Dr. Philippe Boileau
Assistant Professor, Department of Epidemiology, Biostatistics, and Occupational Health and Department of Medicine
I鈥檓 broadly interested in the development of assumption-lean statistical methods and their application to quantitative problems in the health and life sciences. Assumption-lean procedures combine causal inference and machine learning techniques to avoid unjustified assumptions about the data-generating process, encouraging dependable statistical inference. I鈥檓 also committed to the development open-source statistical software and, more broadly, to the adoption of reproducible research practices. My recent work has focused on causal machine learning methods for the analysis of clinical trial data.
1. Assessing treatment effect heterogeneity in clinical trials through differential variability analyses:
Clinical trials are generally underpowered for the detection of heterogeneous subgroups, patient subpopulations that respond differently to investigatory therapies. We aim to demonstrate that heterogeneous subgroups are reliably detected in clinical trials using recently developed causal machine learning methods for differential variability estimation.
2. Treatment effect modifier discovery in clinical trials with time-varying treatment:
Causal machine learning methods were recently developed for treatment effect modifier discovery in randomized control trials with parallel group designs. Methods have not yet been developed for more complex clinical trial designs; reliable effect modifier quantification remains challenging in these settings. We aim to extend existing methodology to permit treatment effect modifier discovery in trials with time-varying treatments.
3. Identifying heterogeneous subgroups in administrative health data to inform future trial design:
Causal machine learning methods were recently developed for treatment effect modifier discovery in randomized control trials with parallel group designs. These methods are also well-suited for identifying drivers of treatment effect heterogeneity in observational data. We aim to demonstrate this in a pharmacoepidemiologic analysis of administrative health data.
Dr. Boileau is currently looking to recruit graduate students.