A General Statistical Approach for Personalized Medicine

• Activity Number: 442
• Type: Invited
• Date/Time: Wednesday, August 1, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Lu Tian*+
• Companies: Stanford University
• Keywords: Personalized Medicine ; Causal Inference ; Interaction

Abstract

The randomized clinical trial provides important information on the comparison between a treatment of interest and its alternative in a target population. However, the ``superiority'' of a treatment established by a randomized clinical trial is only on the average sense, which implies that the superior treatment identified by clinical trial is not necessarily the optimal choice for every individual patient or every subgroup of patients. Ideally, a guideline can be provided to help clinicians to prescribe the treatment to patients who may be benefited from it and avoid to give it to those who may be harmed by the treatment. This requires constructing a scoring system based on personal characteristics and biomarker profile to stratify the population into different subgroups with group-specific treatment effects. In this paper, we propose a unified framework to rigorously evaluate and compare existing scoring systems. The validity of the method does not rely on specific models used to derive the scoring system and the extension of the method can be used to construct, calibrate and combine different scoring systems. Extensive numerical studies will be presented to support the methods.