ABSTRACT
A diffuse reflectance near-infrared (NIR) spectrometer was evaluated as a noninvasive alternative to breath and blood measurements for in vivo alcohol testing. A hybrid partial least squares (PLS) calibration was constructed using a combination of in vivo and in vitro spectral data. This model was subsequently evaluated for its performance in quantifying alcohol concentrations in vivo using a prospective validation study involving subjects who did not participate in the calibration. The validation study entailed induction of alcohol excursions in ten human subjects and comparison of the noninvasive NIR alcohol measurements to blood and breath alcohol measurements. Blood and breath alcohol measurements were performed at the time of each noninvasive NIR measurement (N = 372), establishing the noninvasive NIR measurement standard error relative to blood alcohol at 4.9 mg/dL (0.0049%). Assessment of the hybrid calibration model's sensitivity and selectivity provided strong evidence that the hybrid calibration yielded measurements that were both sensitive to alcohol and independent of other absorbing analytes in human tissue.