Abstract

Background

Meta-analyses of trials of 3-hydroxy-3-methyl-glutaryl-CoA reductase inhibitors or statins for cardiovascular disease prevention have failed to show any statistically significant benefit of statins for cancer prevention. However, these trials included relatively young participants, who develop few cancers, and their follow-up periods may have been too short to detect an association between statin use and cancer incidence. We investigated this association in a population of veterans.

Methods

We identified patients using antihypertensive medications but no cholesterol-lowering medications (n = 25594) and patients using statins (n = 37248) who were enrolled in the Veterans Affairs New England Healthcare System between January 1, 1997, and December 31, 2005. Age- and multivariable-adjusted Cox proportional hazards models were used to calculate the hazard ratio (HR) and its 95% confidence interval (CI) for cancer incidence, excluding nonmelanoma skin cancer, among patients taking statins compared with patients taking antihypertensive medications and among patients grouped by statin dose (as equivalent simvastatin dose). All statistical tests were two-sided.

Results

The absolute incidence of total cancers was 9.4% among statin users and 13.2% among nonusers (difference = 3.8%, 95% CI = 3.3% to 4.3%, Pdifference < .001). Statin users had a statistically significant lower risk for total cancer than nonusers after adjustment for age (HR = 0.76, 95% CI = 0.73 to 0.80) and multiple potential confounders (HR = 0.74, 95% CI = 0.70 to 0.78). After multivariable adjustment, a statistically significantly decreased risk of all cancers was also associated with increasing statin use ( Ptrend < .001).

Conclusions

Patients using statins may be at lower risk for developing cancer. Additional observational studies and randomized trials of statins for cancer prevention are warranted.

CONTEXT AND CAVEATS
Prior knowledge

In meta-analyses of trials of statins for cardiovascular disease prevention, statins failed to show any statistically significant cancer prevention activity. However, these trials included relatively young participants, who develop few cancers, and their follow-up periods may have been too short to detect an association between statin use and cancer incidence.

Study design

Retrospective cohort study of veterans using statin and veterans using antihypertensive medications (but not statins) who were enrolled in the Veterans Affairs health-care system. Information on the statins prescribed was available, and statin doses were converted to equivalent simvastatin doses.

Contribution

The absolute incidence of total cancer was 9.4% among statin users and 13.2% among nonusers. Statin users had a statistically significantly lower risk of total cancer than nonusers, after adjustment for age and multiple potential confounders. In a multivariable analysis, decreased risk of all cancers was statistically significantly associated with increasing statin doses.

Implications

Statin users may be at a lower risk for developing cancer. Additional observational and randomized studies investigating the relationship between statin use and the development of cancer are warranted.

Limitations

Patients may not have been first-time statin users. All new cancer diagnoses could not be verified. Quantitative information on smoking or alcohol exposure was not available. There were limited numbers of women and minorities in this cohort.

Cancer-related morbidity and mortality are clinically significant. More than 1 million new cancer cases and more than 500000 cancer-related deaths are estimated to have occurred in the United States in 2007 ( 1 ). As the population of the United States ages, the incidence of cancer is expected to increase. Current recommendations for cancer prevention include maintaining a healthy lifestyle with good nutrition and moderate exercise and avoiding carcinogens such as tobacco. Recent laboratory and clinical data raise the possibility that cholesterol-lowering drugs, in particular 3-hydroxy-3-methyl-glutaryl-CoA reductase inhibitors or statins, could play a role in cancer prevention.

Several laboratory studies ( 2–4 ) have shown that statins may inhibit cancer cell progression, although results from population studies ( 5–17 ) have been mixed. Randomized trials that evaluated statins as chemoprevention agents for cardiovascular disease have not found that statins decreased the risk of developing cancer. In fact, three recent meta-analyses of these trials ( 5–7 ) concluded that statins did not prevent cancer. These trials, however, were not designed to evaluate statins as cancer chemoprevention agents, and incident cancer was not the primary outcome for any trial included in these meta-analyses. Therefore, new cases of cancer may not have been reported and any new cancer that was reported may not have been confirmed. Furthermore, in most of these trials, the mean age of participants was younger than 65 years and the follow-up period was less than 5 years, which, in these relatively healthy populations, may not have been adequate to detect differences in cancer incidence between statin users and nonusers.

Results from some observational studies ( 8–12 ) have indicated that statins may prevent cancer, whereas those from other observational studies ( 13–17 ) have not. Differences in the results of these studies may, however, be secondary to differences in study methodologies and study populations. Some of these observational studies ( 9 , 11 ) compared statin users with the general population without accounting for the healthy user effect ( 18 ). A few observational studies ( 8 , 11 , 14–16 ) considered statin dose or duration in their analyses. In this analysis, we used the national Veterans Affairs (VA) administrative and clinical databases along with the VA New England Veterans Integrated Service Network-1 (VISN-1) pharmacoepidemiology database ( 19 , 20 ) to investigate the relationship between statins use and cancer incidence, excluding nonmelanoma skin cancer.

