![IconGems-Print](images/IconGems-Print.png)
512 – Solving Complex Statistical Problems in Network Meta-Analysis
Solving Complex Statistical Problems in Network Meta-Analysis: Discussion
David C. Hoaglin
University of Massachusetts Medical School
i) Speaker: Mireille Schnitzer (mireille.schnitzer@umontreal.ca) Affiliation: Université de Montréal, Faculté de pharmacie Title: Defining model-independent meta-analytical effects of interest using the causal inference perspective ii) Speaker: Jing Zhang (jzhang86@umd.edu) Affiliation: University of Maryland, School of Public Health Title: Bayesian hierarchical methods for meta-analysis combining randomized-controlled and single-arm studies iii) Speaker: Dungang Liu Affiliation: Lindner College of Business, University of Cincinnati Title: Efficient network meta-analysis: A confidence distribution approach 6. Description of session: Modern drug development for the treatment of any disease is fast-paced and highly competitive. The increased speed and level of competition creates a situation where several competing drugs may be compared in various combinations in randomized trials that are conducted within a relatively short window of time. It is often impossible to include every plausibly beneficial drug in every randomized trial. Therefore, there is an urgent need to synthesize information from all available trials rather than restrict to only those trials that have arms for every one of a small set of available treatments. Network meta-analysis methods have been developed to try to leverage all relevant trial data to improve comparisons across all studied treatments. However, the complex nature of the problem immediately presents statistical problems due to missingness, confounding, and selection bias which must be simultaneously and efficiently addressed in order to make well-calibrated decisions. In this way, the proposed program links directly to the conference theme. The conference organizer and one of the speakers are based in Canada. CIHR's Drug Safety and Effectiveness Network (DSEN) has three primary collaborating centres devoted to statistical methodology, one of which is devoted completely to network meta-analysis. Therefore, we believe this session is of great interest to Canadian biostatisticians and graduate students who would be attending JSM. The session will be of particular interest to applied statisticians, as the speakers and the discussant are all members of faculties of medicine or schools of public health and strongly integrate applications with their methods work, so relevance to substantive problems will be emphasized by all four speakers. The audience size could be substantial for this session (no fewer than 30, possibly as many as 60-80), depending on the level of attendance of statisticians collaborating with or working in the pharmaceutical industry.