|
Abstract:
|
For pharmaceutical interventions, it's well known that the strategy of "one-size fits all" is hardly applicable to most common diseases. With recent advances in biological science, personalized medicine has become a very hot topic in pharmaceutical industry. Our work is motivated by a real randomized, double-blind, placebo-controlled phase II dose-ranging study of a novel treatment for asthma patients. Biomarkers for individual patients, including genetic markers and patient baseline demographics, were collected. Compared to the traditional two-arm design, such a dose-ranging study raises a number of analytical issues that cannot be resolved by existing methods. In this paper we propose a novel statistical framework for the dose-ranging study to explore the relationship between the dose, biomarker and clinical outcome and consequently identify the right patient population as well as appropriate doses for future study. The dose-biomarker-outcome relationship is estimated by nonparametric methods, with the constraint of the outcome being monotonic in dose for given biomarkers. Our approach outperforms the existing methods in numerical studies.
|