Participants and Methods

Data Source and Definition of Outcome

We assembled a retrospective cohort of patients aged 18 years or older in the VA VISN-1 health-care system between January 1, 1997, and December 31, 2005, by use of VA national and regional patient care, health-care utilization, and claims databases. We excluded all patients with a cancer diagnosis defined by International Classification of Diseases, Ninth Revision, Clinical Modification ( ICD-9-CM ) codes 140.XX–208.XX on the cohort entry date. We defined the cohort entry date as the first recorded date that a prescription was filled for the medication of interest. An observation period for each patient was defined as beginning 2 years after his or her entry date and continuing until the first occurrence of the first appearance of an ICD-9-CM cancer code in the patient's medical record, 1 year after the last date that a prescription was filled for a medication of interest, death, or the end of the analysis period (ie, December 31, 2005). To diminish any potential effects of latent cancer on our predictor variables, we excluded patients who were diagnosed with cancer within 2 years after their potential entry date. Because long-term exposure would likely be required for any medication to reduce cancer incidence, we also excluded patients who discontinued their medication of interest within 2 years after their potential entry date. We performed sensitivity analyses that included patients who developed cancer and/or discontinued their medication of interest within 2 years.

The primary outcome for our analysis was cancer incidence, excluding nonmelanoma skin cancer. Cancer incidence was defined by a new ICD-9-CM code for cancer ( ICD-9-CM code range = 140.XX–208.XX) in the VA electronic medical record of a participant during an inpatient or outpatient encounter. To validate the diagnosis, a physician blinded to medication use reviewed 300 random charts with a new ICD-9-CM code for cancer from patients who were included in our analyses. A diagnosis of cancer was considered to be confirmed if the date of a new ICD-9-CM code for cancer matched a clinical or pathologic notation of a new cancer in the electronic medical record. Of these new cancers, 69% were confirmed with available medical data by chart review. No statistically significant difference was found in the confirmation rate between statin users (66%) and nonusers (69%) for cancer incidence. Confirmation rates varied by type of cancer reviewed. Some cancers had higher rates of confirmation, such as lung cancer, with a confirmation rate of 83%; other cancers had lower rates, such as melanoma with a rate of 62%. The difference in confirmation rates for each sample group was not statistically significant for any specific cancer.

Predictor Variables

Patients were selected among active users of the VA VISN-1 health-care system who filled at least two prescriptions (generally for a 90-day prescription) for any antihypertensive medication or statin within 1 year, continued filling prescriptions for an identified medication of interest at least yearly, and were seen at least one time per year in an outpatient VA clinic. The referent group was defined as patients who never filled a prescription for any cholesterol-lowering medication but did fill at least one prescription from the following classes of antihypertensive medications: beta-blockers, angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, calcium channel blockers, alpha-blockers, loop diuretics, thiazide diuretics, or centrally active antihypertensive medications. Statin users were defined as patients who filled prescriptions for any of the following medications: atorvastatin, fluvastatin, lovastatin, pravastatin, or simvastatin. Statin users may have been prescribed antihypertensive therapies in addition to their cholesterol-lowering therapy.

Several comorbid conditions and potential confounders were documented before or at the start of the observation period and coded as present or absent in the analysis on the basis of ICD-9-CM codes. These conditions included thyroid disease (240.XX–246.XX), diabetes mellitus (250.XX), hypertension (401.XX, 405.XX), renal failure (584.XX–586.XX), chest pain (413.XX, 786.5X), mental illness (295.XX, 296.XX, 300.XX, 301.XX, or 309.81), alcohol or drug abuse (303.XX, 304.XX), cardiovascular disease (410.XX–412.XX, 414.XX, 428.XX–438.XX, or 441.XX–444.2X), lung disease (491.XX–492.XX or 496.XX), gastrointestinal disease (531.XX–534.XX, 555.XX–556.XX, 570.XX–573.XX, or 577.XX), and prostate disease (600.XX–601.XX or 604.0X). We defined aspirin use (yes or no) as an active prescription at date of cohort entry for any of the following agents: aspirin; aspirin, buffered oral; aspirin, oral enteric-coated; or aspirin suppository. Current Procedural Terminology codes 45.23 for colonoscopy or 45.24 for sigmoidoscopy defined colorectal endoscopy and were selected as markers of screening, not as methods of diagnosis. We extracted information from the electronic medical record on sex (male or female), history of smoking (yes, no, or unknown), age (in years), and weight (in pounds) at entry into the observation period. We identified measured serum values for total cholesterol, low-density lipoprotein cholesterol, and high-density lipoprotein cholesterol within 6 months of each patient's entry date for the cohort.

Statistical Analysis

Characteristics of each group at the beginning of the observation period were compared by use of t tests for continuous variables and chi-square tests for categorical variables. We constructed age- and multivariable-adjusted Cox proportional hazards models for total cancer incidence and the incidence of each of the five most frequently diagnosed cancers in our cohort to calculate the hazard ratio (HR) and 95% confidence interval (CI) for risk in statin users compared with nonusers. Formal tests were conducted to confirm the assumption of proportionality, which was found to hold. Multivariable models included several available historic and statistically significant confounders, including age, weight, thyroid disease, diabetes mellitus, hypertension, cardiovascular disease, renal failure, chest pain, aspirin use, mental illness, alcoholism, lung disease, gastrointestinal disease, prostate disease, history of colonoscopy or sigmoidoscopy, smoking history, and total cholesterol. We also calculated a propensity score of being prescribed a statin ( 21 , 22 ). We constructed models that included the propensity score for our entire cohort; the population between the 5th and 95th percentile of propensity score; and the population within quartiles of propensity score (defined as <0.506, 0.506–0.580, 0.581–0.705, and >0.705).

