Abstract
The International Agency for Research on Cancer (IARC) conducted a quantitative bias analysis (QBA) of talc use and ovarian cancer in Monograph 136. While the inclusion of a QBA was an important improvement compared to prior monographs, it was based on "best guesses" of sensitivity and specificity, rather than available data on talc recall. IARC incorporated some uncertainty in its analysis, but did not consider uncertainty around each sensitivity and specificity value. IARC concluded that a positive association between talc and ovarian cancer was credible but, even setting aside methodological issues, the QBA clearly showed that cohort study results were very similar before and after adjustment with various assumptions about sensitivity and specificity, and that case-control studies results were greatly attenuated. Thus, IARC's conclusions are inconsistent with the analyses presented in the Monograph, which clearly demonstrate that exposure misclassification could fully explain associations in case-control studies, and that epidemiology evidence does not support an association between talc and ovarian cancer. We propose that future IARC QBAs rely on empirical data rather than expert guesses (when possible), are fully transparent, consider all relevant information (including dose-response data), and use probabilistic and Bayesian analyses to address uncertainties.