Statistical Assessment of Analytical Method Transfer∗

∗This article represents the authors' opinion only and not the official position of U.S. FDA.

Abstract

Analytical method transfer is an important part of analytical method development and maintenance. The current common practice of analytical method transfer is based on the equivalence of the means between the original laboratory and the receiving laboratory. However, for some assays the most scientifically sound approach would be to show the equivalency of individual sample readings between the two laboratories. In this study, statistical approaches such as limit of agreement, total deviation index, and tolerance interval approach to address individual equivalence between laboratories are discussed. Using a tolerance interval approach that is equivalent to the two one-sided hypothesis testing is proposed. These approaches are compared with each other and also with the mean equivalence approach on their statistical properties. Examples are presented to illustrate each analysis approach and the comparisons.

Key Words:

• Analytical method transfer
• Concordance correlation coefficient
• Double one-sided test
• Tolerance interval

ACKNOWLEDGMENTS

This article was written with most of the materials presented at 2007 FDA Visitation Day of PhRMA CMC Statistical Expert Team, 2007 ICSA Applied Statistical Symposium, 2007 Joint Statistical Meetings and 2007 FDA/Industry Statistical Workshop. The authors thank the organizers, Drs. Timothy Schonfield, David Christopher, James R. Schwenke, and Stella Machado, for their encouragement and constructive discussions. The authors also thank Dr. Meiyu Shen of FDA for the many discussions on tolerance interval. The article is part of the project on applications of tolerance interval in pharmaceutical quality control and assurance supported by the 2007 CDER/FDA Regulatory Science Research Fund RSR #07-08.

Notes

∗This article represents the authors' opinion only and not the official position of U.S. FDA.