A Marginal Structural Model for Normal Tissue Complication Probability

• Activity Number: 351 - Contributed Poster Presentations: Statistical Society of Canada
• Type: Contributed
• Date/Time: Tuesday, August 9, 2022 : 2:00 PM to 3:50 PM
• Sponsor: SSC (Statistical Society of Canada)
• Author(s): Thai-Son Tang* and Zhihui (Amy) Liu and Olli Saarela
• Companies: University of Toronto and Princess Margaret Cancer Centre, University Health Network and Dalla Lana School of Public Health, University of Toronto
• Keywords: marginal structural models; dose-volume histograms; normal tissue complication probability; multiple monotone regression; radiotherapy treatment planning; stochastic interventions

Abstract

The goal of radiation therapy is conformed delivery of radiation dose to target organs while minimizing exposure to surrounding tissue to avoid normal tissue complications. As such, dose-volume histograms (DVHs), which characterize the functional relationship between radiation dose and organ volume, are focal in guiding treatment planning. Normal tissue complication probability (NTCP) modelling has centered around making patient-level predictions with DVHs, but few have considered adapting a causal framework to evaluate the comparative effectiveness of treatment plans. We present causal estimands for functional DVH exposures based on the stochastic interventions framework and propose estimators based on marginal structural models that parametrize bivariable monotonicity between dose, volume, and NTCP to reflect the natural biological mechanisms between these quantities. The properties of these estimators are studied through simulations, along with an illustration of their use in the context of anal canal cancer patients in Ontario.