The Author’s Reply

Excerpt

Our recently published article Parametric and Nonparametric Population Methods: Their Comparative Performance in Analysing a Clinical Dataset and Two Monte Carlo Simulation Studies,[1] which appeared in this journal, has prompted a Letter to the Editor from Drs Proost and Eleveld. They assert that their KinPop (MW\Pharm software package, Mediware BV, Zuidhorn, The Netherlands) iterative two-stage Bayesian (ITSB) algorithm software “performed much better” than the parametric iterative two-stage Bayesian (IT2B) software used in our study; however, both use the first-order conditional estimation (FOCE) approximation to compute the likelihoods. They cite the blind comparison of ten different population modeling methods done by Girard and Mentré.[2] They state that KinPop performed “reasonably well” in that analysis, and “provided the most precise estimates of the correlations between parameters”. They describe a small bias of about 1% in the estimation of the population mean of the elmination rate constant (k_e), not far from what our IT2B found (page 381, paragraph 2[1] ). …