Classification of lyophilised mixtures using multivariate analysis of NIR spectra

Excipient selection is critically affecting the processing and the stability of a lyophilised product. Near infra-red (NIR) spectroscopy was applied to investigate freeze-dried samples containing varying ratios of the commonly used excipients mannitol and sucrose. Further variation in the formulation was achieved by adding NaCl, CaCl2 and histidine and by exposing the samples to different conditions. Untreated NIR spectra are strongly affected by the physical nature of samples and can thus be useful for detecting production outliers. Applying standard normal variate (SNV) transformation highlights chemical information. The obtained NIR spectra of the freeze-dried samples were clustered by principal component analysis (PCA) after applying SNV correction in the range from 4200 to 7400 cm−1 (1350–2380 nm). Relative humidity under storage and the mannitol/sucrose ratio were clearly represented in the first two principal components, while influence of other excipients was observed in the 3rd and 4th principal component. It was investigated whether this could be due to an influence of the excipients on the mannitol crystallisation behavior. Performing PCA with two principal components of SNV-corrected spectra in the range 4200–4500 cm−1 (2220–1380 nm) led to the following observation: while the 1st principal component closely resembled the spectra of β-mannitol, the 2nd principal component contained additional features that were not attributable to β-mannitol but correlated well to the main absorbance band of δ-mannitol and mannitol hemihydrate. Therefore, it seems feasible that NIR can analyse versatile freeze-dried samples and classify these according to composition, water content and solid-state properties.