Evaluation of infrared spectroscopy as a screening tool for serum analysis: Impact of the nature of samples included in the calibration set

Abstract The application of attenuated total reflectance Fourier transform infrared (ATR-FT-IR) spectroscopy to the determination of clinical parameters in serum using partial least squares (PLS) has been evaluated as a point-of-care diagnostic tool. In this study the effect of using an increased size of the calibration set and the influence of the origin of samples and their interyear variation on the prediction capability of the method were considered. PLS-ATR-FT-IR provides a green, fast and cheap point-of-care tool for the determination of total protein. Albumin, glucose, urea, HDL, LDL and total cholesterol were predicted with relative errors between 15 and 32%. The analytical prediction capability of models built from an increased number of samples, from 100 till 750, was evaluated with independent sample sets. The evolution of the relative root mean square standard error of prediction (RRMSEP) as a function of the number of samples employed for calibration was different among different analytes, being the prediction capability strongly dependent of the concentration level of each analyte in the sample. Two sets of 750 each were built in two successive years, including samples from different origins (primary care, pre-dialysis and hospital). Multivariate analysis of variance (MANOVA) and principal component analysis (PCA) evidenced the strong influence of the aforementioned factors on the sample spectra.