To further investigate the relationship between statin dose and cancer incidence, we defined patient groups by tertiles of equivalent simvastatin doses (≤10, 11–39, and ≥40 mg), as previously described ( 19 ). Briefly, equivalent simvastatin doses were calculated by dividing lovastatin and pravastatin doses by 2, dividing the fluvastatin dose by 4, and multiplying the atorvastatin dose by 2. We then determined the hazard ratio and 95% confidence interval of each tertile of equivalent simvastatin dose, compared with nonuser referent group, for total cancer incidence and each of the five most frequently diagnosed cancers in our cohort. We controlled for the same comorbidities and potential confounders listed for our models described above. We calculated tests of trend across tertiles of equivalent simvastatin dose, with the median equivalent simvastatin dose in each tertile as an ordinal variable. In a sensitivity analysis, we compared the hazard functions during the follow-up periods of our statin and referent groups.

All statistical tests were two-sided; P values of less than .05 were considered to be statistically significant. All tests were performed with SAS, version 9.1 (Cary, NC). The study protocol was reviewed and approved by the Institutional Review Board of the VA Boston Healthcare System.

Results

We identified a cohort of 62842 VISN-1 patients who met our entry criteria. The mean age of patients in the cohort was 66.5 ± 10.9 years (±SD), and median total follow-up time was 5.0 years (minimum = 2.0 years and maximum = 7.2 years). The proportion of each statin agent in the statin user group was 61.8% for simvastatin, 34.8% for lovastatin, 2.9% for atorvastatin, 0.3% for pravastatin, and 0.1% for fluvastatin. The mean equivalent simvastatin dose among the statin user group was 22.1 ± 19.2 mg (±SD). Several statistically significant differences between the nonuser referent group and patients using statins were identified ( Table 1 ). Statin users were more likely than nonusers to have thyroid disease, diabetes mellitus, cardiovascular disease, or lung disease and to have been prescribed an aspirin. Statin users were less likely than the referent group to have hypertension or a history of alcohol or drug abuse. Total cholesterol was higher but high-density lipoprotein cholesterol was lower among statin users than among nonusers.

Table 1

Cohort characteristics at the beginning of the observation period *

Variable Referent group (n = 25594) Statin medication users (n = 37248) P value † 
Age, mean ± SD, y 66.3 ± 12.0 66.6 ± 10.0 .009 
Weight, mean ± SD, pounds 193.5 ± 43.3 196.5 ± 38.0 <.001 
Female, % 3.3 2.4 <.001 
Thyroid disease, % 5.3 6.3 <.001 
Diabetes mellitus, % 24.3 32.3 <.001 
Hypertension, % 78.1 72.8 <.001 
Renal failure, % 2.9 3.1 .41 
Mental illness, % 22.3 20.0 <.001 
Alcohol and/or drug abuse, % 10.9 6.5 <.001 
Cardiovascular disease, % 35.2 58.3 <.001 
Lung disease, % 9.4 12.8 <.001 
Gastrointestinal disease, % 7.4 5.2 <.001 
Prostate disease, % 4.0 6.0 <.001 
History of colorectal endoscopy, % 4.6 4.3 .11 
Aspirin use, % 33.4 41.5 <.001 
Smoking history, %   <.001 
Yes 28.9 35.2  
No 8.3 9.2  
Unknown 62.7 55.7  
Cholesterol, mean ± SD, mg/dL 188.1 ± 35.4 205.5 ± 46.9 <.001 
LDL-C, mean ± SD, mg/dL 109.3 ± 33.4 121.6 ± 41.0 <.001 
HDL-C, mean ± SD, mg/dL 45.2 ± 15.9 43.4 ± 12.1 <.001 
Variable Referent group (n = 25594) Statin medication users (n = 37248) P value † 
Age, mean ± SD, y 66.3 ± 12.0 66.6 ± 10.0 .009 
Weight, mean ± SD, pounds 193.5 ± 43.3 196.5 ± 38.0 <.001 
Female, % 3.3 2.4 <.001 
Thyroid disease, % 5.3 6.3 <.001 
Diabetes mellitus, % 24.3 32.3 <.001 
Hypertension, % 78.1 72.8 <.001 
Renal failure, % 2.9 3.1 .41 
Mental illness, % 22.3 20.0 <.001 
Alcohol and/or drug abuse, % 10.9 6.5 <.001 
Cardiovascular disease, % 35.2 58.3 <.001 
Lung disease, % 9.4 12.8 <.001 
Gastrointestinal disease, % 7.4 5.2 <.001 
Prostate disease, % 4.0 6.0 <.001 
History of colorectal endoscopy, % 4.6 4.3 .11 
Aspirin use, % 33.4 41.5 <.001 
Smoking history, %   <.001 
Yes 28.9 35.2  
No 8.3 9.2  
Unknown 62.7 55.7  
Cholesterol, mean ± SD, mg/dL 188.1 ± 35.4 205.5 ± 46.9 <.001 
LDL-C, mean ± SD, mg/dL 109.3 ± 33.4 121.6 ± 41.0 <.001 
HDL-C, mean ± SD, mg/dL 45.2 ± 15.9 43.4 ± 12.1 <.001 
*

SD = standard deviation; LDL-C = low-density lipoprotein cholesterol; HDL-C = high-density lipoprotein cholesterol.

