tissue. The absorption ofthe pulse causes a microscopic rise in the tem-perature of the tissue, which induces anacoustic pressure impulse that propagates tothe tissue surface. The time and intensity of thepressure impulse carry information regardingthe optical properties within the tissue, whichin turn can be analyzed to determine the typeand concentration of substances inside the tis-sue. The photoacoustic technology creates anultrasonic image of the optical properties of tis-sue directly beneath the sensor. The ultrasonicimage resolves the blood vessel from the tissuelayers around it, enabling a separated analysisof changes in optical properties of the blood inthe vessel and of surrounding tissue. (Figure 1shows the image of the blood vessel used forthe inference of glucose values.) Glucose con-centration varies between physiological com-partments and is known to affect the opticalproperties of the blood (absorption, scattering,index of refraction). By separating the changesin the blood from those in the surrounding tis-sue compartments, we achieved a reliable cor-relation between the photoacoustic signal thatoriginates in the blood vessel and changes inglucose concentration. A novel software is usedto analyze several features of the photoacousticsignal and extracts additional information suchas vein shape, size, etc. This combination pre-cisely localizes the source of the emission (theintravascular compartment) and allows the de-vice to screen out the relevant signals fromother parameters/analytes. The photoacousticimage also monitors mechanical and structuralvariations of the examined tissue, providingadditional information used to monitor andcompensate for motion artifact and physiolog-

ical changes. Thus, the differences of blood andtissue properties are strongly impacted by glu-cose concentration and are used to mathemat-ically derive the systemic glucose levels. Thesensor used in these studies was 35 mm � 38mm � 15 mm in size, yet future prototypes areplanned to be significantly smaller.

For each pair of results, we calculated the fol-lowing parameters: (1) difference (sensor valueminus reference); (2) relative difference (differ-ence divided by the reference and converted topercentage); and (3) mean absolute relative dif-ference (MARD) (absolute difference dividedby the reference and converted to percentage).The reference and sensor-derived glucose mea-surements were plotted using a Clarke errorgrid. We further tested the paired results forISO (International Organization for Standard-ization) compatibility (for reference glucosevalues �75 mg/dL the sensor should detect aresult within �15 mg/dL, and for referenceglucose values �75 mg/dL, the sensor shouldbe within �20% of the reading).

RESULTS

The analysis included 979 pairs of referencevalues and sensor-derived glucose determina-tions. Three representative studies are shownin Figure 2. The mean of the relative absolutedifference (RAD) of the sample was 19.9%, witha median RAD of 13.2% (Table 1). The mean

NONINVASIVE CONTINUOUS GLUCOSE MONITORING 71

TABLE 1. DESCRIPTIVE ACCURACY STATISTICS OF ASPIRE READINGS

Mean 95% confidence interval of mean Median

Difference (mg/dL) �5.55 �8.33 to �2.56 �5.0Absolute difference (mg/dL) 35.40 33.6 to 37.3 28.0Relative difference 2.6% 0.66 to 4.52% �2.8%RAD 19.9% 19.2 to 22.2% 13.2%

TABLE 2. RAD AND PERCENT MEETING ISO CRITERIA

ACROSS GLUCOSE RANGES

Glucose range Median Met ISO criteria

�150 mg/dL (n � 254) 20% 51%151–250 mg/dL (n � 660) 13% 67%�251 mg/dL (n � 65) 9% 86%

and median differences were both around �5to �6 mg/dL, indicating a slight overestima-tion of the sensor. We divided the samples tothree glucose ranges: �150 mg/dL, 151–250mg/dL, and �250 mg/dL. As shown in Table2, the median RAD was 20% in the lower range,13% in the middle range, and 9% in the highglucose range. The ISO criteria for glucosemonitors were met in 51%, 67%, and 86% of thesamples in the low, middle, and high glucoserange, respectively. When all results were plot-ted using the Clarke error grid (Fig. 3), 66.5%,28.1%, 1%, 4.4%, and 0% were within the A, B,C, D, and E ranges, respectively.

When the three study protocols were ana-

lyzed separately (Table 3), the mean RAD forthe OGTTs and meal tolerance tests was supe-rior to the mean RAD of the glucose infusionstudies (17% vs. 19% vs. 22%, respectively).Similarly, the prevalence of range D errors inthe Clark error grid was smaller in OGTTs ver-sus meal tolerance tests and glucose infusionstudies (2.1% vs. 5% vs. 8.6%, respectively).

DISCUSSION

The importance of tight glucose monitoringfor the achievement of optimal glycemic levelsin patients with diabetes and in intensive care

WEISS ET AL.72

TABLE 3. ACCURACY ANALYSIS IN THE DIFFERENT STUDY PROTOCOLS

Intravenous glucoseOGTT (n � 30) Meal test (n � 17) infusion (n � 15)

Paired points 519 238 222Glucose range (mg/dL) 62–352 52–351 43–343RAD mean 17% 19% 22%Clarke error grid range

Range A 69.7% 61.3% 65.8%Range B 27.0% 33.2% 24.3%Range C 1.2% 0.4% 1.4%Range D 2.1% 5.0% 8.6%Range E 0.0% 0.0% 0.0%

FIG. 3. Clarke error grid of 979 paired points.

settings cannot be overemphasized. A majorlimitation for the achievement of such moni-toring is the need for invasive measures likefrequent finger-pricking and transdermal or in-travenous devices. In this study, we demon-strate the successful performance of a nonin-vasive glucose sensor in the inpatient setting.The median RAD reported in this manuscript(13.2%) is comparable to the performance ofother state-of-the-art invasive continuous glu-cose sensors yet achieves these results nonin-vasively.

