Abstract
Lipid-lowering medications, particularly statins, may protect against aggressive prostate cancer. Fatal prostate cancer, the most clinically relevant outcome, remains understudied for this association. We prospectively studied lipid-lowering medication use and both incident and fatal prostate cancer in black and white men in the Atherosclerosis Risk in Communities (ARIC) study. A total of 6,518 men without cancer at visit 2 (1990–1992), the start of the statin era, were followed for prostate cancer incidence and death through 2012. Medication use was collected during study visits and telephone calls at up to nine time points during follow-up. Cox regression was used to estimate HR and 95% confidence intervals (CI) of total (white N = 541, black N = 259) and fatal (white N = 56, black N = 34) prostate cancer overall and by race. Lipid-lowering medication use was modeled as time-dependent current use or duration (never, <10, and ≥10 years). By visit 4 (1996–1998), 21% of white and 11% of black men had used a lipid-lowering medication, mostly statins. There was a suggestion that current users were less likely to die from prostate cancer than nonusers (HR = 0.67, 95% CI = 0.42–1.07) after multivariable adjustment. We observed no statistically significant differences between black and white men. Current use was not associated with incident prostate cancer, although long-term use was statistically significantly inversely associated with incidence (HR = 0.68; 95% CI = 0.50–0.92). Long-term lipid-lowering medication use was associated with lower risk of prostate cancer. Current use was possibly associated with fatal prostate cancer.
Introduction
Accumulating evidence supports that lipid-lowering medications, particularly statin drugs, the most widely prescribed class since the 1990s, may preferentially protect against aggressive prostate cancer (1). For example, a recent meta-analysis reported a modest inverse association between statin drug use and incident prostate cancer [relative risk (RR) = 0.93, 95% confidence interval (CI) = 0.87–0.99] based on 27 studies, but a more pronounced inverse association for advanced disease (RR = 0.80; 95% CI = 0.70–0.90) based on seven studies (2). Some studies have observed that advanced/aggressive disease risk decreases with increasing duration of use (2), but these studies were able to only investigate shorter-term use. Although these results are intriguing as we seek new prevention strategies, the definition of advanced (stage-based) or aggressive (grade, or stage plus grade) disease is not consistent among these studies, making it difficult to determine whether such drugs might be useful in preventing disease that is the most likely to progress to being fatal. Furthermore, fatal prostate cancer, the most clinically relevant outcome for this cancer with a heterogeneous prognosis, remains understudied for statin drugs including for long-term duration of use; its study requires long follow-up of large cohorts of men in the statin era. The two studies that have examined statin drugs and fatal prostate cancer among men without the diagnosis at the start of follow-up report their use to be inversely associated with fatal prostate cancer (3, 4), but questions remain about the needed duration of use for maximal benefit and which subpopulations, including black men, will benefit.
Thus, to address these gaps, we undertook an analysis of current and long-term use of lipid-lowering medications and risk of incident, lethal (i.e., first primary prostate cancer case that either had distant metastasis at diagnosis or that led to death with prostate cancer as the underlying cause), and fatal prostate cancer (i.e., death due to prostate cancer as the underlying cause, regardless of whether prostate cancer was the first primary cancer diagnosis) in the Atherosclerosis Risk in Communities (ARIC) study overall and separately in white and black men.
Materials and Methods
Study population
We conducted this analysis in the ARIC study, a community-based, multicenter, prospective cohort study of 15,792 men (45%) and women aged 45–64 years who were recruited from four U.S. communities: Forsyth County, NC; suburban Minneapolis, MN; Washington County, MD; and Jackson, MS. Twenty-seven percent of the participants are black. Study enrollment began in 1987 and continued through 1989. Participants returned every 3 years for follow-up study visits in 1990–1992, 1993–1995, and 1996–1998, and for a fifth study visit in 2011–2013. At each visit, participants received an extensive examination, which included measurement of anthropometric variables by trained study personnel and interviews asking detailed questions about their medical history, healthcare access, and utilization, as well as demographic and lifestyle factors including medication use. In addition, blood and urine were collected and stored or shipped to central laboratories for analyses. Follow-up telephone calls have been conducted annually between study visits and semiannually since 2012. Local Institutional Review Boards approved the ARIC protocol and informed consent was obtained from the participants. The majority (99.7%) of participants gave approval for follow-up for the occurrence of noncardiovascular diseases including cancer.
