HMG-coA reductase inhibitors, commonly known as statins, account for the great majority of cholesterol-lowering drug use. However, little is known about the association between long-term statin use and incidence of most types of cancers. We examined the association between long-term use of cholesterol-lowering drugs, predominantly statins, and the incidence of ten common cancers, as well as overall cancer incidence, among 133,255 participants (60,059 men and 73,196 women) in the Cancer Prevention Study II Nutrition Cohort during the period from 1997 to 2007. Multivariate Cox proportional hazards regression was used to estimate relative risks (RR). Current use status and duration of use were updated during follow-up using information from biennial follow-up questionnaires. Current use of cholesterol-lowering drugs for five or more years was not associated with overall cancer incidence (RR = 0.97, 95% CI = 0.92–1.03), or incidence of prostate, breast, colorectal, lung, bladder, renal cell, or pancreatic cancer but was associated with lower risk of melanoma (RR = 0.79, 95% CI = 0.66–0.96), endometrial cancer (RR = 0.65, 95% CI = 0.45–0.94), and non-Hodgkin lymphoma (NHL; RR = 0.74, 95% CI = 0.62–0.89). These results suggest that long-term use of statins is unlikely to substantially increase or decrease overall cancer risk. However, associations between long-term statin use and risk of endometrial cancer, melanoma, and NHL deserve further investigation. Cancer Res; 71(5); 1763–71. ©2011 AACR.
HMG-coA reductase inhibitors, commonly known as statins, were introduced in the United States in 1987 and rapidly became the most commonly used type of cholesterol-lowering medication (1, 2). By 1997, statins accounted for approximately 86% of cholesterol-lowering drug use in the United States (2). Statins are used by an estimated 24 million adults in the United States alone (3) and are proven to reduce risk of cardiovascular disease (4). Because statins are often used indefinitely, the prevalence of long-term use is likely to increase over time. It has been hypothesized that statin use could reduce risk of certain cancers given that they inhibit carcinogenesis in a variety of rodent models (5) and have some anti-inflammatory effects in humans (6–8). However, it should also be noted that for many types of cancer, adverse effects of long-term statin use cannot be ruled out on the basis of currently available evidence.
A meta-analysis of data from randomized trials designed to examine cardiovascular outcomes found no association between statin use and overall cancer incidence or incidence of colorectal, breast, lung, or prostate cancer (9). However, these trials are not well suited to address the potential effects of long-term use on cancer incidence because of relatively short intervention and follow-up periods. Observational studies that examined associations between long-term use (typically defined as more than 4 or 5 years) and cancer incidence generally do not support associations with the incidence of breast, colorectal, or overall prostate cancer (10). However, very few studies have examined associations between long-term use and incidence of other cancers.
The most comprehensive study of long-term statin use and cancer to date is an analysis of computerized pharmacy data from the Kaiser Permanente Medical Care Program in northern California that examined risk of more than 25 types of cancers (11). Among women, there were no associations between 5 years or more of statin use and incidence of any type of cancer. Among men, 5 years or more of statin use was associated with slightly higher overall cancer incidence (RR = 1.09, 95% CI 1.02–1.17), due in part to a statistically significant increase in bladder cancer incidence. Limitations of the Kaiser Permanente analysis included the absence of information on potential confounders, including smoking and obesity, and limited statistical power for less common cancers. In addition, most statin prescriptions in the Kaiser Permanente analysis were for lovastatin, the earliest commercially available statin, which was largely replaced in the United States by more potent statins in the mid-1990s (2).
We examined the association between long-term use of cholesterol-lowering drugs and incidence of 10 common cancers among participants in the Cancer Prevention Study-II (CPS-II) Nutrition Cohort. The CPS-II Nutrition Cohort is well-suited to examine this association due to a relatively high prevalence of long-term use, regularly updated information on use of cholesterol-lowering drugs, and detailed information on many potential confounders.
