The Association of Prediagnostic Statin Use with Aggressive Prostate Cancer from the Multiethnic Cohort Study

Prostate cancer is the most common cancer and the second leading cause of cancer deaths among men in the United States (1). Well-established risk factors for prostate cancer include increasing age, African ancestry, family history of prostate cancer, and inherited susceptibility. There is also strong evidence to support an association of excess body weight and abdominal adiposity with the risk of advanced or fatal disease (2). The prolonged carcinogenesis of prostate cancer makes it an attractive target for chemoprevention. However, in large-scale randomized trials to date, no lifestyle modifications, dietary changes, and/or nutrient supplements have been demonstrated to substantially reduce a man's risk of prostate cancer (3).

Research in the past decade has highlighted the potential benefits of statins in reducing prostate cancer incidence and/or mortality. Statins (3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors) are a class of medications that can effectively reduce low-density lipoprotein cholesterol. Statins are currently one of the most widely prescribed medications in the United States. According to the National Health and Nutrition Examination Survey, statin use increased from 16.3% in 2003–2004 to 23.2% in 2011–2012 among adults ages 40 or over (4). During 2005–2012, an estimated 55.5% of adults ages |$ \ge|21 years eligible for cholesterol treatment were taking cholesterol-lowering medications, and this proportion was higher in Whites (58.0%) than in Latinos (47.1%) or Blacks (46.0%; ref. 5). Given the potential chemoprevention effects of statins observed in various cell lines, numerous epidemiologic studies have been conducted to investigate the association of statin use and the risk of breast cancer, colorectal cancer, lung cancer, and prostate cancer. In general, few strong or consistent associations between statin use and cancer incidence had been detected among the examined cancer sites (6). Although studies on statin use and total prostate cancer have reported conflicting findings (7–16), a consistent relationship between statin use and reduced risk of aggressive prostate cancer has been observed in several recent studies (7, 8, 11, 15), especially among long-term statin users (9, 14, 15, 17). Moreover, statin use, prediagnostic or postdiagnostic, has both been associated with a large reduction in all-cause mortality and prostate cancer-specific mortality in patients with prostate cancer (17–22).

Epidemiologic studies of statin use and prostate cancer were primarily conducted in White populations. In studies that included African American participants, the associations were mostly nonsignificant, which may partially be due to smaller sample sizes (9, 13, 23). Studies in Asian populations reported mixed findings regarding the potential inverse association of statin use with prostate cancer risk (8, 10, 14). No studies in Latino or other racial/ethnic minority groups have been published. To better understand the potential benefits of statins in lowering prostate cancer incidence and/or mortality in multiple racial and ethnic groups, we investigated the association of prediagnostic statin use and the risk of prostate cancer among 31,062 men over an average of 11.4 years of follow-up in the Multiethnic Cohort Study (MEC).

The MEC is an ongoing prospective cohort study established in 1993–1996 to study risk factors for cancer (24). The MEC recruited over 215,000 residents of Hawaii and Los Angeles at age 45 to 75, primarily from five racial/ethnic groups (Japanese American, Native Hawaiian, African American, Latino, and White). Upon entry into the cohort, participants completed a 26-page self-administered questionnaire, including information regarding demographics, medical history, family history of cancer, cigarette smoking, height/weight, and physical activity, as well as a quantitative food frequency questionnaire. Follow-up questionnaires were sent every 5 years to update this information and to collect information on the history of prostate-specific antigen (PSA) testing from male participants. From 2001 to 2006, about 70,000 men and women in the MEC contributed blood and urine specimens to the biorepository and completed a medication questionnaire at specimen collection. At this visit, participants were asked to gather all medications they had taken during the past 2 weeks. Trained study personnel then recorded the name of each medication taken. Of the 36,860 men who completed the medication questionnaire, a total of 31,062 were included in this analysis. Participants were excluded if they (i) had a prostate cancer diagnosis prior to the medication questionnaire (N = 4,596), (ii) had missing information on statin use (N = 748), date of cancer diagnosis (N = 210), or date of medication questionnaire (N = 19), or (iii) self-reported as “other” race/ethnicity (N = 574). This study was approved by the Institutional Review Boards overseeing research on human subjects at the University of Hawaii (Honolulu, Hawaii) and the University of Southern California (Los Angeles, CA). All individuals included in this analysis provided informed consent.

