Reply to D. Tural et al.

Abstract

We thank Tural et al for their interest in our recently published study on the use of statins and prostate cancer death. We would like to clarify some of the inaccuracies stated in their letter regarding our study. As acknowledged in our article, there were missing data on T stage, Gleason score, and prostate-specific antigen levels. However, Tural et al claim that such missing information may have biased the observed effects. Although we agree that missing data can lead to residual confounding, our models were adjusted for a wide range of potential confounders, which included prostate cancer–related treatments that are likely good proxies of prostate cancer aggressivity. Moreover, adjustment for these and other potential confounders had little effect on the hazard ratios (crude: 0.75 v adjusted: 0.76). This is reassuring and suggests that confounding had a minimal role. Tural et al also state that our models did not consider the use of antihypertensive and antidiabetic drugs, such as metformin. This is inaccurate, as shown in Table 1 of our article, in which the models were clearly adjusted for all known antihypertensive and antidiabetic drug classes for the same reasons noted by the authors. Finally, the authors claim that residual confounding explains why the upper limit of one of the statin cumulative duration categories for all-cause mortality is close to 1.00 ( 36 months, hazard ratio, 0.82; 95% CI, 0.69 to 0.97). Although residual confounding is always possible in observational studies, the width of confidence intervals (ie, precision of the point estimates) plays no role in confounding. Indeed, it is possible to have a biased estimate with narrow confidence intervals, and vice versa. These are two distinct concepts that Tural et al appear to confuse. To conclude, we agree with Mucci and Stampfer that carefully designed observational studies using large electronic databases have the potential to provide much-needed information on the long-term effects of statins on prostate cancer outcomes. These studies can then be used to inform the design of randomized controlled trials that assess the antineoplastic effects of statins in patients with prostate cancer.