Statin Use and Risk of Colorectal Cancer

Abstract

Background

Statins have anticancer activity in various cell types, including colon cancer cells. Although epidemiologic data on the relationship between statin use and the risk of colorectal cancer are limited, one case– control study reported a 50% reduction in colorectal cancer risk among statin users. We conducted a population-based case–control study to assess this association with respect to statin type, dose, and duration of use.

Methods

Case patients with adenocarcinoma of the colon or rectum were ascertained from participating hospitals in Massachusetts and the Massachusetts Cancer Registry from January 1, 2001, through November 30, 2004. Age-, sex-, and precinct-matched control subjects were chosen from Massachusetts town lists. Information on statin use and other relevant data were obtained by telephone interview. We used multivariable conditional logistic regression models to estimate odds ratios (ORs). All tests for statistical significance were two-sided.

Results

Among 1809 case patients and 1809 matched control subjects, regular use of statins for at least 3 months was not associated with the risk of colorectal cancer (OR = 0.92, 95% confidence interval [CI] = 0.78 to 1.09). There was no consistent trend across dose or duration of use (e.g., for ≥10 years of use, OR = 0.86, 95% CI = 0.51 to 1.45). The risk of stage IV cancer was, however, statistically significantly lower among statin users than among nonusers (OR = 0.49, 95% CI = 0.26 to 0.91). There was no evidence of an interaction between statin use and nonsteroidal anti-inflammatory drug use.

Conclusions

Overall, use of statins did not appear to be associated with reduced risk of colorectal cancer. The reduced risk of stage IV cancer observed among statin users requires confirmation.

Statins are inhibitors of 3-hydroxy-3-methyl glutaryl coenzyme A (HMG-CoA) reductase, the enzyme that is required for the conversion of HMG-CoA to mevalonic acid ( 1 ) . Mevalonic acid is a precursor of cholesterol and some nonsterol isoprenoid derivatives that are important for various cellular functions, including proliferation, differentiation, and survival ( 1 – 4 ) . Statins have been shown to arrest cell cycle progression ( 5 , 6 ), to induce apoptosis ( 2 ), and to alter the adhesion and migration of tumor cells ( 7 , 8 ). Anticancer effects for statins have been reported in colon cancer cells in vitro ( 8 – 11 ) and in rodent models ( 12 , 13 ). In one study ( 14 ), lovastatin augmented sulindac-induced apoptosis in colon cancer cells.

There is little epidemiologic data on the relationship between statin use and the risk of colorectal cancer, other than one case–control study that found that the use of statins for at least 5 years reduced the risk of colorectal cancer by 50% ( 15 ). This reduction was not observed in other studies ( 16 – 19 ), but these studies were limited by small numbers of subjects and lack of detailed information on statin use. We undertook this population-based case– control study to assess the association of statin use with colorectal cancer risk, with specific attention to statin type, dose, and duration of use. We also investigated the potentially confounding or synergistic effects of nonsteroidal anti-inflammatory drug (NSAID) use.

Patients and Methods

Case Patients and Control Subjects

Case patients with adenocarcinoma of the colon or rectum were ascertained from the tumor registrars of participating hospitals in Massachusetts and from the Massachusetts Cancer Registry from January 1, 2001, through November 30, 2004. The institutional review boards of Boston Medical Center and all participating institutions approved the study. Case patients were eligible if they were Massachusetts residents, aged 50–74 years, spoke English, were physically and cognitively able to complete a telephone interview and a self-administered food-frequency questionnaire, and had received a first diagnosis of primary colorectal cancer. Case patients were not contacted unless we first received permission from their physician to do so.

We were notified of 4287 potentially eligible case patients. Of the 4014 patients for whom we could identify and contact a physician, 365 were not eligible, 117 were too ill, and the physician refused permission for 417. Permission to contact was given for 3115 patients, of whom we were able to contact 2879. Of those 2879 patients, 336 were ineligible, 105 were too ill to participate, 597 refused to participate, and 32 did not satisfactorily complete the interview. Consequently, we successfully interviewed 1809 case patients by telephone.

