“all sources of variability”
FDA recently published a GMP Warning Letter issued January 15th, mainly focused on “failure to investigate all critical deviations”, among other issues. Warning Letters such as this give us an insight into current regulatory expectations for compliance with 211.192, which are important, as there is little guidance available regarding the expectation that manufacturing events be “thoroughly investigated”. What constitutes a “thorough” investigation in 2025? What does a good root cause analysis look like, and what tools should we use to determine product impact?
This warning letter gives us some strong insight. Although the remediation instructions for the company are not necessarily new, they are an excellent reminder, considering the “C” in CGMP always moves forward - never backward. Here are some takeaways to keep in mind, and may be useful to your organization as an internal maturity check:
Understanding process variability is key to demonstrating root causes.
The letter recommends that for issues where no determinate root cause has been identified during the failure investigation, the company perform a “thorough” review of potential production factors that could have led to the failure. This is no easy task, considering a review of documentation (e.g., batch record) is unlikely to identify an obvious cause. Once the low hanging fruit has all been picked, the next step requires identifying potential causes via “scientific” rationale (the hard part). Listing potential causes here based on opinion or best efforts will be considered insufficient, as it is not scientific. The scientific part here means speaking the language of QRM, but without an understanding of process risk and a knowledge management program to turn to, the team performing the investigation is stuck with nowhere to turn, the hard part is actually impossible!
Finding the root cause and therefore the product impact in these situations is unlikely to happen. The best the team can do is to:
Point to the most likely sources of the failure (via reference to a documented process & data risk assessment)
Evaluate worst case (patient) scenarios for each element
Make the conclusion, being transparent that this is not objective, but as objective as possible considering the circumstances and residual patient risk. Every decision to close an investigation involves some level of subjectivity. That’s OK, the regulators are simply requesting that the decision be as objective as possible, following good scientific practices
Decisions should be “data driven”.
Understanding “all sources of variability” via a process-level granular risk assessment and subsequent targeted data monitoring program (we might call this CPV) greatly enhances the ability of the team to increase the level of objectivity in the final impact statement, not to mention result in meaningful CAPAs, rather than the traditional expansion of work instructions/training and more best efforts at the floor level.
The letter goes on to describe some elements of this transformation, many of which have been outlined in this blog (hence the statement above that this is not necessarily anything new).
In summary, it might be wise to reflect on your current organization’s maturity regarding these two points. Are your investigation teams set up for success in 2025? Do they have a knowledge management program (see ICH Q8/9/10) to turn to when difficult situations arise? Solving the difficult problems requires a quality system transformation, this is clearly evident in the letter. Although you may not be able to transform an entire organization, can you transform your department or division? Start small.
Map out your workflow and identify “all sources of variability”.
Perform a risk qualitative risk assessment using the concepts of severity and vulnerability.
Trace them back to your current process.
Do you have the data you need to monitor variability, if yes - create a dashboard
Don’t get hung up with perfection, just do what you can do within your department, and before you know it you will have a division that serves as the model for the organization! If you are looking for a solution that will achieve full objectivity, I’m afraid I have bad news… Just do the best you can with what you have, and let the future be the judge.
Pete