Materials and Methods
Men (n = 60,059) and women (n = 73,196) in this analysis were participants in the CPS-II Nutrition Cohort, a prospective study of cancer incidence in the United States established in 1992 and described in detail elsewhere (12). The Emory University Institutional Review Board approves all aspects of the CPS-II Nutrition Cohort. At enrollment in 1992 or 1993, participants completed a mailed self-administered questionnaire including information on demographic, medical, and lifestyle factors. Follow-up questionnaires to update exposure information and to ascertain newly diagnosed cancers were sent in 1997 and every 2 years thereafter. The response rate for each follow-up questionnaire was at least 88%.
Follow-up for this analysis began on the date of completion of the 1997 questionnaire, rather than at enrollment into the cohort in 1992 or 1993, because data from retail pharmacies in the United States indicate that statins accounted for approximately 86% of cholesterol-lowering drug use in 1997, compared with 62% in 1992 (2). A total of 65,047 men and 78,747 women from the Nutrition Cohort completed the standard version of the 1997 questionnaire, which included questions on use of cholesterol-lowering drugs. All analyses excluded participants who provided incomplete or uninterpretable information on use of cholesterol-lowering drugs (n = 3,759), smoking status (n = 2,824), or who were lost to follow-up because they did not return any of the follow-up questionnaires (n = 3,956). For analyses of each cancer, participants with a history of that type of cancer were excluded. Analyses of endometrial cancer excluded an additional 28,943 women who had had a hysterectomy at baseline and also censored women who subsequently reported a hysterectomy.
Ascertainment of cancer cases
Of the 15,839 first cancers included in this analysis, 13,497 were initially identified by self-report on the follow-up questionnaires and were subsequently verified by obtaining medical records or through linkage with state cancer registries when complete medical records could not be obtained (12). Ascertainment of cancer by self-report is estimated to have a sensitivity of 93% in the Nutrition Cohort (13). An additional 2,342 cases of cancer were identified through linkage with the National Death Index (14), of which 1,875 were verified through subsequent linkage with state cancer registries. Analyses of specific cancer sites also included some cases that were the first occurrence of cancer at that site, but were not first cancers. Analyses did not include in situ prostate, colorectal, and cervical cancers, but it did include in situ cancers at other sites.
Assessment of use of cholesterol-lowering drugs
The 1997 questionnaire asked participants to report whether they had taken any “cholesterol-lowering drugs” regularly during the past year, and provided as examples the brand names for 4 statins commonly used at that time (lovastatin, pravastatin, simvastatin, and fluvastatin; ref. 2), as well as the brand names for the fibrate drug, gemfibrozil, and the bile-acid binding resin, cholestyramine. Although this question did not distinguish between statins and nonstatin cholesterol-lowering drugs, statins accounted for approximately 86% of cholesterol-lowering drug use in the United States by 1997 (2) The 1999, 2001, and 2003 follow-up questionnaires included a similar question about use of cholesterol-lowering drugs. A more detailed question about type and dose of cholesterol-lowering drug was included on the 2005 questionnaire. In 2005, 93% of all users of cholesterol-lowering drugs reported using a statin. Information on use of cholesterol-lowering drugs before the start of follow-up for this analysis was available from the 1992 questionnaire, which included a question similar to that on the 1997 questionnaire. We considered use of cholesterol-lowering drugs to be a reasonable surrogate measure for use of statins because, during the study time period, statins accounted for the great majority of cholesterol-lowering drug use.