For this analysis, we excluded women (N = 8,710), men with a history of cancer at visit 1 or were not otherwise eligible for studies on incident cancer (N = 374), men who were not white or black (N = 23), and men who did not consent to follow-up for noncardiovascular diseases (N = 1). Given that the first approved statin drug reached the market in 1987 and use did not become common until the early 1990s, we used visit 2 (1990–1992) as the start of follow-up and thus, also excluded men who had a cancer diagnosis or died before visit 2, or who attended visit 1 but not visits 2, 3, or 4 (N = 166). These exclusions left 5,007 white and 1,511 black men in the analytic cohort.
Lipid-lowering drug use and duration
At each study visit, participants were asked to bring all medications they had taken during the previous 2 weeks to the visit. Trained study personnel recorded the name and dose of each medication taken. Beginning in 2006, during each follow-up telephone call, participants were asked to retrieve any prescription medications they had taken in the previous 2 weeks and read the labels to the telephone interviewer. Thus, use of lipid-lowering drugs was ascertained from information collected at each study visit through visit 4 (1996–1998) and from information collected during the follow-up telephone calls beginning in 2006 through the end of follow-up. The validity of self-reported lipid medication use in ARIC has been found to be highly accurate (5). We classified the following medications as lipid-lowering: statins, fibrates, bile acid sequestrants, and niacin. By visit 4 (1996–1998), 76% of lipid-lowering medication use was of a statin, and by 2011, 85% was a statin (5). Men were categorized as nonusers of a lipid-lowering medication until the first report of lipid-lowering medication use at a visit or on a follow-up call. If information on lipid-lowering medication use was missing at any assessment, medication use from the prior assessment was carried forward into the next time period. Once a participant became a lipid-lowering medication user, use status was assumed to be unchanged until the date of censor. We made this conservative assumption to minimize the bias that might occur if men who discontinued medications near the end of life were classified as not current users; such a bias would make lipid-lowering medication use appear protective against fatal prostate cancer. With the conservative assumption we made, if men discontinued use for other reasons, and given that we have prospectively collected data, any misclassification should be nondifferential by disease status and the bias in the association should be toward the null. Duration of use was calculated as time of first reported use beginning at visit 1. As described in detail below in the statistical analysis section, duration was modeled as a time-dependent variable.
Covariate assessment
We used data collected at the study visits, including age (years, visit 1), race (visit 1), field center (visit 1), attained education (visit 1), height (cm, visit 1), body mass index (BMI in kg/m2; each visit), cigarette smoking status (each visit), aspirin use (self-reported use in the past 2 weeks, each visit), regularity of routine physical examinations self-reported (visit 1), and health insurance status (self-reported, visit 1) and the type of health insurance (self-reported, visit 3). We also categorized men (at each visit) as having diagnosed diabetes if they self-reported a physician diagnosis or used pharmacologic treatment for diabetes. Men without a diagnosis of diabetes were categorized as, (i) having undiagnosed diabetes if they had a fasting glucose ≥126 mg/dL or nonfasting glucose ≥200 mg/dL, (ii) being at risk for diabetes if they had a fasting glucose of 100 to <126 mg/dL or nonfasting glucose of 140 to <200 mg/dL, or (iii) as not diabetic and not at risk for diabetes. We used total cholesterol concentration that was previously measured in blood collected during study visits (each visit) and categorized participants as having borderline high or high cholesterol (≥5.2 nmol/L) or normal (<5.2 nmol/L).
