Hello good people of the world! This post is related to one of my favorite topics: data crunching! Specifically, using statistics to determine sampling and acceptance criteria for Process Performance Qualification (PPQ), as defined by the FDA in their 2011 guidance (found here). Along with a risk-based, scientific approach to validation, agencies are expecting more and more the use of statistics to rationalize GMP decisions. That means using methods such as Statistical Process Control (SPC): X-bar/R chart, I-MR chart, etc. and Process Capability (Cpk, Ppk, Six-Sigma, etc.). With PPQ, you are not going to be doing 100% testing in most cases, which is where the use of statistics comes in. How do you determine the sampling requirements? How do you know what the sample results mean to the entire batch. Statistics provides you with a specific amount of confidence that a given requirement has been met. For instance, for batch attribute with a defined Acceptable Quality Limit (AQL), you’ll want the sampling plan to demonstrate a, e.g. greater than 90% confidence that the non-conformance rate is less than the AQL. For statistical parameters like means and standard deviations, you’ll want to calculate, e.g. a Pkp of greater than 1.0. How do you use statistics in your validation program? Leave a comment below and please share this post with whomever you think would benefit.