Use of cholesterol-lowering drugs was categorized as never, former, or current, using a time-dependent variable initially defined by use in 1997 and updated by use reported in 1999, 2001, 2003, and 2005. Use of cholesterol-lowering drugs was relatively consistent over time. Among participants with complete data on use of cholesterol-lowering drugs in both 1997 and 1999, 91% of current users in 1997 reported current use in 1999, and 90% of never users in 1997 reported no use in 1999. Therefore, if a participant did not provide information on use of cholesterol-lowering drugs on a follow-up questionnaire, they retained the same status (never, former, or current use) reported on the immediately preceding questionnaire approximately 2 years earlier. Participants who did not provide information on use of cholesterol-lowering drugs on 2 consecutive questionnaires were censored from further follow-up. Current users were further categorized as having less than 5 years of use and 5 or more years of use, based on use reported on previous questionnaires. Specifically, participants were categorized as current users of 5 or more years during the 1997–1999 follow-up interval if they reported current use in both 1992 and 1997; during the 1999–2001 follow-up interval if they reported current use in 1992, 1997, and 1999; and during subsequent follow-up intervals if they reported current use on the 3 immediately preceding questionnaires (e.g., reported current use in 1997, 1999, and 2001 for the 2001–2003 follow-up interval).
Cox proportional hazards regression models (15) were used to estimate relative risks (RR) and 95% confidence intervals (CI) for incident cancer. Follow-up time for Cox models began on the date of completion of the 1997 questionnaire. All models were adjusted for age, sex, race, education, smoking status, use of nonsteroidal anti-inflammatory drugs (NSAID, including aspirin) in 1992, use of NSAIDs in 1997, body mass index (BMI), physical activity, and history of elevated cholesterol, diabetes, heart disease, and hypertension. Models for prostate, breast, and colorectal cancer were also adjusted for history of PSA testing, history of mammography, and history of colorectal endoscopy, respectively, using time-dependent variables and updated information from follow-up questionnaires. History of PSA testing was categorized as never, in the last 2 years, more than 2 years ago, or unknown. Those that reported having PSA testing in the last 2 years were further categorized on indication for PSA testing (routine or for symptoms). History of mammography was categorized as never, in the last 2 years, more than 2 years ago, or unknown. History of colorectal endoscopy was categorized as never or ever. Models including women were also adjusted for hormone therapy use categorized as never, current estrogen, former estrogen, current estrogen and progesterone, former estrogen and progesterone, or unknown or other type. Age was adjusted for using the stratified Cox procedure (16) with one year strata. Unless otherwise noted previously, all covariates were based on status in 1997 and were modeled using the categories shown in Table 1. Further adjustment for alcohol use and geographic region had negligible effects on results and these factors were not included in final models.
Cross-product terms between long-term use of cholesterol-lowering drugs and follow-up time (continuous) were modeled to examine whether the association between long-term use and incidence of each cancer varied by follow-up time. Multiplicative interaction terms between long-term use and continuous variables for attained age and BMI, and categorical variables for sex, smoking status (never, current, or former), NSAID use (< 15, or ≥ 15 pills per month in 1997), and history of other cancer were also modeled. Because this analysis includes a considerable number of different cancer outcomes, these interactions were considered noteworthy only if P < 0.01. All P values are 2-sided.
Participants in this analysis were predominantly white and over age 60 at baseline, regardless of use of cholesterol-lowering drugs (Table 1). Among women, current long-term users of cholesterol-lowering drugs (hereafter referred to simply as long-term users) were more likely than never users to have a lower level of education. Among men, long-term users were more likely than never users to be former smokers. Nearly all long-term users reported having been told by a physician that they had elevated cholesterol, although many never users also had a history of elevated cholesterol. Long-term users were more likely than never users to be overweight or obese, to have a history of diabetes, heart disease, or hypertension, and to take at least 30 NSAID pills a month, likely reflecting daily use of aspirin for prevention of cardiovascular disease. At baseline in 1997, approximately 20.8% of men and 15.2% of women included in this analysis reported current use of cholesterol-lowering drugs. By 2005, this prevalence had increased to approximately 52.9% among men and 40.3% among women.
