Recent observational studies have suggested that statins have a protective effect against cancer. However, long-term statin users were shown to be healthier, less frail, and more adherent to therapy and screenings. We conducted a cohort study to evaluate the effect of statins on several common cancers in a large elderly population.
Statins have been shown to reduce cardiovascular morbidity and mortality in many studies, leading to their extensive use.1 Several recent observational studies have suggested that the incidence of certain cancers may be decreased by the use of statins, which has led to clinical trials being conducted to evaluate their effects on cancer prevention.2 However, little reduction in cancer incidence, if any, has been shown in most randomized trials and meta-analyses.3-6 Compared with non—statin-users, long-term statin users have been shown to be more adherent to therapy, less frail, and healthier than those who do not take statins.7-9 This could explain the apparent protective effects of this class of drugs on other outcomes in observational studies. To evaluate the effect of statins on the incidence of colorectal, lung, and breast cancer, we performed a cohort study among a large population of elderly patients.
Patients and methods
Over a period of 9 years, we linked patient information from a large state prescription benefit program with Medicare health care utilization data and cancer registry data. We identified all patients who filled a prescription for a statin drug and included only new statin users, that is, patients who had not filled a prescription for a statin at least 12 months before the first prescription. New users of a glaucoma medication served as a comparison group. To ensure that patients were adherent to their medication regimen, we required statin initiators to fill 3 or more prescriptions for any statin drug (lovastatin [Altocor, Altoprev, Mevacor], pravastatin [Pravachol], simvastatin [Zocor], fluvastatin [Lescol], atorvastatin [Lipitor], or cerivastatin [Baycol]) during the first 180 days after the first prescription and every 180 days thereafter. The same criteria for adherence were applied to glaucoma users. Outcomes included all registry-identified cases of colorectal, lung, and breast cancer. For colorectal and breast cancer, we excluded patients with in-situ cancer from the primary analyses to avoid screening bias. Multivariable Cox proportional models were used to adjust for health care utilization measures, prevention-related activities, known risk factors for the selected cancers, race, age, and sex. We also identified statin and glaucoma drug users from the 1999-2001 Medicare Current Beneficiary Survey (MCBS) data and estimated the prevalence of factors that were unmeasured in our data. To determine the effect of unmeasured factors such as a family history of cancer, aspirin use, and smoking, we performed sensitivity analyses.
We identified 24,439 statin initiators and 7284 glaucoma drug initiators. Statin users were slightly younger than glaucoma users; however, their characteristics regarding risk factors for the cancers of interest, use of preventive services, and health service utilization were similar. The incidence of cancer in our cohort was comparable to the age-sex-standardized incidence for the general US population based on Surveillance Epidemiology End Result (SEER) data. The mean follow-up period was 2.9 years, and the longest follow-up period was 8.4 years. Most cases of cancer in statin users (41% of lung cancer, 61% of colorectal cancer, and 77% of breast cancer) occurred after 3 years of drug use.
Multivariate adjusted hazard ratios (HRs) were 1.11 (95% confidence interval [CI], 0.77-1.60) for lung cancer, 0.96 (95% CI, 0.70-1.31) for colorectal cancer, and 0.99 (95% CI, 0.74 - 1.33) for breast cancer. Compared with glaucoma users, there was no significant increase or decrease in the risk of cancer for statin users. The results of secondary analyses, which included in-situ cases for the outcome for colorectal and breast cancers, were similar to the main analyses: 0.97 (95% CI, 0.74 - 1.28) for colorectal cancer and 0.93 (95% CI, 0.68 - 1.26) for breast cancer. The HR estimates comparing the hydrophobic statins (simvastatin, lovastatin, fluvastatin, and atorvastatin10) or pravastatin alone to glaucoma drugs were not significantly different from the overall results (point estimates ranged from 0.87 - 1.18).
Results of the proportional hazard test were significant for breast cancer ( = .003), suggesting that the effect of statin use might be different over time. The short-term (≤3 years) effect of statin use tended to be protective when we estimated the duration effect of statins (adjusted HR for statins among short-term users = 0.66 [95% CI, 0.42 - 1.05]; adjusted HR for statins among long-term users = 1.28 [95% CI, 0.90 - 1.84]).
