Updated: Nov 13
When an API batch is OOS, several processes get initiated in parallel.
We need to exclude the possibility of any lab errors (OK, this is the 1st step every time, but if it takes much time, it is worth to start with manufacturing investigation before concluding the QC investigation).
The manufacturing investigation should be able to determine the root cause generating the OOS batch.
But next to finding the root cause, it's also crucial to find a way to 'save' the batch, if possible. But how?
ICH Q7 gives us two scenarios:
Reprocessing (14.20): Introducing an intermediate or API, including one that does not conform to standards or specifications, back into the process and reprocessing by repeating [...] steps (e.g., distillation, filtration, chromatography, milling) that are part of the established manufacturing process is generally considered acceptable.
Reworking (glossary): Subjecting an intermediate or API that does not conform to standards or specifications to one or more processing steps that are different from the established manufacturing process to obtain acceptable quality intermediate or API (e.g., recrystallizing with a different solvent).
What is this 'established manufacturing process'? that's the process description submitted in your file (CEP, ASMF, etc.). It already feels the reworking is something heavier, isn't it? In both scenarios, you have to evaluate the same topics, but with reworking everything is going to be a bit more difficult.
After either reprocessing or reworking, the impurity profile should be evaluated (the best way is to do it through smaller-scale batches). Not only the 'usual' (LC) impurities, but the residual solvents, water content, elemental impurities, etc. Don't forget that your product specification is built up to be suitable for your 'established manufacturing process', so you have to be prepared to understand what will the product be like. In case of reworking, e.g., with using a new solvent, you may need to develop and validate new analytical methods, as well.
Since reprocessing and reworking is only applied when you have an OOS batch*, and hopefully you don't have that many OOS cases, the usual PPQ-like validation is out of the picture. However, you want to have a detailed evaluation on your reprocessing/reworking process. What's the solution: in case of rework, ICH Q7 spells it out for us (14.31): concurrent validation. You write your validation plan for that one OOS batch, and you compile the report immediately you have all necessary data. I personally would (and used) the same approach for the reprossings steps, as well.
Yes, this is where it gets exciting: how can you make sure the reprocessed/reworked API behaves like the 'usual' batches in terms of retest/expiration date?
There are two components to answer with to this question.
Knowledge about the API: if you had reprocessings/reworkings in the past, use those data to establish your assessment on the product's stability
Actal stability study: if you have some doubts in your heart that something can go south with the batch, you can just go ahead and initiate a stability study.
Since it's not going to be a common ICH Q1-proof study, you can do whatever fits your purpose: either a single accelarated study for 3 months to quickly check the equivalence of the stability trend against the usual stability trend, or a 'full' primary study for 6 months - again, whatever that makes you sure you can release the batch with a proper retest/expiration date.
However, let's not forget ICH Q7 doesn't require stability studies in case of all reprocessed/reworked batches. Even in case of reworking, it says (14.31): 'Batches that have been reworked should be subjected to [...] stability testing if warranted'.
ICH Q1F emphasizes the same(2.2.13): 'In certain situations additional batches should be included in the ongoing stability programme. [...] Any reworking, reprocessing or recovery operation should also be considered for inclusion.'
There may be several different approaches to make this clear in front of the authorities, however, I'm a Q guy, I wouldn't dare to suggest any scenarios on those. If I get any comments how to best solve these cases through the CEPs/ASMFs, I'm open to extend this post :)
*. as a great colleague and friend of mine reminded me: there is another scenario where you may need reprocessing: the blending of batch tailings. Thx, Friend! :)