Biomedical Research Building room 105
Presenter: Kathryn Menta, Epidemiology & Biostatistics PhD trainee
Title: Developing Real-World Dynamic Treatment Regimens in Multiple Sclerosis
Abstract:
Multiple sclerosis (MS) is a chronic, immune-mediated neurological disease requiring lifelong, adaptive management. Despite the availability of more than 20 disease-modifying therapies (DMTs), clinicians and patients continue to face uncertainty about how and when to adjust treatment in response to evolving disease activity. Current guidelines rely largely on expert consensus rather than data-driven evidence on optimal sequencing.
This dissertation proposes to develop and evaluate statistical frameworks for estimating dynamic treatment regimens (DTRs) - data-driven decision rules that specify how therapy should change over time based on patient characteristics, treatment history, and outcomes - using observational data. Aim 1 applies and compares three complementary methods, the parametric g-formula, inverse-probability weighting of marginal structural models (IPW-MSMs), and Q-learning, across two distinct MS datasets: the Cleveland Clinic Mellen Center registry and the multi-institutional TriNetX EHR network. By contrasting a specialty registry rich in disease-specific measures with a large, diverse EHR cohort, Aim 1 will evaluate how data granularity and representativeness influence the validity and generalizability of estimated treatment strategies.
Aim 2 employs simulation studies calibrated to the NIH All of Us Research Program to quantify how violations of key causal assumptions, sequential exchangeability, consistency, and positivity, affect bias and precision in DTR estimation. The simulations will yield practical diagnostic tools and empirically grounded thresholds for assessing robustness in applied analyses.
This seminar will present the design and aims of the proposed dissertation research. The focus will be on methodological framework, simulation design, and anticipated contributions; no empirical data or analytic findings will be presented.
If unable to attend in person in Biomedical Research Building room 105, you may join via Zoom at
Meeting ID: 958 2937 2435
Passcode: 087450