Response: Re: Fruit and Vegetable Intake and Overall Cancer Risk in the European Prospective Investigation Into Cancer and Nutrition

Extract We agree with Bouchardy et al. on the fact that a relatively weak association of fruits and vegetables with reduced overall cancer risk might mask a stronger association with some specific cancers. In our study, we mentioned the analyses that were performed on specific cancers using the same study population; however, we believe that an estimate of the association with all cancers combined could be interesting from a scientific and a preventive viewpoint. The comments by Ioannidis and Siontis are very pertinent, but we do not think they contradict our interpretation of the data. We did not claim that our results provide decisive support to the hypothesis that high intake of fruits and vegetables is associated with reduced cancer risk. Epidemiological research in cancer etiology is increasingly facing the challenge of weak associations, which could be explained by residual confounding and other sources of bias, and particular care should be used in their interpretation. Considerations about residual bias and confounding, as well as exposure misclassification, which could attenuate an existing association, are as important in the interpretation of epidemiological findings as are the estimate of the strength of the association and the associated level of statistical significance, no matter whether they are interpreted within a frequentist or Bayesian framework. We do agree with Ioannidis and Siontis, however, that a Bayesian approach in the interpretation of epidemiological associations is particularly useful when the studies are large in size, as it is the case of the European Prospective Investigation into Cancer and Nutrition.