For each interviewed case patient, we selected a potential control subject of the same sex, 5-year age-group, and town precinct from Massachusetts town books (i.e., lists of adult residents). We replaced potential control subjects who could not be contacted, refused, or were ineligible, until we had interviewed an appropriately matched control subject. We attempted to contact 4046 potential control subjects, of whom we were able to locate and contact 3466. Of the 3466 contacted, 463 were ineligible, 157 were too ill to participate, 1003 refused, and 34 did not satisfactorily complete the interview. Consequently, we successfully interviewed 1809 control subjects by telephone also.

Data Collection

Information was obtained by telephone interview on demographic factors, medical history, reproductive history (for women), alcohol consumption, cigarette smoking, leisure and occupational physical activity, and family history of cancer. Information on use of selected medications was obtained, including the use of statins and NSAIDs and, for women, female hormone supplements. Data were obtained in reference to an index date that was defined as the date of diagnosis for case patients or to a date that was selected so that the interval between the index date and the interview of the control subject was the same as the interval between diagnosis and interview for the matched case patient. The median interval between index date and interview was 6 months.

Statin use was ascertained by asking whether the participant had ever used a drug to lower levels of blood cholesterol and by listing specific statins (atorvastatin, simvastatin, lovastatin, pravastatin, cerivastatin, fluvastatin, and rosuvastatin) by trade name. NSAID use was ascertained by asking whether the participant had ever used a drug for pain, headache, or other relevant indications and by listing the most commonly used NSAIDs. Only regular use—defined as use for at least three times a week for at least 3 months—was recorded. For such episodes, we recorded the drug name, the frequency of use, the number of tablets taken per day or week, the dose, the date of first use, and the duration of use. We created a standardized dose scheme according to dose equivalencies of the various statins ( 20 ). Statins that achieved a reduction in low-density lipoprotein of 31%–35%, equivalent to treatment with 10 mg of atorvastatin, were assigned a score of 2; those that achieved less of a reduction were assigned a score of 1; and those that achieved more of a reduction were assigned a score of 3. We calculated cumulative standardized statin dose by multiplying the total days that a statin was used at a given standardized dose and summing across all doses; average daily standardized statin dose was calculated as the cumulative standardized dose divided by the total number of days used.

Intake of dietary nutrients was calculated from the self-administered Block Brief 2000 Questionnaire (Block Dietary Data Systems, Berkeley, CA), which included 70 food items. For case patients, we obtained a copy of the discharge summary and pathology report to confirm the diagnosis and to classify the tumor stage ( 21 ). Case patients provided written informed consent; verbal consent was accepted for control subjects.

Statistical Analysis

We evaluated regular statin use relative to a reference category that included irregular use and nonuse. People whose use began in the year before the index date were placed in a separate category (i.e., recent use) because such use would not have preceded the onset of cancer in most case patients. We further subdivided regular statin use into use that continued into the year of the index date and use that was discontinued a year or more before the index date. NSAID use was similarly classified.

