Sterile Medical Packaging and Shelf Life Validation Sample size

P

phloQS

Dear all,

hope anyone can give me good advise with this one:

We need a shelf-life validation for our new packaging system for sterile medical packaging. Previous designs have been validated on a sample base of 5 samples for each test. As most of the test (peel, dye, burst etc.) are destructive and prices for transportsimulation, accelerated aging and real time aging depend on volume a large sample size is very expensive. The testlab which we chose gave us the advise to test five samples again but they have no rationale on what they decide so. They only say it is common use. our daily production rate differs from day to day and we produce some very small batches of one product, so i can not find any samplesize based on batchsize. My question is: Does FDA accept shelf-life validaion reports based on the five samples or do they want to see larger sample sizes? As the project manager form our lab told us, they never had any problems with their reports but many customers ask them for the rationale.

Regards

phloQS
 

Statistical Steven

Statistician
Leader
Super Moderator
Re: Packaging validation Sample size

Dear all,

hope anyone can give me good advise with this one:

We need a shelf-life validation for our new packaging system for sterile medical packaging. Previous designs have been validated on a sample base of 5 samples for each test. As most of the test (peel, dye, burst etc.) are destructive and prices for transportsimulation, accelerated aging and real time aging depend on volume a large sample size is very expensive. The testlab which we chose gave us the advise to test five samples again but they have no rationale on what they decide so. They only say it is common use. our daily production rate differs from day to day and we produce some very small batches of one product, so i can not find any samplesize based on batchsize. My question is: Does FDA accept shelf-life validaion reports based on the five samples or do they want to see larger sample sizes? As the project manager form our lab told us, they never had any problems with their reports but many customers ask them for the rationale.

Regards

phloQS

Are you testing sterility? In other words, are you doing a sterilization validation on the packaging? If you have some additional risk mitigation tests, you can use a risk based rationale for a sample size of 5 (especially for continuous data).
 

pkost

Trusted Information Resource
I seem to recall warning letters issued by the FDA relating to inadequate justification of a sample size during design validation - I'm sure someone here will be able to dig one out for you.
 
L

lfrost

Here is a FDA Warning letter concerning the need for verification of processes that cannot be verified.

http://www.fda.gov/ICECI/EnforcementActions/WarningLetters/2011/ucm255098.htm

It is only one of several from the FDA Website. I searched the Warning letters for "sterile medical packaging" and the results that were returned, "90" this was one from 2011. I think you would be safe to assume that the FDA considers this to be a very important aspect of the QS Regulation (21CFR820).

:2cents:
 
L

Lucasmf

Whatever the sample size you select, they will expect to see some justifiable rationale.

In my experience a good approach is to require a high reliability level for your variable tests such a peel force or pouch burst and then allow a lower requirement for attribute tests especially if you combine several of them together to get a 'worst-case' test. Environmental, transit, plus accelerated aging is a common combination preconditioning tests that can then be used as a rationale for a smaller sample set.

I know large med companies that use n=29, c=0 sample sizes for packaging validation and I also know of small med companies who go as low as n=6, c=0 for packaging validation. But it is usually coupled with demonstrating 95%-95% pouch seal or pouch burst for sterile barrier validation. Peeling or bursting pouches is pretty cheap compared to the chamber time and everything else you'll need to do through your packaging test house.
 
R

Ron Boumans

In addition to Lucasmf:
If I see a company that has a clear rationale regarding sampling size, I don't look too much further into the exact details. This means that normally I would look at the intended use and the organisation of the distribution to see how this is translated into the risk analysis. Those risks need to be translated into worst case conditions for testing. Then we come to the bit where statistics usually blur the whole picture, but with some thumb of rule figures I can often see if it looks solid. In doubt I take those figures back to the office to have them checked properly. We may have some discussion about sample sizes and I try to be flexible and open minded. I presume the manufacturer knows very well what he is doing. But the bottom line is that he needs to be able to explain to me what he is doing. If not, he has a problem.

So as usual: go back to the basics to define what should be necessary. Use the standards only as a guide after you have established what you need to know.
 

Ronen E

Problem Solver
Moderator
Dear all,

hope anyone can give me good advise with this one:

We need a shelf-life validation for our new packaging system for sterile medical packaging. Previous designs have been validated on a sample base of 5 samples for each test. As most of the test (peel, dye, burst etc.) are destructive and prices for transportsimulation, accelerated aging and real time aging depend on volume a large sample size is very expensive. The testlab which we chose gave us the advise to test five samples again but they have no rationale on what they decide so. They only say it is common use. our daily production rate differs from day to day and we produce some very small batches of one product, so i can not find any samplesize based on batchsize. My question is: Does FDA accept shelf-life validaion reports based on the five samples or do they want to see larger sample sizes? As the project manager form our lab told us, they never had any problems with their reports but many customers ask them for the rationale.

