Dr. Gabrielle Simoneau, Senior Principal Biostatistician, Biogen Digital Health
Hosted by Dr. Joanna Mills Flemming, CANSSI Associate Director, Atlantic Region
Friday, November 18 | 12:00–12:45 p.m. ET
Biogen Digital Health, Montreal
This event is past. WATCH THE VIDEO RECORDING
Presentation Abstract: Propensity score matching is among the most popular confounding adjustment methods for comparative effectiveness research in non-randomized settings, and these methods are increasingly being used for comparisons of more than two treatments. When pairwise matching is applied to estimate treatment effects for multiple pairs of treatments, the underlying target population changes across the comparisons. This is attributable to the differences in covariate distributions across treatment groups. Consequently, this results in different treatment effect interpretations for the different underlying populations across the multiple pairwise comparisons. The interpretation of treatment effect estimates in relation to the matching (i.e., the target estimand) is rarely clarified and consequently might lead to erroneous conclusions about real-world effectiveness of different treatments. Based on empirical research, we illustrate that multiple pairwise matching for the investigation of comparative effectiveness of more than two treatments can lead to targeting different estimands. We propose visualization tools to illustrate the problem and clarify the connection between estimand, target population and pairwise matching to avoid misinterpretations and treatment decision-making errors in clinical practice.
About Gabrielle Simoneau: Dr. Gabrielle Simoneau obtained her PhD in Biostatistics from McGill University in 2019 under the supervision of Drs. Erica Moodie and Robert Platt. In early 2020, she joined Biogen, a biotechnology company in neurosciences, as a biostatistician in a group focused on analytics and methodological innovations for real-world evidence. Her current responsibilities include methods and strategies for comparative effectiveness research, precision medicine, and digital health data.