Clinical Chemistry 49: 924-934, 2003;
10.1373/49.6.924
(Clinical Chemistry. 2003;49:924-934.)
© 2003 American Association for Clinical Chemistry, Inc.
Monitoring Blood Glucose Changes in Cutaneous Tissue by Temperature-modulated Localized Reflectance Measurements
Shu-jen Yeh1,
Charles F. Hanna1 and
Omar S. Khalil1,a
1 Abbott Laboratories, Diagnostics Division, 100 Abbott Park Rd., Abbott Park, IL 60064.
aAuthor for correspondence. Fax 847-938-7072; e-mail Omar.Khalil{at}abbott.com.
 |
Abstract
|
|---|
Background: Most proposed noninvasive methods for glucose measurements do not consider the physiologic response of the body to changes in glucose concentration. Rather than consider the body as an inert matrix for the purpose of glucose measurement, we exploited the possibility that noninvasive measurements of glucose can be approached by investigating their effects on the skin’s thermo-optical response.
Methods: Glucose concentrations in humans were correlated with temperature-modulated localized reflectance signals at wavelengths between 590 and 935 nm, which do not correspond to any near-infrared glucose absorption wavelengths. Optical signal was collected while skin temperature was modulated between 22 and 38 °C over 2 h to generate a periodic set of cutaneous vasoconstricting and vasodilating events, as well as a periodic change in skin light scattering. The method was tested in a series of modified meal tolerance tests involving carbohydrate-rich meals and no-meal or high-protein/no-carbohydrate meals.
Results: The optical data correlated with glucose values. Changes in glucose concentrations resulting from a carbohydrate-rich meal were predicted with a model based on a carbohydrate-meal calibration run. For diabetic individuals, glucose concentrations were predicted with a standard error of prediction <1.5 mmol/L and a prediction correlation coefficient 0.73 in 80% of the cases. There were run-to-run differences in predicted glucose concentrations. Non-carbohydrate meals showed a high degree of scatter when predicted by a carbohydrate meal calibration model.
Conclusions: Blood glucose concentrations alter thermally modulated optical signals, presumably through physiologic and physical effects. Temperature changes drive cutaneous vascular and refractive index responses in a way that mimics the effect of changes in glucose concentration. Run-to-run differences are attributable to site-to-site structural differences.
|
Introduction
|
|---|
Noninvasive (NI)1
optical methods for the monitoring of glucose concentrations have attracted considerable interest (1)(2). Most proposed NI methods for glucose determinations are based on monitoring extremely weak near-infrared glucose absorption spectral features (3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15), which overlap with other overtone and combination bands of water, hemoglobin, protein, and fats (1). A second approach is to use the effect of glucose on the tissue scattering coefficient (16)(17)(18)(19)(20), which is a nonspecific effect that is common to other soluble analytes. Some reported methods for glucose determinations have problems, such as unexplained biological noise, the possibility of chance correlation with other biological or instrumental events, data overfitting, and the lack of appropriate control experiments.
We describe a novel NI optical method for monitoring changes in blood glucose concentrations by correlating the temperature-modulated localized reflectance signal with glucose values in a series of modified meal tolerance tests (MTTs). The method is based on the assumption that glucose concentrations in tissue affect cutaneous microcirculation and light scattering. Temperature modulation is used to drive/enhance these effects above inherent biological noise. The optical properties of skin depend on temperature changes (21)(22)(23)(24). In previous NI glucose monitoring studies, tissue temperature was not controlled or was kept constant. In this study optical signal was collected while skin temperature was modulated between 22 and 38 °C over of 2 h. Modulating the measurement temperature generates a periodic set of cutaneous vasoconstricting and vasodilating events, as well as a periodic change in skin light scattering.
We studied the optical response of humans to carbohydrate-rich meals and high-protein/no-carbohydrate meals under temperature-modulation conditions. The wavelengths used were shorter than those reported in other studies (3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15) and did not correspond to any glucose infrared absorption bands. Probe design allowed measurement of optical properties down to a depth of 2 mm in the skin, where temperature can be controlled and varied (23)(24)(25)(26).