Histologically confirmed invasive prostate cancer was identified by linkage of the cohort to the Surveillance Epidemiology and End Results (SEER) cancer registries covering the state of Hawaii and California. The linkage also provides additional clinical information about the diagnosis such as cancer stage and grade. The cohort was also linked to death certificate files in Hawaii and California and the National Death Index. The current analysis includes all prostate cancer cases diagnosed among eligible participants beginning on the date the medication questionnaire was completed through December 31, 2017. A total of 2,748 incident prostate cancer cases were diagnosed during an average follow-up period of 11.4 years, including 2,034 localized, 425 advanced (T4 or N1 or M1), 1,945 low-grade (Gleason score ≤ 7), 639 high-grade (Gleason score ≥ 8), 261 fatal (prostate cancer as the cause of death), and 880 aggressive (T4 or N1 or M1 or Gleason score ≥ 8) prostate cancer cases.

In this prospective cohort analysis, we assessed the association of prediagnostic statin use to prostate cancer risk considering all cases, cases classified by stage or grade, fatal cases, and aggressive cases. In all analyses, men contributed person-time at risk from the completion date of medication questionnaire until invasive prostate cancer diagnosis, death, or end of follow-up (December 31, 2017). The HR and 95% confidence intervals (CI) were estimated using Cox proportional hazards models with age as the time metric. The proportionality assumption was tested by Schoenfeld residuals and found to be met. The full model adjusted for age at cohort entry, history of prostate cancer in father or brother(s) (positive, negative), body mass index (BMI; < 25 kg/m², 25–30 kg/m², > = 30 kg/m²), history of diabetes (yes, no), smoking (never, former, current), physical activity (quartiles), education (≤ 8th grade, 9th–12th grade, vocational school/some college, graduated college or higher), and history of PSA testing (ever, never). Ethnicity was adjusted as a strata variable in the model. Physical activity included in the model was the metabolic equivalents for a 24-hour day (MET-hour/day) of moderate and vigorous activity per day, parameterized as quartiles based on the distribution of the variable in the overall study population. Education was considered an indicator of socioeconomic status. Missing values of these variables were coded as an “unknown” category in the multivariate analysis. History of diabetes was summarized from all questionnaires completed before the medication questionnaire while self-reported family history of prostate cancer and PSA testing were derived from all questionnaires completed before the event of interest or end of follow-up. The remaining variables were from the baseline cohort questionnaire. We performed these analyses in all men and separately by race/ethnicity (White, African American, Japanese American, Latino, and Native Hawaiian), family history of prostate cancer (negative, positive), education (≤ 12th grade, > 12th grade), history of diabetes (no, yes), BMI (<25 kg/m², ≥ 25 kg/m²), physical activity (below or above median), smoking status (never, ever) and history of PSA testing (never, ever). The likelihood ratio test (LRT) was performed to evaluate the interactions between statin use and these covariates. We also performed sensitivity analyses that (i) additionally adjusting for self-reported history of colonoscopy (ever/never) as the proxy variable for access to care, and (ii) limited to participants with no missing data for any covariates (N = 28,799). All statistical tests were two sided, and P < 0.05 was considered statistically significant. Data analysis was performed using R (25).

