Any change in the cleaning validation program requires an assessment as part of a company’s change control program. For this month we will consider a relatively simple way to deal with changes we might want (or might be forced to make) in your swab sampling or rinse sampling SOP. In most cases (but not all) we would prefer to avoid having to evaluate the recovery percentage for every possible API molecule and every possible surface type that the sampling procedure is used for. So, what can be done?
We’ll first consider a simple case where all you are doing is changing only one element in a swabbing procedure. Such examples might include a change from the existing Swab AB to a proposed Swab CD, with each swab having different characteristics such as swab head composition, swab head shape, swab head dimensions or swab handle length/flexibility. In this particular simple example, one possible lab study is a comparison of percentage recoveries between use of the two different swabs with several (and not all) API molecules. Those API molecules evaluated may include the least soluble molecules, and where possible different chemical natures (by chemical nature, I mean something like “ionic versus nonionic” molecules or like molecular weights). The number of molecules selected for testing will depend situationally on the specific mix of molecules, but probably should be at least two molecules, but most likely no more than four. This type of study should be done on the most common material of construction (such as stainless steel). In this simple case of only two different swabs, it is probably not required to evaluate other additional surface types, since it is not expected that just a change in the swab itself will be a significant factor in recovery from those other surfaces.
So, what is the experimental design? The basic concept is to do “side by side” experiments comparing (for each selected API molecule) the percent recovery using three replicates for each swab type. All replicates should be done by the same person within a similar time period (such as no more than one hour apart). Preferably, if there are to be three replicates for Swab AB and three replicates for Swab CD, the order of swabbing coupons should be somewhat randomized (to deal with possible operator “fatigue” in moving from the first coupon to the sixth in a series). The swabs are extracted in the same manner and analyzed for the API molecule by the same analytical procedure at roughly the same time on the same analytical instrument. The percent recoveries are calculated in the same manner as done by your company procedures. The three percentages for each swab type are then averaged, and averages from the two different swab types are compared.
Example A: If the average recovery percentages are within 5% absolute of each other, some (including me) would most likely conclude that the change in swab type was not significant, and that the recovery percentage previously established is still valid and its use can be continued. Here is an example (in tabular form) of this type of data.
Note that the absolute difference between the two average results for the two different swabs is 1.5%. In this example, the new swab (CD) gave a slightly higher percent recovery, but that difference in this limited data is not significant. Therefore with this data I should continue with the historical data previously established for Swab AB even though I am switching to Swab CD.
| Run 1 | Run 2 | Run 3 | Average | % RSD | |
| Swab AB | 79.2 | 76.5 | 81.6 | 79.1 | 3.2 |
| Swab CD | 78.2 | 80.3 | 83.3 | 80.6 | 3.2 |
Example B: What should be done if the new swab type (Swab CD) gives a lower recovery of more than 5% absolute, as shown in the data from the example below?
| Run 1 | Run 2 | Run 3 | Average | % RSD | |
| Swab AB | 79.2 | 76.5 | 81.6 | 79.1 | 3.2 |
| Swab CD | 74.8 | 73.4 | 71.9 | 73.4 | 2.0 |
In this case, the absolute difference between the two sets of data is 5.7%. So, I might consider using a correction factor applied to the previous recovery percentage (done with Swab AB). In this example, I might apply a correction factor of 0.734/0.791, or 0.928, to the previous recovery percentage to deal with the slightly lower recovery with Swab CD. That is, if my previous recovery factor for this active was found to be 81.1%, going forward with the use of Swab CD I would use a recovery of 0.928*81.1%, or 75.3%.
