Optimizing The Entire Drug Development Process Using Pharmacometric Tools: From Preclinical To Marketing
Serge Guzy
Pop_Pharm
Objectives: This study concentrates on T2D patients where the main goals were to characterize the PK correlation with a PD biomarker, to propose the optimal trial conditions for the upcoming Phase 2 trial, to estimate using the PK/PD model results the probability of technical success (PTS) for Phase 2 using the target Product Profile criteria and finally modifying the current decision analysis based estimation of the product value (Expected Net Present value, ENPV) using the model based estimation of the Phase 2 PTS.
Methods: The Monte Carlo Expectation Maximization algorithm combined with importance sampling [1, 2] was the tool from which we perform all the fitting procedures. The Precision tree software (Precision Tree) was used to build an interactive program of the product market forecast. The risk analysis software (RISK) was combined with Precision Tree to quantify the uncertainty in all the business related parameters of the decision tree program .
Results: The Preclinical analysis helped defining the PK model and the dosing conditions for the subsequent Phase 1 trial. The Phase 1 PK/PD model included semi mechanistic as well as engineered based processes that could mimic in particular a special Bell-shape dose response relationship showing that the response increases first with dose but then starts to be inhibited at large doses. The PK/PD model analysis was used to predict the optimal conditions for the Phase 2 trial. The criteria for Phase 3 go/no go (PTS)was defined from which the PK/PD model combined with parametric bootstrap and uncertainty analysis resulted in a distribution assessment for the Phase 2 PTS.Decision tree analysis was finally performed and resulted in an updated ENPV for the product that took into consideration the information retrieved from the data using the PK/PD population modeling approach.
Conclusions: PK/PD modeling combined with decision analysis allow changing dramatically the forecast of the Product value from the sponsor's side (part of the pie the sponsor would get in the deal). The modeling of the PK/PD correlation leads to an estimate of the PK/PD parameters as well as their uncertainty. These sets of parameters were used to estimate the PTS distribution for Phase 2 and how it would result in a new estimate of the all the possible ranges of the ENPV for the product. The new average ENPV increased by a almost 80% compared to the original estimate that did not consider the PK/PD modeling output results.
References:
[1] Robert J. Bauer , 1 Serge Guzy , 1 and Chee Ng 2:" A Survey of Population Analysis Methods and Software for Complex Pharmacokinetic and Pharmacodynamic Models with Examples", AAPS Journal 2007; 9 (1) Article 7
[2] Ng CM, Guzy S, Bauer RL. "Validation of Monte-Carlo parametric expectation algorithm in analyzing population pharmacokinetic/pharmacodynamic data", Clin Pharmacol Ther 2006;79(2):P28