Behavior within a Clinical Trial and Implications for Mammography Guidelines

Abstract

I unite the economics and medical literatures by examining behavior within a clinical trial to inform treatment guidelines. I use data from the Canadian National Breast Screening Study, an influential clinical trial on mammography. During the active study period of the trial, a substantial fraction of women in the control group received mammograms, and some women in the intervention group did not. Using this mammography behavior, random assignment within the trial, and a standard model from the economics literature, I divide participants into three groups that differ in how likely they are to receive mammograms. Making comparisons across these groups, I find two important relationships. First, I find heterogeneous selection into mammography: women more likely to receive mammograms are healthier. I find this relationship using a marginal treatment effect model that assumes no more than the local average treatment effect assumptions. Second, I find treatment effect heterogeneity along the margin of selection into mammography: women more likely to receive mammograms are more likely to experience harm from them. I find this relationship using an ancillary assumption that builds on the first empirical relationship. I find additional empirical support for the ancillary assumption using baseline covariates. My findings contribute to the literature concerned about harms from mammography by demonstrating variation across the margin of selection into mammography. This variation is problematic for current mammography guidelines for women in their 40s because it implies that they unintentionally encourage mammography for healthier women who are more likely to experience harm from them.