Residual moisture in lyophilizates is a critical quality attribute of a drug and has an impact on the long-term stability. Therefore, the moisture concentration needs to be verified. The Karl Fischer titration is a widely used and established method for that application. However, the method requires the sample to be destroyed in order to test and therefore does not allow a 100% inspection. NIR spectroscopy, on the other hand, is a non-destructive analytical technology process that allows real time testing at production speeds.

The chemometric NIR model contains specific algorithms and mathematical pre-treatment to perform the high speed analysis. The chemometric model is based on spectra of products with different residual humidity in the lyo cake. The model is created to achieve an NIR prediction for the water content of every product passing the sensor. The NIR prediction is used to identify all products which do not comply with the humidity specification range for automatic reject. The NIR spectra is analysed according to a hierarchical model (HM).

1. Identification of the product

The PCA qualitative analysis is used for the identification of different products. PCA shows the ability to discriminate between the different populations of the sample classes.

2. Moisture evaluation

The PLS quantitative analysis is used for the determination of water content in the cake. The correlation between spectral data and the corresponding water content values (Karl Fischer method as reference analysis) of the individual components is created using PLS algorithm.

3. Cake defect classification

Classification is the separation, or sorting, of a group of materials into one or more classes based on distinguishing characteristics in the materials for example, the separation between good and bad classes of the cake defect calculation.

During the method development process, the product specific spectra are analyzed and build the basis for the chemometric model. Each model is tailor-made to the individual product allowing the highest level of sensitivity. As a reference method the Karl Fisher method is used.