, rmontes@alnylam.com, director of CMC Statistics, Process, and Analytical Sciences at Alnylam Pharmaceuticals, Cambridge, MA, has more than 20 years of experience in the bio/pharmaceutical industry. He received his PhD in Chemical Engineering (ChE) from the Georgia Institute of Technology, his MSChE from Texas Tech University, and his BSChE from St. Louis University.

Richard Montes

Submitted: Aug. 25, 2020. Accepted: Sept. 8, 2020

Trending stability data to identify out-of-trend (OOT) results is a critical part of avoiding out-of-specification (OOS) events for drug substances and products. Analytical OOT is usually caused by an invalid analytical measurement, while process control OOT is typically due to some production- related event. To date, there is still no regulatory guidance on which statistical techniques are best to use for identifying OOT conditions. 

Methods such as regression control chart (RegCC), by time tolerance interval (ByTimeTI), slope-by-lot Control Chart (SlopeCC), prediction interval (PI), and Z-score have been described in the literature. However, there had been no systematic evaluation of the effectiveness of each of these methods.

The author has used simulation to compare these methods under varying scenarios of sample size, relative lot-to-lot and within-lot variation, and extent of analytical or process control OOT. Results verified that PI, Z-score, and RegCC were related methods with some slight variation in how each one worked. ByTimeTI was the least effective in detecting either types of OOT. SlopeCC was effective in detecting process control OOT, but, overall, RegCC performed the best at detecting both analytical and process control OOT, although its false alarm rates (Type I error) slightly exceeded 5%. In addition, the method is relatively simple to use, and can be implemented by quality control staffers who have not had in-depth statistical training, or who cannot confer with a statistician.


Vol. 45, No. 4
April 2021
Pages: 34–43

When referring to this article, please cite it as R. Montes, “Comparing Methods for Determining Out-of-Trend Stability Test Results,” 

Articles

This article describes in detail how simulation was used to compare the statistical techniques used to determine out-of-trend (OOT), which is crucial to avoiding out-of-specification (OOS) events for drug substances and products. 