Proper sample size determination is critical for medical device validation. Too few samples may fail to detect problems, while too many wastes resources. FDA and ISO standards require statistical justification for sample sizes in validation protocols.
Under-powered Study:
Over-powered Study:
Calculate required sample sizes for validation studies and statistical tests
FDA typically requires 95% confidence for medical device validation
Expected success rate or pass rate (e.g., 95% for most device validations)
Acceptable precision (typical: ±5%)
For finite populations, enables finite population correction
95% Confidence Interval:
90.0% to 100.0%
Interpretation:
With 73 samples, you can estimate the true success rate within ±5.0% with 95% confidence.
Requirement: 99% of devices pass leakage current test
Settings: p=0.99, margin=±2%, confidence=95%
Result: n ≈ 96 devices
Requirement: 95% test cases pass
Settings: p=0.95, margin=±5%, confidence=95%
Result: n ≈ 73 test cases
Note: Sample size calculations should be documented in your validation protocol. Consult with a statistician for complex study designs. FDA guidance documents (e.g., "Statistical Guidance on Reporting Results from Studies Evaluating Diagnostic Tests") provide additional requirements.
Use when measuring binary outcomes (pass/fail, yes/no)
Examples:
Typical: 95% confidence, ±5% margin
Use when measuring continuous variables (measurements with units)
Examples:
Typical: 95% confidence, ±2σ margin
Use when comparing two groups or testing hypotheses
Examples:
Typical: 80-90% power, 95% confidence
FDA QMSR (ISO 13485 7.3.6) requires design validation with "actual or simulated use conditions" and demonstration that device meets user needs. Sample sizes must be "adequate" and statistically justified.
FDA guidance: "The number of samples tested should be sufficient to demonstrate that the device will consistently meet predetermined specifications under actual or simulated use conditions."
Clause 8.1.2 requires documented rationale for sample sizes used in validation. Statistical methods must be appropriate for the data type and study objectives.
ISO guidance: "When sampling is used as a means of acceptance, sample size shall be based on recognized statistical principles and appropriate for the intended use of the product."
Clause 7 recommends minimum 15 subjects per user group for usability validation, with higher numbers if use errors are observed or statistical analysis is required.
IEC guidance: "The sample size shall be sufficient to identify use errors and patterns of use errors."
Sterilization validation requires specific sample sizes based on sterility assurance level (SAL). Minimum 30 samples per cycle for initial qualification.
Standard requirement: "Three consecutive successful production runs with n≥30 per run for process validation."
The "n=30" rule is a rough guideline, not a statistical requirement. Actual sample size depends on variability, confidence level, and margin of error.
Sample size formulas assume specific distributions (e.g., normal). If data is skewed, you may need larger samples or non-parametric methods.
In clinical studies, plan for 10-20% dropout rate. Calculate sample size, then increase by expected dropout percentage.
Sample size must be determined BEFORE conducting the study. Post-hoc calculations are statistically invalid and rejected by FDA.
Margin of error relates to precision of estimate. Statistical power relates to detecting true effects. These are different concepts.
Document all assumptions (expected proportion, standard deviation, effect size) and their sources (pilot data, literature, engineering judgment).
Include in validation protocol:
Small preliminary studies help:
When uncertain:
Have a statistician review: