Chao Chen and Yin Yin
GlaxoSmithKline, Greenford, Middlesex, UK and Research Triangle Park, North Carolina, USA
Objective: To assess the potential for bias in dose-response relationships obtained from titration trials.
Methods: Using an Emax pharmacodynamic model, clinical studies were simulated to allow dose in individual patients to increase up to a pre-specified maximum amount at log-equal increments or until a cut-off response was observed. Multiple simulations were performed to assess the estimation sensitivity of ED50, Emax and Gamma to the following model properties or trial design factors: steepness of the Emax curve, inter-patient variabilities in Emax, ED50 and Gamma, intra-patient residual variability in response measurement, maximum dose, dose-escalation rate and cut-off response. Data were generated by SAS and modeled by NONMEM.
Results: ED50 and Emax were underestimated and Gamma was overestimated in all trials simulated. The extent of the bias in these parameters was a function of the trial design and the pharmacostatistic properties of the dose-response model.
Conclusions: Emax type of dose-response relationships obtained from titration trials may be biased. When feasible, forced dose-escalation design, which is expected to produce less biased results, should be implemented instead.
Reference: PAGE 10 (2001) Abstr 172 [www.page-meeting.org/?abstract=172]
Poster: poster