To investigate the association between statin use and the risk of colorectal cancer, we calculated odds ratios (ORs) from multivariable conditional logistic regression models that accounted only for the matching factors (age, sex, and town precinct) (basic model) and from models that included other potential confounders, including education in years (<12, 12, 13–15, or ≥16 years); vigorous or moderate exercise (none, 1–3 hours/week, >3 hours/week, or unknown); occupational physical activity (sedentary, medium, intense, or unknown); family history of colorectal cancer (yes, no, or unknown); alcohol use (never, former, current, or unknown); smoking (never, former, current, or unknown); race (white or nonwhite); body mass index (<20, 20–24.9, 25–29, or >29 kg/m ² or unknown); use of vitamin E (yes, no, or unknown), multivitamin (yes, no, or unknown), or calcium supplements (yes, no, or unknown); a cholecystectomy before the index date (yes, no, or unknown); NSAID use (no regular use, continuing regular use, discontinued regular use, or recent regular use); and use of screening colonoscopy (reporting at least one screening colonoscopy at least 1 year before the index date: yes, no, or unknown). We tested the effect of confounders by entering them into the model simultaneously. Two covariates—NSAID use and reporting at least one screening colonoscopy at least 1 year before the index date—appreciably changed the estimates, and these estimates from the basic and adjusted models are presented. The other variables did not change the odds ratio by more than 10%. We tested for the confounding effect of dietary variables among the 853 matched case–control sets, in which both the case patient and control subject returned useable dietary data. The addition of daily intake of calories, fiber, folate, cholesterol, and total fat (in quartiles) or the number of servings per day of fat (<1, 1–2, 3, or ≥4), vegetables (<2, 2–3, 4, 5–6, or ≥7), meat (<1, 1–2, or ≥3), fruits (<1, 1–2, or ≥3), or dairy products (<1, 1 to <2, or ≥2) did not change the adjusted odds ratio by more than 10%. We calculated odds ratios separately by sex, age (<65 or ≥65 years), cancer site (colon or rectum), cancer side (left or right), and cancer stage (I, II, III, or IV).

We tested for trend in the risk of colorectal cancer across dose and duration by including the dose and duration variables as continuous variables in models that included only regular statin users and assessing the Wald chi-square statistic associated with those variables. We assessed interaction on a multiplicative scale between statin and NSAID use with the likelihood ratio test ( 22 ). All tests of statistical significance were two-sided.

Results

Ninety-seven percent of the study population were white, 55% were male, and approximately 50% were 65 years old or older ( Table 1 ). Case patients reported a family history of colorectal cancer more often than control subjects and were also more likely to be current or former smokers. Control subjects were more likely to be college graduates and to have more frequent visits to a health care provider than case patients. Control subjects were also more likely to have had a screening colonoscopy and to have used oral contraceptives, female hormone supplements, statins, NSAIDs, or multivitamins than case patients.

Regular statin use was associated with a statistically significantly reduced risk of colorectal cancer in the unadjusted model (OR = 0.82, 95% confidence interval [CI] = 0.70 to 0.95). However, when we adjusted for NSAID use and reporting at least one screening colonoscopy at least 1 year before the index date, the association was no longer statistically significant (OR = 0.92, 95% CI = 0.78 to 1.09) ( Table 2 ). In fact, in the adjusted model, risk of colorectal cancer was not associated with continuing use (i.e., use that continued into the year of the index date), discontinued use (use that was discontinued at least 1 year before the index date), or recent use (use that began within 1 year of the index date). The risk of colorectal cancer did not vary over categories of duration of use (e.g., for ≥10 years of regular statin use, OR = 0.86, 95% CI = 0.51 to 1.45). No consistent trend was observed across categories of average daily or quartiles of cumulative standardized dose. Thus, among 1809 case patients and 1809 matched control subjects, regular use of statins was not associated with the risk of colorectal cancer.

In the adjusted model, risk of colorectal cancer was not associated with the three most commonly used statins (atorvastatin, simvastatin, and pravastatin) ( Table 3 ); however, the risks associated with 5 or more years of use of these statins were slightly, although not statistically significantly, reduced. In this model, increased risk of colorectal cancer was associated with lovastatin use (OR = 1.73, 95% CI = 1.03 to 2.88), but decreased risk was associated with use of the other statins (primarily fluvastatin) (OR = 0.67, 95% CI = 0.34 to 1.32). The combined risk of colorectal cancer was not associated with the use of simvastatin and lovastatin, which are less bioavailable than the other statins (OR = 0.97, 95% CI = 0.74 to 1.28), or with the use of the remaining statins combined (OR = 0.86, 95% CI = 0.70 to 1.07).