P values (two-sided) were from t tests or chi-square tests. All statistical tests were two-sided.

During the observation period, 3381 (13.2%) of the 25594 patients in the nonuser referent group developed cancer, compared with 3515 (9.4%) of the 37248 patients using statins (difference = 3.8%, 95% CI = 3.3% to 4.3%, Pdifference < .001). Compared with the referent group, the risk of cancer incidence was 24%–26% less among patients using statins after adjusting for age only (HR = 0.76, 95% CI = 0.73 to 0.80) or for age and other potential confounders (HR = 0.74, 95% CI = 0.70 to 0.78) ( Table 2 ). Placing a propensity score in the model, rather than each potential confounder, did not change the results (HR = 0.74, 95% CI = 0.70 to 0.78). Furthermore, results from sensitivity analyses and analyses limited to patients between the 5th and 95th percentile of the propensity score and in each quartile of propensity score did not differ markedly from the results of our multivariable model, which included all available potential confounders (data not shown). For each of the five most common cancers in our cohort (prostate, lung, colorectal, and bladder cancer and melanoma), rates were generally lower among statins users than among nonusers. In particular, the age-adjusted risks for prostate (HR = 0.85, 95% CI = 0.78 to 0.93), lung (HR = 0.75, 95% CI = 0.66 to 0.86), and colorectal (HR = 0.62, 95% CI = 0.54 to 0.73) cancers were lower among statin users than nonusers ( Table 2 ). After adjustment for multiple potential confounders or the calculated propensity score, the point estimate for each cancer remained relatively constant.

Table 2

Age-adjusted, multivariable-adjusted, and propensity score–adjusted Cox proportional hazards models for first incidence of all cancer and the five most common cancer subtypes *

Outcome No. of incident cancers in referent group No. of incident cancers in statin group Age-adjusted HR (95% CI)  Multivariable-adjusted , HR (95% CI)  
All cancers 3381 3515 0.76 (0.73 to 0.80) 0.74 (0.70 to 0.78) 
Top five cancers     
   Prostate 1001 1164 0.85 (0.78 to 0.93) 0.90 (0.81 to 0.99) 
   Lung 431 436 0.75 (0.66 to 0.86) 0.70 (0.60 to 0.81) 
   Colorectal 371 316 0.62 (0.54 to 0.73) 0.65 (0.55 to 0.78) 
   Bladder 258 326 0.94 (0.80 to 1.11) 0.94 (0.77 to 1.13) 
   Melanoma 236 304 0.94 (0.79 to 1.12) 0.84 (0.70 to 1.02) 
Outcome No. of incident cancers in referent group No. of incident cancers in statin group Age-adjusted HR (95% CI)  Multivariable-adjusted , HR (95% CI)  
All cancers 3381 3515 0.76 (0.73 to 0.80) 0.74 (0.70 to 0.78) 
Top five cancers     
   Prostate 1001 1164 0.85 (0.78 to 0.93) 0.90 (0.81 to 0.99) 
   Lung 431 436 0.75 (0.66 to 0.86) 0.70 (0.60 to 0.81) 
   Colorectal 371 316 0.62 (0.54 to 0.73) 0.65 (0.55 to 0.78) 
   Bladder 258 326 0.94 (0.80 to 1.11) 0.94 (0.77 to 1.13) 
   Melanoma 236 304 0.94 (0.79 to 1.12) 0.84 (0.70 to 1.02) 
*

HR = hazard ratio; CI = confidence interval. Multivariable-adjusted models controlled for the following variables: age (in years), weight (in pounds), thyroid disease (yes or no), diabetes mellitus (yes or no), hypertension (yes or no), cardiovascular disease (yes or no), renal failure (yes or no), chest pain (yes or no), aspirin use (yes or no), mental illness (yes or no), alcoholism (yes or no), lung disease (yes or no), gastrointestinal disease (yes or no), prostate disease (yes or no), history of colonoscopy or sigmoidoscopy (yes or no), smoking history (yes, no, or unknown), and total cholesterol (mg/dL).