Ascertainment of prostate cancer cases and deaths
We identified incident prostate cancers diagnosed from 1987 through 2012 by linkage with state cancer registries (MD, MN, MS, and NC) supplemented by active surveillance of the cohort, including for cases diagnosed before the registries were established and for cases among participants who moved out of an ARIC state (6, 7). Cases diagnosed before the registries that were established were ascertained from routinely collected hospital discharge summaries and confirmed with medical record abstraction. Beginning in 2014, we contacted participants who ever reported at a visit or on an annual or semiannual follow-up telephone call that they had a cancer diagnosis for more information on their diagnosis, and collected and reviewed pertinent medical records. For prostate cancer that was self-reported, identified by a prostate cancer code on a hospital discharge summary or as the underlying cause of death assigned to a death certificate, all available sources of data, including from the registry and medical records, were reviewed to confirm or refute the diagnosis. Date of diagnosis, pathologic and clinical tumor–node–metastasis (TNM) stage, and Gleason sum were abstracted from the combination of registry linkages and medical record abstractions. All of these sources of data were adjudicated using standardized protocols described in detail previously (8). We defined incident prostate cancer as a first primary diagnosis of prostate cancer through 2012. We defined fatal prostate cancer as a death due to prostate cancer as the underlying cause through 2012, irrespective of whether the diagnosis of prostate cancer was a first or subsequent primary. For fatal prostate cancer cases, men were censored at their date of death. We defined lethal prostate cancer as a first primary prostate cancer case that either had distant metastasis to any organ at diagnosis [pathologic TNM stage IV or Surveillance Epidemiology and End Results (SEER) summary stage I, IV, or VII] or that led to death with prostate cancer as the underlying cause. For lethal prostate cancer cases, men were censored at their date of diagnosis. Among the analytic cohort, we ascertained 750 incident cases, 91 lethal cases, and 90 fatal cases from visit 2 through 2012. Median follow-up for fatal prostate cancer was 20.6 years (maximum 22.9 years).
Statistical analysis
We calculated means and proportions of demographic, lifestyle, and health-related characteristics by lipid-lowering drug use, overall and stratified by race, for men attending visit 4, after statins had been available on the market for several years. We used Cox proportional hazards regression to estimate the HR and 95% CI of incident, lethal, and fatal prostate cancer for current lipid-lowering medication use (yes/no) or for duration of use (never, <10, ≥10 years). For the incident and lethal prostate cancer analyses, men contributed person-time at risk from visit 2 (when most participants with an indication would have had the opportunity to be prescribed a statin) until prostate cancer diagnosis, diagnosis of another cancer, death, or end of follow-up in 2012, whichever came first. For the fatal prostate cancer analyses, men contributed person-time at risk from visit 2 until prostate cancer–specific death, death due to other causes, or end of follow-up in 2012, whichever came first. To test for trend across duration of use, we included a single continuous variable with values equal to the cumulative duration of use, the coefficient for which we evaluated using the Wald test. All models were adjusted for age (continuous) and joint terms for race and field center (Minneapolis White; Jackson Black; Forsyth County and Washington County White; Minneapolis, Forsyth County, and Washington County Black). Multivariable models were further adjusted for risk and protective factors for fatal prostate cancer in this cohort and others: height (continuous), BMI (time dependent, continuous), cigarette smoking status (time dependent; current/quit <10 years ago, quit ≥10 years ago, never smoker), diabetes status (time dependent), aspirin use (time dependent; yes, no), and education level (baseline; less than high school, high-school graduate/vocational school/some college, college graduate or graduate school, missing). We conducted a sensitivity analysis examining current use of statin drugs separately from current use of other lipid-lowering medications. We also conducted these analyses separately in white and black men, and evaluated statistical interaction using the Wald test.
