How to respond to 483 validation finding we disagree with?

yodon

Leader
Super Moderator
Ok, made me go back and look. The regs say:

(a) Where the results of a process cannot be fully verified by subsequent inspection and test, the process shall be validated with a high degree of assurance and approved according to established procedures. The validation activities and results, including the date and signature of the individual(s) approving the validation and where appropriate the major equipment validated, shall be documented.

I guess "cannot" and "are not" are arguable here. My experience is that inspectors interpret that as "are not." Clearly, I had my blinders on. :)

There is an appeals process - I believe it's outlined in the 483. I think I would still do a nominal validation (documented) and not try to fight it.
 

Ronen E

Problem Solver
Moderator
Don, I'm with you on this. I think you're not wrong.

Regardless of any one person's take on the legal language, the FDA has followed the approach you described from time immemorial - it's not an individual investigator's take. If a process requires assurance at its exit point (and this needs to be clearly established), they expect either a 100% verification or a process validation, and it's been spelled out abundantly in official guidance or otherwise over the years. From a pedantic perspective "can not" and "are not" are obviously not the same; however, from a practical QA perspective, if something can theoretically be done (and provide substantial value), but in fact ISN'T being done, it can't be used as a solid base for a waiver.
 

Ronen E

Problem Solver
Moderator
We don't 100% inspect all dimensions on every single part
Verification (in this context) means inspecting every single part. Yes. In some cases (not often) you might be able to get away with something close to 100%, but there would have to be a very solid reasoning to it and a robust statistical justification on record, and you'd still have to have the FDA convinced, because it would be the exception. Please also note that even then the percentage inspected would, most likely, need to be in the 90s (probably high 90s).

Having said that, there's no requirement to inspect "all dimensions" (BTW, there are more properties to a work piece than dimensions). It's up to you to establish, through engineering analysis and risk management, which part properties / outcomes of the production step (whether intentional or not) are significant to part/article quality, and of these which pose a significant risk (and have all this documented clearly). These are the only properties of ongoing concern and the ones that need to be verified (i.e. inspected 100%), or - if possible and desirable - validated, so that you could omit verification. Properties that are not of ongoing concern are irrelevant to the verification vs. process validation debate.
 

Ronen E

Problem Solver
Moderator
so you machine a shaft and you can easily measure the diameter and the length of the shaft.
It could or could not be as simple as that. In this example, suppose that the turning process is not optimized and thus the surface or subsurface properties of the shaft are deteriorated. This, in turn, could be significant or not - depending on the application, the requirements, the magnitude of the deterioration etc. That's why an engineering analysis and at least a basic risk analysis are required (typically an early activity in every process validation, but never mind that - it wasn't yet established that a process validation is necessary in that example). Once you know what's important, you ask whether it can be measured/inspected etc., and whether doing so is practical and desirable. It's also a question of economics - if you're only making 10 units a year it would normally make more sense to just verify them and be done with it.
 

Tidge

Trusted Information Resource
I am aligned with @yodon and @Ronen E on this one. Comments 41 and 43 of the Preamble to 21 CFR 820 speak directly to this, and comment 146 makes some reinforcement (with clarity and not wiggle room, IMO).

Personally: I would not at all be surprised to find that a similar machining process could be validated such that the as-implemented inspection routine provides reasonable (but known!) confidence that the outputs are conforming. There are a few slight subtleties(*1) in such validations but my experience with 'QMS auditors' has been that they are rather 'binary' in their thinking about "validated or not validated" (mostly by necessity, but I don't rule out mathematical discomfort) that I would be sure to speak to, even if others might see it as practically trivial.

(*1) An example of a 'slight' subtlety that immediately springs to mind is this: It is common for machinists to have the attitude that "I look at the parts to convince myself there is no problem with my machine: the 1st tells me if I set it up correctly, the last tells me that nothing changed... the middle is just an early chance to catch a problem with the machine so I don't waste a whole batch of material." There is nothing wrong with such a practical approach, but it is explicitly worrying about the setup of the machine and NOT the outputs. If the process has been validated, and these inspections are checking on the key-process indicators of the validated process, then it is very likely that such inspections are well-motivated.