The risk of cancer incidence by tertiles of equivalent simvastatin dose from the multivariable-adjusted model is presented in Table 3 . All categories of equivalent simvastatin doses (≤10, 11–39, and ≥40 mg), compared with nonuse, were statistically significantly associated with a decreased risk of cancer after controlling for age and other potential confounders ( Ptrend < .001). For the five most frequent cancers, a dose of 40 mg or higher was associated with statistically significantly decreased incidences of lung cancer (HR = 0.73, 95% CI = 0.54 to 0.97), colorectal cancer (HR = 0.59, 95% CI = 0.41 to 0.85), and melanoma (HR = 0.64, 95% CI = 0.44 to 0.94) but not of prostate or bladder cancer. The trend across dose categories was statistically significant for lung ( Ptrend = .002) and colorectal ( Ptrend = .001) cancer and melanoma ( Ptrend = .004) but not for prostate ( Ptrend = .20) or bladder ( Ptrend = .25) cancer.

Table 3

Multivariable Cox proportional hazards models (HRs, with 95% CIs) for incidence of all cancers and of the five most common types of cancer by tertiles of equivalent simvastatin doses *

 Equivalent simvastatin dose  
Outcome ≤10 mg (n = 15598) 11–39 mg (n = 13852) ≥40 mg (n = 7410) Ptrend† 
All cancers 0.76 (0.71 to 0.81) 0.72 (0.66 to 0.77) 0.67 (0.60 to 0.75) <.001 
Top five cancers     
   Prostate 0.89 (0.79 to 1.01) 0.89 (0.78 to 1.02) 0.93 (0.77 to 1.12) .20 
   Lung 0.70 (0.58 to 0.84) 0.70 (0.57 to 0.86) 0.73 (0.54 to 0.97) .002 
   Colorectal 0.66 (0.54 to 0.82) 0.63 (0.50 to 0.81) 0.59 (0.41 to 0.85) .001 
   Bladder 0.99 (0.79 to 1.23) 0.93 (0.72 to 1.19) 0.81 (0.56 to 1.18) .25 
   Melanoma 0.88 (0.71 to 1.09) 0.63 (0.48 to 0.83) 0.64 (0.44 to 0.94) .004 
 Equivalent simvastatin dose  
Outcome ≤10 mg (n = 15598) 11–39 mg (n = 13852) ≥40 mg (n = 7410) Ptrend† 
All cancers 0.76 (0.71 to 0.81) 0.72 (0.66 to 0.77) 0.67 (0.60 to 0.75) <.001 
Top five cancers     
   Prostate 0.89 (0.79 to 1.01) 0.89 (0.78 to 1.02) 0.93 (0.77 to 1.12) .20 
   Lung 0.70 (0.58 to 0.84) 0.70 (0.57 to 0.86) 0.73 (0.54 to 0.97) .002 
   Colorectal 0.66 (0.54 to 0.82) 0.63 (0.50 to 0.81) 0.59 (0.41 to 0.85) .001 
   Bladder 0.99 (0.79 to 1.23) 0.93 (0.72 to 1.19) 0.81 (0.56 to 1.18) .25 
   Melanoma 0.88 (0.71 to 1.09) 0.63 (0.48 to 0.83) 0.64 (0.44 to 0.94) .004 
*

HR = hazard ratio; CI = confidence interval. Multivariable-adjusted models controlled for the following variables: age (in years), weight (in pounds), thyroid disease (yes or no), diabetes mellitus (yes or no), hypertension (yes or no), cardiovascular disease (yes or no), renal failure (yes or no), chest pain (yes or no), aspirin use (yes or no), mental illness (yes or no), alcoholism (yes or no), lung disease (yes or no), gastrointestinal disease (yes or no), prostate disease (yes or no), history of colonoscopy or sigmoidoscopy (yes or no), smoking history (yes, no, or unknown), and total cholesterol (mg/dL).

Ptrend (two-sided) was from multivariable Cox proportional hazards models. All statistical tests were two-sided.

As an additional sensitivity analysis, we investigated the association between statin use and the risk of cancer incidence across periods of follow-up. Statin use was statistically significantly associated with lower risk of all cancers than nonuse throughout the entire follow-up period ( Ptrend < .001).

Discussion

Statin use appeared to be associated with a lower risk for developing cancer, especially lung and colorectal cancers, than nonuse. A dose–response relationship may exist between statins and cancer incidence because the risk of cancer incidence appeared to decrease as the equivalent simvastatin dose increased in our cohort.

Our study builds on previous observational studies. Poynter et al. ( 9 ) and Khurana et al. ( 12 ) recently reported that statin use was associated with reduced risks of colorectal and lung cancers, respectively. However, both of these studies used a case-control design and had limited information on statin dose. Using a retrospective cohort study design, we found similar crude results for the incidence of both colorectal and lung cancers. Furthermore, we were able to explore a potential dose-response relationship between statin use and the incidence of all cancers, as well as of several specific cancers.

Inhibition of intracellular cholesterol production may be one mechanism by which statins prevent the development of cancer. The phosphorylation of Akt, which is an important step in intracellular signaling, is regulated by caveolin-1, a protein that is regulated by cholesterol and that is the main structural protein of caveolae ( 23 , 24 ). Elevated cholesterol increases the expression of mRNA for caveolin-1 ( 25 ). In a cancer cell model system, addition of a statin to the medium surrounding cancer cells decreased the number of caveolae in cell membranes, whereas addition of cholesterol to the medium increased the number of caveolae ( 26 ). Thus, both cholesterol and statins may affect caveolae expression. The level of caveolin-1 expression may be associated with the development of cancer through sustaining Akt activation ( 24 ). Furthermore, overexpression of caveolin-1 has been associated with metastasis in some studies of lung and colon cancer ( 27 , 28 ). However, little evidence supports an association between the reduction of cancer incidence and use of other cholesterol-lowering medications, including fibric acid derivatives and bile acid sequestrants.

