Residential 222Rn exposure and lung cancer: testing the linear no-threshold theory with ecologic data

In most rigorous epidemiologic studies, such as case-control and cohort studies, the basic unit of analysis is the individual. Each individual is classified in terms of exposure and disease status. However, in ecologic epidemiologic studies, the unit of analysis is some aggregate group of individuals. Summary measures of exposure and disease frequency are obtained for each aggregate, and the analyses focus on determining whether or not the aggregates with high levels of exposure also display high disease rates. The ecologic study design has major limitations, including ecologic confounding and cross level bias. Cohen has attempted to circumvent these limitations by invoking the linear no-threshold theory of radiation carcinogenesis to derive aggregate "exposures" from individual-level associations. He asserts that, "while an ecologic study cannot determine whether radon causes lung cancer, it can test the validity of a linear-no threshold relationship between them." Cohen compares his testing of the linear no-threshold relationship between radon exposure and lung cancer to the practice of estimating the number of deaths from the person-rem collective dose, dividing the person-rem by the number of individuals in the population to derive the individual average dose, and then determining individual average risk by dividing the number of deaths by the number of individuals in the population. We show that Cohen's erroneous assumptions concerning occupancy rates and smoking effects result in the use of the wrong model to test the linear no-threshold theory. Because of these assumptions, the ecologic confounding and cross level bias associated with Cohen's model invalidate his findings. Furthermore, when more recent Iowa county lung cancer incidence rates are regressed on Cohen's mean radon levels, the reported large negative associations between radon exposure and lung cancer are no longer obtained.