A Bayesian approach to hazard identification. The case of electromagnetic fields and cancer

This paper discusses certain issues related to uncertainty in hazard identification. Research on the hypothesis that exposure to 50-60-Hz magnetic and electric fields (EMF) increases the risk of cancer has been ongoing for two decades. Epidemiological studies provide a somewhat consistent pattern indicating an increased risk for childhood leukemia and adult chronic lymphatic leukemia and possibly also for other leukemias and brain cancer. However, there is still no good candidate for a mechanism. Epidemiological studies have throughout the two decades been interpreted with great caution, and final evaluations as to carcinogenicity have been deferred. The reason for this carefulness may be the lack of knowledge about a plausible mechanism. The purpose of this paper is to discuss the process of weighing epidemiological data, experimental data, and other background information into a synthesis such that the evaluation can be based on all data combined. A Bayesian approach to this weighing is discussed along with some alternatives. The Bayesian approach provides a structure for the pooling of evidence and points out where subjective judgments come into play.