Designing, implementing, and interpreting clinical trials of therapies for the treatment of unusual but devastating acute illnesses (e.g., toxic exposures, envenomations) is difficult: eligible subjects are scarce; continuing to randomize patients evenly while comparative evidence on efficacy accumulates may not be ethical; and the proper interpretation of limited data may require expert opinion or external information. Further, the decision to terminate the trial at an interim analysis may be best addressed using decision analysis. Our broad objective is to develop Bayesian decision-theoretic clinical trial designs which use partially adaptive allocation to maximize patient benefit.
Specific aims are: (1) to delineate the operational characteristics of these decision-theoretic trial designs; (2) to investigate numerical methods for more computationally efficient trial design; and (3) to construct web-enabled applications for the elicitation and synthesis of expert opinion and for the real-time distributed conduct of adaptive trials. The characteristics of trial designs generated will be tested by both Monte Carlo simulation and through the use of publicly-available software (WinBUGS). Optimal trial designs will be compared to designs obtained using more computationally efficient approximate methods (e.g., forward search methods). When this work is complete, the modular software system will be publicly available for the design and implementation of adaptive Bayesian clinical trial designs. Dr. Lipsky will concurrently serve as a statistical consultant for interdepartmental clinical trials, under the guidance of Dr. Lewis, to further his abilities in trial design and analysis. ? ? ?
| Lipsky, Ari M; Lewis, Roger J (2013) Response-adaptive decision-theoretic trial design: operating characteristics and ethics. Stat Med 32:3752-65 |
| Lipsky, Ari M; Greenland, Sander (2011) Confounding due to changing background risk in adaptively randomized trials. Clin Trials 8:390-7 |
| Lipsky, Ari M; Gausche-Hill, Marianne; Vienna, Muna et al. (2010) The importance of ""shrinkage"" in subgroup analyses. Ann Emerg Med 55:544-552.e3 |
