Applying spectral peak area analysis in near-infrared spectroscopy moisture assays

Spectral peak area analysis has in this study been shown to be a viable method in near-infrared spectroscopy (NIRS) moisture assays. The study also shows that the required number of calibration samples can be minimized, and the method is, therefore, especially suitable for moisture assays in early formulation development and in-situ process monitoring. Diffuse NIRS was utilized in the development of moisture assays for the model compounds polyvinylpyrrolidone and hydroxypropyl-beta-cyclodextrin and also for a lyophilized formulation. Reference data were obtained using coulometric Karl Fischer titration. The NIRS measurements were performed through the bottoms of the sample vials using either a Fourier Transform-Near-Infrared (FT-NIR) spectrometer fitted with a diffuse reflectance probe or a dispersive single beam spectrometer. The ratios of the peak areas of a water peak at 5200 cm(-1) and a reference peak were evaluated using linear regression analysis. The spectral peak area analysis method was compared with a conventional partial least squares regression method. The moisture assays were verified using independent test sets. The investigated moisture range was 0-22% for the samples of PVP, 0-8.5% for the samples of hydroxypropyl-beta-cyclodextrin and 0.5-8.5% for the samples of the lyophilized formulation. The results of the spectral peak area analysis and the conventional partial least squares regression were similar, but the peak area method was more robust and could also make accurate predictions for lyophilized PVP samples, although the calibration set consisted of non-lyophilized samples. The peak area method required fewer calibration samples than the conventional partial least squares regression method.