Statins may have additional antineoplastic effects that depend on their hydrophilic or lipophilic characteristics. Previous laboratory studies (29, 30 ) have shown that hydrophilic statins, such as pravastatin, do not appear to have antineoplastic effects, and several other laboratory studies ( 31–35 ) have described mechanisms through which lipophilic statins, such as simvastatin and lovastatin, may reduce cancer incidence. Prenylation of G proteins by isoprenoid products is important for cell signaling, survival, and migration ( 31 , 32 , 36–38 ). For example, geranylgeraniol, an intermediate in the mevalonate pathway, was able to overcome the antiproliferative effects of lovastatin in a cell model of prostate cancer ( 33 ). Furthermore, lipophilic statins may promote cell cycle arrest by inhibiting the proteasome, a mechanism that is independent of their effect on 3-hydroxy-3-methyl-glutaryl-CoA reductase ( 33 ). Inhibition of the proteasome by the lactone moieties of statins and isoprenoid inhibitors has been shown to arrest cancer cells in G 1 phase of the cell cycle ( 34 , 35 ). It is important to note that the most commonly used statins in our cohort, simvastatin and lovastatin, were lipophilic.

Our study population and design have several strengths. The large size of our cohort provided sufficient power to investigate cancer incidence in a population free from overt cancer. Because of our large sample size, we were also able to control for several potential confounders, including age, history of cardiovascular disease, history of colonoscopy or sigmoidoscopy, smoking history, and total cholesterol for total cancer and each specific cancer. We had detailed pharmacy data that allowed us to investigate a potential relationship between statin dose and cancer incidence. Compared with the general population, statin users may receive more counseling regarding a healthy lifestyle, which may lead to the increased prevention of some cancers. Statin users may also receive more cancer screening than the general population and, therefore, may have higher cancer incidence than the general population. Previous observational studies have failed to account for these potential healthy user effects in their analyses ( 8 , 9 ). Statin users may also be more compliant with drug therapy than the general population and compliance with drug therapy has been found to be a predictor of maintaining good health ( 39 ). In our study, we compared two groups with similar compliance who likely received similar information about healthy lifestyles and cancer screening.

Some limitations of our study should be considered. Not all patients in our study were first-time statin or antihypertensive medication users. However, each patient had at least 1 year of documented medication use before entering the study and 2 years of documented medication use before entering the study's observation period. In addition, patients in our cohort were examined by outpatient VA providers and filled their prescriptions on a regular basis. We could not verify that all new cancer diagnoses were identified. However, any absence would likely be random and probably few in number because patients who receive outpatient medications through the VA system are required to have a primary care provider in the VA system for follow-up. We relied on unconfirmed ICD-9-CM codes for diagnoses of potential confounders. However, any misclassification of our confounders would likely be random misclassification because the method that health care providers use to assign ICD-9-CM codes does not vary by the presence or absence of a statin or antihypertensive medication. Therefore, random misclassification would bias our results toward the null. If misclassification were not random, the level of confounding required to overcome the risks that we observe in our crude analyses would need to be extremely large and is therefore unlikely. We also relied on ICD-9-CM codes for our outcome. We did not find statistically significant differences in classification rates between statin users and nonusers by chart validation. Therefore, any misclassification would likely be random misclassification and bias our results toward the null. We did not have information on lifestyle variables such as diet and exercise. However, it is unlikely that any difference in lifestyle variables between our two groups would be large enough to account for the statistically significant findings in our results. We did not have quantitative information on smoking or alcohol exposure. Because of the limited numbers of women and minorities in our veteran population, care should be taken before generalizing our results.

Our findings support the hypothesis that statins may reduce the risk of cancer, in particular lung and colorectal cancers. This relationship may be affected by the dose of statin. Observational studies and randomized trials to evaluate statins as cancer preventive agents are needed to confirm or refute these findings.

Funding

Cooperative Studies Program, Department of Veterans Affairs Office of Research and Development (601 and 96).

We would like to acknowledge all of the dedicated staff at the Massachusetts Veterans Epidemiology Research and Information Center for their assistance in this project.