To assess the effect of NSAID use on the association between statin use and colorectal cancer, we compared subjects who used statins only, NSAIDs only, and both statins and NSAIDs with subjects who had used neither statins nor NSAIDs. In the adjusted model, risk of colorectal cancer was not associated with statin use only (OR = 0.84, 95% CI = 0.65 to 1.09), but reduced risk was associated with NSAID use only (OR = 0.79, 95% CI = 0.67 to 0.95) ( Table 4 ). In subgroup analyses, the reduced risk associated with statin use was found only among subjects whose NSAID use continued into the year of the index date (OR = 0.67, 95% CI = 0.53 to 0.85); this risk was statistically significantly increased among subjects whose use was discontinued before the year of the index date (OR = 1.91, 95% CI = 1.31 to 2.78). A reduced risk of colorectal cancer was associated with use of both statins and NSAIDs (OR = 0.71, 95% CI = 0.58 to 0.88); however, again in subgroup analyses, this risk was found only among those whose NSAID use continued into the year of the index date (OR = 0.55, 95% CI = 0.41 to 0.74). There was no evidence of statistical interaction between statin and NSAID use ( P = .12) or between statin use and use of cyclooxygenase 2 inhibitors in particular ( P = .45).

To assess the effect of female hormone supplement use on the association between statin use and colorectal cancer, we compared women who used statins only, female hormone supplements only, and both statins and female hormone supplements with women who had used neither. In the adjusted model, the risk of colorectal cancer was not associated with the use of statins only (OR = 1.03, 95% CI = 0.73 to 1.45); however, a statistically significantly decreased risk of colorectal cancer was associated with the use of female hormone supplements only (OR = 0.75, 95% CI = 0.58 to 0.96). The risk associated with the use of both statins and female hormone supplements (OR = 0.74, 95% CI = 0.52 to 1.06) was similar to that observed among users of female supplements alone.

The risk of colorectal cancer was not associated with statin use among men, among women, among subjects younger than age 65 years, among subjects aged 65 years or older, or among those with colorectal cancer on the left or right side ( Table 5 ). Statistically significantly reduced risks for cancer of the rectum (OR = 0.70, 95% CI = 0.51 to 0.95) and for stage IV cancers (OR = 0.49, 95% CI = 0.26 to 0.91) were associated with statin use. However, in subset analyses for cancer of the rectum, reduced risk was statistically significantly associated with less than 5 years of use (data from 85 case patients and 141 control subjects, OR = 0.63, 95% CI = 0.44 to 0.90) but not with 5 or more years of use (data from 49 case patients and 59 control subjects, OR = 1.09, 95% CI = 0.66 to 1.81), indicating that the overall reduction in risk was due to chance. The inverse pattern was found for stage IV cancers; risk was not associated with less than 5 years of use (data from 25 case patients and 37 control subjects, OR = 0.64, 95% CI = 0.33 to 1.27), but reduced risk was statistically significantly associated with 5 or more years of use (data from 8 case patients and 18 control subjects, OR = 0.18, 95% CI = 0.05 to 0.62). The distribution of stage was similar among colon and rectal cancer case patients. Statin use was associated with reduced risks of both stage IV colon cancer (OR = 0.51, 95% CI = 0.24 to 1.09) and stage IV rectal cancer (OR = 0.33, 95% CI = 0.10 to 1.12).

We assessed the association between the use of other antihyperlipidemic drugs (e.g., gemfibrozil or cholestyramine resin) and colorectal cancer risk. The risk of colorectal cancer was not associated with use of other antihyperlipidemics (data from 107 case patients and 105 control subjects who had used these drugs, OR = 1.33, 95% CI = 0.92 to 1.91).

