Combined statistical analyses for long-term stability data with multiple storage conditions : a simulation study

Research output: Contribution to journalArticleAcademicpeer-review

2 Citations (Scopus)

Abstract

Shelf-life estimation usually requires that at least three registration batches are tested for stability at multiple storage conditions. The shelf-life estimates are often obtained by linear regression analysis per storage condition, an approach implicitly suggested by ICH guideline Q1E. A linear regression analysis combining all data from multiple storage conditions was recently proposed in the literature when variances are homogeneous across storage conditions. The combined analysis is expected to perform better than the separate analysis per storage condition, since pooling data would lead to an improved estimate of the variation and higher numbers of degrees of freedom, but this is not evident for shelf-life estimation. Indeed, the two approaches treat the observed initial batch results, the intercepts in the model, and poolability of batches differently, which may eliminate or reduce the expected advantage of the combined approach with respect to the separate approach. Therefore, a simulation study was performed to compare the distribution of simulated shelf-life estimates on several characteristics between the two approaches and to quantify the difference in shelf-life estimates. In general, the combined statistical analysis does estimate the true shelf life more consistently and precisely than the analysis per storage condition, but it did not outperform the separate analysis in all circumstances. Keywords: Linear regression analysis, Long-term stability, Multifactor stability study, Multiple degradation profiles, Poolability, Shelf-life estimation
Original languageEnglish
Pages (from-to)493-506
JournalJournal of Biopharmaceutical Statistics
Volume24
Issue number3
DOIs
Publication statusPublished - 2014
Externally publishedYes

Fingerprint

Dive into the research topics of 'Combined statistical analyses for long-term stability data with multiple storage conditions : a simulation study'. Together they form a unique fingerprint.

Cite this