Clinical assessment of near-infrared spectroscopy for noninvasive diabetes screening

Current diabetes screening techniques comprise the fasting plasma glucose (FPG) and oral glucose tolerance tests. Both tests demand patient compliance, and neither test has ideal performance. Near-infrared (NIR) spectroscopy is a noninvasive means of interrogating characteristics of a sample and is evaluated as a novel screening method for type 2 diabetes. One hundred fifty-four patients with and without type 2 diabetes were recruited. Their forearm skin was measured with the NIR spectroscopic system, and a capillary blood glucose measurement was also taken. Sixty-six patients returned for a second visit at a later date. A multivariate model, generated from a separate training study, was employed to produce a quantitative risk marker of disease for each NIR spectrum. Sensitivity and specificity (the probabilities that the NIR method will correctly identify a subject as having diabetes or as not having diabetes, respectively) were calculated. As the NIR method produces a continuous rather than categorical classification, various thresholds were evaluated to give several sensitivity and specificity pairs. Test reproducibility was also determined. At a false-positive rate of 70%, the NIR test had a sensitivity of 77.7%, which is comparable to the 77.3% sensitivity for the FPG test as reported for the Third National Health and Nutrition Examination Survey (NHANES III) study. The reproducibility of the NIR test was also similar to the FPG test (inter-day agreement rates of 84.2% and 79.2%, respectively). A noninvasive NIR spectroscopic measurement of the volar forearm was shown to have comparable performance characteristics with the FPG test. The source of the spectroscopic signal is still uncertain and is the subject of ongoing research.