Glucose determination in simulated plasma solutions using infrared spectrophotometry

Primary blood constituents which may interfere with the infrared (IR) spectrophotometric measurement of glucose in plasma were identified. Blood plasma solutions were simulated by mixing glucose and the interfering constituents in their actual physiological concentrations in phosphate buffered saline (PBS). The feasibility of accurate prediction of physiological glucose concentration in a specially designed set of simulated plasma solutions with pH of 7.4, maintained at 37°C, was assessed by applying univariate and multivariate statistical and artificial neural network methods to their mid-IR spectra (1180–930 cm−1). Multivariate methods, Partial Least Squares, Principal Components Regression and artificial neural networks produced calibration models with small standard errors of prediction (SEP) (20.6, 20.6 and 19.8 mg/dl, respectively) compared with peak height and area based univariate methods (smallest SEP = 40.1 mg/dl). We conclude that it is feasible to predict physiological glucose concentration in multicomponent aqueous mixtures containing varying concentrations of interfering constituents, using multivariate methods.