This is the third in a series of Memos on bioburden issues. It covers measuring bioburden during routine manufacture, and specifically what is expected of that data and how we might respond to that data. Routine monitoring may be every cleaning event, or it may be based on elapsed time (quarterly, for example) or based on number of cleaning events (every fifth cleaning process, for example).
While we might focus on the monitoring data as an indication of passing bioburden results, a better interpretation is as a measure of process consistency, including process drift. That is, we would like to collect data that provides an alert of possible drift before unacceptable bioburden results appear.
To be clear, I would certainly be glad to see bioburden data during routine monitoring meeting my original bioburden limit. However, I would be even happier to collect data showing no drift or change in the cleaning process. Assuming I am taking bioburden samples at the end of the cleaning process, I might always achieve my acceptance limits but the robustness of my cleaning process might “hide” any underlying process changes.
We will cover bioburden monitoring for two cases. One case is where my data is based on sampling the final portion of the last process rinse, such as in CIP cleaning. The other case is swab sampling, typically after completion of the cleaning process (including the drying step).
For both types of sampling, it should be realized that variation in the measured bioburden may be due to the cleaning process itself (such as process times, dirty hold time, temperature, and bioburden content of cleaning chemicals used) and the manufacturing process of the drug product (such as bioburden of raw materials, processing temperatures, and openness of the process). I would be negligent if I didn’t mention that a basic source of variability is the bioburden measurement itself, particularly if traditional plate count methods are used.
For process rinse sampling, as suggested previously just taking a sample of the final portion of the final process rinse may not provide relevant data as to possible changes in the cleaning related to bioburden. One way to address this (and “Yes”, it will involve more work) is to take a rinse sample of an earlier portion of that final process rinse. In other words, if I had data from my original validation protocols on bioburden at that earlier time, I could compare the data during routine manufacture at that same time point in the final process rinse to see if the results are comparable. If my cleaning process is robust (and more so if the final process rinse is hot), I might just find that in both cases (original protocol and routine manufacture monitoring) that all samples are less than 1 CFU. However, if the original protocol data was less than 1 CFU but the monitoring data were 10 CFU, I might do a simple investigation to see if that difference was significant. As long as the result for the final rinse was unacceptable, I might wait until I had some data point before I do a more extensive investigation to bring the rinse process into the previous state of control. Note that taking a rinse sample for the same sample port at different times might be problematic unless the external surfaces of the port are “sanitized” and the external surfaces of the port are well flushed before taking a sample.
For swab sampling, the situation is somewhat different. I cannot easily perform an earlier swab sample as a monitoring check (although conceivably it might involve interrupting the rinse step before completion and then without drying swabbing a worst case location – ugh!!). An approach analogous to what was offered for rinse sampling might be to collect one or more swab sample from locations that are easy to access. By “easy to access”, I mean a location that can be easily sampled without tank entry, or without the need for an extension pole to reach the location, or without the need for the sampler to be a contortionist to adequately sample the location. This selected location might be an easy to clean location which had already been sampled in the original protocol (so that I have some baseline results for comparison). If it were an easy to clean location where previous data was less than 10 CFU/swab, then any results above that baseline might be of concern that there has been a change in my cleaning process and might lead to an investigation including more extensive sampling.
Now some might look at the additional testing to detect changes or drift in your cleaning process and think it overkill, which it might be in many facilities. Collecting this additional data is also problematic because of the inherent variability of microbial measurements using traditional plate count methods. So if you have designed and implemented a robust cleaning process for preventing microbial contamination, your effort might be better placed elsewhere. At a minimum for microbial monitoring, however, it is prudent as you collect microbial monitoring sample in routine manufacture to trend results and have that trend data evaluated by QA.
Next month we’ll cover a final bioburden topic, which is the issue of objectionable organisms.
Copyright © 2025 by Cleaning Validation Technologies
