Comparison of non-compartmental to individual pharmacokinetic modeling approach in the assessment of bioequivalence for 2 formulations of trastuzumab

J Stevens, JAA Reijers, J Burggraaf. Centre for Human Drug Research, Leiden, The Netherlands

Non compartmental analysis (NCA) is considered the gold standard to demonstrate bioequivalence (BE), even though a NCA is highly dependant on times of sampling and known to be less appropriate in cases of nonlinear pharmacokinetics (PK). In many clinical circumstances, some or even all of the individuals may be sparsely sampled, making the individual evaluation difficult. In such cases, the use of models, in particular population models, becomes appealing. However, regulatory guidelines on the use of population pharmacokinetics (popPK) for demonstrating BE are currently lacking. The objective of this study is to evaluate the use of PK parameters obtained by popPK for demonstrating BE compared to standard NCA.

The dataset was obtained from a study designed to demonstrate BE between two trastuzumab formulations (test and reference [Herceptin®]). In a dose escalation part healthy male volunteers received single doses of 0.49, 1.48 or 2.96 mg/kg test product in consecutive cohorts. In the BE cohort, 92 subjects were evenly randomized and received 5.95 mg/kg test- or 6.44 mg/kg reference product. Blood samples were collected up to 64 days after administration. A standard NCA was performed on the BE cohort, using the software package R. PopPK analyses included data from the dose escalation part and were performed on three subsets of the data; test product, reference product and data from both formulations combined. Nonlinear mixed effects modelling (NONMEM 7.2) was applied for model development. In all popPK analyses, the effects of weight, height, age, body mass index, body surface area, lean body weight (LBW) and shed antigen were investigated on all PK parameters. Additionally, in the combined dataset, formulation was evaluated as covariate. The three popPK models were structurally similar to allow comparison between individual empirical Bayes estimates obtained by the three models. After model development, individual profiles were simulated (R-packages deSolve and MASS) with time steps of 0.1 hour. Consecutively, individual areas under the curve (AUC; 0 to 64 days) were calculated and maximal concentrations (Cmax) were obtained. Linear content-correction was applied, to correct for the dose difference between products. The ratio of geometric means (GMR, test/reference) of the AUCs and Cmax of both formulations were compared to the GMR and 90%-confidence interval (CI) of the NCA.

The GMRAUC of the NCA was 88.6% (CI: 84.1-93.3%) and the GMRCmax was 89.4% (CI: 83.4-96.0%) thus within the set BE limits (i.c. 80-125%). Three compartmental models parameterized in terms of volumes of distributions, intercompartmental clearances with combined linear and nonlinear (Michaelis Menten) clearance best described the data. Interindividual variability was identified for central volume of distribution, the concentration which produces half of the maximum rate of elimination and the elimination rate constant. A combined additive and proportional error model best described the residual variability. Normalised LBW was identified as covariate on the central volume of distribution. Implementing formulation as covariate in the combined dataset did not improve the descriptive properties of the model, proving no statistical significance of formulation on the population parameter estimates. The separate test and reference models resulted in a GMRAUC of 86.8%, the combined model in 89.3%. The GMRCmax values were 86.7% and 92.5%. All values are well within the CIs of the NCA and within BE limits.

In contrast to the raw data used for the NCA, the models described the full concentration-time profiles and captured the nonlinearity in population parameters that have predictive properties for other dosing regimens in further drug development. The results from the population models are indistinguishable from the standard NCA, confirming the usefulness of popPK methods for the evaluation of BE.