A four-term linear regression model based on temperature-modulated reflectance data monitored changes in glucose concentrations. A model based on carbohydrate meal calibration data predicted glucose values in carbohydrate MTTs. There were a run-to-run differences that were attributable to skin site-to-site structural differences.
|
Materials and Methods
|
|---|
instrument
A schematic diagram of the temperature-controlled optical system used in this study is shown in Fig. 1
. The instrument comprised a light source module, a human interface module, and a signal detection module. These three modules were interconnected through a branched optical fiber bundle. The light source module had four light-emitting diodes (LEDs) mounted in a circular holder (A in Fig. 1
). Light from the LEDs was focused on the input end of an illuminating fiber (G in Fig. 1
) in the source tip (F in Fig. 1
) by an objective lens (B in Fig. 1
; 28-mm focal length RKE precision eyepiece, part no. 30787; Edmund Scientific). The LEDs were modulated by drive circuits (not shown), with each LED modulated at a different frequency. A portion of the light was diverted by a beam splitter (C in Fig. 1
) and focused on a silicon photodiode reference detector (D in Fig. 1
), which was amplified by amplifier (E in Fig. 1
) to generate a reference signal, which was used to correct for fluctuations in the intensity of the LEDs. A previously described optical system used up to six wavelengths and six source-detector distances that were varied sequentially through the use of stepper motor-controlled wheels (23)(24). This optical system used a quadrant photodiode to allow simultaneous detection at four distances. Frequency modulation of the LEDs and demodulation of each detected signal allowed the simultaneous detection of four wavelengths.
The wavelength and frequency at which each LED was modulated were 590 nm at 819 Hz, 660 nm at 1024 Hz, 890 nm at 455 Hz, and 935 nm at 585 Hz. The half-bandwidth of the first two LEDs was 25 nm, and that of the last two was 50 nm. The power at the end of the illumination fiber was in the range 1.4–5.0 µW for each individual LED, as measured by a power meter. The four photodiodes and the four LEDs combined for the 16 combinations of wavelengths and source-detector distances.
The out end of the illuminating fiber and the input ends of the light-collecting fibers were mounted in a common tip. All fibers were 400-µm diameter, low-OH silica. The distances between the source fiber and the light-collecting fibers were 0.44, 0.90, 1.21, and 1.87 mm, respectively. The common tip was situated at the center of a 2-cm diameter temperature-controlled disc (I in Fig. 1
) located in the human interface module. The disc temperature was controlled by a thermoelectric element (H in Fig. 1
). A thermocouple (J in Fig. 1
) embedded in the disc provided feedback to the temperature controller (K in Fig. 1
; Marlow Industries). The disc and common tip touched the skin surface (L in Fig. 1
). The light re-emitted from the skin was collected by four detection fibers (M in Fig. 1
), and the output end of each fiber in the detection tip, (N in Fig. 1
) was imaged on one quadrant of a quadrant silicon photodiode (O in Fig. 1
; Hamamatsu S4349). The signal from each quadrant of the detector was amplified separately by one of a set of transimpedance detection amplifiers (P in Fig. 1
) and measured with a Hewlett-Packard 3458A multimeter. The amplifiers and the frequency-modulated drive boards were designed and built in house. The optical signals were collected every 3 s. Data were transferred to the computer every 30 s.
The signal from each quadrant of the detector represented light from one detector fiber, which corresponded to the signal at one source-detector distance. The signal at each modulation frequency corresponded to one wavelength.
The body interface module was mounted on a cradle (not shown), which was mounted on an arm of a standard clinical reclining chair (not shown). Individuals sat in the chair with one forearm resting on the cradle. The optical probe located in the body interface module was pressed against the dorsal side of the individual’s forearm at a constant force of 160 g (
45 g/cm2). A personal computer running LabViewTM software (National Instruments) managed data acquisition and set the temperature of the disc via the Marlow temperature controller.
experimental protocol
The Abbott Laboratories Institutional Review Board approved the test protocol and the informed consent form. Each volunteer read and signed an informed consent form before the experiment. Three male volunteers, designated A, B, and C, participated in the experiment. They included two individuals with diabetes (A and B) and one nondiabetic individual (C). All volunteers were of similar age (50–58 years), body build (body mass index <27), and ethnicity. Volunteers A and B had type 2 diabetes and were on oral diabetes medication (no insulin injections). Volunteer A had diabetes for 15 years, and volunteer B had been diagnosed with diabetes less than 5 years previously. The diabetic volunteers used their regular medication doses. Tests were performed at least 2 h after a regular meal and use of the medication. A total of 11 runs were performed on volunteer A, 11 runs were performed on volunteer B, and 8 runs were performed on volunteer C.
Each NI test session lasted for 2 h. During a NI test session, each participant sat in a semi-reclining position and rested his left forearm on the armrest. The probe was in continuous contact with the skin. NI optical data were collected for 15 min, then the volunteer ingested a meal and data collection was continued for the rest of the 2 h. The meals used were a high-sugar liquid (340 mL of Ocean-Spray cranberry-grape cocktail, containing 59 g of sugar; Pepsi Cola Company), 36–67 g of jellybeans (27), a doughnut, or an apple jelly sandwich. Carbohydrate meals increased the participants’ blood glucose concentrations. The range of change in glucose values was 5–15 mmol/L (90–270 mg/dL). None of the participants had hypoglycemic glucose values in any of the experiments.
