讲座题目：An Adaptive Trial Design to Optimize  Dose–Schedule Regimes with Delayed Outcomes for Ordered  Subgroups
讲座时间：2018年10月16日14:00-15:00

讲座地点：综合大楼601会议室

主讲人简介：Dr. Ruitao Lin received his Ph.D.  degree in 2016 from the Department of Statistics & Actuarial Science at the  University of Hong Kong. Currently, he works at the Department of Biostatistics  at the University of Texas MD Anderson Cancer Center as a postdoctoral research  fellow. His research interests are focused on Bayesian adaptive design, Bayesian  modeling, Clinical trials, Empirical likelihood, Meta-analysis and Missing data.  Most of his research has been published in top journals such as Clinical Cancer  Research[IF:10.2 ], Biometrics, Statistica Sinica, Biostatistics, Statistics in  Medicine, Annals of Applied Statistics, JRSSC, Statistical methods in Medical  research etc.    

摘要：In  this talk, I will introduce a two-stage phase I-II clinical trial design to  optimize dose–schedule regimes of an experimental agent within ordered disease  subgroups. The design is motivated by settings where prior biological  information indicates it is certain that efficacy will improve with ordinal  subgroup level. We formulate a flexible Bayesian hierarchical model to account  for associations among subgroups and regimes, and to characterize ordered  subgroup effects. Sequentially adaptive decision making is complicated by the  problem, arising from the motivating application, that efficacy is scored at day  90 and toxicity is evaluated within 30 days from the start of therapy, while the  patient accrual rate is fast relative to these outcome evaluation intervals. To  deal with this in a practical way, we take a likelihood-based approach that  treats unobserved toxicity and efficacy outcomes as missing values, and use  elicited utilities that quantify the efficacy-toxicity trade-off as a decision  criterion. Adaptive randomization is used to assign patients to regimes while  accounting for subgroups, with randomization probabilities depending on the  posterior predictive distributions of utilities. A simulation study is presented  to evaluate the design’s performance under a variety of scenarios, and assess  its sensitivity to the amount of missing data, the prior, and model  misspecification.