Abstract / Summary
This paper examines the inferential framework used to attribute the positive association between acid-suppressive therapy (AS) and esophageal adenocarcinoma (EAC) to confounding by gastroesophageal reflux disease (GERD), a common indication for AS. The experimental literature has proposed several mechanisms by which AS may increase the risk of EAC—either by altering the properties of refluxate or enhancing proliferation in reflux-conditioned tissue—and reviews of the in vivo literature have suggested that the AS–EAC association may be causal. At the same time, clinical guidelines have long maintained that this association reflects confounding by GERD, and cite observational studies that either estimate the AS–EAC association among those without GERD or derive the treatment odds ratio upon adjusting for GERD as a main-effect covariate. This paper first outlines the causal hypotheses articulated in the in vivo literature and establishes that the patient’s underlying reflux burden is an effect modifier under each of these hypotheses such that no causal effect would exist in the absence of esophageal reflux. It then argues that the epidemiologic findings presented in clinical guidelines as reassuring evidence against causality have been widely misinterpreted as they do not distinguish confounding by GERD from any of these hypotheses.