In patients with prostate cancer diagnosed after the medication questionnaire, we conducted a survival analysis to describe the effect of prediagnostic statin use, overall and by ethnicity, on all-cause and prostate cancer–specific mortality. Survival time was modeled as years starting at the date of the invasive prostate cancer diagnosis and ending at the first of the following endpoints: (i) date of prostate cancer–specific death, (ii) date of nonprostate cancer–specific death, or (iii) the end of follow-up (December 31, 2017). HRs and 95% CIs were estimated using Cox proportional hazards models. In the analysis of prostate cancer–specific mortality, the model adjusted for age at prostate cancer diagnosis, SEER summary stage (local, regional/metastatic), Gleason score (≤7, >7), ethnicity, history of PSA testing (ever/never), and history of colonoscopy (ever/never). To assess all-cause mortality, the models were further adjusted for smoking status (never, former, current), BMI (<25 kg/m², ≥25 kg/m²), history of diabetes (yes, no), and physical activity (quartiles in MET-hour/day). We also evaluated the association of any statin use with the two mortality outcomes in all patients with prostate cancer diagnosed at any time since the MEC enrollment.

For the association of prediagnostic statin use with prostate cancer incidence and mortality, we calculated an E-value to assess potential bias resulting from an unknown or unmeasured confounder (26). The E-value was calculated directly from the estimated HR and the 95% CIs.

This study included 31,062 male MEC participants that completed the medication questionnaire, of which 2,748 incident prostate cancer cases were identified during an average of 11.4-year follow-up. A total of 9,739 men reported statin use, which represents 90.6% of men reporting use of any lipid-lowering medication. Compared with Whites (29.9%), the prevalence of statin use was significantly lower in African Americans (21.3%) and Latinos (25.5%), and higher in Japanese Americans (38.6%) and Native Hawaiians (34.0%). The observed differences in the prevalence of statin use were independent of the age differences across ethnic groups. Men that were overweight (BMI, 25–30 kg/m²) or obese (BMI ≥ 30 kg/m²) at cohort baseline were also more likely to report statin use. Statin use appeared to be more prevalent in men with a higher education level or those that had ever undergone PSA testing. History of diabetes was strongly correlated with statin use, with 45.0% of diabetic men taking statins compared with only 27.9% of nondiabetic men. Men who were more physically active at cohort baseline (i.e., fourth quartile) had a significantly lower frequency of statin use in comparison with men with low levels of physical activity (i.e., first quartile; Table 1).

The associations of prediagnostic statin use with different prostate cancer phenotypes are presented for all men and by race/ethnicity in Table 2. Statin use had no impact on the incidence of total prostate cancer or nonaggressive prostate cancer (low-grade or localized). However, we did observe a suggestive inverse association of statin use with risk of high-grade (HR = 0.86, 95% CI = 0.72–1.03, P = 0.10) or advanced (HR = 0.82, 95% CI = 0.66–1.03, P = 0.09) prostate cancer. These inverse associations were consistently suggested across the racial/ethnic groups (P_LRT, 0.18–0.47) and appeared to be greater in Latinos and Whites (HR, 0.64–0.77). It was suggested that Japanese American statin users had a lower risk of localized prostate cancer than nonusers (HR = 0.87, 95% CI = 0.75–1.02, P = 0.08) whereas a significantly higher risk of localized prostate cancer in statin users was observed in African Americans (HR = 1.35, 95% CI = 1.08–1.69, P = 0.01) and Native Hawaiians (HR = 1.54, 95% CI = 1.08–2.19, P = 0.02). This directionally opposite association with localized prostate cancer across ethnic groups was statistically significant (P_LRT = 0.005; Table 2).

In our study cohort, statin use was significantly associated with a 32% reduction in the risk of fatal prostate cancer (HR = 0.68, 95% CI = 0.50–0.91, P = 0.01; Table 2). This inverse association was consistently suggested across racial/ethnic groups (P_LRT = 0.95). The E-value was estimated to be 2.33, indicating that an unknown confounder would have to be associated with both statin use and fatal prostate cancer by a relative risk of 2.33 to explain away the observed inverse association (Supplementary Table 1). When aggressive forms of prostate cancer were combined, the risk of aggressive prostate cancer was suggestively lower in statin users than nonusers in all men (HR = 0.89, 95% CI = 0.76–1.03, P = 0.11). There were no appreciable differences in the results from the sensitivity analysis (Supplementary Table 2).

