Precision from drug stability studies: Investigation of reliable repeatability and intermediate precision of HPLC assay procedures

Abstract: A multi-company investigation is presented to obtain and compare precision results for LC assay procedures. Forty-four drug substances and drug products of various types subjected to 156 stability studies, with 2915 assay values in total, were included. This provides an excellent source of real long-term precision estimates, as the same analytical procedure was applied during the whole stability study, extending from 12 to 60 months. Intermediate precision was calculated either using the residual standard deviation of the regression line or applying an analysis of variances, depending on whether there was a significant degradation of the analyte or not. The results show impressively the large intervals where the individually calculated parameters scatter. Distribution ranges and averages for repeatability, intermediate precision, and the ratio between the two precision levels are mainly dependent on the type of drug product. Repeatabilities were found up to 0.8% for solutions, 1.6% for drug substances, 1.9% for tablets, 2.3% for creams, and 3.4% for a bath. For intermediate precision, which includes additional variability factors due to the reference standard, operator, equipment, reagents, etc., a similar dependency was obtained with a slightly changed order: up to 1.1% for drug substances, 2.2% for solutions, 2.3% for tablets, 3.1% for creams, and 3.2% for a bath. The ratio between the precision levels is up to 2.5 and similar for all investigated drug product types, apart from solutions with up to 5.3. These differences for the types of drug product may be explained by the influence of the sample and/or the sample preparation: the more complex, the higher the variability contribution. For the investigated examples, the impact of the analyte and of the concentration (dosage) seems to be of less importance. Therefore, a classification of drug product types for orientation on acceptable precision (ranges) for LC assay seems to be possible. [Copyright &y& Elsevier]