Long-term use of cholesterol-lowering drugs was not associated with overall cancer incidence or with risk of prostate, breast, colorectal, lung, bladder, pancreatic, or renal cell cancer but was associated with statistically significantly lower risks of melanoma (P = 0.02), endometrial cancer (P = 0.02), and non-Hodgkin lymphoma (P < 0.001; Table 2). Shorter-term use of cholesterol-lowering drugs also was associated with lower risk of NHL. Results for NHL shown in Table 2 are based on a broad definition of NHL recently recommended for epidemiologic research by the International Lymphoma Epidemiology Consortium (17). This definition includes lymphocytic leukemias, multiple myeloma, and plasma cell tumors other than multiple myeloma. Using a narrower definition of NHL that does not include these cancers, the RR associated with long-term use of cholesterol-lowering drugs was 0.76 (95% CI 0.62–0.93). Because statin use has been reported to be associated with lower risk of advanced prostate cancer in several studies (10), including a previous analysis of the CPS-II Nutrition Cohort (18), we examined results specifically for advanced prostate cancer (stage III or IV (19)). Long-term use of cholesterol-lowering drugs was not associated with a statistically significant reduction in risk of advanced prostate cancer.
No notable interactions between long-term use of cholesterol-lowering drugs and potential effect modifiers (follow-up time, attained age, BMI, smoking status, NSAID use, or history of other cancer) were observed in analyses of any of the cancer outcomes examined.
Results were similar in analyses examining 7 years or more of cholesterol-lowering drug use, rather than 5 years or more. Compared with never use, use of cholesterol-lowering drugs for 7 years or more was not significantly associated with overall cancer incidence or with risk of prostate, breast, colorectal, lung, bladder, pancreatic, or renal cell cancer but was associated with lower risk of melanoma (RR = 0.70, 95% CI 0.55–0.88), endometrial cancer (RR = 0.53, 95% CI 0.32–0.87), and NHL (RR = 0.77, 95% CI 0.62–0.94). There were insufficient numbers to examine longer duration use.
In analyses by NHL subtype, point estimates for RR for long-term use were below 1 for all NHL subtypes except follicular lymphoma (Table 3).
In this large prospective study, long-term use of cholesterol-lowering drugs, predominantly statins, was not associated with overall cancer incidence, or with incidence of most common cancers. However, long-term use of cholesterol-lowering drugs was associated with lower incidence of melanoma, endometrial cancer, and NHL.
Our null result with respect to long-term use of cholesterol-lowering drugs and overall cancer incidence is generally consistent with results from an analysis of computerized pharmacy data from Kaiser Permanente (11), to our knowledge the only previous study of this association. Long-term statin use was not associated with overall cancer incidence among women in the Kaiser Permanente analysis but was associated with a small increase in incidence among men (RR = 1.09, 95% CI 1.02–1.17). It is possible that the higher overall cancer incidence observed among men in that study was due to confounding by smoking. RR for 4 common smoking-related cancers (lung, esophagus, bladder, and kidney) were reduced in a sensitivity analysis that used external adjustment to account for the higher prevalence of ever smokers among statin users observed among responders to a Kaiser Permanente member health survey (11, 20). Externally adjusted results for overall cancer incidence were not reported.
Our null results with respect to long-term use and risk of prostate, breast, and colorectal cancer are consistent with results from most previous studies of long-term use of statins. Of studies that included at least 20 cases of breast (11, 21–24), prostate (11, 18, 24–28), or colorectal cancer (11, 29–32), with 4 or more years of use, only 2 (26, 29) reported a statistically significant reduction in risk.
Few previous studies examined the association between long-term statin use and risk of bladder, renal cell, pancreatic, or lung cancer. For bladder cancer, our null results for long-term use differ from those of the Kaiser Permanente analysis (11), the only previous study of this association, which showed a statistically significantly higher risk for men, and a similar, though not statistically significantly, higher risk for women. Our null results for long-term use and risk of renal cell, pancreatic, and lung cancer are consistent with results from the Kaiser Permanente analysis. However, use of statins for more than 4 years was associated with a highly statistically significant and approximately 80% reduced risk of both pancreatic and lung cancer in analyses of a large Veteran's Administration database (33, 34). However, the reduced risk associated with long-term use observed in these analyses is difficult to interpret and may have been largely an artifact of differential classification of exposure. Duration of use for cases was based on prescriptions received between 1998 and diagnosis date (between 1998 and 2004), whereas duration of use for controls appears to have been based on all prescriptions received during the 1998–2004 study period.