Several control experiments were conducted. Two test sessions were conducted during which each volunteer did not ingest any food or fluid. In one test session, each volunteer drank 300 mL of water. In two additional tests, each of the diabetic volunteers was given a high-protein/no-carbohydrate meal (200 g of scrambled eggs and sausages) to induce the digestion process without changing blood glucose concentrations. The total experimental protocol spanned a period of 4 weeks.
Capillary blood glucose concentrations were monitored every 15 min by lancing a finger and measuring the blood glucose with a home glucose meter (Bayer Elite®; Bayer Corporation). Blood glucose values were interpolated to generate concentrations at time points corresponding to the NI measurements. The interpolated line through the capillary glucose values was used for the MTT data plots.
data analysis
The distance between the center of the illumination fiber and the center of each light-collecting fiber was the source-detector distance. For each source-detector distance (r), wavelength (
), and temperature (T), the reflectance [R'(r, , T)] was defined as:
 | (1) |
The least-squares regression method was used to establish calibration and validation models. Assuming that glucose linearly affects the optical variables, independent of any other analyte, its concentration of was expressed as:
 | (2) |
The coefficients a0, a1, a2, [ellipse] were determined from the regression of a calibration run, and Pi is any of the reflectance values [R'(r, , T)] as defined by Eq. 1
. The value of T was 22 or 38 °C, and the number of non-zero coefficients a1, a2, ... , in Eq. 2
was limited to 4 to avoid overfitting. The two temperature limits selected in this experiment were 22 °C (close to room temperature) and 38 °C (slightly higher than the body core temperature). Skin temperature measured by an infrared thermometer was in the range 28–32 °C. Thus, the two temperature limits straddled the upper and lower skin temperature range by approximately ± 6 °C. Temperature modulation was assumed to drive the physiologic response. The use of multiple source-detector distances and wavelengths allowed collection of signals from different depths in cutaneous tissue.
During the NI measurement experiment, the temperature of the aluminum disc surrounding the optical probe was repetitively modulated between 22 and 38 °C at 6 min/cycle. The output of the thermocouple embedded in the aluminum disc is plotted in Fig. 2
. Slightly more than 60 s were required for the temperature to reach the targeted value when the disc was in contact with the human body. Measurements at any given time provided data at all four wavelengths, source-detector distances, and at temperature T. The signals were converted to reflectance [R(r, , T)] by dividing the skin reflectance signal by the signal detected from a scattering glass disc taken before each NI test. An example of the temperature-modulated reflectance data is shown in Fig. 3
.
View larger version (34K):
[in this window]
[in a new window]
|
Figure 3. Example of the response of human skin to temperature modulation.
Shown is the localized reflectance for volunteer A during a control run. The three-digit numbers in the legend are the wavelengths in nm. The distances between the illumination fiber and the detection fibers are 0.44 mm (r1), 1.21 mm (r2), and 1.87 mm (r3).
|
|
Thirty-two sequences of reflectance data at temperatures T1 and T2 were obtained. These were R' (at T1) = loge[R(r, , T1)] and R' (at T2) = loge[R(r, , T2)]. For each MTT over the 2-h period, the temperature sequences encompassed 20 data points (6 min/point).
The calibration relationship is a four-term linear equation as described in Eq. 2
. All combinations of 4 of 32 possible values, 16 of R'(r, , T1) and 16 of R'(r, , T2), were tested, and the regression model with the highest correlation coefficient was adopted.
The established four-term multiple least-squares linear regression relationship was then applied to other test data to predict glucose concentrations in other NI experimental runs. The mean-adjusted standard error of the predicted glucose concentrations (SEP) and the correlation coefficient between predicted and blood glucose values (Rp) are defined in Eqs. 3 and 4 :
 | (3) |
where Grefi is the measured capillary glucose value, Gprdi is the predicted NI glucose value, and n is the number of data points in a MTT:
 | (4) |
Data from one meal day in which the SD of reference glucose values were >2.5 mmol/L (45 mg/dL) were used to establish a calibration relationship between the optical signal and glucose concentration. NI glucose values for a particular run were considered successfully predicted when Rp was >0.72 and the SEP was <2.0 mmol/L (36 mg/dL).
|
Results
|
|---|
mtt experiments
The mean-adjusted predicted glucose values for the carbohydrate meals of the two diabetic volunteers are shown in Figs. 4
and 5
and are summarized in Table 1
. On the basis of the results for diabetic volunteer A, and using day 3 as the calibration data set, the calibration equation was:
 | (5) |
R' is as defined by Eq. 1
. The specified source-detector distance (r, in mm), wavelength (, nm), and temperature (T, °C) are in parentheses. On the basis of an independent calibration run (day 3), the glucose concentration was predicted with mean adjusted SEP <1.5 mmol/L (27 mg/dL) and Rp >0.73. Regression results for all five of the carbohydrate meal days (days 2, 4, 7, 8, and 9) are given in Table 1
. All five runs met the success criteria of Rp >0.72 and SEP <2.0 mmol/L (36 mg/dL). There was a day-to-day difference that ranged from 13.72 to -7.50 mmol/L (247 mg/dL to -135 mg/dL) glucose.
