Farkad Ezzet (1) Earvin Liang (2) Jonathan Wagg (1) Margaret Cooney (2) Susan Abushakra (2)
(1) Pharsight Corporation, St. Louis, MO, USA (2) Elan Pharmaceuticals, Inc. South San Francisco, CA, USA
Introduction: Sample size of a clinical trial is aimed at appropriately powering the study on the basis of one or more primary efficacy and/or safety endpoints. The calculation typically assumes a given effect size and variance of the response variable. For a phase 3 clinical trial investigating ELND005 250 mg twice daily (BID) versus placebo for 78 weeks, various PK sampling designs were compared and selected on the basis of the predicted accuracy of the PK model and impact on PK/PD model estimates.
Objective: To determine and compare power of alternative sampling designs in a Phase 3 clinical trial to detect the difference in clinical endpoints between ELND005 and placebo
Methods: Sample size was fixed at 450 patients per treatment arm, assuming 30% dropout rate by the end of the study. The primary clinical endpoints were expressed as a function of the drug exposure model previously developed. Power was calculated by simulating a large number of clinical trials (1000) and counting the proportion of trials with a statistically significant difference from placebo. For each simulated trial, response was calculated using the PK/PD model, incorporating uncertainty in its estimates as well as uncertainty in the PK model due to its design. Using a base design as a reference: 1 sample collected at pre-dose and 2 post-dose collected with at least 1 hour apart from each other during baseline visit, and 2 blood samples drawn at approximately 5 to 9 hours post-dose with at least 1 hour apart from each other post dose during follow-up clinical visits at weeks 6, 12, 24, 48. Three alternatives A, B and C revised to 2 post dose samples with one soon after dosing (0.5 hours) and the other at ~2 (Design A), ~3 (Design B), or ~4 (Design C) hours post-dose during baseline visit and to 1 sample post dose sample during the rest of the visits were compared.
Results and Discussion: Base design yielded power of ~74-79% for clinical endpoints. Although design C, followed by B then A, had the smallest variance of clearance, they achieve very similar power. Power calculated from all 3 Designs A, B, C was higher by at least 10% in all designs in a slightly increasing order (84-91%, 90-95%, and 90-97%, respectively).
Conclusion: Design B provides an efficient and practical sampling plan with a higher power than the base design, and hence preferred for future Phase 3 study of ELND005.
Reference: PAGE 21 () Abstr 2312 [www.page-meeting.org/?abstract=2312]
Poster: CNS