605 – Innovative Data Analysis of Clinical Trials in the VA Cooperative Studies Program
The Best Treatment Modality Selection for Multiple Binary Endpoints
Stephen F. Bingham
VA Cooperative Studies Program
Joseph Collins
VA Cooperative Studies Program
Zhibao Mi
VA Cooperative Studies Program
Dimitri Novitzky
Haley VA Medical Center
The management of brain dead organ donors (BDOD) is complex. The use of inotropic agents and replacement of depleted hormones known as hormonal replacement (HR) is crucial for multiple organ procurements. Yet the optimal HR modality has not been identified, and the statistical adjustment for the best selection is not trivial. Traditional pair-wise comparisons between every pair of treatments, multiple comparisons to all (MCA), are statistically conservative. Hsu's multiple comparison with the best (MCB)-adapted from the Dunnett's multiple comparisons with control (MCC)-has been used for selecting the best treatment with a single continuous outcome. We selected the best HR modality for the success of multiple organ procurements using a two step approach. First, we estimated the predicted margins of the logits by constructing generalized linear models (GLM) or generalized linear mixed models (GLMM), and then we applied the multiple comparison methods to identify the best HR modality given that the testing HR modalities are independent. Among sixteen HR modalities, we found that the combination of thyroid hormone, steroid, vasopressin, and insulin was the best HR modality for multiple organ procurement for transplantation based on the 95% simultaneous confidence intervals or decision limits.
