Speaker
Description
Over the last decades, considerable scientific efforts have been made to determine whether exposure to radiofrequency electromagnetic fields (RF-EMF) below guideline levels may affect cancer risk. The widespread use of handheld mobile phones in the general population, that developed from none to essentially 100% in less than two decades, makes this a potentially very important public health issue. The research field has to a large extent been driven by epidemiological studies, some of which reported increased risks of brain tumours, while others found no associations. Notably, raised risk estimates have been reported in some studies with a case-control design, while the few cohort studies found no increased risks. Case-control studies with retrospectively collected exposure information are subject to several sources of bias which may have influenced their findings, such as differential recall bias and selection bias from non-participation. These biases are especially problematic for case-control studies of brain tumours, as the disease often affects memory, progress rapidly and has a poor prognosis. Cohort studies with prospectively collected exposure information are not affected by differential recall or selection bias but may instead be subject to non-differential exposure misclassification, especially if they lack quantitative exposure data. The COSMOS cohort study was initiated to address these limitations, through prospective collection of exposure information to achieve the same level of detail in exposure data as the case-control studies, but without differential recall bias because all participants are blind to their disease status when the information is collected (brain tumours will occur in the future), and with no selection bias as all participants can be followed in nationwide well-established cancer registers and population registers. This presentation will discuss the theoretical principles of epidemiological case-control and cohort designs and highlight similarities and differences, particularly with regard to exposure misclassification and selection bias, and implications for the interpretation of findings and future prospects.