This study was a 28-laboratory collaborative study to determine the Apparatus 1 and Apparatus 2 PVT acceptance criteria for
USP Lot P Prednisone RS Tablets (3). Participants were given the then-current Lot O tablets for assembly suitability testing
and two sets of blinded tablets, one of which was Lot O and the other Lot P. Results reported here are for the two sets of
blinded tablets. Results from this study, conducted in 2005, are compared with those obtained for Lots N and O in a collaborative
study conducted in 2003.
Each laboratory was instructed to run two experiments of six tablets each. For each apparatus, two analysts were told to perform
experiments using different equipment. After dropping experiments as outliers for failure to follow protocol or failure of
assembly suitability (see Reference 3 for explanations), 38 experiments for Lot P using Apparatus 1 and 40 using Apparatus
2 were available for statistical analysis.
These data were analyzed in a random-effects ANOVA similar to that for Studies 1 and 2. Random effects were laboratory, analyst,
and experiment within analyst. The %CVs reported are from the residual variance from the ANOVAs. These intra-experiment %CVs
include contributions from, at least, the tablets tested, variability among positions within the assemblies, assay, and position
of the tablet when dropped in the vessel. The data shown in Figure 1 are based on pooling the intra-experiment results for
each laboratory, as done for Study 1.
Figure 1. Lot P intra-experiment %CVs from Study 5.
Intra-experiment variability results from Study 1 are shown in Table I. For the USP Prednisone RS Tablets, the %CVs were consistently
at or below 5% and had a combined value of 3.6%. The %CVs for the two commercially available tablets were much higher (ranitidine
hydrochloride = 13.9% and metformin = 12.8%). Results from Study 2 are shown in Table II. Variability differs by assembly
and shows fairly consistent results across site and analyst. Results from Study 3 are shown in Table III. In this study, again
variability differed by assembly, even though assemblies alpha, gamma, and epsilon of Study 1 were the same as in Study 3.
By switching the vessels used in assembly alpha with those in assembly gamma, analysts found that the high variability in
assembly alpha was due to the vessels (2). Of note, the authors did see some low %CV for Apparatus 2 corresponding to those
of Table I. Eight of 24 assembly–analyst combinations (before switching vessels) had a %CV <6% for Apparatus 2 (data not shown).
Study IV's intra-experiment CVs for the vessels from the three manufacturers are shown in Table IV. Results vary considerably
by vessel manufacturer.
Table I: Intra-experiment variability of USP Lot P Prednisone and commercial tablets (Apparatus 2, Study 1).
Table V shows the results from Study V for the intra-experiment %CV combined across all participating laboratories. Results
remain fairly consistent across the three lots. Table V does not show the variability among laboratories in the %CV. This
is shown in Figure 1 for Lot P. In Figure 1, the median values are 8.0% for Apparatus 1 and 6.5% for Apparatus 2. That is,
half the laboratories had an intra-experiment %CV with Apparatus 2 of no more than 6.5%. The larger values of Table V are
a result of the inclusion of the results from the laboratories with the larger values at the top of Figure 1, many of which
are greater than 10% and thus are considered highly variable (16).
Table II: Intra-experiment %CV (Apparatus 2, Study 2).