All control subjects and 87% of the case patients had listed telephone numbers. We repeated the analysis among case patients with listed telephone numbers and their matched control subjects to assess whether the inclusion of cases without listed telephone numbers introduced bias in the main analyses. In the adjusted model, risk of colorectal cancer was not associated with regular statin use (OR = 0.96, 95% CI = 0.80 to 1.15), compared with irregular and nonuse. We also had dietary data for 853 matched sets. Among these subjects, risk was not associated with statin use (OR = 1.08, 95% CI = 0.83 to 1.40), compared with irregular use and nonuse, in the adjusted model.

Discussion

In this large case–control study, we found no association between the use of statins and risk of colorectal cancer. There was no statistically significant decrease in risk associated with statin use of long duration or in the highest statin dose categories, including use of the three most commonly used statins. A statistically significant 73% increase in risk was associated with lovastatin use, but this increase may have been due to chance; no other data suggest that lovastatin, in particular, increases the risk of colorectal cancer. There was no statistical interaction between statin and NSAID use, and the reduced risk associated with the use of both reflects the influence of continuing NSAID use. Also, there was no statistical interaction between the use of statins and hormone replacement therapy.

Statin use was statistically significantly associated with reduced risk for cancer of the rectum, but the inconsistency of risk associated with duration of use indicates that this finding was probably due to chance. We also found that a statistically significantly reduced risk for stage IV cancers was associated with statin use, raising the possibility that statins may inhibit cancer progression rather than cancer initiation. However, some studies have found an inverse relation between cholesterol levels and colon cancer risk ( 23 – 26 ), indicating that lowered cholesterol may be a consequence of preclinical disease ( 26 ). The reduced risk for stage IV cancer that we observed may also reflect this inverse relationship.

Our results conflict with those from a population-based case–control study of similar size conducted in Israel ( 15 ). In adjusted multivariable analyses in that study, a reduced risk of colorectal cancer was found to be associated with the use of statins for at least 5 years, compared with less than 5 years of use (OR = 0.50, 95% CI = 0.40 to 0.63). This analysis was adjusted for NSAID use, physical activity, hypercholesterolemia, family history of colorectal cancer, ethnic group, and vegetable consumption. The authors of that study subsequently reported ( 27 ) a reduced risk for colorectal cancer associated with statin use after adjustment for the prevalence of cancer screening, smoking, and vitamin supplementation (OR = 0.54, 95% CI = 0.41 to 0.72). This risk differed little from the original risk ( 15 ), but risks associated with duration of use or dose were not presented. A reduced risk for both colon and rectal cancers was observed, but this risk did not differ by stage. It is not clear why our results and results from that study differ. In the Israeli study, patients were recruited between May 31, 1998, and March 31, 2004, and control subjects were identified from Israel's largest health care provider. Case patients and control subjects also had similar health insurance and access to health services. In our study, the most commonly used statin was atorvastatin, whereas in the Israeli study, the most commonly used statins were simvastatin and pravastatin. No reduction in risk was found among users of these drugs in our study. Furthermore, the age-standardized incidences between the United States (33.1 cases per 100 000 people) and Israel (34.9 cases per 100 000 people) of colorectal cancer are similar ( 28 ).

In the Cancer Prevention Study II Nutrition Cohort ( 18 ), current use of cholesterol-lowering drugs for 5 or more years was not associated with the incidence of colorectal cancer (relative risk = 1.09, 95% CI = 0.83 to 1.43). Although the subjects in that cohort were not specifically queried about statin use, an assumption was made that, because drug use was ascertained in 1992 and 1997, 53% of users would have been using statins. If statins reduced the risk of colorectal cancer by approximately 50%, the authors argued, a reduced rate ratio of approximately 0.75 would be expected.

