As described in one of my earlier posts, one can easily perform an analytical method validation by spiking the dissolution medium using a solution of a drug (API, link). The suggested approach is scientifically correct and valid. However, I do see where dissolution scientists, in general, will face difficulties. Let me explain:
Suppose an analyst follows the suggestion made and obtains a %RSD of “X” for the method (say less than 5) and the analyst is happy with it. Then, the analyst proceeds to the next step, test the tablets, and gets a %RSD of “Y”. Under normal circumstances, this will reflect the %RSD of the product, including %RSD of the method. Usually, there are no concerns as this value of “Y” often comes out acceptable, between 5 and 10.
However, dissolution testing, particularly using paddle/basket, faces a unique problem, that it introduces one additional variability component, which is very well documented in the literature. This is because of the positioning effect of the tablet/capsule i.e., where it settles at the bottom of the vessel (link1, link2, link3). Unfortunately, people do not realize how such a minor variation can cause a big problem, but it does. As one cannot control this positioning effect, one cannot control variability due to this effect. It is totally random. The contribution from this random effect is reported to produce very high RSD, up to 37% (link). So, when it is asked what should be the expected variability for drug dissolution testing of a product, a safe bet/estimate is 37%. A product may have excellent repeatability/reproducibility of its drug dissolution characteristics (with extremely low %RSD). However, dissolution results may or may not reflect this low variability.
It is just like any other biased but random phenomenon, where one may or may not succeed. However, one always sees advertisements of some examples of big winnings/successes. In dissolution terminology, one may observe some low %RSD values at random, but overall variability using paddle/basket apparatuses will always be high. There are many publications available describing this high variability aspect, which may be useful. In addition, some posts may also be useful in this regard e.g., see link.