CMStatistics 2021: Start Registration
View Submission - CMStatistics
Title: The subtype-free average causal effect for disease heterogeneity studies Authors:  Daniel Nevo - Tel Aviv University (Israel) [presenting]
Amit Sasson - Tel Aviv University (Israel)
Abstract: A common goal in molecular pathological epidemiology studies is to evaluate whether the effects of risk factors on disease incidence vary across different disease subtypes. A popular approach implements a multinomial regression in which each of the non-zero values corresponds to a bona fide disease subtype. Then, heterogeneity in the exposure effects across subtypes is examined by comparing the coefficients of the exposure between the different subtypes. We explain why this common approach does not recover causal effects, even when all confounders are measured, due to a built-in selection bias in the multinomial regression model. We further develop the Subtype-Free Average Causal Effect (SF-ACE), a well-defined causal effect inspired by the Survivor Average Causal Effect (SACE). We propose identification and estimation approaches for the SF-ACE under different sets of assumptions. Similar to the SACE, the assumptions underlying the identification of the SF-ACE from the data are untestable and can be too strong in some scenarios. Therefore, we also develop a sensitivity analysis to relax some of these assumptions. Finally, we apply our methodology to data from two large cohort studies to study the heterogeneity in the causal effect of smoking on colorectal cancer subtyped by microsatellite status.