BACKGROUND: Results to date for the association between use of statins and prostate cancer (PCa) death in observational studies are inconsistent. We investigated the application of causal inference methods, which aim to address observational data as if they were from a randomised clinical trial (RCT).
MATERIAL AND METHODS: We examined the association between statins and PCa-death in 14,926 men in PCBaSe Sweden. We used inverse probability weighted (IPW) estimation of marginal structural models (MSM), as well incorporating truncated IPW in the presence of time-dependent confounders (TDC) (e.g., disease severity), affected by the exposure.
RESULTS: The baseline adjusted odds ratio (OR) was 0.62 (95%CI: 0.51-0.75), which compared risk of PCa-death between men on statins and men not on statins. The calculated IPW for the MSM were highly variable, with the smallest weight at 0.0019 and the largest at 13,574, resulting in an OR of 0.89 (95%CI: 0.69-1.14). Truncating the weights improved variability, reducing the largest weight to 13.16. The truncated MSM OR was 0.86 (95%CI: 0.81-0.91).
CONCLUSION: An association of statins and risk of PCa-death could not be reliably discerned, due to lack of data on essential confounders, namely serum cholesterol levels and disease severity. No observational studies on statin-use to date present information on serum cholesterol levels and disease severity in one setting, highlighting the need for careful interpretation of investigations into drugs in relation to diseases other than their intended purpose in observational settings.
2015. Vol. 116, no 5, 681-683 p.