Process Validation — Stage 2 (PPQ)
Confidence/Reliability Sample Size Calculator
Calculate the statistically justified minimum number of validation runs (e.g., PPQ batches) needed to demonstrate a target reliability at a given confidence level, assuming zero failures.
Beyond the "3 Batches" Default
Regulatory guidance does not mandate a fixed number of PPQ batches — three is a common industry default, but a scientifically defensible number should be tied to the reliability and confidence you want to demonstrate. The confidence/reliability (C/R) method provides one statistically grounded way to justify a minimum sample size when assuming zero failures are observed.
This is the same logic behind commonly cited "95/95" sampling — 95% confidence that the process meets at least 95% reliability — assuming all sampled runs pass.
Common Confidence/Reliability Combinations
| Confidence | Reliability | Minimum n (Zero Failures) |
|---|
Interactive Calculator
Important Caveats
- This method assumes zero failures across all n runs — a single failure invalidates the assumption and requires either root cause resolution and a fresh run count, or a different statistical model (e.g., binomial with allowed failures).
- Higher confidence/reliability targets increase n significantly — 95/95 might require far more runs than intuition suggests for high reliability targets.
- This is one of several acceptable approaches to sample size justification; process complexity, risk assessment (Stage 1), and prior knowledge should also inform the final PPQ batch count.
This calculator illustrates the statistical logic behind confidence/reliability-based sample sizing for educational and planning purposes. The final number of PPQ (or other validation) batches must be documented and justified in the approved validation protocol, considering process risk, complexity, and applicable regulatory expectations.
No comments:
Post a Comment