David Hermann (1), Wenping Wang (2), Christine Falcoz (3), Daniel Hartman (1), Jaap Mandema (4)
(1) Pfizer Global Research and Development, Ann Arbor, USA; (2) formerly with Pharsight Corporation, now at JNJ PRD; (3) Pharsight Corporation; (4) formerly with Pharsight Corporation, now at Quantitative Solutions
Objectives: To assess the utility of a novel PK/PD-based modeling and simulation strategy as well as the utility of the Drug Model Explorer™ (DMX™) technology for decision-making during early clinical development of CI-1027.
Background: CI-1027 was developed as a low-density lipoprotein cholesterol (LDL-C) lowering compound. The team was interested in assessing early the effect of CI-1027 plus statin combination compared with statin monotherapy or a key competitor plus statin combination. Given the LDL-C lowering effect across the CI-1027 plus statin doses range, should clinical development continue?
Strategy: A single Phase IIA trial was planned along with a dose-response surface meta-analysis of literature data on key competitors and CI-1027 data for several efficacy and safety endpoints. DMX software provided to the team an interactive, easy to use, query tool to compare treatments and make trade-off based on all endpoints.
Methods: The Phase IIA trial was a single 8-week, double-blind, study in hypercholesterolemic patients with placebo, three CI-1027 doses, three atorvastatin doses, and their respective combination. Summary data from the trial were combined with CI-1027 Phase I data and literature data from ezetimibe and statin trials. A nonlinear mixed effects regression analysis was undertaken to describe (1) the mono-therapy dose-response for the non-statins, CI-1027, and ezetimibe, and (2) the dose-response for 5 statins as mono-therapy and in combination with a non-statin. Summary data from 21 clinical trials (~10,000 patients) were included for LDL-C. Emax models described the relationship between percent change in LDL-C and CI-1027, ezetimibe, and statin (mono-therapy) dose. Combinations were well described by adding a simple interaction term to the model.
Results: The predictive distribution of the dose-response surfaces was obtained from the models covariance matrix and uploaded into DMX. After selecting an endpoint, population, and treatment of interest the DMX system immediately displayed the corresponding quantitative result, including likely differences between CI-1027 and competitors. For LDL, the CI from the ANCOVA analysis of the Phase IIA trial overlaps that of ezetimibe. The CI from the meta-analysis does not overlap the ezetimibe CI clearly suggesting that CI-1027 is unlikely to lower LDL-C sufficiently to compete with ezetimibe.
Conclusion: In this case, the availability of integrated dose-response models for CI-1027 and competitors guided informed decision-making during early development. Based, in part, on the quantitative knowledge obtained through modeling all relevant data and made accessible via DMX, the development of CI-1027 was discontinued after one Phase IIA trial in the target population.
Reference: PAGE 14 (2005) Abstr 805 [www.page-meeting.org/?abstract=805]
Poster: poster