The last point in my footnote will be familiar with those who work with injection molded components ( and who also care to guarantee conforming products). Once the tooling and process has been validated, there is typically a small number of inspection points of the manufacturing outputs that give confidence that the process remains in its validated state.
 

mrsrobinson1110

Starting to get Involved
Verification (in this context) means inspecting every single part. Yes. In some cases (not often) you might be able to get away with something close to 100%, but there would have to be a very solid reasoning to it and a robust statistical justification on record, and you'd still have to have the FDA convinced, because it would be the exception. Please also note that even then the percentage inspected would, most likely, need to be in the 90s (probably high 90s).

Having said that, there's no requirement to inspect "all dimensions" (BTW, there are more properties to a work piece than dimensions). It's up to you to establish, through engineering analysis and risk management, which part properties / outcomes of the production step (whether intentional or not) are significant to part/article quality, and of these which pose a significant risk (and have all this documented clearly). These are the only properties of ongoing concern and the ones that need to be verified (i.e. inspected 100%), or - if possible and desirable - validated, so that you could omit verification. Properties that are not of ongoing concern are irrelevant to the verification vs. process validation debate.

So, basically, we should start inspecting 100% of the critical dimensions on all parts for the machining processes that need to be validated, and need to show risk assessment that outlines what "critical" is?
 

mrsrobinson1110

Starting to get Involved
I am aligned with @yodon and @Ronen E on this one. Comments 41 and 43 of the Preamble to 21 CFR 820 speak directly to this, and comment 146 makes some reinforcement (with clarity and not wiggle room, IMO).

Personally: I would not at all be surprised to find that a similar machining process could be validated such that the as-implemented inspection routine provides reasonable (but known!) confidence that the outputs are conforming. There are a few slight subtleties(*1) in such validations but my experience with 'QMS auditors' has been that they are rather 'binary' in their thinking about "validated or not validated" (mostly by necessity, but I don't rule out mathematical discomfort) that I would be sure to speak to, even if others might see it as practically trivial.

(*1) An example of a 'slight' subtlety that immediately springs to mind is this: It is common for machinists to have the attitude that "I look at the parts to convince myself there is no problem with my machine: the 1st tells me if I set it up correctly, the last tells me that nothing changed... the middle is just an early chance to catch a problem with the machine so I don't waste a whole batch of material." There is nothing wrong with such a practical approach, but it is explicitly worrying about the setup of the machine and NOT the outputs. If the process has been validated, and these inspections are checking on the key-process indicators of the validated process, then it is very likely that such inspections are well-motivated.

The last point in my footnote will be familiar with those who work with injection molded components ( and who also care to guarantee conforming products). Once the tooling and process has been validated, there is typically a small number of inspection points of the manufacturing outputs that give confidence that the process remains in its validated state.

We currently do a C=0 sampling for everything in process, maybe we could write up a document for validating the machining processes with a risk assessment and then pointing to the years-long records of in-process sampling reports to show we've checked the outputs on this process on various applications and they are in tolerance.....
 

yodon

Leader
Super Moderator
We currently do a C=0 sampling for everything in process ... and then pointing to the years-long records of in-process sampling reports

This is not a validation, even with historical context and I wouldn't expect that proposing this would close your 483. Is there a particular reason you're opposed to doing a validation? Given the risk controls, I think it would be a pretty simple validation.
 

mrsrobinson1110

Starting to get Involved
This is not a validation, even with historical context and I wouldn't expect that proposing this would close your 483. Is there a particular reason you're opposed to doing a validation? Given the risk controls, I think it would be a pretty simple validation.

If checking dimensions of parts coming from that machining process is not validation, then what would validate it? We've done equipment validation, so other than inspecting parts, what would you check?
 
Top Bottom