Abstract
Recent experience has shown that the regulatory authorities try to narrow specifications for a new drug product closely around stability data that have been generated in the course of registration stability testing. This should ensure that the quality of the marketed product is produced within tight limits, very similar to the quality which is registered. A strategy is proposed to set scientifically based, statistically justified specifications. The long-term stability data available at the time point of registration application are extrapolated to a target shelf life (usually 36 months) of the new pharmaceutical product by linear regression. Batch-to-batch heterogeneity is tested as prescribed by the International Conference on Harmonization (ICH) guideline on stability testing of new drug substances and drug products. The data of different batches are combined in appropriate statistical models for further evaluation. The one-sided 99% confidence limit for individuals is used for the determination of release and shelf life specifications. Special attention is given to power calculations by which patients' risk of receiving material which does not fulfill the requirements can be controlled.