Because statin drugs are sometimes prescribed to patients for indications other than an elevated serum cholesterol level and one mechanism by which statin drugs may influence prostate cancer is via cholesterol lowering (9), we stratified by measuring circulating total cholesterol concentration (< vs. ≥5.2 nmol/L) prior to the report on use of a lipid-lowering drug. To address the possibility of biases related to uptake of screening for elevated cholesterol and screening for prostate cancer, we (i) stratified by frequency of routine physical examinations (at least every 5 years vs. less frequently), (ii) restricted to participants who had health insurance at visit 1 (91% of the study population), and (iii) further restricted to participants who had private health insurance and/or Medicare at visit 3 (74% of the study population, excluded men without health insurance or on Medicaid only).
Because statin drugs are prescribed to reduce risk of death from cardiovascular diseases in persons with indication, we accounted for competing risks of death from other causes (10) in a sensitivity analysis for fatal prostate cancer.
Statistical analyses were conducted using SAS 9.4. All statistical tests were two sided and P < 0.05 was considered to be statistically significant.
Results
At visit 4, when statins had been available on the market for several years, 20% of men were current users of a lipid-lowering medication (76% of which were statins), which differed between black (11%) and white (21%) men. Compared with men not taking lipid-lowering medications, men taking a lipid-lowering medication were less likely to be current smokers (black men), and more likely to have diabetes, ever to have hypertension, or ever to have had elevated cholesterol (Table 1).
Prostate cancer incidence
Current lipid-lowering medication use was not associated with incident prostate cancer after adjusting for age and jointly for race and field center, or after further multivariable adjustment (Table 2). Throughout our analysis, the multivariable adjusted results were similar to those adjusted only for age and jointly for race and field center; thus, we present multivariable adjusted results in text throughout the Results section. Compared with never use, shorter-term use (<10 years) was not associated with risk of incident prostate cancer, whereas long-term use (≥10 years) was associated with a statistically significantly reduced risk of incident prostate cancer (HR = 0.68; 95% CI = 0.50–0.92, Table 2). We observed similar patterns of association when we considered specifically statins. Current use of a statin was not associated with incident prostate cancer (HR = 1.03; 95% CI = 0.86–1.22). Compared with never use, shorter-term use (<10 years) was not associated with risk of incident prostate cancer (HR = 1.07; 95% CI = 0.90–1.28). The inverse association observed for use of long-term (≥10 years) lipid-lowering medications was inverse, but attenuated and not statistically significant when examining statins specifically (HR = 0.78; 95% CI = 0.56–1.10).
Current lipid-lowering medication use was not associated with incident prostate cancer in either white or black men (Pinteraction = 0.14; Table 3). There was no association with incident prostate cancer when stratifying by circulating total cholesterol concentration prior to starting a lipid-lowering medication (Pinteraction = 0.44; Table 3).
In a sensitivity analysis to minimize the potential for detection bias, the null association between lipid-lowering medications use and incident prostate cancer did not differ by frequency of routine physical examination (Supplementary Table S1, Pinteraction = 0.76). In addition, this null association persisted among the 91% of men who had health insurance and among the 74% of men who had private health insurance and/or Medicare (Supplementary Table S1).
Fatal and lethal prostate cancer
Current lipid-lowering medication users were borderline statistically significantly less likely to die from prostate cancer than nonusers after adjusting for age and jointly for race and field center (HR = 0.64; 95% CI = 0.41–1.02; Table 4). This finding was similar after further multivariable adjustment (HR = 0.67; 95% CI = 0.42–1.07; Table 4). Current use of a statin (rather than any lipid-lowering medication) also was inversely associated with fatal prostate cancer (HR = 0.65; 95% CI = 0.40–1.05). Risk of fatal prostate cancer was similar for both shorter- and longer-term use of lipid-lowering medication (Table 4), and similar results were found with increasing duration of use of a statin (vs. never use: <10 years: HR = 0.64; 95% CI = 0.38–1.08; ≥10 years: HR = 0.74; 95% CI = 0.39–1.40). We observed no association for lethal prostate cancer (i.e., incident metastatic prostate cancer or death from a first primary prostate cancer; current use vs. no use: HR = 1.11; 95% CI = 0.65–1.89; duration of use vs. never use: <10 years: HR = 1.17; 95% CI = 0.68–2.01, ≥10 years: HR = 0.83; 95% CI = 0.28–2.51; Ptrend = 0.92).
