For ethical and economic reasons, many Phase III cancer clinical trials incorporate group sequential interim monitoring to permit appropriate early stopping in the presence of a clear treatment effect. Frequently, staggered entry of patients into these studies is an operational necessity which can complicate the distributional properties of the test statistics used. This difficulty is exacerbated when multiple outcome are being considered, especially if one or more of the outcomes is a failure time. However, a number of significant advances in survival analysis theory and implementation over the last several decades have resulted in a rich variety of statistics which can be sensitive to many different alternative hypotheses: but the flexibility and applicability to the group sequential setting is severely limited by the analytic complexity of the underlying distributions. The proposed research will seek to address these issues through the following interrelated goals: (1) Develop flexible Monte Carlo methods for accurately determining the null distribution of multivariate statistical tests in a group sequential clinical trial with staggered entry of patients; (2) Develop versatile survival analysis test procedures possessing improved flexibility, which can be generalized to permit tied data, stratification, and the comparison of more than two treatments; (3) Evaluate and compare these versatile procedures using analytic, data, and simulation studies so that clear criteria for optimal use can be established; (4) Further extend these results for evaluating power and sample size requirements in group sequential designs; (5) Use these Monte Carlo procedures to construct flexible multivariate group sequential boundaries which correspond to hypotheses which are clinically relevant; (6) Develop a suitable method of assessing multivariate information accrual so that the alpha-spending approach for designing sequential clinical trials can be applied to this setting; and (7) Implement these methods in flexible, well documented, and user-friendly software. The theme for this research is increasing appropriate utilization of multiple endpoint data in cancer clinical trials through development of flexible multivariate test statistics in a group sequential setting.
| Nadkarni, Nivedita V; Zhao, Yingqi; Kosorok, Michael R (2011) Inverse regression estimation for censored data. J Am Stat Assoc 106:178-190 |
| Cao, Hongyuan; Kosorok, Michael R (2011) Simultaneous Critical Values For T-Tests In Very High Dimensions. Bernoulli (Andover) 17:347-394 |
| Zhao, Yufan; Zeng, Donglin; Socinski, Mark A et al. (2011) Reinforcement learning strategies for clinical trials in nonsmall cell lung cancer. Biometrics 67:1422-33 |
| Ma, Shuangge; Kosorok, Michael R (2010) Detection of gene pathways with predictive power for breast cancer prognosis. BMC Bioinformatics 11:1 |
| Kosorok, Michael R (2009) What's So Special About Semiparametric Methods? Sankhya Ser B 71-A:331-353 |
| Zhao, Yufan; Kosorok, Michael R; Zeng, Donglin (2009) Reinforcement learning design for cancer clinical trials. Stat Med 28:3294-315 |
| Song, Rui; Kosorok, Michael R; Fine, Jason P (2009) On Asymptotically Optimal Tests Under Loss of Identifiability in Semiparametric Models. Ann Stat 37:2409-2444 |
| Kosorok, Michael R (2009) Rejoinder on Discussion of: What's So Special About Semiparametric Methods? Sankhya Ser B 71-A:369-371 |
| Kosorok, Michael R (2009) On Brownian Distance Covariance and High Dimensional Data. Ann Appl Stat 3:1266-1269 |
| Ma, Shuangge; Kosorok, Michael R (2009) Identification of differential gene pathways with principal component analysis. Bioinformatics 25:882-9 |
Showing the most recent 10 out of 12 publications
