科学研究
学术报告
当前位置: 学院主页 > 科学研究 > 学术报告 > 正文

Enhanced Double Robust Estimate of Treatment Difference in Nonrandomized Studies

发布时间:2020-12-17 作者: 浏览次数:
Speaker: Ming Tan DateTime: 2020年12月22日上午10:00-11:00
Brief Introduction to Speaker:

Dr. Tan has longstanding research interests in methods for the design, monitoring and analysis of clinical trials and predictive analytics in clinical trials and Big data. One of his current research foci is statistical methods for searching and evaluating multi-drug combinations utilizing both experimental data and system biology, innovative methods to optimally design and efficiently analyze pre-clinical drug combination therapies in cancer by integrating concepts in modern statistical and number-theoretic methods and pharmacology; and high dimensional genomics data analysis in Cancer Epidemiology, all funded by R01 grants from the NCI and NHLBI. Dr. Tan also has extensive collaborative research experience in the design, conduct and analysis of clinical trials (in both multi-center and single institutional settings), laboratory investigations, biomarker evaluation, genomics and epidemiological research. Dr. Tan has served on multiple NIH panels and scientific review boards, Data and Safety Monitoring Boards, and FDA Advisory Committees. He is an elected Fellow of the American Statistical Association and an elected Member of the International Statistical Institute. Dr. Tan is Executive Editor of Molecular Carcinogenesis; and Associate Editor of Statistics in Medicine and Drug Design, Development and Therapy, and has served on the editorial board for Biometrics. Before joined Georgetown in the fall of 2012, he was professor of Epidemiology and Public Health and Head of the Division of Biostatistics and Bioinformatics of the University of Maryland School of medicine and Marlene and Stewart Greenebaum Cancer Center (UMGCC) (2002-2012). He was previously a senior member (faculty) at St. Jude Children's Research Hospital Cancer Center and biostatistics director of St Jude's Developmental Therapeutics for Solid Malignancies Program (1997-2002), assistant and associate staff/professor of Biostatistics and Epidemiology at the Cleveland Clinic (1990-1997).

Place: 腾讯会议(会议号请联系左国新老师索取)
Abstract:We first consider a motivating example where a clinical trial is designed to compare a new therapy with an external control which are obtained using real world data such as historical database, disease registry or electronic health record. This design is at the forefront of drug registration trials when RCT is not feasible. The goal is to compare relapse-free survival (RFS) rate at 3 years between locally treated high-risk ocular melanoma patients on adjuvant combination immunotherapy versus a matched contemporaneous control population. With the non-randomized comparison, causal inference is necessary. However the doubly robust procedure has been shown to several major drawbacks, for example, too sensitive to mild model misspecifications. This motivated us to extend the double robust causal inference procedure to obtain estimate of treatment effect that has enhanced robustness with respect to misspecification of propensity score and the outcome model.