Statins and Cancer Prevention—Association Does Not Mean Causation

Emerging research has focused on the association between statins and cancer risk; however, clinical studies have been both few and inconsistent. Statins are a widely prescribed medication class that inhibit HMG CoA reductase and therefore reduce cholesterol synthesis. Given the key role of cholesterol in cancer, this suggests that statins may have anticancer activities (1, 2). In addition, given statins lower inflammation and alter protein prenylation (i.e., preventing activation of the proto-oncogenes Ras and Rho), there are reasons to believe statins may have activity beyond simply lowering cholesterol (3). While many previous studies assessed the association between statins and cancer risk, many focused on a single cancer (4). Understanding the association between statins and cancer more broadly remains understudied.

To address this, in the current issue of Cancer Prevention Research, Maeda-Minami and colleagues examined >50,000 subjects all with hyperlipidemia from a population-based health insurance claims database in Japan and divided patients into two groups: “statin users” and “drug nonusers” (5). Drug nonusers were patients not prescribed any dyslipidemia drugs. Unlike previously published studies, Maeda-Minami and colleagues analyzed multiple types of cancer, subgrouping them based on anatomic location (e.g., “male genital organs,” “digestive organs,” etc…). They adjusted for multiple confounding factors such as age, sex, smoking status, and body mass index. Overall, they found cancer risk was significantly, but modestly, lower among statin users versus nondrug users, with an overall incidence of 7.26 and 8.35 per 1,000 person-years, respectively [adjusted HR = 0.84; 95% confidence interval (CI), 0.72–0.97]. However, in subgroup analysis, cancer risk was significantly lower only for digestive organ cancers (adjusted HR = 0.79; 95% CI, 0.63–0.99). Further analysis showed that the lower cancer risk was only seen in nonsmokers (adjusted HR = 0.78; 95% CI, 0.65–0.92) while null results were seen in smokers (adjusted HR = 1.11; 95% CI, 0.83–1.49; P_interaction = 0.044). Finally, while there was no interaction by statin type (P_interaction = 0.16), results were slightly stronger for lipophilic (adjusted HR = 0.78; 95% CI, 0.65–0.94) versus hydrophilic statins (adjusted HR = 0.89; 95% CI, 0.74–1.1).

As with all studies, the study by Maeda-Minami and colleagues had notable limitations, several of which they addressed (5). First, the age of the subjects was relatively young for a cancer cohort (median age 52) with the vast majority being <65 years. Given cancer is typically a disease of aging, whether similar results would be seen in an older cohort is unknown. Second, follow-up time was short. With a mean follow-up of only 2 years, the number of events would be modest, and the possible long-term effects of statin use cannot be assessed. Thus, longer follow-up is needed. Third, their definition of statin user required patients to have a statin refill within 6 months of the initial statin prescription. Once people are diagnosed with cancer, they often stop statins, and thus, this requirement may have preferentially eliminated cancers in the statin group thereby making statins look “protective” (6). A lag-time analysis, which was not done, would have addressed this issue. Nonetheless, the use of large population data and adjustment for some key confounders were notable strengths. Furthermore, including only initiators of statins reduces bias present in previous observational studies, such as the Hernán and colleagues study that erroneously suggested a protective effect of hormone replacement therapy on coronary heart disease (7).

Ultimately, the authors conclude, “Our results suggest that statin use may reduce cancer risk in patients with dyslipidemia.” However, our interpretation is that rather than statin use “reducing” cancer risk, instead it is “associated” with reduced cancer risk. As we know, mere association does not mean causation. As such, beyond some of the reasons discussed above, what are potential confounding factors and biases that could explain the results? First, let us consider who is the general statin user. This is someone who had their cholesterol checked. Then followed-up with a physician who perhaps suggested lifestyle interventions first. After the lifestyle change failed to sufficiently change their cholesterol, they returned to the physician for another check. Only then did the subject take a statin, while perhaps maintaining their lifestyle interventions. The fact that they remained on the medicine implies they continued to follow-up and stayed on the medicine despite risk of side effects. As such, we can conclude that these patients likely have health-seeking behaviors, perhaps a better lifestyle overall, and certainly are believers in preventative medicine. Given all of this, it is certainly plausible these “benefits” may be related to lifestyle changes rather than the statin itself. Indeed, effects were only seen in digestive organ cancers, wherein lifestyle factors have already been shown to be strongly predictive of cancer risk (8). Further analyses that would lend credibility to this being a true statin effect would be assessing a dose response, which was not done. In short, while the data are intriguing, they are far from conclusive about cause-and-effect.

