Issue link: https://www.e-digitaleditions.com/i/492796
P 28 April 2015 Tablets & Capsules dissolution testing Does IVIVC make sense? Ben Locwin Healthcare Science Advisors One goal of pharmacokinetic studies is to properly and robustly define in vitro-in vivo correlation of dose dissolution. Do we have adequate evidence? harmacokinetic (PK) parameters typically encompass dose, dosing interval, peak concentration (C max ), time to peak concentration (T max ), half-life (t 1/2 ), and area under the curve (AUC) from time zero to last-measured time- point (AUC 0-t ). There are other parameters of interest— variously used in specific circumstances and/or not con- sidered when they should be—but these parameters establish the foundation for pharmacokinetic rigor (Figure 1). When the FDA evaluates a generic candidate, for example, it may want to see the candidate product pro- duce the same type of curve as the innovator's product to demonstrate that it is released over a certain period. Today's dissolution characteristics are measured at 1-, 2-, and 4-hour intervals and then every subsequent 2 hours until 80 percent or more of dissolution is complete. Those are wide ranges, but it is not necessarily true that a different performance in dissolution is relevant, that it indicates a significant difference in pharmacological effect [1]. Another source of error in our estimates is the misapplication of mono-compartmental analysis (first- order kinetics), multi-compartmental analysis, and non- compartmental analysis. One goal of PK studies is to properly and robustly define in vitro-in vivo correlation (IVIVC) of dose disso- lution. And from that, presumably, we can learn the rate at which the therapy is introduced into the patient. If that really is the point, it needs rethinking. First, pinning down good models to show this has been extremely diffi- cult. Second, maybe it's an invalid question in the first place and not worth asking. The hypothesis IVIVC rests on the hypothesis, based on limited evi- dence, that a correlation exists between in vitro models and in vivo models. That idea provides the basis for seek- ing in vitro tests that can mimic in vivo results—a lab test to show what the drug product will do in a mouse or in a human being. Statistically speaking, this leads us to pre- sume that the null hypothesis (H 0 ) is not rejected and that our in vitro models are correct approximations and we can keep collecting data using this assumption as a backbone. The first comes from Sir Ronald Fisher, who devised today's formulations for hypothesis testing: "Every experi- ment may be said to exist only in order to give the facts a chance of disproving the null hypothesis" [2]. The second is courtesy of George E.P. Box, an eminent statistician: "Remember that all models are wrong; the practical ques- tion is how wrong do they have to be to not be useful" [3]. That's just what we're doing with IVIVC, attempting to apply data to different models to see if one holds sway. That's not a foolish pursuit, since models can be enor- mously useful across a range of scientific fields, allowing us to posit potential effects from certain well-described causes. This usually means that models only work under