View larger version (26K):
[in this window]
[in a new window]
|
Figure 4. Mean-adjusted predicted glucose values (in mg/dL) for the carbohydrate meal runs for volunteer A, using day 3 as a calibration model.
The symbols ( ) represent the predicted values, and the dotted line represents the interpolated capillary blood glucose values.
|
|
View larger version (28K):
[in this window]
[in a new window]
|
Figure 5. Mean-adjusted predicted glucose values (in mg/dL) for the carbohydrate meal runs for volunteer B, using day 4 as a calibration model.
The symbols () represent the predicted values, and the dotted line represents the interpolated capillary blood glucose values. Day 8 data correspond to the day when the individual had a fever.
|
|
The results for volunteer B were similar to those for volunteer A. The calibration day was day 4, and the equation had the following terms:
 | (6) |
One exception for volunteer B was that on day 8, he had a fever attributable to the flu. Experiments on him were stopped and resumed 3 days later, after he returned to work. This run was repeated later and appears in Table 1
as "Repeat for volunteer B, day 8". The predicted glucose values for day 8 trended in the opposite direction of the invasive data: Rp (-0.69) was large and negative, and SEP was >5.46 mmol/L (98.36 mg/dL). Volunteer B had three of five carbohydrate meal runs successfully predicted (excluding run 8) by the independent calibration run 4. The intercept varied between 13.89 and 6.56 mmol/L (250 mg/dL to -118 mg/dL) between runs.
The nondiabetic volunteer (C) had a narrow glucose change that led to erratic prediction results, i.e., it was impossible to establish a calibration relationship from the data that would allow prediction of glucose values for other days/runs with use of the same criteria as for volunteers A and B.
control tests
The results of control runs for volunteer A are shown in Fig. 6
and are summarized in Table 2
. The day 1 run for volunteer A was a control run, where no meal was used, but he showed a steady decrease in glucose concentration over a range of 5.6 mmol/L (100.8 mg/dL). The glucose concentration was predicted with a SEP of 1.28 mmol/L (23 mg/dL) and Rp of 0.65, using the day 3 calibration model described in Eq. 6
. For no-meal control days 5 and 6, SEP values were <1.78 mmol/L (32 mg/dL), but Rp was negative, indicating no correlation between measured signal and glucose concentration.
View larger version (18K):
[in this window]
[in a new window]
|
Figure 6. Mean-adjusted predicted glucose values for volunteer A, using day 3 as a calibration model.
(A), no-meal and water runs; (B), protein meal runs. The symbols () represent the predicted values, and the dotted line represents the interpolated capillary blood glucose values.
|
|
Similar to the case of no-meal runs, protein meal control runs for volunteer A had a SEP <1.39 mmol/L (25 mg/dL) when predicted by the model of Eq. 6
(day 3 carbohydrate meal model). One of the two days had a Rp of 0.17 and the other had a Rp of 0.83. Glucose values for the control runs predicted by the carbohydrate meal run on day 3 had a large amount of scatter. The control runs for volunteer B are shown in Fig. 7
.
View larger version (17K):
[in this window]
[in a new window]
|
Figure 7. Mean-adjusted predicted glucose values for volunteer B, using day 4 as a calibration model.
(A), no-meal and water runs; (B), protein meal runs. The symbols () represent the predicted values, and the dotted line represents the interpolated capillary blood glucose values.
|
|
When the predicted glucose values for volunteer A were plotted on a Clarke Error Grid, all of the data points from the carbohydrate, control, and protein meal runs fell in zones A and B, as shown in Fig. 8
. The Clarke Error Grid presentation did not show the scatter in the predicted glucose values for the control and protein meal runs that was evident in Fig. 6
. Only one of five control runs for volunteer B had a high correlation coefficient for glucose that was predicted by the carbohydrate meal calibration model. The mean-adjusted predicted glucose values fell in zones A and B on the Clarke Error Grid.
View larger version (33K):
[in this window]
[in a new window]
|
Figure 8. Clarke Error Grid for all mean-adjusted predicted glucose values (in mg/dL) for the carbohydrate meals and control runs.