One interesting observation was that inverse associations were only seen in nonsmokers. Given multiple subanalyses and lack of controlling for multiple testing, it is certainly possible this is a type I error. Nonetheless, under the assumption that this is a true association, possible explanations include (i) the carcinogenic effects of smoking outweigh the “protective” effect of statins on cancer risk in smokers, (ii) smokers are generally less healthy and are less health motivated, such that their risk of developing cancer is greater regardless of statin usage. Ultimately, whether smoking truly modifies the association between statins and cancer risk requires further study.

An alternative approach to studying cancer prevention is to study delaying cancer progression. Indeed, though current research is limited regarding statin use and cancer prevention—as described above and noted by the authors—there is promising evidence of statin usage being associated with cancer progression and survival. A randomized control trial (RCT) showed that pravastatin doubled survival in patients with hepatocellular carcinoma (18 vs. 9 months; ref. 9). As well, an observational meta-analysis found statin use combined with radiotherapy was associated with a 30% reduced mortality in patients with prostate cancer, whereas an observational cohort study showed statin use after androgen deprivation therapy initiation was associated with a 27% decreased risk of PSA relapse and 18% decreased prostate cancer death (10, 11). In an RCT of patients with colorectal carcinoma, statins were associated with improved colorectal cancer–specific survival, but not progression-free survival in patients receiving standard-of-care chemotherapy (12). In a breast cancer meta-analysis of seven observational studies, lipophilic statins were associated with significantly increased recurrence-free survival and overall survival (13). Thus, we propose a more fruitful approach to assess the benefits of statins in cancer is rather may be well-controlled RCTs combined with standard cancer treatment to delay progression in select cancer populations versus a large cancer prevention trial.

Thus, where do we go from here? Should all patients be prescribed statins to prevent cancer? How about only those at increased risk of digestive cancers? We strongly believe, given modest findings from this study, conflicting results in the literature, and importantly null results from RCTs to date, that “healthy” people should not be given statins to reduce cancer risk (4). Ultimately, what is needed to prove statins have cancer prevention effects are large well-conducted clinical trials. These trials would need to be large RCTs with long follow-up periods, such that the median age at the end of follow-up is well passed the median age of that cancer's incidence. However, is it really wise to devote resources to prove a modest 15% reduction in cancer risk? This is especially poignant as there are proven medications that reduce cancer risk and yet are not routinely used. Notably, tamoxifen reduces breast cancer risk by 31%–35% (14). However, tamoxifen is not routinely used in practice to reduce breast cancer risk. Similarly, 5-alpha-reductase inhibitors lower prostate cancer incidence by 23%–25% (15). However, again, these medications are not used clinically as effects were only seen in less aggressive prostate cancer subtypes. Importantly, this suggests that any cancer prevention study for a cancer with decent survival rate, may need an overall cancer-specific survival endpoint, making future large RCTs nearly impossible.

Cancer is a devastating disease that affects millions of patients every year. There is a constant push for novel approaches for prevention and treatment, especially as cancer therapies become more personalized. Statins are an interesting approach as emerging research suggests it may have “benefits” as an adjunctive cancer treatment. What remains unclear is whether statins can help prevent cancer in the first place. And even if statins do help prevent cancer, will such a benefit outweigh the possible side effects? While the study by Maeda-Minami and colleagues helps shed some light on this, the limited study design coupled with only modest “benefits” lend little support to moving statins forward as a cancer prevention tool, given the heavy resource and financial involvement that would be needed versus investing in other emerging agents and targets. Ultimately, what is needed is better biomarkers to identify the subset of people who stand the most to benefit from statins and which cancers are most statin “sensitive.” Only after this is known, can we start to design the RCT that will truly answer whether the associations seen in epidemiologic studies of statins are due to causation or not.