A modified error model for the assessment of stability of pharmaceutical products

The purpose of the stability study of pharmaceutical products is to establish the shelf life based on the change of the quality with time. For the statistical evaluation, the ICH Guideline recommends the ANCOVA model applying the batch factor as fixed effect and the degradation rate as covariate. That approach leaves the interval-to-interval variability of an analytical method out of consideration; however, generally, the reproducibility is the most significant source of the fluctuation of the pharmaceutical stability data. literature, Norwood proposes an extended model which includes also the factor of the date of analysis. This additional factor is random and defined within batches, namely, each batch–time combination is a different level of the date factor. We criticize this nested parameterization because it does not utilize the information on the identity of the analytical circumstances at a certain date point. To solve this problem, we propose a two-way crossed design between the batch and the date factor. In our model, two time scales are used, one for storage period and the other one for date of analysis. The storage time belongs to the degradation rate and the date scale joins the random effect attributed to the interval-to-interval variability. The two time scales coincide only in case when the batches are put to storage at the same time. paper, an example is also given illustrating the difference among the three models.