Between men with a positive and negative history of PSA testing, we observed directionally different associations of statin use for high-grade prostate cancer (P_LRT = 0.06) and aggressive prostate cancer (P_LRT = 0.03) while the associations with other prostate cancer phenotypes were similar between the two groups (Supplementary Table 3). In men reported a positive history of PSA testing (63.5% of the study population), prediagnostic statin use was significantly associated with a lower risk of high-grade prostate cancer (HR = 0.78, 95% CI = 0.63–0.96, P = 0.02), fatal prostate cancer (HR = 0.60, 95% CI = 0.42–0.87, P = 0.006) and aggressive prostate cancer (HR = 0.79, 95% CI = 0.66–0.95, P = 0.01) and no association was found with total prostate cancer and nonaggressive prostate cancer (localized or low-grade). Although no significant association was observed among men without previous PSA testing, it was suggested that statin users had a slightly higher risk of localized prostate cancer (HR = 1.17, 95% CI = 0.97–1.40, P = 0.09) compared with nonusers. We found no evidence of effect modification by other covariates (Supplementary Table 3).

Among the 2,748 men with prostate cancer, 971 died during follow-up (average years after diagnosis = 5.7) and 261 had prostate cancer as the primary cause of death (average years after diagnosis = 3.8). The most common causes of death were prostate cancer, other cancers, and cardiovascular diseases, accounting for 29.1% (N = 203), 18.6% (N = 130), and 23.6% (N = 165) of deaths in nonusers and 21.1% (N = 58), 19.6% (N = 54), and 32.7% (N = 90) of deaths in statin users, respectively. Compared with nonusers, there was a higher proportion of advanced prostate cancer and a similar proportion of high-grade and low-grade disease among statin users (Supplementary Table 4) The risk of dying from prostate cancer was substantially lower in statin users than in nonusers (HR = 0.81, 95% CI = 0.59–1.10, P = 0.17; Table 3). Prediagnostic statin use also showed a suggestive inverse association with all-cause mortality (HR = 0.92, 95% CI = 0.79–1.07, P = 0.28), and this association appeared to be greater in African American men (HR = 0.78, 95% CI = 0.58–1.05, P = 0.10) and Japanese American men (HR = 0.79, 95% CI = 0.59–1.05, P = 0.11).

Similar results were observed in all patients with prostate cancer (N = 6,999) diagnosed since the MEC enrollment (Supplementary Table 5). Statin use, prediagnostic or postdiagnostic, was significantly associated with a 20% lower prostate cancer–specific mortality (HR = 0.80, 95% CI = 0.67–0.96, P = 0.02) and an 8% decrease in all-cause mortality (HR = 0.92, 95% CI = 0.85–0.99, P = 0.03), especially in African Americans (HR = 0.80, 95% CI = 0.69–0.94, P = 0.005) and Japanese Americans (HR = 0.81, 95% CI = 0.70–0.94, P = 0.007).

In this multiethnic, population-based cohort, prediagnostic statin use was associated with a 32% lower risk of fatal prostate cancer adjusting for other risk factors. This inverse association was consistently suggested across the five racial/ethnic groups included in the analysis. Statin use was also related to a 14%–18% reduction in risk of aggressive disease (advanced and/or high-grade). Statin use showed no impact on the risk of total prostate cancer or nonaggressive prostate cancer phenotypes (low-grade or localized) in all men, though in Japanese Americans a 10%–13% risk reduction was suggested and among African Americans and Native Hawaiians, statin users appeared to have a higher risk of localized prostate cancer than nonusers. Among patients with prostate cancer, prediagnostic use of statin or any statin use was associated with a lower risk of dying from the disease and reduced all-cause mortality.