References

1.
Jemal
A
Siegel
R
Ward
E
, et al.  . 
Cancer statistics, 2007
CA Cancer J Clin
 , 
2007
, vol. 
57
 (pg. 
43
-
66
)
2.
Kusama
T
Mukai
M
Iwasaki
T
, et al.  . 
Inhibition of epidermal growth factor-induced RhoA translocation and invasion of human pancreatic cancer cells by 3-hydroxy-3-methylglutaryl-coenzyme a reductase inhibitors
Cancer Res.
 , 
2001
, vol. 
61
 (pg. 
4885
-
4891
)
3.
Khanzada
UK
Pardo
OE
Meier
C
Downward
J
Seckl
MJ
Arcaro
A
Potent inhibition of small-cell lung cancer cell growth by simvastatin reveals selective functions of Ras isoforms in growth factor signalling
Oncogene
 , 
2006
, vol. 
25
 (pg. 
877
-
887
)
4.
Mehta
N
Hordines
J
Sykes
D
Doerr
RJ
Cohen
SA
Low density lipoproteins and lovastatin modulate the organ-specific transendothelial migration of primary and metastatic human colon adenocarcinoma cell lines in vitro
Clin Exp Metastasis
 , 
1998
, vol. 
16
 (pg. 
587
-
594
)
5.
Dale
KM
Coleman
CI
Henyan
NN
Kluger
J
White
CM
Statins and cancer risk: meta-analysis
JAMA
 , 
2006
, vol. 
295
 (pg. 
74
-
80
)
6.
Bonovas
S
Filioussi
K
Tsavaris
N
SitarasStatins
NM
cancer risk: a literature-based meta-analysis and meta-regression analysis of 35 randomized controlled trials
J Clin Oncol
 , 
2006
, vol. 
24
 (pg. 
4808
-
4817
)
7.
Browning
DRL
Martin
RM
Statins and risk of cancer: a systematic review and metaanalysis
Int J Cancer
 , 
2006
, vol. 
120
 (pg. 
833
-
843
)
8.
Graaf
M
Beiderbeck
AB
Egberts
ACG
Richel
DJ
Guchelaar
HJ
The risk of cancer in users of statins
J Clin Oncol
 , 
2004
, vol. 
22
 (pg. 
2388
-
2394
)
9.
Poynter
JN
Gruber
SB
Higgins
PDR
, et al.  . 
Statins and the risk of colorectal cancer
N Engl J Med
 , 
2005
, vol. 
352
 (pg. 
2184
-
2192
)
10.
Blais
L
Desgagne
A
Lelorier J. 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors and the risk of cancer. A nested case-control study
Arch Intern Med
 , 
2000
, vol. 
160
 (pg. 
2363
-
2368
)
11.
Friis
S
Poulsen
AH
Johnsen
SP
, et al.  . 
Cancer risk among statin users: a population-based cohort study
Int J Cancer
 , 
2005
, vol. 
114
 (pg. 
643
-
647
)
12.
Khurana
V
Bejjanki
HR
Caldito
G
Owens
MW
Statins reduce the risk of lung cancer in humans: a large case-control study of US veterans
Chest
 , 
2007
, vol. 
131
 (pg. 
1282
-
1288
)
13.
Jacobs
EJ
Rodriguez
C
Brady
KA
Connell
CJ
Thun
MJ
Calle
EE
Cholesterol-lowering drugs and colorectal cancer incidence in a large United States cohort
J Natl Cancer Inst
 , 
2006
, vol. 
98
 (pg. 
69
-
72
)
14.
Cauley
JA
McTiernan
A
Rodabough
RJ
Statin use and breast cancer: prospective results from the Women's Health Initiative
J Natl Cancer Inst
 , 
2006
, vol. 
98
 (pg. 
700
-
701
)
15.
Kaye
JA
Jick
H
Statin use and cancer risk in the General Practice Research Database
Br J Cancer
 , 
2004
, vol. 
90
 (pg. 
635
-
637
)
16.
Eliassen
AH
Colditz
GA
Rosner
B
Willett
WC
Hankinson
SE
Serum lipids, lipid-lowering drugs, and the risk of breast cancer
Arch Intern Med
 , 
2005
, vol. 
165
 (pg. 
2264
-
2271
)
17.
Setoguchi
S
Glynn
RJ
Avorn
J
Mogun
H
Schneeweiss
S
Statins and the risk of lung, breast, and colorectal cancer in the elderly
Circulation
 , 
2007
, vol. 
115
 (pg. 
27
-
33
)
18.
Majumdar
SR
McAlister
FA
Eurich
DT
Padwal
RS
Marrie
TJ
Statins and outcomes in patients admitted to hospital with community acquired pneumonia: population based prospective cohort study
BMJ
 , 
2006
, vol. 
333
 
