The United States Pharmacopeial Convention provides reference standard tablets for use in performance verification testing of dissolution Apparatus 1 and 2 (1–3). The performance verification test (PVT) is specified in United States Pharmacopeia (USP) General Chapter Dissolution ‹711› (4) with specific elements provided by USP in its web-based Dissolution Toolkit (5). In the PVT, USP sets acceptance criteria (tolerances) for each tablet tested based on a collaborative study conducted for each lot of Reference Standard (RS). For the assembly to pass, each tablet must pass. The overall approach speaks to a horizontal standard (i.e., the dissolution procedure) and the various ways in which such a procedure is verified based on sound metrologic principles (2, 6).
MethodsReanalyses. This paper summarizes data from four of five published studies (2, 3, 12, and 13) and provides additional detail about a fifth study that was reported only briefly (15). To enhance comparability of results across the studies, data were reanalyzed in a consistent manner. As a consequence, some data reported in this article differ slightly from the data originally reported in References 3 and 14. The method of analysis was first to transform the data to the natural log scale because previous work has found that the normality assumption is better satisfied in this scale. Variances, S 2 , are determined from the log-scaled data and then transformed back to the original scale by the log-normal formula,
where %CV equals percent coefficient of variation.
Because this paper focuses on the contribution of the prednisone tablets to variability, the authors are reporting the lowest-level variance available for each study. That is, we are reporting the variance calculated removing whatever other sources of variance (e.g., analyst and assembly) that can be removed depending on the design of the study. The goal is to obtain a variance that includes the variability due to the tablet but with as little else as possible. When the variance is determined by an analysis of variance (ANOVA), the lowest-level variance is that usually termed the residual variance. When the variance is determined directly from the log-scale data, the variance is first determined within each experiment of six tablets and then pooled over experiments in the study. The magnitude of each variance depends on sources of variability—such as the assay—other than the tablets. These sources differ by study and are described for each study.