Questions: Recovery of swab for cleaning validation

  1. Surfaces type selection.
    If we have aroud 12-20 different type of surfaces from the production line. Do we do recovery study and validate for all of surface?
  2. Placebo consideration
    When doing the recovry study, spiking the known amout of API on the coupon, do we have to include placebo in the spiking solution, if we have different formalations, do we test all of them?
    3.spiking level selection
    generally how many levels will need for the recovery test? (eg. if limit level=5ppm)
    Thank you very for reply.

Dear b0b,

1- Unfortunately you need a recovery test for every type of surface, but it is accepted practice to bracket, e.g. for stainless steel of different finish, recovery test for mirror polish and rough finish is sufficient, worst case recovery factor is used. After doing the recovery test for all the surfaces and two or three analytes, you will realize that the impact of the sampling technician on the value and variability of the recovery factor is much higher that the surface itself, and that recoveries from rough groups of materials (stainless steel, teflon®, silicone, PVC and anodized aluminium) are enough.

2- As to my knowledge, placebo testing is not a current requirement, because:

  • It is assumed that the excipients are washed away more easily than the active, and thus do not interfere with the analytical method. The “visually clean” condition ensures that the common excipients are eliminated to levels that won’t interfere with the method.
  • Usually the analytical method for cleaning validation is a range extension of the analytical method for dosing the active in the product, and in the validation of that method the robustness has been proved with the placebo of the product where the “contaminating” analyte comes from.
  • In our specific case, part of the cleaning validation is a risk assessment matrix for contamination and interference of all the excipients that come in contact with the equipment, and normally the excipients pose a low risk of (negative) interference with the method.

3- In our case we perform the recovery test at five spiking levels (50%, 80%, 100%, 150% and 200% of the target level), triplicate for three sampling technicians (that is 45 samples for each surface). Nevertheless, we do not repeat this for every type of surface but only for the main surfaces (normally for the rough groups of materials stated above), and for each surface that is sampled by rinsing. (Fortunately, out of our 110 different active ingredients only 7 are the worst case in the different equipment trains!).

Best regards


Thank you so much for your answer.
For #3 when you performing at 5 levels, what the the recovery factor going to use for calculation (mean of 45) ? And what createria will use for consider as validated?
(RSD < ? for the 45 determinations)

When the batch size change or new product intruduced to the same equipment trains, the limit may change, Do you do new validation for the method?


Dear b0b,

the recovery factor we use is the worst result among the five levels, of the worst result among analysts, of the averages per analyst. In other words:
Among three replicates of each analyst: acceptance criteria: RSD < 5%, data used: average.
Among three analysts: acceptance criteria: RSD < 10%, none below 50%. Data used: lowest recovery percentage.
Among the five levels spiked: RSD< 20%, none below 50%, data used: Lowest recovery percentage.(Sorry, it is not too clear, but can’t find a better way!).

To make it clear: these figures are arbitrary, and based solely on the experimental data accumulated. So don´t stick to this approach, there may be better ways to do it. A lot of companies simply set a minimum recovery percentage regardless of the variability among replicates and analysts.

When the batch size changes, it impacts twofold: The equipment train may change (because of minimum processing capability of the equipment), and the accepted contamination level may change. In our case we revalidate the method, because our product/equipment matrix is extremely complex and uses a lot of “worst case” considerations, which simplify recalculations when product portfolio, batch sizes, formulas change, but in turn artificially lower the accepted contamination level to extremely low levels, so that there is no room for additional worst case considerations. My personal recommendation is to use, for each piece of equipment, the minimum processing capability (should be an outcome from the PQ of the equipment), and the highest figure of daily intakes of the whole product portfolio. With this approach you are on the safe side regarding the impact of batch reductions on the accepted contamination level, although changes in the equipment trains still may impact. An even safer approach is to validate each analytical method from scratch for a lower level (e.g. 10% of the target level), assuming that a minimum batch size reduction to 1/10th is beyond any technical feasibility.

For the case of equipment with no batch size limitation (tablet presses, dedusters, metal detectors, blister packaging lines), which in theory could accomodate batch sizes from 1 to infinite dosage units, our policy, stated in the validation master plan, is to set a minimum batch size of 1/10th of the bulk production batch size, which means that e.g. out of 1.000.000 tablets, a minimum packaging batch size of 100.000 tablets is allowed.

Best regards


It is very helpful for me. Thanks a lot.