ABSTRACT
As mentioned in previous chapters, the labeled expiration dating period or shelf-life of a drug product is usually established based on the primary stability data obtained from long-term stability studies that are conducted under ambient conditions. For the determination of a labeled shelf-life, the FDA stability guidelines require that at least three batches, and preferably more, be tested to allow for a reliable estimate of batchto-batch variability and to test the hypothesis that a single expiration dating period for all batches is justifiable. In addition to individual shelf-lives estimated from each batch, it is desirable to establish a single shelf-life for a drug product. As indicated by the FDA stability guidelines, this single labeled expiration dating period should be applicable to all future batches. Before one can combine stability data from all batches, it is required by the FDA stability guidelines to perform preliminary tests for batch similarity. Batch similarity is usually evaluated by testing the equality of intercepts and the equality of slopes of degradation lines among different batches. For testing the hypotheses of the equality of intercepts and the equality of slopes among batches, the FDA stability guidelines suggest the 0.25 level of significance be used. If the hypotheses of equal intercepts and equal slopes are not rejected at the 0.25 level of significance, a single expiration dating period can be estimated using the methods described in Chapter 3 by fitting a single degradation curve based on pooled stability data of all batches under the assumption that batch effects are fixed. If the hypotheses of equal intercepts and equal slopes are rejected at the 0.25 level of significance, the FDA suggests that a single expiration dating period of the drug product be determined based on the minimum of the individual shelf-lives obtained from each batches. This method, however, lacks statistical justification (Chow and Shao, 1991). Under the assumption of fixed batch effects, as an alternative, Ruberg and Hsu (1992) proposed a method for estimating an expiration dating period using multiple comparison techniques for pooling stability data with the worst batches.