The prediction model was generated on day 3 and is described by Eq. 5
.
|
|
The mean-adjusted predicted glucose values for the nondiabetic volunteer (C) and the control runs for the same individual are shown in Fig. 9
, A and B, respectively. Volunteer C had a narrow glucose change that led to more erratic predictions than was the case of volunteers A and B.
View larger version (27K):
[in this window]
[in a new window]
|
Figure 9. Mean-adjusted predicted glucose values (in mg/dL) for volunteer C, using day 4 as a calibration model.
(A), carbohydrate meal runs; (B), no-meal and water runs. The symbols () represent the predicted values, and the dotted line represents the interpolated capillary blood glucose values.
|
|
|
Discussion
|
|---|
Several observations can be made on the results of the MTT plots and the data in Tables 1
and 2
. The first observation is that the temperature-modulated optical signal monitored the change in glucose concentration during five of five runs for volunteer A and three of five runs for volunteer B, using the stated criteria. The second observation is that the calibration models had reflectance terms at both the low and high limits (22 and 38 °C). The final observation is that, when we used calibration data from a carbohydrate-meal day, the data predicted glucose concentration changes in other carbohydrate meal tests but showed scatter in the control runs.
Generally, correlating glucose concentration to the optical signal in MTTs assumes that the glucose concentration is the only time-dependent variable. Other factors, such as temperature, blood perfusion, tissue compression, blood oxygenation, cutaneous water, and other metabolites or medications that affect tissue blood dynamics are not considered. The wavelengths used in this study (590–935 nm) do not correspond to any near-infrared glucose absorption bands as did those used in previous studies (3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15). Spectral bands in this wavelength range correspond to blood absorption and thus reflect hemodynamic changes in cutaneous tissue. Localized reflectance data in the wavelength range 590–950 nm correlate with hemoglobin concentration (25).
Diabetes causes changes in cutaneous vascular response to external stimuli (28). Glucose has a vasoconstricting effect, and insulin causes vasodilatation in human circulation (29)(30). Glucose also affects tissue light scattering by changing the refractive index of intercellular fluid (16)(17)(18)(19)(20). Light scattering by human skin is linearly dependent on temperature (21)(22)(23)(24)(25). Heating of the skin causes increased cutaneous blood flow and vasodilation (31). Cooling causes vasoconstriction and closing of capillary shunts. Considering all these observations, the effects of temperature and a change in glucose concentration on human skin are twofold: (a) they affect cutaneous vascular circulation (physiologic effect) and (b) they affect cutaneous light scattering (physical effect). Temperature modulation may thus drive one or both of the vascular response and cutaneous light scattering in a way that mimics or enhances the physiologic or physical effect of changes in glucose concentration on optical properties of cutaneous tissue. The correlation between glucose concentration and optical signal in this spectral range may thus be attributed to the effect of glucose on cutaneous hemodynamic response and refractive index changes.
Insulin, as well as increased temperature, has a vasodilating effect. It is a possible that the optical signal measured in this experiment may also correlate with insulin concentration. Because the concentration of insulin is mush lower than that of glucose, it is expected to have a smaller effect on the refractive index of intracellular fluids in cutaneous tissue. It therefore is possible that temperature-modulated optical signals are related to the combined effect of glucose and insulin on cutaneous vascular response. Changes in insulin concentrations were not followed in these experiments because monitoring would require frequent venous blood sampling and the participation of trained medical personnel. No attempt was made to separate these two circulation and scattering effects in this study. The use of a laser Doppler flowmetry measurement probe in addition to, or in place of, the localized reflectance probe in this experimental set-up may be useful in delineating the effect of glucose on microcirculation.
The day-to-day predicted glucose values differed, leading to a shift in the response curve from day to day. The magnitude of the signal varied with the repositioning the probe on the skin, when the arm was reseated with respect to the probe. Maruo et al. (32) observed similar differences in the predicted glucose values in a fiber optics-based near-infrared absorption experiment. An illuminating optical fiber in touch with the skin delivered polychromatic light, and reflected light was collected at a distance of 0.65 mm by a second fiber, the output of which was analyzed by a spectrometer as absorption spectra in the wavelength range 1500–1800 nm (32).
