Process / Signoff / Orphan Data?

Hi, I was wondering if some one can share experiences with FDA related to unprocess chromatograms.  Are they expect everything processed and/or signoff?  1. For example a sample set that failed system suitability, 2. An instrument failure that stops the sequence, 3. A 3D channel acquired by error (I will expect that for #3 they may accept that channels were locked having an appropriate explanation).

Thanks in advance!

Comments

  • I don't know how exactly FDA reacts to such events but the policy at our organisation is the following (we've been FDA review quite a few times):
    1. Process & sign off. Comment about the failed SST criteria. Mentioned in the lab notebook
    2. Process & sign off. Comment about the failed SST criteria. Mentioned in the lab notebook
    3. If this was acquired additional (2D and 3D) we would leave it as is. If it's 3D instead of 2D we'd extract the channel and that would be that. No exact mention needed.
  • All depends on the company to be honest. Smart companies have a written SOP as to how to deal with data from acquisition to processing and reviewing results etc. My company instructs us to always lock channels from unprocessed data or from original sample sets that had to be rejected because of some error. You can usually add a detailed Default String with a relevant reason when rejecting results and locking channels. 
    Also, some companies only require you do trial or test injections in a specialized project to keep them all together. Which is a good idea as long as methods for all instruments are contained within!
  • I also can't speak for the FDA, but our policies are similar to DavidHPLC:

    1) We process to the point of failure, demonstrating any numeric criteria failure. We only require signoff on results that are valid/reported, so if it failed such that there are no valid results, we don't bother with a signoff.  We just document it in our lab notebook and start re-injecting then.

    2) Same as #1 basically...it all depends on if a sample bracket was completed successfully.

    3) This seems very situational...just one random sample injection vs the entire analysis, a single injection where both were collected vs a duplicate injection where the 3D was inadvertently collected.  For any regulatory-submission level work, we'd potentially go so far as to issue a non-conformance and either justify acceptance or negation of the analysis...we do have some allowance if it is grossly wrong in that we could negate everything as an identified analyst error and re-perform the step/analysis correctly with simple documentation in the lab notebook that it was identified as incorrect and was therefore being treated differently.  If it was, however, a case where something was injected twice, then we'd likely either have to go the non-conformance route and/or report both values.
Sign In or Register to comment.