Endemic statistical paradoxes in epidemiologic studies distort knowledge on prostate cancer: mitigation and caution...

Many studies on epidemiology of prostate cancer (PCa) are based on a diagnosis of PCa using PSA (prostate-specific antigen) level. However, biases can distort the interpretation of the results, which in turn limits policy and decision making on public health prevention strategies or clinical guidelines. The main confusion is to interpret the posterior probability of the outcome following the exposure as a change in the prevalence of the disease outcome, whereas this change reflects only the predictive values of the PSA test induced by the exposure of interest.

Many studies report potential causal factors involved in PCa risk. However, the lack of integration of how physiological changes in PSA values are associated with the exposures being investigated, they explain in part contradictory and controversial results on PCa risk factors in the literature.

A strategy to perform case--control studies based on PSA stratification is suggested to avoid misinterpretation related to PSA misclassification. Real data are analysed, and we show that we can exploit the mechanism of selection biases using different modalities of controls recruitment based on biomarker stratification to distinguish real from false causal factors.