The source-detector distances of the probe used in our study sampled cutaneous layers down to a depth of 2 mm (23)(24)(25)(26). This depth includes the cutaneous vascular system (upper plexus, lower plexus and connecting arterioles, venules, and shunts), which contributes to the control of the body temperature (33)(34). Cutaneous blood flow depends on temperature (34). Cutaneous red blood cell flux varies between comparable sites in the same individual, and 1-mm2 areas of vascular "territories" surrounded in part by relatively avascular areas have been identified in human skin (35)(36)(37). An optical coherence tomography study of human skin showed pockets of different refractive index values in the skin (38). Thus, even if the shape of the MTT curve is tracking the change in glucose concentration, a constant-term difference may result from positioning of the illumination and detection fibers with respect to vascular territories and avascular areas in the skin, and/or with respect to pockets with different refractive index values.
Different meal conditions were selected in these experiments to avoid chance correlation resulting from the effect of the circadian rhythm on body temperature and blood flow. Ingestion of protein meals showed a change in the optical signal that led to erroneous glucose values when predicted by a carbohydrate-meal calibration model. This shows that examining the carbohydrate meal data alone without running control experiments where variables other than glucose are allowed to change ignores the effects of the underlying physiologic variables on the optical signal. Correlating carbohydrate meal data curves to each other, especially if they have the same shape, may not be sufficient to establish the ability of a NI method to monitor changes in glucose concentration.
Each NI test yields results corresponding to one instantaneous physiologic condition and may not predict data under different conditions. A model that covers multiple physiologic conditions should be applied, assuming that they are not widely different, such as the case of volunteer B’s fever on day 8. Because the cutaneous vascular system plays a role in controlling body temperature, changes in that system between calibration run 4 and prediction run 8 led to an erroneous correlation. The effects of disease states and body response seem to be important. However, incorporating them in calibration models may make these models too complex. Disease states and body response were not incorporated in previous studies (3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(19)(20).
The observed correlation was interpreted as attributable in part to tissue vascular physiologic response to changes in glucose concentration. This interpretation has some bearing on patented methods for the determination of glucose in this spectral range that claim quantification of glucose but do not address the body’s physiologic response.
Use of control runs should be applied to all NI measurements, whether based on optical or other types of physical measurements. The purpose of the control experiments is to isolate underlying physiologic effects from glucose-induced physical effects and/or intrinsic glucose signals.
There are several possible improvements that can be applied to this method. Only one set of temperature limits was selected in the present study. Use of other temperatures and other rates of temperature change may overcome some of the contributions of other vasodilating species. Use of multiple probes can average out site-to-site variations. Variations in initial skin temperature can be overcome by contacting the skin in touch with the temperature-controlled probe for 60 s before the measurement to attain temperature equilibration within the 2-mm depth sampled by the light beams (24). Other ways to improve signal tracking of glucose concentrations are to minimize probe-skin interface variations by skin conditioning such as shaving the measurement site, as is used for the minimally invasive iontophoreses measurement (GlucoWatch®; Cygnus Corporation); use of other body sites, such as the inner lip (12); selection of wavelengths; use of a thermal-coupling agent to improve heat transfer into the skin (39); and the possible use of skin stratum corneum-clearing agents (40).
In conclusion, the physiologic effects of glucose on cutaneous circulation and its physical effect on refractive index have been suggested as the source of the localized reflectance signal tracking of glucose concentrations under temperature modulation. Use of a calibration model from a run involving consumption of a carbohydrate meal enabled the prediction of changes in glucose concentration with a Rp >0.72 and SEP <2.0 mmol/L (36 mg/dL) for 8 of 10 NI carbohydrate-meal MTTs performed with diabetic individuals. Cutaneous structural effects caused differences in the signal. Site-to-site variations were accounted for by mean-adjusting the response curve. In the case of the non-carbohydrate meal control runs, cutaneous hemodynamic changes attributable to digestion affected the optical signal. The extent and rate of change in the cutaneous hemodynamic response to protein meals and its direction may be different from that caused by the change in glucose concentration. As a result, a carbohydrate meal-based calibration model predicts glucose values in the control runs with a high degree of scatter.
|
Acknowledgments
|
|---|
We acknowledge the help of our Abbott colleagues, Stan Kantor for signal processing, Ron Hohs for design of the electronics, Jim Forsyth for machine shop work, and Dr. John Lindberg for optical system design. We thank our colleagues who volunteered for these experiments. We also thank Drs. S-T. Wong and T-W. Jeng for helpful discussions. We thank Drs. Hans Kanger, Rene Bolt, and Frits deMul (University of Twente, The Netherlands) for their help in starting the tissue optics research program at Abbott. Finally, we acknowledge the encouragement of Dr. James Babb.
|
Footnotes
|
|---|
1 Nonstandard abbreviations: NI, noninvasive; MTT, meal tolerance test; LED, light-emitting diode; and SEP, standard error of predicted glucose concentrations. 