7576
pg. 
999
 
19.
Scranton
RE
Young
M
Lawler
L
Solomon
DH
Gagnon
DR
Gaziano
JM
Statin use is associated with fewer fractures: study from a US veterans population
Arch Intern Med
 , 
2005
, vol. 
165
 (pg. 
2007
-
2012
)
20.
Hermos
JA
Young
MM
Gagnon
DR
Fiore
LD
Characterizations of long-term oxycodone/acetaminophen prescriptions in veteran patients
Arch Intern Med
 , 
2004
, vol. 
164
 (pg. 
2361
-
2366
)
21.
Rubin
DB
Estimating causal effects from large data sets using propensity scores
Ann Intern Med
 , 
1997
, vol. 
127
 (pg. 
757
-
763
)
22.
Braitman
LE
Rosenbaum
PR
Rare outcomes, common treatments: analytic strategies using propensity scores
Ann Intern Med
 , 
2002
, vol. 
137
 (pg. 
693
-
695
)
23.
Zhuang
L
Lin
J
Lu
ML
Solomon
KR
Freeman
MR
Cholesterol-rich lipid rafts mediate Akt-regulated survival in prostate cancer cells
Cancer Res.
 , 
2002
, vol. 
62
 (pg. 
2227
-
2231
)
24.
Li
L
Ren
CH
Tahir
SA
Ren
C
Thompson
TC
Caveolin-1 maintains activated Akt in prostate cancer cells through scaffolding domain binding site interactions with and inhibition of serine/threonine protein phosphatases PP1 and PP2A
Mol Cell Biol.
 , 
2003
, vol. 
23
 (pg. 
9389
-
9404
)
25.
Fielding
CJ
Bist
A
Fielding
PE
Caveolin mRNA levels are up-regulated by free cholesterol and down-regulated by oxysterols in fibroblast monolayers
Proc Natl Acad Sci USA
 , 
1997
, vol. 
94
 (pg. 
3753
-
3758
)
26.
Zhuang
L
Kim
J
Adam
RM
Soloman
KR
Freeman
MR
Cholesterol targeting alters lipid raft composition and cell survival in prostate cancer cells and xenografts
J Clin Invest
 , 
2005
, vol. 
115
 (pg. 
959
-
968
)
27.
Ho
CC
Huang
PH
Huang
HY
Chen
YH
Yang
PC
Hsu
SM
Up-regulated caveolin-1 accentuates the metastasis capability of lung adenocarcinoma by inducing filopodia formation
Am J Pathol
 , 
2002
, vol. 
161
 (pg. 
1647
-
1656
)
28.
Patlolla
JM
Swamy
MV
Raju
J
Rao
CV
Overexpression of caveolin-1 in experimental colon adenocarcinoma and human colon cancer cell lines
Oncol Rep
 , 
2004
, vol. 
11
 (pg. 
957
-
963
)
29.
Campbell
MJ
Esserman
LJ
Zhou
Y
, et al.  . 
Breast cancer growth prevention by statins
Cancer Res.
 , 
2006
, vol. 
66
 (pg. 
8707
-
8714
)
30.
Sivaprasad
U
Abbas
T
Dutta
A
Differential efficacy of 3-hydroxy-3-methylglutaryl CoA reductase inhibitors on the cell cycle of prostate cancer cells
Mol Cancer Ther
 , 
2006
, vol. 
5
 (pg. 
2310
-
2316
)
31.
Mo
H
Elson
CE
Studies of the isoprenoid-mediated inhibition of mevalonate synthesis applied to cancer chemotherapy and chemoprevention
Exp Biol Med
 , 
2004
, vol. 
229
 (pg. 
567
-
585
)
32.
Graaf
MR
Richel
DJ
VanNoorden
CJF
Guchelaar
H
Effects of statins and farnesyltransferase inhibitors on the development and progression of cancer
Cancer Treat Rev.
 , 
2004
, vol. 
30
 (pg. 
609
-
641
)
33.
Ghosh
PM
Ghosh-Choudhury
N
Moyer
ML
, et al.  . 
Role of RhoA activation in the growth and morphology of a murine prostate tumor cell line
Oncogene
 , 
1999
, vol. 
18
 (pg. 
4120
-
4130
)
34.
Rao
S
Porter
DC
Chen
X
Herliczek
T
Lowe
M
Keyomarsi
K
Lovastatin-mediated G1 arrest is through inhibition of the proteasome, independent of hydroxymethyl glutaryl-CoA reductase
Proc Natl Acad Sci USA
 , 
1999
, vol. 
96
 (pg. 
7797
-
7802
)
35.
Efuet
ET
Keyomarsi
K
Farnesyl and geranylgeranyl transferase inhibitors induce G1 arrest by targeting the proteasome
Cancer Res.
 , 
2006
, vol. 
66
 (pg. 
1040
-
1051
)
36.
Zhang
FL
Casey
PJ
Protein prenylation: molecular mechanisms and functional consequences
Ann Rev Biochem
 , 
1996
, vol. 
65
 (pg. 
241
-
269
)
37.
Coxon
JP
Oades
GM
Kirby
RS
Colston
KW
Zoledronic acid induces apoptosis and inhibits adhesion to mineralized matrix in prostate cancer cells via inhibition of protein prenylation
BJU Int
 , 
2004
, vol. 
94
 (pg. 
164
-
170
)
38.
Neville-Webbe
HL
Rostami-Hodjegan
A
Evans
CA
Coleman
RE
Holen
I
Sequence- and schedule-dependent enhancement of zoledronic acid induced apoptosis by doxorubicin in breast and prostate cancer cells
Int J Cancer
 , 
2005
, vol. 
113
 (pg. 
364
-
371
)
39.
Simpson
SH
Eurich
DT
Majumdar
SR
, et al.  . 
A meta-analysis of the association between adherence to drug therapy and mortality
BMJ
 , 
2006
, vol. 
333
 
7557
pg. 
15
 

Author notes

The study sponsor had no role in the experimental design; the collection, analysis, or interpretation of the data; or in the writing and submission of the manuscript.