Document resource
Objective Describe the framework for cancer hazard identification and how bias assessment can be used to draw causal inferences from observational epidemiology studies in the International Agency for Research on Cancer (IARC).Methods The framework for IARC Monographs cancer hazard identification and the role of observational studies of human cancer were reviewed. The history of false positive and false negative evaluations (based on re-evaluations) in the Monographs was evaluated. Past monographs were examined to select representative agents and summarize approaches Working Groups have used to rule out (or not) bias or confounding as explanations for positive findings.Results IARC’s hazard identification uses principles of systematic review to conduct synthesis within and integration across evidence streams, with primary attention given to avoidance and management of conflicts of interest. Only 1 of ~100 agents (human papillomavirus type 66) moved from sufficient (to limited) evidence for cancer in humans upon re-evaluation. Of the 15 agents that moved between limited and inadequate evidence, 14 increased in classification and only 1 (coffee) decreased. In the four selected examples (night shift work, red meat consumption, radiofrequency electromagnetic field radiation, and opium consumption), a variety of approaches, including use of directed acyclic graphs, indirect adjustments, and worst-case adjustments, were used to assess uncontrolled confounding, information bias, and selection bias, but there was a lack of common tools described for such analyses in the Monographs in the context of hazard identification.Conclusions The IARC Monographs programme uses a replicable, robust framework for cancer hazard identification, which could benefit from the availability of more tools for bias assessment.