A Unified Framework for Detecting Out-of-Trend Results in Stability Studies. (2nd October 2018)

• Record Type: Journal Article
• Authors: Yu, Binbing; Zeng, Lingmin; Ren, Pin; Yang, Harry
• Is Part Of: Statistics in biopharmaceutical research. Volume 10:Number 4(2018)
• Page Start: 237
• Page End: 243
• Publication Date: 2018-10-02
• Subjects: Alert limits -- Mixed-effects model -- Out-of-trend (OOT) -- Specification range -- Stability data; Pharmacy -- Statistical methods -- Periodicals; Pharmaceutical biotechnology -- Statistical methods -- Periodicals; Biopharmaceutics -- Periodicals; Biometry -- Periodicals; Pharmacy -- Statistical methods; Periodicals; 615.190727
• DOI: 10.1080/19466315.2017.1371070
• Languages: English
• ISSNs: 1946-6315

Abstract

ABSTRACT: Stability studies are required for the registration of all pharmaceutical products to ensure the quality, safety, and efficacy of the products throughout the shelf life. Moreover, stability data are used to evaluate the change of critical quality attributes and to assess batch-to-batch consistency for quality monitoring. The goal of the trend analysis of stability data is to identify atypical trends and to assess the potential impact of abnormal analytical results on product quality. The out-of-trend (OOT) results or data series may be within the specification limits, but do not follow the expected trend. The OOT results may have a significant business and regulatory impact. The PhRMA CMC Statistics and Stability Expert Team has reviewed several important aspects of OOT stability data and discussed three types of OOT results related to analytical testing and product quality. The team proposed various statistical approaches for calculating alert limits for OOT identification. Currently, separate statistical methods are used for the three types of OOT results, and the methods are disconnected. In addition, certain OOT results may be missed and some within-trend results may be incorrectly identified as OOT, due to the limitation of the current methods. Based on the mixed-effects model, we propose a unified framework for deriving the alert limits for OOT detection. The proposed method takes both the batch-to-batch and within-batch variabilities into account and is flexible to detect all three types of OOT results.