Example C: What should be done if the new swab type (Swab CD) gives a higher recovery of more than 5% absolute, as shown in the data from the example below?
| Run 1 | Run 2 | Run 3 | Average | % RSD | |
| Swab AB | 74.8 | 73.4 | 71.9 | 73.4 | 2.0 |
| Swab CD | 79.2 | 76.5 | 81.6 | 79.1 | 3.2 |
In this case, the absolute difference between the two sets of data is 5.7%. Analogous to what was done for Example B, I might consider using a correction factor applied to the previous recovery percentage (done with Swab AB). In this example, I might apply a correction factor of 0.791/0.734, or 1.077, to the previous recovery percentage to deal with the slightly higher recovery with Swab CD. That is, if my previous recovery factor for this active was found to be 71.8%, going forward with the use of Swab CD I would use a recovery of 1.077*71.8%, or 77.3%.
Another option for this situation would be not to apply a correction factor to increase the previously established percent recovery. The rationale might be that if the difference between a recovery of 77.3% and a recovery of 71.8% were likely to cause a failure in meeting the swab acceptance limit, then I probably should have designed a more robust cleaning procedure. So I could just continue with the previously established recovery percentage with the use of the Swab CD. Some might wonder why this option is presented for Example C and not for Example B. Even though the argument could be made that with a robust cleaning process, the Example B difference between 81.1% and 75.3% is not significant, the reason it is not presented for Example B is a matter of optics (or of perception). I want to avoid the situation where regulators or auditors perceive that by not applying a correction factor I am making it easier to get passing results.
This same simple evaluation could be used if there is only one other type of change to the sampling procedure (but using the same swab type). An example with only one change might be to a larger swabbed area (such as from 25 cm2 to 100 cm2). A second example might be a change to a lower extraction volume (such as from 10 mL to 5 mL). A third simple change might be the change in solvent used to wet the swab head. The same principles previously discussed could be used for these examples.
Life becomes more difficult if there are multiple changes in the swabbing process. For example, I might still want to use the same swab type, but I might want to change both the swabbed area (from 25 cm2 to 100 cm2) and the extraction volume (from 10 mL to 5 mL). In that case, I should be more careful in my evaluation. With two variables changing, one variable might have the effect of lowering the recovery percentage while the other might raise the recovery percentage, but the net result might be that the recovery percentage is essentially the same. In one sense I am fine with that result. However, at a later time, I should not assume that the recovery percentage will be the same if I decide to keep the swabbed area at 100 cm2 but to change the extraction volume back to 10 mL.
Additional comments—
In the examples given, only data from one API was presented (for simplicity of illustration of the principles involved). As suggested, it is preferable to have data from at least two different actives to confirm adequacy of any conclusion as to what action should be taken. If there is significant divergence between recovery results for multiple actives, then further investigations should be performed.
Common sense should be used in evaluating such data. Vastly different recovery percentages by the same operator on the three coupons with one swab type should require some level of investigation. Some may also want more statistical analysis; if that is the case, I would recommend that there be at least six replicates for each condition to obtain more reliable statistics.
Furthermore, some appropriate controls should be put in place. For example, one control might be just a wetted swab placed in the extraction solution, vortexed (or sonicated), and then analyzed; this would be to run just to provide an assurance of adverse effect from the new swab. Note that controls should be run for both the new swab and for the old swab.
Note that the comparison here is only between these two situations evaluated “side by side”. I am not comparing the results from these new experiments to the results obtained in the recovery evaluation that was done several years ago. That certainly could be done. However, care should be used in that comparison. If the results on the existing swab (Swab AB is the examples above) done now are significantly different from the results several years, there may be several causes. One might be that the comparison done now uses three replicates using one “operator” now, and the previous results might have been from data from two or three persons several years ago. Furthermore, if the official recovery percentage previously established was based on the lowest average of any one operator, or on the lowest single data point among several operators, then any comparison between new data and the previous data may be suspect. I should add, however, that if the difference between the new data and old data (each using the identical swabbing procedure and identical analytical method) is more than 15% absolute, further investigation should be considered. Such an investigation should focus on possible differences that could have been overlooked. For example, if the analytical method has been changed from TOC to HPLC without evaluating the effect on swabbing recovery, the introduced variability may account for the observed difference.
These principles could also be used for rinse sampling recovery (although the extent of possible changes is more limited).
Copyright © 2024 by Cleaning Validation Technologies