In the original clinical trials of statins, no statistically significant difference in colorectal cancer incidence was observed between treatment and placebo groups in the initial follow-up periods of 5–6 years ( 29 – 32 ) or in a 10-year follow-up of participants in the Scandinavian Simvastatin Survival Study ( 33 ), although the number of cases of colorectal cancer was small. A recent meta-analysis of clinical trials found no effects of statin use on cancer overall or on colorectal cancer specifically ( 34 ). Likewise, three studies that linked pharmacy and cancer registry databases found no associations between statin use and colorectal cancer risk ( 16 , 17 , 19 ). These studies were limited by short follow-up times (i.e., 3–7 years), lenient definitions of use, small numbers of colorectal cancer cases, small numbers of statin users, and no information on duration of statin use.

In the prospective Women's Health Initiative, a reduced risk of breast cancer was observed among users of hydrophobic statins (simvastatin, lovastatin, and fluvastatin) but not among users of the other statins ( 35 ). Results of some cell culture studies ( 36 , 37 ) indicate that anticancer effects are limited to the hydrophobic statins. However, simvastatin and lovastatin have the lowest bioavailability of the statins. In our study, the risk for colorectal cancer was not associated with the use of simvastatin and lovastatin combined. Likewise in the meta-analysis of statin trials ( 34 ), no difference in the associations between hydrophilic and hydrophobic statins and cancer incidence was found.

Our study had many long-term regular users of statins, which provided more than 99% power to detect a protective effect of the magnitude observed in the Israeli study (i.e., 50% reduction in risk). Our study had several limitations. Although we cannot rule out selection bias, associations of colorectal cancer with NSAID use, with colonoscopy screening, and with use of oral contraceptives or female hormones were in the expected directions, providing reassurance as to the validity of the data. No associations with other antihyperlipidemic drugs were found, as expected. In addition, results from analyses limited to cases and controls with listed telephone numbers, and to those with dietary data, were similar to those from the main analysis. Another limitation was the potential for confounding. In our analyses, adjustment for NSAID use and for screening colonoscopy shifted estimates adjusted only for the matching factors toward the null, and adjustment for other potential confounders had no effect. In nonrandomized studies, control of confounding can rarely be considered complete. However, within the limits of what is now known about the epidemiology of colorectal cancer, we consider it unlikely that uncontrolled confounding in our data masks an important preventive effect of statins because such a confounder would have to be a strong risk factor for colorectal cancer that was associated with statin use.

Another limitation in our study was that we classified exposure on the basis of self-reported drug use, and accuracy of recall is of concern. Recall accuracy is greater when drug use is elicited by asking about both indication and specific drug names, when drugs have been used on a regular basis and for longer periods of time, and when the time between reporting of use and actual use is short ( 38 ). We used both indication and specific drug names to elicit use. Most use was daily, long term, and current. Two validation studies have evaluated statin use. Among the 286 self-reported statin users for whom pharmacy records were available in the Israeli study ( 15 ), records confirmed use for 276 (96.5%). Among women enrolled in a case–control study of breast cancer and medication use ( 39 ), self-reported statin use at three times before the index date was compared against pharmacy records. Sensitivities for use 6 months before the index date were 83% for case patients and 93% for control subjects; corresponding sensitivities for use 2 years before the index date were 75% and 86% and for 8 years before this date were 67% and 75%. Sensitivities did not differ statistically significantly between case patients and control subjects. Specificity was approximately 100% for both case patients and control subjects for the three times. If recall bias were operating in the present study, one would expect it to also have affected the associations between colorectal cancer risk and the use of NSAIDs, oral contraceptives, and hormone replacement therapy, but these associations were in the expected direction.

In conclusion, we found no association between the use of statins and colorectal cancer risk. The suggestive evidence of a decrease in risk of stage IV cancer that we found requires confirmation in other studies. This study was initiated because there is compelling evidence that statins have anticancer effects in vitro and in rodent models ( 12 , 13 ). Although the various statins have different pharmacokinetics, overall there is an important hepatic first-pass effect (i.e., metabolism by the liver), and they have low systemic bioavailability ( 40 , 41 ). Thus, statins may not reduce cancer risk at sites distant to the liver.

References