Experience With NONMEM Applied To Phase III Clinical Trial With Many Covariates
P. Burtin, P. Girard, A. Leizorovicz, J.P. Boissel
Unité de Pharmacologie Clinique, Hôpital Neuro-Cardiologique, 162 av. Lacassagne, 69424 Lyon Cedex 3, France.
Pharmacokinetic modeling is widely used in phase 1 clinical trials, in order to explain the variability between the dose-concentration relationships. Performed on healthy volunteers, its extrapolation is not straightforward for patients. Furthermore full pharmacokinetic experiences are either not ethically possible with patients, or are poorly informative, since only few patients and few hypotheses can be considered in a single experiment. That is why population pharmacokinetics has been developed, which estimates the statistical distribution of the pharmacokinetic parameters of the model using few points from many patients, and tries to explain interindividual variability using physiopathologic covariates.
As an illustration, we present the analysis of data from a phase III clinical trial, consisting of 663 mexiletine plasma concentrations collected in 252 post-myocardial infarction patients. The data were fitted to a one-compartment model with first order absorption, using the Nonlinear Mixed Effect Model NONMEM)(Sheiner et al,1977, J. Pharmacokin. Biopharm., 5:445-478). The method computes, with first-order approximation, the first two moments of each parameter. The typical parameters for a male individual under sixty, not smoking, taking no other drug, were : renal clearance = twice creatinine clearance, metabolic clearance = 20.1 L.h-1.m-2 , volume of distribution = 600 L/m2, absorption rate constant = 0.54 h-1. The residual interindividual variabilities were estimated as 80%, 41%, 80% and 277% respectively for the four parameters. In addition, each parameter was expressed as a function of relevant covariates. Various models were tested using a descending stepwise strategy. The best model included proportionality between body surface area and volume of distribution and a 20% decrease of metabolic elimination for patients taking digitalis medication or long-acting nitrates.
The importance of clinical trial pharmacokinetic data lies in the pre-existence of a controlled experimental design, the central determination of drug concentration, the controlled drug compliance and blood sampling and the large number of covariates, which lead to unpredictable findings but the limit is essentially the great number of unusable data : only 663 concentrations out of 1020 could be used, for various reasons : samples corresponded to stopped treatment, missing values, undetectable levels... Considering the multicentric nature of the trial with North American and European centers this is as expected. Another limit is the lack of precision of the various times of blood sampling and drug intake. For these reasons, and also because the study was not designed for this purpose, such results need to be confirmed by further specific studies, if they prove to be of interest for clinical practice.