Sample size strategy for shelf life testing

vivianis7

Registered
Hi all,

My question concerns determining the sample size for medical device shelf-life testing.
My company is developing a neurovascular guidewire (class II) with expected shelf life of 1 year, for which we will age the devices in two stages: 12 months accelerated aging and 12 months real time aging.
This is the first device the company is developing, and as we are working on the production of the very first batch of wires we don't think AQL is applicable to determine sample size for design verification testing. For our verification strategy, we intend to use a confidence/reliability plan (using 29 samples for attribute testing).

When it comes to repeating part of the design verification tests on aged samples to evaluate the shelf life, we were advised to reduce our sample size from 29 to 3 or 5 samples, as when testing at T0 we are already building statistical confidence, but we would like to learn more about the feasibility of this strategy.

Any suggestion will be highly appreciated, thank you in advance!
 

planB

Super Moderator
Question ahead: is your aim to validate the stability of your sterile barrier system over your claimed shelf life, or the stability of the packaged device or both?
 

vivianis7

Registered
Question ahead: is your aim to validate the stability of your sterile barrier system over your claimed shelf life, or the stability of the packaged device or both?
Hi PlanB, thanks for the prompt feedback!

Currently our focus is on the final and sterilized device, as the appointed CMO for packaging and sterilization is suggesting leveraging data from packaging validation of an older device for evaluating the stability and integrity of the SBS.
But if you have any comment regarding shelf life of packaging, I'd be happy to also hear it!
 

planB

Super Moderator
If you really want to repeat all design verification tests on samples after aging without further considerations, you would take the very same sample size as chosen for T=0.

If you are in a position to tie in risk management and have no indication that you device is susceptible to degradation over time, you might be in a position to perform your tests in a confirmatory manner, consequently reduce your sample size and maybe even omit e.g. accelerated aging of products altogether.

HTH,
 

Steve Prevette

Deming Disciple
Leader
Super Moderator
When it comes to repeating part of the design verification tests on aged samples to evaluate the shelf life, we were advised to reduce our sample size from 29 to 3 or 5 samples, as when testing at T0 we are already building statistical confidence, but we would like to learn more about the feasibility of this strategy.

Any suggestion will be highly appreciated, thank you in advance!

29 is certainly rigorous and well-accepted. If you have 29 successful samples with no failures, you are 90% confident that no more than 10% are "defective".

If you have prior data, that can help, as long as the units there are representative of the final design. Depends on how many data values you had, and were there any failures. This can be done as a simple binomial (counting towards the 29) or using Bayesian calculations.

What is the definition of "degradation"? Can it be measured as a continuous variable? You can greatly reduce the sample size if you can shift from go-no go to some measurement. Perhaps a simple example is an indicator of failure is a high resistance value. If you can measure that, you can then analzye that data on a t-test and likely reduce the sample size, depending on the variability (standard deviation) of the data.
 

planB

Super Moderator
"Degradation" in the context of biocompatibility is the "decomposition of a material" (ISO 10993-9:2019, section 3.1). Typically, these chemical and physical material changes over time could be quite complex to be characterised/measured as continuous variables .
 

vivianis7

Registered
Thanks both of you for the help!

We are currently evaluating the option of performing all the tests on accelerated aged samples (Taa), and executing simulated use and PME at T=0 to assess baseline respondence to performance requirements.
Our rationale is that we want to confirm requirements are met at the end of product life, but still confirm that general performance was there all along. We expect time to affect materials in a negative manner, therefore we are testing in worse-case conditions (i.e. aged samples).
Does this sound familiar, or pose any red flag to you?

Thank you for any input!
 

planB

Super Moderator
... and which of these materials would you evaluate as being susceptible to degradation/decomposition under storage conditions?
 
Top Bottom