Aims Publications Software Investigators Led by: Anastasios (Butch) Tsiatis, PhD
In clinical practice, treatment of cancer is a dynamic process involving a series of therapeutic decisions over time. However, most cancer clinical trials focus on effects of treatments given at a single decision point in the course of the disease, e.g., the selection of a first-line chemotherapeutic option for patients with Stage IIIB/IV non-small cell lung cancer. Conclusions on the best overall strategy over the series of key decision points in the disease are consequently cobbled together from the results of many such single-decision studies, and, due in part to the possibility that the treatment given at one point in time may have delayed effects on the efficacy of future treatment, may be misleading and, indeed, deleterious. In some chronic disease/disorder areas, notably behavioral disorders and infectious diseases, there has been a growing recognition that this myopic point of view may not result in patients receiving the best sequence of treatments and that the entire sequential decision-making process must be studied as a whole in order to identify strategies that are the most beneficial.
This perspective has led to considerable recent interest in methodology for developing and studying dynamic treatment regimes. A dynamic treatment regime is a set of sequential decision rules dictating at each decision point the selection of the next treatment for a patient based on information on the patient, including measures of disease progression, biomarkers, and previous treatment, up to that point, thereby individualizing each step of treatment to the patient. With more than one option at each decision point and numerous possibilities for synthesizing the available information at each into decision rules, many regimes may be conceived, and identifying the optimal regime, that leading to the most benefit if followed over the course of the disease by the population of cancer patients, presents many challenges. We propose four specific aims that will lead to advances in methodology for discovering and evaluating dynamic treatment regimes:
Aim 1: To develop and evaluate learning methods for optimal dynamic treatment regimes.
Aim 2: To develop methods for identifying optimal dynamic treatment regimes from a restricted, feasible set of prognostic variables.
Aim 3: To develop and evaluate methods of statistical inferential for dynamic treatment regimes.
Aim 4: To develop methods for the design of sequentially randomized cancer clinical trials for dynamic treatment regimes.
The overarching goal of this project is to catalyze a paradigm shift in the way cancer therapies are conceived and evaluated that has the potential to make evidence-based, individualized treatment strategies a reality.