In-Process Inspection Raw Data

[irfank]

Hello,

We are setting up a vision system in-process to "pass/fail" units.
Is it enough that we capture the "pass/fail" result on the DHR, or do we need to keep the raw measurement data in the DHR as well.

Note: the vision system collects the raw measurement data and spits it into an excel file. We just don't know if by virtue of the data's existence it requires us to include it in the DHR?

 

[lgoryoka]

Hello,

We are setting up a vision system in-process to "pass/fail" units.
Is it enough that we capture the "pass/fail" result on the DHR, or do we need to keep the raw measurement data in the DHR as well.

Note: the vision system collects the raw measurement data and spits it into an excel file. We just don't know if by virtue of the data's existence it requires us to include it in the DHR?

1st, you should have the vision system validated. If you are ISO 13485 certified, the std requires that you validated the software and/or excel datasheet for the collection of QMS data. My recommendation is post validation; you should print and attach the excel copy for each work order to the DHR and "Pass/fail" units can be justified. Otherwise, FDA or NB auditors will challenge and will require to see objective evidence of the results.

 

[Steve Prevette]

From strictly a statistical perspective - if you only record pass/fail - you lose a lot of information. Let's say you end up detecting a big increase in failures. It would be helpful to know if these were just 'Close but no Cigar' failures, or are way out there (and on what specific measurement) and is the margin from the requirement increasing or decreasing. That information can help a lot when it comes to troubleshooting (and correcting) the problem.

 

[Tidge]

1st, you should have the vision system validated. If you are ISO 13485 certified, the std requires that you validated the software and/or excel datasheet for the collection of QMS data. My recommendation is post validation; you should print and attach the excel copy for each work order to the DHR and "Pass/fail" units can be justified. Otherwise, FDA or NB auditors will challenge and will require to see objective evidence of the results.

Fair warning: 13485:2016 doesn't have text that explicitly calls out the required validation of Excel datasheets. If the Excel software itself is being used as the record retention system, 4.2.5 "Control of Records" (and 7.5.6) may have something to say about how the greater system is used. I suppose the 'hook' is to 4.1.6, but just because data is recorded using a spreadsheet doesn't necessarily mean that the spreadsheet itself needs to be validated. I definitely recommend making a validation assessment, or minimally having procedural text that forgoes software validation for office applications used like a typewriter.

FDA auditors are extremely unlikely to ask to see validation of any software that falls under the "typewriter rule". This has been repeated (by the FDA) to participants in the CSA ("Computer Software Assurance", supplanting CSV="Computerized System Validation") task force. I am COMPLETELY aware of the emphasis placed by third parties (not to be confused with Notified Bodies or the FDA) to "take no chances" with software systems, and this created a cottage industry of no value-added work for at least a decade. We expect the FDA to finally release the new draft guidance this year, and I expect this part of it to remain (from the early drafts):

It is a common misconception that the use of computer software used in automation necessitates extensive documentation and testing to show that validation of non-product software was completed appropriately; however, this is not always accurate.

The (pre)draft FDA guidance makes it explicit that a critical application of risk management to all quality assurance activities ought to be the foundation of directed work; I don't expect representatives from NBs to see this differently.