From testing to estimation: the problem of false positives in the context of carcinogen evaluation in the IARC monographs.

BACKGROUND Cancer epidemiology has been criticized for producing false-positive associations. The present analysis investigates the frequency of and factors contributing to false-positive findings in cancer epidemiology. METHODS The International Agency for Research on Cancer (IARC) Monographs Group 3 agents were examined to identify potential false-positive findings. Frequency estimates for their occurrence were calculated. Comments of the Working Groups on study quality were recorded for studies with potential false-positives. These were used to determine how many of such studies were criticized for each of the study quality factors that are suspected to contribute to false-positive results. RESULTS Of 509 agents in group 3, 37 agents were found to have potential false-positive associations in the studies reviewed in their respective IARC monograph(s). The overall frequency of potential false-positives among these agents was between 0.03 and 0.10. The individual frequencies ranged from 0.01 to 0.40. The potential false-positive findings were produced by 162 studies. The most common factors contributing to potential false-positive findings were confounding and exposure misclassification. CONCLUSIONS The frequency estimates we have obtained do not suggest that epidemiology is grossly flooded by false-positive findings. The factors for which studies with potential false-positives were most often criticized were factors that are sometimes difficult to address in cancer epidemiologic research and can bias an effect estimate toward or away from the null. IMPACT The low frequency of false-positives in cancer epidemiology restores faith in epidemiologic procedures, making epidemiologic findings a useful guide for public health care measures.