Statins and Colorectal Cancer

Paul Lochhead; Andrew T. Chan

Disclosures

Clin Gastroenterol Hepatol. 2013;11(2):109-118. 

In This Article

Observational Studies of Statins and Colorectal Cancer

Case-Control Studies

Case-control studies are generally able to efficiently examine associations between remote exposures and risk of disease, whereas cohort studies can be valuable in dissecting time-dependent associations between exposures and outcomes. A large number of observational studies have assessed the association between statin use and colorectal cancer ( Table 2 ). Perhaps the most publicized and highly cited of these studies is the analysis by Poynter et al,[26] based on the Molecular Epidemiology of Colorectal Cancer study. This was the first well-powered, case-control study of statins and colorectal cancer, including 1953 cases and 2015 controls derived from the population of northern Israel. After adjustment for multiple potential confounders, Poynter et al[26] observed an impressive 43% reduction in colorectal cancer risk associated with at least 5 years of self-reported statin use. Interestingly, a subsequent genetic association study, performed in the same parent case-control study, showed that a polymorphism in the HMG-CoA reductase (HMGCR) gene modified the association between statins and colorectal cancer and was associated with lower LDL cholesterol levels in both cases and controls.[65] A nested case-control study in veterans with diabetes, including 6080 colorectal cancer cases and 24,320 controls, showed a more modest reduction in colorectal cancer risk (9%) associated with statin use, but without significant dose- or duration-dependent relationships.[29] Furthermore, the protective association was limited to colon (not rectal) cancer in patients older than age 65 years without a history of previous polyps.[29] A German population-based case-control study, comprising 540 cases and 614 controls, showed an inverse association between colorectal cancer risk and statin use of 1 to 4 years, but, interestingly, no additional risk reduction was observed for use beyond 5 years.[27] In contrast, at least 12 case-control studies from the United States, Europe, and Asia have reported no significant associations between statin use and colorectal cancer risk.[23,24,66,67,68,69,70,71,72,73,74,75] Indeed, in a study conducted in the United Kingdom, using an electronic general practice prescribing database, an increased risk of colorectal cancer was associated with statin use of longer than 4 years.[74] A limitation of these data are that case-control studies are particularly susceptible to recall bias, or biases arising from the selection of cases and controls. Moreover, in the case-control study by Poynter et al,[26] more than two-thirds of cases and controls were Ashkenazi Jews, potentially limiting the generalizability of the findings to other populations.[20]

Cohort Studies

Data from cohort studies are more limited. A retrospective cohort study conducted in a large US veterans population, containing more than 37,000 statin users, reported a 35% reduction in colorectal cancer risk associated with statin use, and also found a significant decrease in risk with increasing statin dose.[28] In a combined analysis of prospectively collected data from the Health Professionals Follow-up Study and Nurses' Health Study, statin use was not associated with overall colorectal cancer risk, although an inverse association was observed for rectal cancer risk (relative risk, 0.59).[76] Analysis of incident colorectal cancers over almost 11 years of follow-up in the Women's Health Initiative revealed no difference in incidence according to statin use, but did report a borderline statistically significant risk reduction associated specifically with lovastatin use.[77] Eight additional large cohort studies reported neither beneficial nor harmful associations between statins and colorectal cancer.[25,78–84] Compared with case-control analyses, cohort studies are generally less vulnerable to recall and selection biases but can be limited by sample size and insufficient follow-up time for incident cancers. Some cohort studies encompass a shorter follow-up time than several of the statin RCTs.[78,79,82]

Meta-analyses of Observational Studies

Among 4 meta-analyses that have included data from case-control and cohort studies,[20,22,61,85] 1 study found no association between statin use and colorectal cancer risk,[22] whereas 3 studies reported risk reductions of more modest magnitude (risk estimates, 0.86–0.91).[20,61,85] In 2 of these meta-analyses, a significant association between statin use and colorectal cancer risk was observed only in fixed-effects models but not in random-effects models.[20,61] The case-control study by Poynter et al[26] appears to have been a major source of interstudy heterogeneity in the meta-analysis by Bonovas et al,[61] suggesting that it may represent an outlier. The Poynter et al[26] study also appears to have been the driver of the protective association observed in the meta-analysis by Browning and Martin;[20] because, when the Poynter study[26] was excluded, the association between statin use and colorectal cancer was null.

General Limitations of Observational Data

The main limitation of both case-control and cohort studies is that they are prone to residual confounding by unmeasured exposures or behaviors that are associated with both statin use and colorectal cancer incidence. Indeed, evidence suggests that statin use is associated with higher socioeconomic status,[86] nonsteroidal anti-inflammatory drug (NSAID) use,[69] and health-conscious behaviors, such as multivitamin use and screening colonoscopy.[69] Although most observational studies adjust for use of aspirin and NSAIDs, which are consistently associated with lower colorectal cancer risk, information on cancer screening and other risk modifiers, such as physical activity, are not collected universally. Furthermore, residual confounding is impossible to exclude, even in the best-designed studies.

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....