We read with interest the letter from Grulich and Vajdic related to our article demonstrating a reduction in the relative risk of prostate cancer (PCa) associated with circumcision. We agree that our observational data cannot prove causation, but exposures such as circumcision are unlikely to be tested in a randomized trial. Thus, one has to weigh the evidence based on the available scientific data. In response, we disagree with their suggestion that the evidence is “tenuous” between sexually transmitted infections (STIs) and the risk of PCa. As we describe in our article, a meta-analysis of 29 studies found a summary relative risk of 1.5 (95% confidence interval [95% CI], 1.3-1.7) for PCa in men with a history of STIs, several STIs have been found in the prostate, and higher serum levels of antibodies (herpes simplex virus, Trichomonas) and prostate human papillomavirus DNA have been associated with PCa.1
We do not believe that the reported lower PCa incidence in men with the human immunodeficiency virus (HIV) compared with men without HIV is informative regarding the relation between STIs and PCa. First, the median age of men with HIV is < 40 years2 compared with 67 years for patients with PCa.3 Furthermore, the incidence of PCa is highly dependent on prostate-specific antigen (PSA) screening rates. PSA testing in HIV-positive men is < 20%4 compared with ≥ 50% in the general population.5 In the pre-PSA era, the incidence of PCa in HIV-positive men was similar to that in HIV-negative men, whereas recent reductions in PCa incidence in HIV-positive men is limited to those with locoregional disease, suggesting that the differences observed are related to PSA screening intensity.4
The authors suggest that selection bias due to socioeconomic status or differing participation between cases and controls may have affected the results of our study.1 It is not the participation level in the cases versus controls that is important, but rather whether there is a difference between participating and nonparticipating cases (or between participating and nonparticipating controls) in the exposure of interest (circumcision). Unfortunately, we have no data concerning nonrespondent circumcision status. We do have economic status measures (income and education) and the inclusion of these factors in multivariate regression models did not change the risk estimates for PCa associated with circumcision. In addition, there was no difference with regard to levels of attained education (P < .48) or household income (P < .16) between cases and controls in our data set.
In total, although we agree that our findings are observational and do not prove causality, we believe there is strong biological plausibility for a relation between circumcision and the risk of PCa.