PT  - JOURNAL ARTICLE
AU  - Bar, Raphael
TI  - Statistical Evaluation of Stability Data: Criteria for Change-over-time and Data Variability
DP  - 2003 Sep 01
TA  - PDA Journal of Pharmaceutical Science and Technology
PG  - 369--377
VI  - 57
IP  - 5
4099  - http://journal.pda.org/content/57/5/369.short
4100  - http://journal.pda.org/content/57/5/369.full
SO  - PDA J Pharm Sci Technol2003 Sep 01; 57
AB  - In a recently issued ICH Q1E guidance on evaluation of stability data of drug substances and products, the need to perform a statistical extrapolation of a shelf-life of a drug product or a retest period for a drug substance is based heavily on whether data exhibit a change-over-time and/or variability. However, this document suggests neither measures nor acceptance criteria of these two parameters. This paper demonstrates a useful application of simple statistical parameters for determining whether sets of stability data from either accelerated or long-term storage programs exhibit a change-over-time and/or variability. These parameters are all derived from a simple linear regression analysis first performed on the stability data. The p-value of the slope of the regression line is taken as a measure for change-over-time, and a value of 0.25 is suggested as a limit to insignificant change of the quantitative stability attributes monitored. The minimal process capability index, Cpk, calculated from the standard deviation of the regression line, is suggested as a measure for variability with a value of 2.5 as a limit for an insignificant variability. The usefulness of the above two parameters, p-value and Cpk, was demonstrated on stability data of a refrigerated drug product and on pooled data of three batches of a drug substance. In both cases, the determined parameters allowed characterization of the data in terms of change-over-time and variability. Consequently, complete evaluation of the stability data could be pursued according to the ICH guidance. It is believed that the application of the above two parameters with their acceptance criteria will allow a more unified evaluation of stability data.