B1824
Title: Multivariate network meta-analysis made simple
Authors: Yong Chen - Univ. of Pennsylvania (United States) [presenting]
Abstract: The growing number of treatment options for many conditions has generated an increasing need for scientifically rigorous comparisons of multiple treatments to inform healthcare decision making. There is a steep increase in the publication of network meta-analysis (NMA) in the past decade. NMA expands the scope of conventional pairwise meta-analysis by allowing comparisons of larger sets of treatment options. However, the existing methods development have been focusing on univariate outcome NMA. In practice, efficacy, safety and patient-centered outcomes are crucial for clinical decision making and must be considered simultaneously. Furthermore, different patients can have different preferences (or utility function) on the tradeoff among these multivariate outcomes. We propose a modeling and inference strategy for multiple outcomes network meta-analysis. The proposed method avoids modeling the complex correlation structure among multivariate outcomes, whilst providing the valid inference on personalized utility functions. We also provide a subject-specific treatment ranking based on the surface under the cumulative ranking curve. The proposed method is validated through simulation studies and illustrated by an example of treatment comparisons for bipolar disorder.