Current lipid-lowering medication use appeared to be inversely associated with fatal prostate cancer in both black (HR = 0.76; 95% CI = 0.34–1.70) and white (HR = 0.60; 95% CI = 0.35–1.06; Pinteraction = 0.62; Table 5) men after adjusting for age and field center; results were similar for white men after multivariable adjustment, but somewhat attenuated for black men (Table 5). We observed no differences in the association with fatal prostate cancer when stratifying by circulating total cholesterol concentration prior to starting a lipid-lowering medication (Table 5).
We performed a number of sensitivity analyses. To minimize the potential for detection bias, we took into account access to and uptake of medical care (Supplementary Table S2). We observed no statistical interaction in the inverse association between lipid-lowering medications use and fatal prostate cancer by frequency of having a routine physical examination (Pinteraction = 0.34), although there was a suggestion that the inverse association was stronger among men who reported having a physical examination less frequently than once every 5 years (HR = 0.48; 95% CI = 0.22–1.03) compared with those who had a routine physical at least once every 5 years (HR = 0.85; 95% CI = 0.46–1.57; Supplementary Table S2). In addition, the inverse association between current lipid-lowering medications use and fatal prostate cancer was similar to overall among the 91% of men who had health insurance (HR = 0.63; 95% CI = 0.39–1.02) and among the 74% of men who had private health insurance and/or Medicare (HR = 0.62; 95% CI = 0.35–1.07). To minimize the potential for bias due to competing risks, taking into account all other causes of death (current vs. never use: HR = 0.78; 95% CI = 0.49–1.26) or other noncardiovascular causes of death (HR = 0.76; 95% CI = 0.47–1.23), the inverse association was similar.
Discussion
In this prospective study of black and white men with long follow-up in the statin drug era, we found that long-term use of lipid-lowering medications was associated with a lower risk of developing prostate cancer, and that current use appeared to be associated with a lower risk of fatal prostate cancer among men without the diagnosis at the start of follow up. These findings were not explained by differences in access to and uptake of healthcare or by competing risks of death from other causes. Longer-term use of a statin, which was the most commonly used lipid-lowering drug in this cohort by 2011 (5) also appeared to be inversely associated with incident prostate cancer.
The suggested inverse association between lipid-lowering medications use and fatal prostate cancer in men without the diagnosis at the start of follow-up that we observed is consistent with the two published studies that have addressed this outcome (3, 4). Both of these studies reported an inverse association for fatal prostate cancer (3, 4). Although studying a different point in the natural history of prostate cancer, studies among men with the diagnosis of prostate cancer who were followed to recurrence, metastasis, and prostate cancer–related death (1, 4, 11–14), have found that pre and/or postdiagnostic statin drug use may be inversely associated with these poor outcomes. The fact that studies using both populations, men without and with prostate cancer at the start of follow up, have observed inverse associations suggest that statin drugs may have relevance for both primary and tertiary prevention, but teasing apart causal effects is complex and will require large cohorts or clinical trials of men with well-documented lipid-lowering medication use, and follow-up for prostate cancer diagnosis, recurrence/progression, and prostate cancer–related death.
That we did not observe an association with lethal prostate cancer appears to not be consistent with these previous findings. However, it should be noted that the definition of lethal disease is conceptually different from that of fatal prostate cancer. Although there is considerable overlap in lethal and fatal cases, the person-time denominators differ for these two case definitions with person-time being truncated at the time of diagnosis for lethal cases, and accumulating until the date of death for fatal cases. Thus, with respect to exposure status, if some men began taking a lipid-lowering medication after diagnosis, they would be classified as a never user for lethal analyses, but as a current user from the time of diagnosis forward for fatal analyses. The differences in the findings for lethal and fatal disease may provide additional insight into the timing of the actions of these medications relative to the natural history of prostate cancer from development to progression to death.