Long-term use of cholesterol-lowering drugs was associated with lower risk of endometrial cancer and melanoma in our study. These results should be interpreted cautiously as epidemiologic support for an association between statin use and these cancers is limited. With respect to endometrial cancer, statin use was associated with lower risk of endometrial cancer in one study (35), but not in 4 other studies (11, 24, 36, 37), including the Kaiser Permanente analysis (11), the only previous study to examine long-term use. With respect to melanoma, use of higher doses of statins was associated with lower risk in an analysis of a Veteran's Administration pharmacy database (38). However, no association was observed in an analysis of a large English pharmacy database (39) or in the Kaiser Permanente analysis, the only previous study to examine long-term use (11). The lower risk of endometrial cancer and melanoma associated with long-term use in our study could be due to chance. Additional studies that specifically examine long-term use would be useful to confirm or refute our results.
Use of cholesterol-lowering drugs was associated with modestly reduced risk of NHL in our study, consistent with some (40–42) but not all (11, 24, 43) other studies. The definition of NHL used in most of these previous studies did not include lymphocytic leukemias or multiple myeloma, with the exception of the EPI-LYMPH study, which examined all lymphoid neoplasms. Ever use of statins was associated with statistically significantly lower risk of NHL in the 3 largest (40–42) of the 5 case–control studies published to date (24, 40–43). In the largest of these, a European case–control study (EPI-LYMPH) including 2,362 cases, odds ratio for lymphoma were 0.6 (95% CI 0.5–0.8) for ever use and 0.6 (95% CI 0.3–1.2) for long-term use (≥ 6 years; ref. 42). However, ever use of statins was not associated with NHL in a smaller United States case–control study (24) and was associated with higher risk of NHL in a Japanese case–control study (43). In the Kaiser Permanente analysis, the only other prospective analysis, risk of NHL was not associated with either short-term use or long-term use (for long-term use, RR = 1.0, 95% CI 0.7–1.5 in men based on 32 cases in long-term users, RR = 1.2, 95% CI 0.8–1.8 in women, 25 cases in long-term users; ref. 11). Reasons for these differing results are not clear.
Our analyses by subtype of NHL were limited by small numbers. However, evidence of an association between long-term use and lower risk appeared strongest for diffuse large B-cell lymphoma (DLBCL) and marginal zone lymphomas, whereas there was no suggestion of lower risk for follicular lymphoma. It is possible that the anti-inflammatory effects of statins (6, 7) could be more relevant for DLBCL and marginal zone lymphoma than for follicular lymphoma. Both systemic autoimmune diseases and an established inflammation-related polymorphism in the tumor necrosis factor (TNF) gene are more strongly associated with DLBCL and marginal zone lymphoma than with follicular lymphoma (44–46).
The suggestion of an association between long-term use of cholesterol-lowering drugs and lower risk of multiple myeloma in our study (RR = 0.7, 95% CI 0.5–1.1) is consistent with results of previous studies. Ever use of statins was associated with significantly lower risk of multiple myeloma in a Connecticut case–control study (OR = 0.4, 95% CI 0.2–0.8; ref 47) and the EPI-LYMPH study (OR = 0.5, 95% CI 0.2–1.0; ref. 42). Neither of these studies reported results for long-term use. In the Kaiser Permanente analysis, long-term statin use (≥5 years) was associated with RR of 0.8 (95% CI 0.4–1.6) among men and 0.3 (95% CI 0.1–1.2) among women (11).