Findings from our analysis are in line with previous reports from large population-based cohorts, where current use of statin or lipid-lowering drugs, in general, were inversely associated with aggressive forms of prostate cancer (defined using varying levels of Gleason grade, clinical stage, or a combination of both variables) but not with total prostate cancer. In reports from The Health Professionals Follow-up Study consisting of 44,126 male participants (27, 28), current statin use was inversely associated with lethal (metastatic or fatal) prostate cancer (HR = 0.76, 95% CI = 0.60–0.96) and this association was stronger among those with 5 or more years of statin use (HR = 0.65, 95% CI = 0.47–0.90). Statin use was unrelated to overall, localized, or low-grade prostate cancer risk (28). In another prospective cohort analysis of data from 55,454 male participants of the Cancer Prevention Study-II Nutrition Cohort, investigators found that current use of cholesterol-lowering drugs for 5 or more years was marginally associated with a 40% reduction in risk of advanced prostate cancer (HR = 0.60, 95% CI = 0.36–1.00) but was not associated with overall prostate cancer (29). Results from the Southern Community Cohort (N = 32,091) suggested a 38% lower risk of high-grade prostate cancer in current statin users than nonusers (HR = 0.62, 95% CI = 0.30–1.28; ref. 13). From the single largest cohort study including 249,986 men from the Saskatchewan Ministry of Health, there was an inverse association of statin use with metastatic prostate cancer (HR = 0.69, 95% CI = 0.61–0.79; ref. 7). A recent meta-analysis that pooled results from 36 observational studies and six randomized clinical trials published by the end of 2015 reported a null effect of statins on total prostate cancer [relative risk (RR) = 0.92, 95% CI = 0.82–1.03], localized prostate cancer (RR = 0.98, 95% CI = 0.91–1.06), and low-grade prostate cancer (RR = 0.95, 95% CI = 0.88–1.02), but a significant protective effect of statins on risk of advanced prostate cancer (RR = 0.87, 95% CI = 0.82–0.91) and high-grade disease (RR = 0.83, 95% CI = 0.66–0.99; ref. 30). These associations are further supported by preclinical studies where statins have been found to directly inhibit prostate cancer development and progression through cholesterol-mediated and non–cholesterol-mediated pathways (31).

Most of the abovementioned studies on statin and prostate cancer were conducted in North American cohorts and predominately White populations. Among these studies, several have included African Americans in the analysis, though the sample sizes were usually small. In the Atherosclerosis Risk in Communities study consisting of 5,007 Whites and 1,511 African American men, investigators observed a suggestive inverse association of lipid-lowering medications (76% of which were statins) with fatal prostate cancer in White men (HR = 0.58, 95% CI = 0.33–1.04) and this association appeared to be attenuated in African American men (HR = 0.95, 95% CI = 0.42–2.15), though the difference was not statistically significant (P_interaction = 0.62; ref. 9). A report from the Southern Community Cohort found the association of statin use with high-grade prostate cancer was similar (P_interaction = 0.81) in Whites (HR = 0.56, 95% CI = 0.17–1.83) and African American men (HR = 0.65, 95% CI = 0.29–1.44; ref. 13). In a case-case analysis from the North Carolina-Louisiana Prostate Cancer Project including 344 aggressive and 1,586 nonaggressive Whites (52.4%) or African American (47.6%) cases, statin use was significantly associated with a lower risk of aggressive prostate cancer (OR = 0.74, 95% CI = 0.56–0.96), with similar effect estimates in Whites (OR = 0.64, 95%CI = 0.44–0.95) and in African American men (OR = 0.84, 95% CI = 0.58–1.21; ref. 23). In our analysis, we also found that the association of statins with aggressive prostate cancer phenotypes observed in African American men was comparable in magnitude to those observed in White men.