Regards

phloQS

I find it hard to believe that the FDA will accept a sample size of 5 as adequate for statistical rigour, for any kind of test related to sterile barrier validation. the lowest I've heard of is 6-10 (depends on the specifics).

Cheers,
Ronen.
 
P

phloQS

First of all, thanks for all the responses. Maybe I described my validation approach a bit too short. We will do several test for seal integritiy as well as packaging integrity. The seal integrity is testted trough peel-test, dye-test and burst-test, each with 5 samples where the peeltest is done on each side of the blister so there are four samples for each blister, 20 in alltogether. It is NOT the sterility testing, just the test for packaging (and seal) integrity. We perfom all the described test in five stages: before sterilisation, after sterilisation, after transport simulation after accelerated aging and the last tests after real time aging. What we want to show is, that sterilisation and transport have no bad influence on the packaging. We already have studies for another system with the same study design. The results on each stage are very similar, so the influence of the sterilisation and the transport seem to be very low. The data overall looks very good. Seal strength is far above the limit suggested in standards (DIN EN 868-5:2009). So again from your point of view: would that studydesign be sufficient for an FDA inspection, although we only use five samples for each test (but in total there are 100 samples, twenty in each stage) and have no statisticsal rationale beyond this?

regards

phloQS
 
Last edited by a moderator:

Ronen E

Problem Solver
Moderator
First of all, thanks for all the responses. Maybe I described my validation approach a bit too short. We will do several test for seal integritiy as well as packaging integrity. The seal integrity is testted trough peel-test, dye-test and burst-test, each with 5 samples where the peeltest is done on each side of the blister so there are four samples for each blister, 20 in alltogether. It is NOT the sterility testing, just the test for packaging (and seal) integrity. We perfom all the described test in five stages: before sterilisation, after sterilisation, after transport simulation after accelerated aging and the last tests after real time aging. What we want to show is, that sterilisation and transport have no bad influence on the packaging. We already have studies for another system with the same study design. The results on each stage are very similar, so the influence of the sterilisation and the transport seem to be very low. The data overall looks very good. Seal strength is far above the limit suggested in standards (DIN EN 868-5:2009). So again from your point of view: would that studydesign be sufficient for an FDA inspection, although we only use five samples for each test (but in total there are 100 samples, twenty in each stage) and have no statisticsal rationale beyond this?

regards

phloQS

Hi,

A few remarks:

1. The only thing that statistically counts is repetitions, i.e. the number of test samples from the exact same configuration. The fact that you repeat the process at various stages (or repeat the sampling & testing at different pouch locations) does not add to the statistical rigour, unless you analyse them as one "population" (which will also increase the within-sample variation).

2. From all I've seen from the FDA, "no statistical rationale" is bad practice.

3. When I referred to "sterile barrier validation" I didn't only relate to the "narrow" aspect of sterility tests; I also related to what you call seal testing or seal validation. It has (IMO) the same level of criticality because it directly afects the end result - sterility of the device when put into use.

4. Seal strength (as demonstrated, for example, in peel tests) is important, but it's not the most important factor (understatement:)) when considering sterile barrier quality / reliability. You could have a perfect, consistent strong "seal" (e.g. weld) which is in fact not properly sealing, i.e. not a sterile barrier. I'd put more emphasis on burst / dye tests. BTW, both these last tests are not trivial in terms of establishing the test methods, analysis and pass/fail criteria.

5. Perhaps instead of sampling all 4 sides of the pouch for peel testing, consider getting validation data / certification from the original pouch manufacturer, for the originally sealed 3 sides, and redirect some of your resources into enlarging the sample size from the side you seal (or other tests you perform).

Cheers,
Ronen.
 
Last edited:
P

phloQS

@Ronen E: I know that I have to perform repititions for statistics. This is why I said "no statistical rationale". But all the data collected shows to me that our packaging system is valid. For the old system i talked about, the process control shows that our assumption for that system was right. We never had any problems or bad data for this process, which is very stable. We perfom dyetest and seal-strength-test (sample-based) and 100% visual inspection.
Our new packaging system is a blister with a sealed tyvek lid on top. So all four seals are made in our process. The sealing process itsself is validated (IQ,OQ with critical parameters, PQ with samples from three batches (n=5)), No test shows any significant variation (burst, dye, peel). The peel strength as the only continous data is very consistent. So this is the baseline for our shelf-life validation.

Regards

phloQS
 
Top Bottom