Location
MC 6460
Speaker
Dr. Béliveau is an Associate Professor in the Statistics and Actuarial Science Department at the University of Waterloo.
Title
Finding Order in the Complexity of Medical Treatment Comparisons
Abstract
Decision problems with multiple treatment options induce a large space of possible hierarchy questions—for example, whether a given treatment is among the best, whether one option consistently outperforms another, or whether subsets of treatments rank contiguously. The number of such statements grows combinatorially with the number of treatments, making exhaustive evaluation infeasible.
In medicine, network meta-analysis (NMA) combines evidence from multiple clinical trials to compare many treatments simultaneously. Researchers often summarize NMA results through probabilistic statements about treatment hierarchies (e.g., which treatment is likely best), but in practice only a small set of such questions is examined.
We present an algorithmic approach that systematically identifies all binary treatment hierarchy questionswhose empirical support from NMA data exceeds a chosen credibility threshold (e.g., 95%), followed by a pruning step that removes redundant results. The method is illustrated using NMAs of treatments for diabetes and depression.