Page 203 - Safety Risk Management for Medical Devices

182 Safety Risk Management for Medical Devices

21.4 VARIABLE TESTING
Variable testing of safety requirements produces numerical results along some scale.
Example requirement: “Length of the connector shall be 3.00 6 0.25 mm.”
Using the same method that was described in Steps 1 4of Section 21.3, deter-
mine the Harm severity class for the requirement in question. Identify the chosen
confidence/reliability values. For variable test data, the statistical distribution of the
data matters. Do a normality test on a preliminary dataset, e.g., a confirmation run.
If the data distribution is normal, calculate the average and standard deviation of the
data. Using a statistical software such as Minitab iteratively compute the minimum
sample size needed to obtain a normal-distribution tolerance-interval that is narrow
enough to fit within the specification. In the example of the connector above, that
would be within 6 0.25 mm. If the computed sample size comes out to be very
small, you can increase it to a larger number, e.g., 10 or 15.
If your data is not normal, you can perform a conversion of the data to make it
normal, including the specification limits.
If your data indicates a different distribution, such as Weibull and Lognormal, still
find the minimum sample size such that the tolerance interval falls within the specifi-
cations. You may need to consult a statistician.
Another method that is used is to correlate the risk level to process capability. For
example:
For catastrophic risk class—CpK 2.0
For critical risk class—CpK 1.5
For serious risk class—CpK 1.0
For minor/negligible risk class—CpK 0.8