I sometimes run across situations where a company either wants (or is asked) to perform a sampling recovery study at the LOQ (Limit of Quantitation) of the analytical method. The reasons for doing so may be things like “the LOQ value is closer to the actual values I would measure in a cleaning validation protocol.” In this Cleaning Memo we will address two reasons why this may not be needed or appropriate. One reason involves the fact that doing so may not give us relevant data unless the recovery is actually 100%. The second reason is that in many situations where the LOD is well below the acceptance limit, such a study may not give data that is particularly useful.
For purposes of examples in this Cleaning Memo, I will assume that the acceptance limit is 10.0 units and the LOQ is 2.0 units. The units may be µg/swab, µg/mL, ppm, or µg/cm2; it makes no difference in the illustration of the principles involved, but for simplicity I will just use ppm.
First, let’s assume I do a recovery study with a spiked amount on a coupon (or plate if you prefer that terminology) such that at 100% recovery I would measure 10.0 ppm of the residue (my API, for example). Further, let’s suppose in this swabbing recovery I measured 8.2 ppm. That value of 8.2 ppm is above my LOQ. Therefore I have confidence in that measured value and report out a sampling recovery of 82%. So far, so good.
But, let’s assume that I chose to do a recovery study at the LOQ value. That is, I spike the coupon with an amount such that at 100% recovery, I get a measured value at the LOQ (2 ppm). So if the measured amount in the study were 2.0 ppm, I would say that I have confidence in that value (it is the LOQ, remember), and I would report out a sampling recovery of 100%. But, what if my measured value were 1.6 ppm? I could theoretically do a calculation (1.6/2.0) and say my recovery is 80%. Really?? That measured value of 1.6 ppm is below my LOQ, so what confidence do I have in the reliability of that number such that I can claim a recovery of 80%? In this latter example, it would be stretching things to say the sampling recovery of 80% is a reliable value. Furthermore, note that even in an ideal situation, the precision of a recovery percentage is dependent both on the variability of the analytical method and the variability of the sampling procedure. So I would question the value of using a measured value below the LOQ to establish a recovery percentage.
Some may argue that “the LOQ of 2 ppm is a conservative estimate based on a signal-to-noise ratio in my HPLC method”, and that the “true” LOQ is lower, perhaps as low as 1.6 ppm. My reply would be to run the analytical method on a known sample of 1.6 ppm to confirm that at the level the measured amount is suitably accurate (thus confirming my LOQ). The problem with that approach is that if your LOQ is now 1.6 ppm, then you should do a sampling recovery study at that lower amount. And unless you get 100% recovery, the measured value will be below the LOQ of 1.6 ppm. So you still have the same conundrum.
Now to the second reason for not doing a sampling recovery study at the LOQ value. My recommendation has been that a recovery study done at a higher spiked amount gives a recovery percentage either the same as or a “worst-case” (that is, lower) as compared to the expected recovery at a lower spiked level. I have written Cleaning Memos about this numerous times in the past, starting in 2007, with the latest one being in October 2020. Some may disagree with this assertion, but so far most reliable evidence is consistent with this principle (recognizing, of course, that most sampling recovery studies do not have the accuracy and/or precision that one would get when measuring a known standard in solution).
To illustrate what happens in this second situation, let’s suppose I do a sampling recovery study at a spiked level to give 10.0 ppm (the acceptance limit value), but my measured residue level is 7.4 ppm. In that case I report a sampling recovery of 74%. Now in my actual cleaning validation protocol, I measure 2.7 ppm (above my LOQ), and I apply the recovery percentage of 74% and report out a correct value of 3.6 ppm. I assume that if I actually did a recovery study at a lower level, my sampling recovery percentage would most likely be either the same or slightly higher, meaning that my reporting of 3.6 ppm is a number that is reliably below my acceptance limit.
As a second illustration , let’s suppose I do a sampling recovery study at a spiked level to give 2.0 ppm (the LOQ value), but my measured residue level is below LOQ. In that case, I apply the same recovery percentage of 74% to the LOQ value of 2.0 ppm and report out a corrected value of 2.7 ppm (or no more than 2.7 ppm). In this case I would also assume that if I could actually do a recovery study at a 2.0 ppm (which as discussed above is problematic), my sampling recovery percentage would most likely be either the same or slightly higher; meaning that my reporting of 2.7 ppm is a number that is reliably below my acceptance limit.
So the main take home lesson here is to avoid trying to do a sampling recovery study at the LOQ value; only do the recovery study at the acceptance limit. If you really insist on doing a sampling recovery study at a value closer to the LOQ, do an additional study (in addition to the study at the acceptance limit) at a value 50% greater than the LOQ. This will assure that if the sampling recovery percentage is at least 70%, the measured value in your recovery study will not be below the LOQ.
Let me also clarify that the LOQ I refer to in this Cleaning Memo is the LOQ of the residue in solution; some companies also will refer to a LOQ as the calculated “sampling LOQ” (by taking the solution LOQ and dividing it by the sampling recovery percentage). I am not referring to that latter use.
Also note that (technically speaking) if my measured value in a cleaning validation protocol is less than the LOD (Limit of Detection), then I should consider applying the recovery percentage to my LOD value. For example, if my LOD were 1.0 ppm and my sampling recovery percentage were 78%, then I should really report out that the value is <1.3 ppm (and not <1.0 ppm). In almost all situations this will not materially affect whether I pass or fail in a cleaning validation protocol as long as my LOD is not the same as my acceptance limit. If my LOD were the same as acceptance limit, I would refine my analytical procedure to lower the LOD, refine my sampling parameters (such as extracting my swab with a smaller volume of solvent), increase my limit (such as by making a larger batch of the next product), and/or dedicate the equipment to that one product.
In all cases, consider the real objective, along with the variability and limitations of your analytical method, in designing a recovery study.