A comparison of computational approaches to the population pharmacokinetics. An example of toxicological data

Four different approaches to population pharmacokinetic analysis were applied to routine clinical data on carbamazepine intoxications in epileptic and alcoholic patients. The computational issues were noticed for maximum likelihood based methods, while Markov Chain Monte Carlo approach revealed that the estimates of absorption and lag parameters might be unreliable. Non-identifiability of these parameters may be the source of computational problems with other methods. Simultaneous use of different approaches to population pharmacokinetics is therefore advised, since it allows for verification of the obtained results.