|
References
|
|---|
- Khalil O. Spectroscopic and clinical aspects of noninvasive glucose measurements. Clin Chem 1999;45:165-177.[Abstract/Free Full Text]
- McNichols RJ, Cote GL. Optical glucose sensing in biological fluids. J Biomed Opt 2000;5:5-16.[CrossRef][ISI][Medline]
[Order article via Infotrieve]
- Robinson MR, Eaton RP, Haaland DM, Keep GW, Thomas EV, Stalled BR, et al. Noninvasive glucose monitoring in diabetic patients: a preliminary evaluation. Clin Chem 1992;38:1618-1622.[Abstract/Free Full Text]
- Ward KJ, Haaland DM, Robinson MR, Eaton RP. Postprandial blood-glucose determination by quantitative midinfrared spectroscopy. Appl Spectrosc 1992;46:959-965.[CrossRef]
- Burmeister JJ, Arnold MA, Small G. Noninvasive blood glucose measurement by near-infrared transmission spectroscopy across human tongue. Diabetes Technol Ther 2000;2:5-16.[Medline]
[Order article via Infotrieve]
- Burmeister JJ, Arnold MA. Evaluation of measurement sites for noninvasive blood glucose sensing with near-infrared transmission spectroscopy. Clin Chem 1999;45:1621-1627.[Abstract/Free Full Text]
- Gabriely I, Wozniak R, Mevorach M, Kaplan J, Aharon Y, Shamoon H. Transcutaneous glucose measurement using near infrared spectroscopy during hypoglycemia. Diabetes Care 1999;22:2026-2032.[Abstract/Free Full Text]
- Hiese HM, Lampen P. Transcutaneous glucose measurements using near-infrared spectroscopy: Validation of the statistical calibration models. Diabetes Care 2000;23:1208-1209.[Free Full Text]
- Heise HM, Bittner A, Marbach R. Clinical chemistry and near infrared spectroscopy: technology for non-invasive glucose monitoring. J Infrared Spectrosc 1998;6:349-359.
- Blank TB, Ruchti TL, Lorenz AD, Monfre SL, Makarewicz MR, Mattu M, et al. Clinical results from non-invasive blood glucose monitor. SPIE Proc 2002;4624:1-10.[CrossRef]
- Fischbacher C, Jagemann KU, Danzer K, Muller U, Papenkrodt L, Schuler J. Enhancing calibration models for non-invasive near-infrared spectroscopical blood glucose determinations. Fresenius J Anal Chem 1997;359:78-82.[CrossRef]
- Heise HM, Marbach R, Koschinsky T, Gries FA. Noninvasive blood glucose assay by near-infrared diffuse reflectance spectroscopy of human inner lip. Appl Spectrosc 1993;47:875-881.[CrossRef]
- Malin SF, Ruchti TL, Blank TB, Thennadil SN, Monfre SL. Noninvasive prediction of glucose by near-infrared diffuse reflectance spectroscopy. Clin Chem 1999;45:1651-1658.[Abstract/Free Full Text]
- Yoon G, Amerov AK, Joen KJ, Kim Y-J. Determination of glucose concentration in a scattering medium based on selected use of an overtone absorption band. Appl Opt 2002;41:1469-1475.
- Amirov AK, Jeon KJ, Kim Y-J, Yoon G. Method and device for noninvasive blood glucose measurements. SPIE Proc 1999;3599:33-42.[CrossRef]
- Maier J, Walker S, Fantini S, Franceschini M, Gratton E. Non-invasive glucose determination by measuring variations of the reduced scattering coefficient of tissues in the near-infrared. Opt Lett 1994;19:2062-2066.
- Kohl M, Cope M, Essenpreis M, Boecker D. Influence of glucose concentration on light scattering in tissue simulating phantoms. Opt Lett 1994;19:2170-2172.
- Kohl M, Essenpreis M, Cope M. The influence of glucose concentration upon the transport of light in tissue-simulating phantoms. Phys Med Biol 1995;40:1267-1287.[CrossRef][ISI][Medline]
[Order article via Infotrieve]
- Bruulsema JT, Hayward JE, Farrell T, Patterson M, Heinemann L, Berber M. Correlation between blood glucose concentration in diabetics and noninvasively measured tissue optical scattering coefficient. Opt Lett 1997;22:190-192.[Medline]
[Order article via Infotrieve]
- Heinemann L, Schmelzeisen-Redeker G. Non-invasive continuous glucose monitoring in type I diabetic patients with optical glucose sensors. Non-Invasive Task Force (NITF). Diabetologia 1998;4:848-854.