Possible reasons for the association between use of cholesterol-lowering drugs and lower risk of NHL observed in our study include chance, reverse causation, and causation. Chance appears unlikely to entirely explain this association given the low P value associated with long-term use (P < 0.001) and the fact that inverse associations were observed in some previous studies of NHL. Reverse causation is a possibility, as undiagnosed NHL could have lowered cholesterol levels, reducing the need to initiate cholesterol-lowering therapy, and ultimately resulting in cholesterol-lowering drug use being associated with an apparent reduction in risk of NHL. The possibility that lymphoma progression could reduce cholesterol levels is suggested by results from a large cardiovascular disease trial (48) where low cholesterol was associated with significantly higher risk of mortality from lymphoid cancers within 5 years of blood draw but not during later years of follow-up. However, if undiagnosed lymphoma reduced the need to initiate cholesterol-lowering therapy, any resulting reduction in risk would be expected to be stronger for short-term use of cholesterol-lowering drugs (use initiated close to diagnosis) than for long-term use. In our results, the reduction in risk of NHL appeared slightly weaker for short-term use than for long-term use. In addition, risk of NHL remained significantly reduced even for those who had used cholesterol-lowering drugs for 7 or more years. A causal association between use of cholesterol-lowering drugs and NHL is biologically plausible. Statins have been shown to inhibit lymphoma in mouse models (49–51). Statins also have anti-inflammatory effects, demonstrated by substantial reductions in C-reactive protein (a marker of systemic inflammation) at clinically used doses (6–8), that could be relevant given the role of chronic inflammation in NHL etiology (52).
It should be noted that a 20% to 30% reduction in RR of NHL would probably not be considered sufficient to influence clinical decisions about statin use for people at average risk of this cancer. However, a modest reduction in RR of NHL might be a clinically significant benefit for people at high risk, such as some individuals with Sjögren's syndrome, an autoimmune disease which has been associated with a 2- to 18-fold increased risk of developing NHL (45, 53–55). In addition, a modest reduction in RR of multiple myeloma might be a clinically significant benefit for people with monoclonal gammopathy of unknown significance (MGUS), which is associated with a substantial risk of progression to multiple myeloma (56). A small case–control study did not find an association between statin use and progression to multiple myeloma or related lymphoid malignancies among MGUS patients but included only 15 cases of multiple myeloma who had used statins (57). Information on the association between statin use and risk of NHL remains quite limited. Additional detailed observational analyses of statin use and risk of NHL, particularly in people at high risk, could provide useful further information about the potential role of statins in prevention of NHL. However, randomized trials would ultimately be needed to provide definitive results.
Strengths of this study include its prospective design and relatively large size. Our study included approximately twice as many cases of cancer with long-term exposure to cholesterol-lowering drugs as the only previous analysis to comprehensively examine the association between long-term use of cholesterol-lowering drugs and cancer risk (11). In addition, unlike most previous prospective analyses, we were able to adjust results for many potentially important confounders, including history of comorbidities, smoking, and obesity, although information on these factors was self-reported. Limitations of this study include the fact that use of cholesterol-lowering drugs was self-reported and was used as a surrogate measure for use of statins. If use of cholesterol-lowering drugs other than statins is not associated with risk of the cancers examined, our results may underestimate associations with statin use. For example, if long-term statin use was associated with a 20% lower risk of a cancer, but only 90% of long-term users of cholesterol-lowering drugs in this analysis were truly long-term statin users, we would have observed only an 18% lower risk of that cancer (0.2 multiplied by 0.9). Finally, we were unable to examine if risk varied by drug type or dose.
Results from this large prospective study suggest that long-term statin use does not increase risk of common cancers, providing some reassurance about the safety of long-term statin use with respect to cancer. The observed associations between long-term use of cholesterol-lowering drugs and lower risk of NHL, endometrial cancer, and melanoma require replication, ideally from large prospective studies with detailed information on drug type, duration, and dose.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interests were disclosed.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Received August 12, 2010.
- Revision received November 30, 2010.
- Accepted December 16, 2010.
- ©2011 American Association for Cancer Research.