OOS procedure in pharma GMP labs

[aceto_81]

Hi,

we are currently seeking for a good Out Of Specification result procedure for a pharmaceutical lab which is GMP compliant.

The current proposal in case of an OOS:
analyse the same sample (bottle or flask) again for 5 times, together with a previous sample which has a correct result.
If not ok, analyse another sample (other bottle or flask) of the same batch number for 5 times, together with a previous sample which has a correct result, and also let another analyst prepare both samples.

This sounds complicated and a bit overkill:
Why would one analyse a sample 5 times, 3 times looks enough to verify an incorrect result?

What is your procedure?

Ace

[Johnny Rod]

Are you under FDA or EU regs? There is guidance from both areas on dealing with OOS results. Be warned, this is a regulatory hot topic and just re-analysing a few times isn't the answer, you'll get filled in for it. Basically you need to perform an investigation at different levels to determine if the error is simply an analysis error or a wider problem, and some of this is done BEFORE retesting, reanalysing ro resampling (which are all different).

[aceto_81]

EU regs, no FDA.

Do you have a link to such guidelines?
Hopefully there will be clear guidelines, most of the time there are suggestion, but no practical solution

Ace

[Johnny Rod]

MHRA in the UK:
http://www.mhra.gov.uk/Howweregulate/Me ... /index.htm

There is a fairly comprehensive flowchart under question 1. Re-analysis/retesting should be done at least 5 times to be statistically significant, 3 times is what people used to do but won't be accepted any more, you can't do meaningful stats on three results.

[KM-USA]

So one assay is fine if the results are in spec, but must be assayed five times after the investigation?

It figures.

[Johnny Rod]

Yeah I know, if it passes then no-one notices. I think it is assumed that everything normally works, methods are validated etc. and therefore you cna't accidentally get a false pass over and over. one way to look at it is, if ther was something going on with the method, sampling etc. then you'd get an OOS sooner or later and the following investigation would uncover the problem.

[Consumer Products Guy]

In our business, consumer products can contain mixed fragrances, have viscosity differences, stuff like that and be shipped. But when a product is cloudy or hazy, even imperceptibly, then all heck breaks loose.

[danko]

To aceto_81:

I think you are mixing up two different actions/procedures. Investigation (trouble-shooting) and actual retest.
The investigation should be based on logical and professional thinking – no particular number of samples are wrong of right. It’s the nature of the investigation that dictates the number of determinations.

The retest (statistical based procedure) is typically something like 5 or 6 individual determinations that include the natural variation that might occur in the lab. It could be different days, different HPLC systems, or just columns, different analysts etc.

Both of these cases are best placed in the hands of competent and experienced people, because they could end up in either releasing or rejection of a drug f. ex. And in such a case you need to be sure that the right decision was made.

[JMcK]

So one assay is fine if the results are in spec, but must be assayed five times after the investigation?

I think this follows the logic that you do not test quality into a product. The manufacturing process should be validated to ensure that the material will meet specificiations, and therefore an 'in spec' result is expected. Testing is performed to verify that the process is performing as expected. That's a real concern over an OOS. It's not only that you have a bad lot.
You've also had a breakdown in your process.

[krickos]

I think ECA also has posted a joint QC developed guideline for OOS. Next step is OOT/OOE I think.

[aceto_81]

krickos, I searched for the guideline from the ECA, but it seems not available for download.
But thanks for pointing me to the ECA

Ace

[krickos]

hmm. might be members only. Had membership before after attending a course but experied after a few years. ECA can be informative depending on area of intrest.