Optical Coherence Tomography-Based Continuous Noninvasive Glucose Monitoring in Patients with Diabetes

Abstract

The objective of this feasibility study was to test an optical coherence tomography (OCT)-based system for noninvasive, accurate, and continuous monitoring of blood (as opposed to interstitial) glucose concentration in subjects with diabetes. OCT uses low coherence light with precise depth focusing ability to measure changes in the microvasculature structure for glucose detection. Thirty-three subjects with diabetes had blood glucose concentrations evaluated using simultaneous capillary and OCT-based device (Sentris-100, GlucoLight, Bethlehem, PA) measurements. Subjects received a 50-g carbohydrate load at the start of the study period. Two glucose values with 60 mg separation were used to calibrate the OCT system. Six subjects were excluded (two because of hardware and/or software failures at study initiation, two because of sensor connection problems, and two others because of problems with disposable device positioning on the skin). Of the 27 subjects who completed the study, 12 had type 1 diabetes, and 15 had type 2 diabetes. Statistical analysis included plotting of the Clarke error grid, comparison to International Organization for Standardization (ISO) standards for glucose monitoring systems, and calculation of the Pearson correlation coefficient. Based on a total of 236 matched points, Clarke error grid analysis indicated 83% in zone A, 16% in zone B, and <1% in zones C and D. ISO standards comparison showed 83% of OCT estimates were within 20% of the reference value for blood glucoses >75 mg/dL. No conclusions can be made about blood glucose estimates in the <75 mg/dL range since no low reference values occurred. The Pearson correlation coefficient comparing OCT versus capillary blood glucose was 0.85 (95% confidence interval 0.81-0.88, P < 0.0001). Mean relative absolute deviation was 11.5%, and median relative absolute deviation was 8.2%. This study demonstrated that the OCT system shows acceptable accuracy in the prediction of blood glucose levels in subjects with both type 1 and type 2 diabetes. Future efforts will evaluate the accuracy of the system in the hypoglycemic range.