France Mentré1, John Kovarik2 , Christophe Gerbeau2
1INSERM U436, 75013 Paris, France; 2Novartis Pharma, AG, 4002 Basel, Switzerland and 92506 Rueil-Malmaison, France
Basiliximab is an immunosuppressant chimeric monoclonal antibody directed to the human interleukin-2 receptor, a -chain used for prevention of acute rejection episodes in organ transplantation. The minimally effective serum concentration necessary to saturate receptor epitopes in kidney transplant patients is 0.2 m g/ml. In order to guide dose selection for Phase 3 efficacy trials, a population pharmacostatistical model was fitted to intensively-sampled Phase 2 pharmacokinetic data using NONMEM. This served as a basis from which to examine candidate dose regimens with respect to the duration over which receptor-saturating concentrations would be achieved posttransplant.
Three methods were assessed to estimate the prediction interval of the time to reach a threshold concentration given a nonlinear mixed-effects model: one based on simulations, and two others based on first-order approximation using either inverse regression or inversion of confidence intervals. The two later methods were derived from the inverse regression problem in nonlinear calibration curves. An 80% prediction interval was generated by each method to evaluate its predictive performance against prospectively collected Phase 3 data in 39 renal transplant patients who received two injections of 20 mg basiliximab, one prior to surgery and one on day 4 posttransplant. All methods provided correct prediction of the duration of receptor-saturating concentration. As anticipated, the best performance was obtained from the simulation method which predicted 30 values in the 80% prediction interval [19.7 ; 52.7] days. The actually observed 80% interval from the Phase 3 data was [23.7 ; 58.3] days.
The best method to derive the prediction interval was based on simulation of the individual profiles and therefore avoided first-order approximation. Unfortunately, the proposed approach does not presently take into account the standard error of estimation. It can be extended for other applications in pharmacokinetics but also for inverse regression in dose-response or concentration-response population analyses.
Reference: PAGE 8 (1999) Abstr 161 [www.page-meeting.org/?abstract=161]
Poster: poster