Epidemiologic Studies of Biomarkers and Their Role in Carcinogenesis: The Need for a Formal Causal Inference Approach

In this issue of Cancer Epidemiology, Biomarkers & Prevention, Brantley and colleagues investigated the relationships between estrogen metabolites and postmenopausal breast cancer, using data from a nested case-control study within the Nurses' Health Study. One study aim was to investigate the extent to which estrogen metabolism patterns provided further insights into mechanisms in breast cancer development beyond the role of estradiol. In this editorial, we describe the challenges in interpreting results from observational studies of biomarkers and their role in carcinogenesis due to: (i) a general lack of clarity in the research question, (ii) the limits of current knowledge about the complex underlying causal structure involving interrelated biomarkers, and (iii) the limitations in existing data sources (e.g., biomarkers measured at a single time point). We propose that applying a formal causal inference framework in these studies could be a step forward in improving their rigor, by enabling researchers to be more explicit about the causal effects of interest and the assumptions made, and to advocate for the improvement of future studies. See related article by Brantley et al., p. 375.