This Cleaning Memo focuses on the rationales for selecting appropriate swab sampling locations. It is generally accepted that the major concern is dealing with “worst-case” locations, realizing that locations may be considered a “worst case” for a variety of reasons. Furthermore, sometimes we select locations that are not “worst-cases”, but are merely representative of certain situations, like a “material of construction” (MOC) that is different in some way and perhaps not necessarily covered by sampling on the majority MOC (usually stainless steel). So first we’ll first consider what exactly a “worst-case” location is, focusing on swabbing locations.
As I us the term “worst case” for swabbing locations, I generally mean a location where, after completion of the cleaning process, I am more likely to find higher levels of measured residue residues on the surface at that location. This doesn’t mean that I expect to find unacceptable levels of residue at that location. Rather it generally means that if there were issues with the inadequate performance of the cleaning SOP, I would more likely find unacceptable levels of residue at that worst-case location as compared to other locations not considered “worst case”.
There are several ways a location can be identified as a worst case. One is based on actual residue data that I might develop of equipment surfaces during development runs. For example suppose the following locations (for now I’ll just call them Location A, Location B, and so on) had the follow results for measured residue.
| Location | Level |
| A | 0.8 |
| B | 0.2 |
| C | 0.3 |
| D | 4.5 |
I can look at that data and clearly make a conclusion that “D” is a location that should be sampled. Realize that from a cleaning process design perspective, I might be happy with all those results if my acceptance limit were 12. If my acceptance limit were marginal at 5.0 (or were even lower such as a value of 3.0), I would want to re-design the cleaning process to make it more robust.
Furthermore, with the above example this does not mean I only select Location D for swab sampling and ignore sampling the other locations in my validation protocol. There may be other rationales (discussed below) which would suggest that one or more of those other locations should be sampled.
Lastly I should be careful in trying to provide a rank order of locations just based on historical data. While I may be able clearly to differentiate one or more locations that were shown to be more problematic, the data presented above for locations A, B and C may be practically the same. If you really want to establish a rank order, more statistical analysis is probably required (but I would discourage that unless you have lots of resources and a good rationale to establish an exact rank ordering).
As was mentioned, sometimes I might sample representative locations which are not necessarily worst-case locations as determined by use of the above criteria. This may be useful in the development stage or in the initial validation protocol to make sure that “nothing was overlooked”.
In addition, it should be clear that “worst-case” locations are not just a function of the equipment design and geometry, but could change depending on the MOC, the product processed, and the product processing parameters.
Finally, “actual data” may be very useful after validation is completed and/or as data is collected during routine monitoring to help determine the “worst cases of the worst-case locations”. This may help reduce swab sampling locations going forward based on a risk assessment.
Next month will involve considerations related to more appropriately documenting this selection process to make it more useful for validation protocols.
Copyright © 2024 by Cleaning Validation Technologies