Using M&S to Shorten the Repeating Cycle of the Early Drug Development Process: A Case Study
Susan Willavize(1) , Ruolun Qiu (1), Marc Gastonguay (2)
(1) Pfizer, Inc. (2) Metrum Research Group LLC
Objectives: A case study illustrates use of in silico methods (modeling and simulation (M&S)) to augment or replace in vivo study data and to aid the decision making process in early drug development. Several technical decisions, which proved key to the timely and successful implementation of M&S in this case study, are discussed.
Methods: A population PK-PD model was fitted to data at two dose levels from a Phase 1 study in normal volunteers. The model was validated by comparing the observed mean response to data simulated from the parameter estimates and their uncertainty. Validation for the PK-PD model was also obtained by computing a summary function (pd-AUC) for the simulated data and comparing to observed mean pd-AUC values. The model was then used to extrapolate to doses up to seven fold higher. The relationship between predicted exposure and predicted pd-AUC was plotted and considered in comparison with toxicology limits. Percentiles of the simulated data at the extrapolated doses provided an assessment of the probability of technical success; i.e., the probability that the drug can meet target effect levels on average.
Results: A composite indirect response model fitted the multiple dose data well and provided pd-AUC values that agreed well with observed values. Exposure levels at the toxicology limit have a less than 1 % probability of meeting the PD target on average. To attain an 80% probability of technical success, exposures 2-3 fold higher than the toxicology level would be needed. These doses are more than 3 fold higher than the highest dose previously studied in the clinic under multiple dosing. Among the technical decisions critical to the timely application of M&S to this drug development experience were: 1.) dialog with the development team concerning the hypothesized mechanism of action of the drug and concerning the metric of interest for decision making, and 2.) focus on the probability of technical success rather than the probability of trial success.
Conclusions: Based on PK-PD modeling and uncertainty around the parameter estimates, exposure levels needed to reach PD target are 2-3 fold higher than animal toxicology limits. Doses substantially higher than those previously used would be needed to obtain PD endpoint target with high probability (80% or above). Quantitative information influenced the early decision not to progress further development of the drug candidate under examination.