While this Cleaning Memo deals with rinsing, and utilizes studies similar to those described in the March Cleaning Memo, the focus is significantly different. What is covered here is the situation where the number of rinses has already been established, but where the analytical method has an LOQ or LOD such that we cannot establish that the calculated acceptance limit is being met. For clarification, suppose the LOQ is “1” and the LOD is “0.3”, but the acceptance limit is intermediate between the two values, such as “0.5”. In that situation if the measured value is reported as <LOQ (but also >LOD), then we cannot unequivocally say that the limit of 0.5 is being met. On the other hand, if the measured value is <LOD, then I have clearly established that the tested sample meets the acceptance limit. Now, I might not like that latter situation because I might believe my cleaning program is not robust enough.
Are there ways to deal with situation? Certainly one is to revise and improve the analytical method so the LOD/LOQ is lower. That is something you can pursue on your own. But, what is presented below is another option to provide some evidence that our cleaning process may result in effectively meeting the rinse acceptance limit. In the example below, I am assuming that after the washing step in CIP cleaning there is a series of rinses (sometimes called “pulse” rinses). The final pulse rinse is generally non-recirculitng (once-through), while the earlier rinses are recirculating. Furthermore, I then design a study so that the number of rinses goes beyond the number required to achieve measured values below the LOQ or LOD. How many rinses should be evaluated will depend on the specific situation, but it probably best to include at least one additional rinse beyond what you think is needed so that it is less likely to have study results for which a clear conclusion cannot be drawn.
In this type of rinsing study the “marker” that is chosen should be the API itself. It may be possible to choose either a specific method (like HPLC) or a non-specific method (like TOC), as long as the analytical result can be correlated with the presence of the API. It is definitely preferable to utilize a specific analytical method, because in most situations where one has very low acceptance limits, the background “noise” from TOC may interfere with making a meaningful conclusion.
As discussed last month in studies of this type, measuring that marker can be done by selecting any portion of each recirculitng rinse as it exits the equipment. While there may be minor variations between the first portion and the last portion of each individual rinse, if it is of concern (or just to provide evidence that it doesn’t make a difference), samples from the beginning and the end of the rinse could be taken as the rinse leaves the equipment. Then, in the final (non-recirculating) rinse, the situation is slightly different. For this rinse it is at least theoretically possible that the marker level in the first portion of that once-through rinse could be higher than in the last portion of that same final rinse. The reason this may happen is not unique to this final rinse, but is an expectation of any once-through rinse. While this difference is a possibility, it is not a firmly established conclusion if the value of the both the initial and final portion of that once-through rinse all had reported values below the LOD of the marker. In that case, I still suspect the true value is lower in the final portion, but with my analytical method I can’t clearly establish it. So I would like to do one of two things. One option is to collect the entire amount of that once-through rinse and measure the marker in that “composite” sample to determine a marker value of the “entire” final rinse. A second option is collect samples of the once-through rinse at various intervals and average them (or perhaps integrate them based on the portions of the final rinse each sample might represent). In either option I would have a reasonable estimate of the value of the marker in that once-through rinse.
For this example let’s assume, we perform three 250-liter recirculating rinses followed by a single non-recirculating 250-liter rinse. And, at the end of the fourth rinse the measured value is less than the LOQ of “1” (with an acceptance limit of “0.5”). Here is a tabular presentation of the results.
| Rinse No. | 1 | 2 | 3 | 4 |
| Measured API Value | 500 | 50 | 5 | <LOQ |
| Calculated Relative Reduction | N.A. | 10 | 10 | N.A. |
As discussed above we cannot clearly say the acceptance limit is being met just by looking at the API value at the end of the fourth rinse. However, if the relative reduction has been the same in the first three rinses, we may assume that the same relative reduction holds in going from the third rinse to the fourth rise. In this case, the “extrapolated” value of the fourth rinse would be “0.5”, meaning that we are meeting our acceptance criteria, albeit that I would prefer it be at an even lower value.
So what we could do is add another rinse so that we have a total of five rinses, with only the last being non-recirculating. Here is a possible tabulation of those results.
| Rinse No. | 1 | 2 | 3 | 4 | 5 |
| Measured API Value | 500 | 50 | 5 | <LOQ | <LOD |
| Calculated Relative Reduction | N.A. | 10 | 10 | N.A. | N.A. |
Again, if the relative reduction is the same in the first three rinses, we may assume that the same relative reduction in going from the third rinse to the fourth rinse also holds in going from the fourth rinse to the fifth rinse. In this second case, the “extrapolated” value of the fifth rinse would be “0.05”, meaning that we are well below our acceptance criteria. Therefore, the desired number of rinses should be five rinses. We are happy!
Some may look at this second example and just say “If the LOD is a value of ‘0.3’, why not just depend on that value for the fifth rinse to say the acceptance limit is being met. That certainly is possible, but if an additional argument can be provided for showing that that the actual value that is <LOD is closer to 0.05, then I would prefer that situation as a “belt and suspenders” approach (“belt and braces” for those in the UK) to show the robustness of my cleaning process.
Here are some additional comments that should be considered in utilizing studies of this type.
Comment A: In the two examples given, the relative reduction is the same for each rinse. In some situations, the reduction factors may decrease in the later rinses, most likely due to the marker being “held up” in some way as rinsing processed. This can be due to higher levels of a “difficult to clean” marker being left on surfaces or to “dead legs” in the flow paths. The best remedy for this type of situation is to increase the robustness of the washing step (longer time, higher temperature, and/or higher cleaning agent concentration) or to eliminate unacceptable dead legs. In an ideal situation, the reduction factor should be the same as rinsing proceeds; but as I have taught many times, we don’t necessarily live in an ideal world.
Comment B: If the proposed approach is just not feasible in your particular situation, then some may just decide that they will just depend on a more extensive swab sampling plan. This may be possible in a cleaning validation protocol, but may not be so in the situation for routine monitoring, particularly for highly hazardous APIs, where it is more preferable to do routine rinse sampling for CIP processes. In any case, the possibility of doing swab sampling with no rinse sampling should be appropriately documented, preferably in a high level document.
Comment C: In addition to improving the analytical method to provide quantifiable numbers at a lower level, it may also be possible to increase the acceptance limit so that API residue can be quantified b the existing analytical method. That can be explored by reviewing the assumptions that went into your carryover calculation. While not covered this month, that may be a subject for future Cleaning Memo.
Copyright © 2024 by Cleaning Validation Technologies