The technology utilized by the Aprise glu-cose sensor takes advantage of the photo-acoustic properties of blood and blood vesselsto create an output signal utilized for inferringthe ambient systemic glucose level. Processinga signal derived from a blood vessel combinesthe advantages of directly assessing themedium of interest while avoiding the limita-tions of inferring systemic glucose levels de-rived from glucose measurements from non-vascular compartments, such as the interstitialspace.10 A reliable and accurate noninvasiveglucose sensor opens a broad range of possi-bilities for advancing patient care by improv-ing compliance in the outpatient setting by pa-tients with diabetes who are reluctant toperform multiple finger-pricks or to have an in-vasive device connected continuously.

The Aprise demonstrated relatively superiorperformance in OGTTs and meal tests com-pared with intravenous glucose infusion stud-ies. We believe that the changes of systemicglucose levels that occur in response to an oralglucose load or regular meal reflect the relevantdynamics typical of daily settings of patientswith diabetes. The intravenous glucose infu-sion studies were characterized by more rapidchanges in glucose levels, yet the Aprise wasable to detect them reasonably well. The sen-sor demonstrated a slightly larger MARD in thelow normal glucose range. This is probably dueto the smaller effects of glucose changes onblood and tissue properties in this range. Thepresent protocol was not designed to achievetrue hypoglycemic glucose levels; thus furtherstudies of the sensor performance in this criti-cally important range are warranted. Futureadjustments and fine-tuning of the algorithmpresently used may be able to overcome thelimitations of a rapid change in glucose levels

typically achieved by intravenous infusion andimprove sensor performance in the low normaland hypoglycemic ranges. Moreover, perform-ing the mandatory calibration at close-to-nor-mal glucose levels, unlike the present protocolin which the calibration phase was performedat any glucose level the subjects had, shouldimprove the performance of Aprise in the nor-mal and lower glucose ranges.

The results of this study are preliminary andwere achieved in the inpatient setting. Furtherstudies are needed to confirm the ability of theAprise noninvasive glucose sensor to track sys-temic glucose concentrations in the outpatientsetting, during daily life activities, and underdifferent temperature conditions. Achievementof similar results in more demanding condi-tions may open new attractive possibilities forthe management of diabetes in different set-tings.

ACKNOWLEDGMENTS

This study was funded by an unrestrictedgrant from Glucon Inc.

REFERENCES

1. Nathan DM, Cleary PA, Backlund JY, Genuth SM,Lachin JM, Orchard TJ, Raskin P, Zinman B; DiabetesControl and Complications Trial/Epidemiology ofDiabetes Interventions and Complications (DCCT/EDIC) Study Research Group: Intensive diabetestreatment and cardiovascular disease in patients withtype 1 diabetes. N Engl J Med 2005;353:2643–2653.

2. Retinopathy and nephropathy in patients with type 1diabetes four years after a trial of intensive therapy.The Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Compli-cations Research Group. N Engl J Med 2000;342:381–389. Erratum in: N Engl J Med 2000;342:1376.

3. Weinzimer SA, Doyle EA, Tamborlane WV Jr: Diseasemanagement in the young diabetic patient: glucosemonitoring, coping skills, and treatment strategies.Clin Pediatr 2005;44:393–403.

4. Davidson J: Strategies for improving glycemic con-trol: effective use of glucose monitoring. Am J Med2005;118(Suppl 9A):27S–32S.

5. Landgraf R: The relationship of postprandial glucoseto HbA1c. Diabetes Metab Res Rev 2004;20(Suppl2):S9–S12.

6. Klonoff DC: Continuous glucose monitoring:roadmap for 21st century diabetes therapy. DiabetesCare 2005;28:1231–1239.

NONINVASIVE CONTINUOUS GLUCOSE MONITORING 73

7. Steil GM, Panteleon AE, Rebrin K: Closed-loop in-sulin delivery—the path to physiological glucose con-trol. Adv Drug Deliv Rev 2004;56:125–144.

8. Weiss R, Yegorchikov Y, Shusterman A, Raz I: Noninvasive continuous glucose monitoring—resultsfrom the first 62 subjects [abstract 408-P]. http://sci-entificsessions.diabetes.org/Abstracts/index.cfm?fuseaction�Locator.PreviewAbstract&popup�yes&No-Layout�Yes&AbstractID�13405 (accessed Novem-ber 20, 2006).

9. Nagar R, Pesach B, Ben-Ami U, inventors; GluconInc., assignee: Photoacoustic assay and imaging sys-tem. US Patent 6,846,288. January 25, 2005.

10. Koschinsky T, Jungheim K, Heinemann L: Glucosesensors and the alternate site testing-like phenome-

non: relationship between rapid blood glucose changesand glucose sensor signals. Diabetes Technol Ther2003;5:829–842.

Address reprint requests to:Ram Weiss, M.D., Ph.D.

The Diabetes UnitHadassah Hebrew University Hospital

Ein KeremP.O. Box 12000

Jerusalem, Israel 91120

E-mail: weissr@hadassah.org.il

WEISS ET AL.74