How would you interpret a process capability index calculated from PPQ data?

Why

The NIST engineering statistics handbook defines Cp as the specification width divided by six process standard deviations, and Cpk as the smaller of the distances from the mean to each limit divided by three standard deviations. Cpk therefore penalizes an off-center process. For a centered process, a Cp of 1.00 corresponds to about 0.27 percent outside limits and 1.33 to about 64 parts per million, assuming normal data.

The handbook warns that capability estimates are valid only if the sample is large enough, generally about 50 independent values, and that the indices assume normality. Those conditions are rarely met in PPQ. Three batches might yield many tablets for content uniformity, but those values are not independent across batches, and between-batch variation, often the largest component, is estimated from only three points.

So a Cpk of 1.5 from PPQ is encouraging but its confidence interval may be wide. Interpretation should also ask whether the process was stable during the sampling, since capability for an unstable process is not meaningful. The FDA's 2011 guidance supports using such statistics in PPQ and in Stage 3, and recommends heightened monitoring until variability estimates are reliable.

The useful habit is to report capability with its confidence interval and the data it rests on, and to revisit it as commercial batches accumulate.