Causal mediation analysis on survival data: application on the national march cottort

Prostate cancer (PC) is the most common cancer in Europe and USA, the second reason of death caused by cancer among men in the world and its incidence has increased in developed countries during the last 20 years. The etiology of PC is not entirely clear and epidemiologic results on risk factors are not consistent. Physical activity (PA) and abdominal fat, measured through waist-to-hip ratio (WHR), have been suggested to have a potential role in the development of the disease. We designed an observational study to test if the effect of PA on PC is mediated by WHR: our aim was to decompose a total causal effect of the exposure on the outcome into two different components. The first one reflects a direct pathway from PA to PC and the second one reflects an indirect pathway, which involves the mediator variable. We used the Swedish National March Cohort data and we performed a mediation analysis using the recent theoretical development in causal mediation analysis field. Preliminary results showed that occupational PA has a role in preventing low-risk PC incidence but risk reduction is not mediated through the WHR control. Thus, other hypotheses need to be tested in order to highlight how PA acts in reducing PC incidence.