Towards a robust water content determination of freeze-dried samples by near-infrared spectroscopy Original Research Article

The possibility for determination of the water content in pharmaceutical samples by near-infrared (NIR) spectroscopy has been more widely investigated in the past few years. However, many studies claim that changes in sample composition will require the establishment of a new method. The aim of this study was several fold: firstly to investigate validation aspects of water content determination in samples with varying composition and furthermore to see if a model based solely on freeze-dried mannitol–sucrose mixtures can be established that will be able to predict water contents for samples containing proteins, excipients or having a lower density of freeze-dried solids. Samples were measured by NIR, standard normal variate (SNV) corrected and the obtained spectra were compared with the results from a conventional Karl–Fischer titration by means of multivariate analysis, namely principal component analysis (PCA) and partial least square regression (PLS). For the overall sample set, a highly linear correlation between the NIR and the Karl–Fischer method with a slope of 1.00, an R2 value of 0.98 and a root mean square error of cross-validation (RMSECV) of 0.15% were found. In a second step samples solely consisting of mannitol and sucrose mixtures were used to build a calibration set, which resulted in a RMSECV of 0.16%. The prediction of the remaining samples, which included protein or excipient containing samples, as well as lower density samples, resulted in a root mean square error of prediction (RMSEP) of 0.19%. Thus the present study demonstrated, that a general model for the determination of the water content by NIR could be established, within the limits investigated.