In the United States, Asians and Pacific Islanders have the lowest incidence rate of prostate cancer and highest 5-year survival rate among all racial/ethnic groups. The effect of statins or lipid-lowering medications has not been studied extensively in East Asian populations. A cohort analysis including 143,870 men from the Korean National Health Insurance Service-National Sample Cohort found no association of statins with total prostate cancer (HR = 1.01, 95% CI = 0.85–1.18; ref. 10). In another cohort study of 26,628 men with ischemic heart disease from the National Health Insurance Research Database in Taiwan, statin use was associated with a significantly lower risk of total prostate cancer (HR = 0.72, 95% CI = 0.57–0.91) and advanced prostate cancer (HR = 0.72, 95% CI = 0.53–0.97; ref. 8). Japanese Americans represented the largest racial/ethnic group in our study population (36.0%). In contrast to the directionality of associations found in most other groups, prediagnostic statin use was associated with a reduced risk of localized prostate cancer among Japanese Americans with this association being statistically different from those observed in other groups.

It is known that frequent PSA screening promotes the early detection of localized and indolent prostate cancer. In our cohort, statin users were more likely than nonusers to have a history of PSA testing; however, we did not observe statin use to be associated with a greater risk of low-grade disease. In the separate analysis in men with and without PSA testing, the inverse association of statins with aggressive prostate cancer phenotypes observed in all men is still present when limited to men who reported a positive history of PSA testing (64.6% of the overall study population). These findings are consistent with other reports where the association of statins with fatal or aggressive prostate cancer was found to be similar in populations with low or high PSA screening rates (23, 28). Statins are also known to cause declines in serum PSA levels and were found to be associated with a lower rate of an abnormal PSA screening result (32–34). This could potentially lead to fewer biopsies in statin users and a decreased incidence of early-stage prostate cancer. The delayed prostate cancer diagnosis may be expected to result in an increased incidence of advanced prostate cancer. However, these expectations contradict the observed findings from our analysis and most published studies where statin use was associated with a lower risk of aggressive prostate cancer. Therefore, it is unlikely that the effect of statin on PSA levels introduced any significant bias in our findings.

Our study has strengths and limitations. With a large number of cases and a long duration of follow-up, our study is well powered to detect an association of statins with the risk of prostate cancer, including the less prevalent but clinically important aggressive types of prostate cancer. Although this study included multiple race/ethnicity groups there was limited power in each group, and the ethnic-specific results need to be interpreted with caution. Another limitation of our analysis was the lack of detailed information on the duration of statin use, dosage, and type of statin prescriptions. Men were classified as statin users and nonusers based on a single assessment of current prescribed use at the interviewer-administered medication questionnaire. Although several studies have suggested a greater effect of statins on prostate cancer risk among long-term users and/or high-dose users (8, 9, 14, 28), we did not have sufficient information to test these hypotheses in our study population. Moreover, the health care–seeking behavior in statin users may introduce bias to the observed associations with prostate cancer incidence and mortality. However, further adjustments on the proxy variables of access to health care (i.e., history of PSA testing and history of colonoscopy) did not make any appreciable difference in our findings. Finally, although there may be residual confounding in the observed associations due to unknown or unmeasured factors, the E-values suggested that any residual confounding was unlikely to fully explain the observed inverse association between statin use and aggressive prostate cancer phenotypes.

In conclusion, results from this prospective MEC analysis support the potential beneficial effects of prediagnostic statin use on reducing the risk of aggressive forms of prostate cancer (high-grade, advanced, or fatal prostate cancer). Our results also suggest potential heterogeneity of statin effects on the risk of nonaggressive prostate cancer phenotypes (low-grade or localized) in certain racial/ethnic populations. It is unclear whether these findings were due to the small sample size in ethnic-specific analysis or other unmeasured or unknown factors. Future studies are warranted to validate these results.

No disclosures were reported.

F. Chen: Conceptualization, formal analysis, investigation, writing–original draft, writing–review and editing. P. Wan: Data curation, writing–review and editing. L.R. Wilkens: Resources, writing–review and editing. L. Le Marchand: Resources, writing–review and editing. C.A. Haiman: Conceptualization, resources, supervision, writing–review and editing.

This work was supported by NCI at the NIH: U01CA164973 (to C.A. Haiman, L.R. Wilkens, and L. Le Marchand) and T32CA229110 (to F. Chen).

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.