Multivariate calibration for assays in clinical chemistry using attenuated total reflection infrared spectra of human blood plasma.

A spectral analysis of human blood plasma was undertaken by use of a Fourier-transform infrared spectrometer with a circular attenuated total reflection cell. The concentrations of total protein, glucose, triglycerides, total cholesterol, urea, and uric acid were measured by chemical or enzymatic methods. For these constituents the partial least-squares (PLS) algorithm was used for a multivariate calibration including the infrared fingerprint region of the plasma spectra. Best results were achieved for total protein with an average prediction error (PRESS based on cross validation) of 2.1 g/L; other PRESS results were for glucose 22 mg/dL, triglycerides 33 mg/dL, cholesterol 31 mg/dL, urea 4.4 mg/dL, and uric acid 1.6 mg/dL.