Determination of Capsaicin in the Antirheumatical Plasters by near Infrared Reflectance Spectroscopy: A Comparison of Statistical Methods

Abstract

The evaluation of near infrared (NIR) reflectance spectroscopy as a method for the determination of capsaicin (8-methyl-N-vanillyl-6-nonenamide)—an active ingredient in the antirheumatical plasters was examined. The analytical procedure for determining the capsaicin was carried out by conventional, time-consuming colorimetric method. Spectra of the 76 plaster samples were recorded in reflectance mode at 2 nm intervals in the range 1100–2500 nm using InfraAlyzer 500 (Bran+Luebbe GmbH). A comparison is made between two regression methods, stepwise multiple linear regression (MLR) and partial least squares regression (PLS). MLR and PLS regression were used for calibrations, with the aid of the software SESAME ver. 2.10 (Bran+Luebbe GmbH). The PLS method showed consistently lower standard error of calibration and higher R values with first and second difference equations. The first difference PLS regression equation resulted in standard error of calibration of 0.018 %, with an R of 0.95. Generalizability of both methods for prediction of capsaicin contents on independent data sets is discussed. Prediction accuracy for independent data sets was increased using PLS regression, but was poor for sample sets with laboratory-measured concentration ranges beyond those of the calibration set. The results in this study indicate that NIR technique has a high applicability to quantitative analysis of capsaicin content in antirheumatical plasters.