- Bruulsema JT, Hayward JE, Farrell TJ, Essenpreis M, Patterson MS. Optical properties of phantoms and tissue measured in vivo from 0.9–1.3 µm using spatially resolved diffuse reflectance. SPIE Proc 1997;2979:325-334.[CrossRef]
- Laufer J, Simpson R, Kohl M, Essenpreis M, Cope M. Effect of temperature on the optical properties of ex vivo human dermis and subdermis. Phys Med Biol 1998;43:2479-2489.[CrossRef][ISI][Medline]
[Order article via Infotrieve]
- Yeh S-j, Khalil OS, Hanna CF, Kantor S, Wu X, Jeng T-W, et al. Temperature dependence of optical properties of in-vivo human skin. SPIE Proc 2001;4250:455-461.[CrossRef]
- Khalil OS, Yeh S-j, Lowery M, Wu X, Hanna CF, Kantor S, et al. Temperature modulation of the visible and near infrared absorption and scattering coefficients of human skin. J Biomed Opt 2003;8:191-205.[CrossRef][ISI][Medline]
[Order article via Infotrieve]
- Wu X, Yeh S-j, Jeng T-W, Khalil OS. Noninvasive determination of hemoglobin and hematocrit using a temperature-controlled localized reflectance tissue photometer. Anal Biochem 2000;287:284-293.[CrossRef][ISI][Medline]
[Order article via Infotrieve]
- Bevilacqua F, Piguet D, Marquet P, Gross JD, Tromberg BT, Depeursinge C. In vivo local determination of tissue optical properties: applications to human brain. Appl Opt 1999;38:4939-4950.
- Lamar ME, Kuehl TJ, Cooney AT, Gayle LJ, Holleman S, Allen SR. Jelly beans as an alternative to a fifty-gram glucose beverage for gestational diabetes screening. Am J Obstet Gynecol 1999;181:1154-1157.[CrossRef][ISI][Medline]
[Order article via Infotrieve]
- Wiernsperger N. Defects in microvascular heamodynamics during prediabetes: contributor or epiphenomenon?. Diabetologia 2000;43:1439-1449.[CrossRef][ISI][Medline]
[Order article via Infotrieve]
- Frost T, Kunt T, Pohlmann T, Goitom K, Lobig M, Engelbach M, et al. Microvascular skin blood flow following the ingestion of 75 g glucose in healthy individuals. Exp Clin Endocrinol Diabetes 1998;106:454-459.[ISI][Medline]
[Order article via Infotrieve]
- Yki-Jarvinen H, Utriainen T. Insulin-induced vasodilatation: physiology or pharmacology. Diabetologia 1998;41:369-379.[CrossRef][ISI][Medline]
[Order article via Infotrieve]
- Hassan AAK, Tooke JE. Effect of change in local skin temperature on postural vasoconstriction in man. Clin Sci 1988;74:201-206.[Medline]
[Order article via Infotrieve]
- Maruo K, Chin J, Tamura M. Noninvasive blood glucose monitoring by novel optical fiber probe. SPIE Proc 2002;4642:20-27.
- Tuchin V. Tissue optics: light scattering methods and instruments for medical diagnosis 2000:41 SPIE Press Bellingham, WA. .
- Tenland T, Salerud EG, Nilsson EG, Oberg PA. Spatial and temporal variations in human skin blood flow. Int J Microcirc Clin Exp 1983;2:81-90.[ISI][Medline]
[Order article via Infotrieve]
- Braverman IM, Keh A, Goldminz D. Correlation of laser Doppler wave patterns with underlying microvascular anatomy. J Invest Dermatol 1990;95:283-286.[CrossRef][ISI][Medline]
[Order article via Infotrieve]
- Braverman IM. Anatomy and physiology of cutaneous microcirculation. Berardesca E Elsner P Maibach HI eds. Bioengineering of the skin: cutaneous blood flow and erythema 1995:3-22 CRC Press Boca Raton, FL. .
- Braverman IM. The cutaneous microcirculation: ultrastructure and microanatomical organization. Microcirculation 1997;4:329-340.[ISI][Medline]
[Order article via Infotrieve]
- Knuttel A, Boehlau-Godau M. Spatially confined and temporally resolved refractive index scattering evaluation of human skin performed with optical coherence tomography. J Biomed Opt 2000;5:83-92.[CrossRef][ISI][Medline]
[Order article via Infotrieve]
- Wu X, Khalil OS, Yeh S-j, Hanna CF, Jeng T-W, inventors. Method for improving the non-invasive determination of concentration of analytes in a biological sample. US patent no. 6,241,663, 2001..
- Vargas G, Chan EK, Barton JK, Rylander HG, Welch AJ. Use of an agent to reduce scattering in skin. Lasers Surg Med 1999;24:133-141.[CrossRef][ISI][Medline]
[Order article via Infotrieve]
Top
Abstract
Introduction