In MCBS data, statin users were more likely to take aspirin, were more likely to have been smokers, had a slightly higher body mass index, were less functionally limited, and were more educated than glaucoma users. As indicated by the 2 circled points in the Figure, sensitivity analyses showed that the corrected HR was 1.02 for aspirin use and 0.90 for smoking, according to the estimated prevalence of aspirin use and smoking in the MCBS data. It can be assumed that the prevalence of family history of cancer in statin users would be in the range of 5% to 30% based on the estimated prevalence (15%) in the general population because there is no evidence that a family history of cancer is associated with statin or glaucoma use. Based on this assumption, the corresponding range of corrected HR would be 0.87 - 1.07 (Figure).
Figure. Sensitivity analysis: effect of unmeasured confounding. The curves represent how
the "true" or corrected relative risk (RR) changes with various values of prevalence of
unmeasured confounders in statin users. For glaucoma drug users, the prevalence of
aspirin use and smoking is fixed using the estimates for glaucoma users in Medicare
Current Beneficiary Survey (MCBS) data, and the prevalence of family history of lung and
breast cancer is assumed to be 15% based on estimates from the general population.
The 2 circled points represent the corrected RR for the values of prevalence of smoking
and aspirin use in statin vs glaucoma users in MCBS data. No estimates for family history
of cancer were available from MCBS data, but we expect that the prevalence in statin
users is similar to 15%, which is the estimate for the general population.
Compared with elderly glaucoma users, there was no significant increase or decrease in the risk of cancer among elderly statin users. One of the strengths of this study was that it was based on a very large group of elderly patients with a high risk of cancer. In addition, analyses were limited to new statin users, removing the possibility of biases that can occur when including prevalent statin users.11 Registry-validated cancer diagnosis was used to identify cancer cases. We also used a comparison group that had adherence rates for screening procedures and health-seeking behaviors that were similar to the statin users.7
Contrary to the results of previous studies,12,13 our study showed that statin use does not have an effect on the risk of cancer in elderly patients. A 68% decrease in the risk of breast cancer was shown in one study.12 Another population-based case-control study showed that patients aged 60 years or older who had used statins for more than 5 years had a 47% decrease in colorectal cancer.13 In an earlier study,14 however, we showed that patients receiving long-term statin therapy are different from those who do not take statins. Colorectal cancer risk may be decreased by a healthier lifestyle and the health-seeking behavior of longterm adherent statin users. Furthermore, statin users may be more likely to undergo screening for precancerous colorectal polyps, resulting in their removal, giving the impression that statins may have a protective effect. Previous studies may not have considered this possibility in their analyses. In our current study, we adjusted for possible confounding factors related to a healthier lifestyle by comparing statin users with patients who were taking a different kind of preventative medication—glaucoma drugs—which, like statins, are less frequently prescribed to elderly patients.7,9
The findings from our study suggest that the short-term effects of statin use on breast cancer may be different from the long-term effects. There was a slight protective effect from short-term use, whereas there appeared to be a slight but nonsignificant increase in cancer risk with long-term use. The possibility that long-term statin users may have an increased risk of cancer cannot be excluded based on the results of our study.
One of the limitations of our study was that some of the possible confounders, such as obesity and smoking, were incompletely measured or (as with family history of cancer or aspirin use) were not measured at all. By the choice of the comparison group, however, we may have been able to decrease the bias due to this lack of data. Although some of the possible confounding factors were not measured, the imbalance was not significant enough to cause a major bias in our calculations as shown in our sensitivity analyses. We considered glaucoma drug users a more suitable comparison group than nonstatin users; however, the group of glaucoma drug users was comparatively small, which may have been responsible for arriving at less precise estimates in our study. Finally, although the mean follow-up period of 2.9 years in our study was still fairly short, it was somewhat longer than that of randomized controlled trials.3-6 The follow-up period was more than 3 years for 40% of patients, however, and 60% of cancers occurred after 3 years of follow-up.
The results of our study showed that in elderly patients, the risk of common cancers is not increased or decreased by statin use within in the first several years of taking the drugs. These results, however, cannot exclude the possibility that statins may increase the risk of cancer with longterm use.
and an exam are available with this article