Previous reports have highlighted the importance of distinguishing between overall prostate cancer and prostate cancer that has the potential to be fatal when studying the association with lipid-lowering medications (1, 2). Although a recent meta-analysis reported a summary RR < 1.0 for statin use and incident prostate cancer, associations have ranged from weakly positive to null to weakly inverse (2). Positive associations with incident prostate cancer could be due to detection bias (15). Similar to some previous studies (2), we found no overall association between current use of lipid-lowering medications and risk of incident prostate cancer. However, we did observe a lower risk of prostate cancer among long-term users (≥10 years) compared with never users. Although the meta-analysis concluded that longer-term use was not associated with incident prostate cancer risk, most previous studies have only been able to address ≥5 years with a few studies being able to examine ≥10 or more years separately (2). Given that long-term users of lipid-lowering medications may be systematically different from men who take these drugs for shorter periods of time and from never users, we (i) controlled for multiple potential confounding factors, (ii) restricted to those with health insurance (likely able to access medical care for screenings), and (iii) stratified by frequency of medical check-ups, and the results remained the same. Nevertheless, we cannot rule out residual confounding and/or detection bias as explanations for the inverse association for long-term users.
To our knowledge, this is the first study to examine the association between lipid-lowering medication use and risk of fatal prostate cancer in black men. We did not observe meaningful differences in the association between lipid-lowering medication use and either incident or fatal prostate cancer by race, although our sample size was small for fatal prostate cancer and, thus, our ability to detect differences for this outcome, in particular, was limited. One previous study found that statin drug use was associated with a reduced risk of prostate cancer, particularly high-grade disease, in both white and black men (16). Interestingly, a recent analysis of nationally representative data in the United States concluded that, among individuals with coronary heart disease, black men are approximately half as likely as white men to take a statin (17). Meanwhile, the prostate cancer mortality rate is 2.6 times higher in black men than in white men in the United States (18). Increasing the prescription of clinically indicated lipid-lowering drugs in black men might have an impact on prostate cancer mortality in this high-risk group.
Plausible biological mechanisms, both cholesterol and noncholesterol related, by which lipid-lowering medications, in particular statin drugs may preferentially prevent prostate cancer with an aggressive phenotype have been extensively reviewed (9). Prostate cancer cells tend to overaccumulate cholesterol in their cell membranes, forming large lipid rafts, which, in the case of cancer cells, may facilitate procarcinogenic cell signaling (9). Therefore, having lower cholesterol may inhibit cholesterol-sensitive pathways, such as sonic hedgehog and Akt, which are important in carcinogenesis. Furthermore, high cholesterol, particularly high low-density lipoprotein cholesterol, is associated with increased inflammation, which may be associated with increased risk of prostate cancer (9). Perhaps most relevant to prostate cancer mortality is the evidence suggesting that cholesterol is a substrate for intratumoral steroidogenesis, which could lead to castration resistance and disease progression (9). A role for cholesterol in prostate cancer is also supported by a recent review of epidemiologic studies, which concludes that men with lower cholesterol have a lower risk of more aggressive prostate cancer (9). Finally, a recent Mendelian randomization analysis examined genetic risk scores for lipid traits in relation to risk of prostate cancer. This study concluded that there was evidence that the genetic scores for higher low-density lipoprotein cholesterol and triglycerides were associated with an increased risk of aggressive prostate cancer, whereas a genetic variant in HMGCR that mimics the cholesterol-lowering effect of statin drugs was associated with a lower risk of prostate cancer (19). Taken together, this literature provides strong evidence for a protective biologic mechanism of lower cholesterol on prostate cancer mortality.
Our study has many important strengths, including the prospective design, detailed, long follow-up time in the statin era, updated information on lipid-lowering medication use including duration of use, and estimation of the association specifically in black men. In addition, information was available on important covariates, allowing us to control for many potentially confounding factors of the association we studied here, including the fatal prostate cancer risk factors like cigarette smoking (20) and obesity (21).
Several aspects of our study warrant discussion. First, based on analyses we performed and other evidence, we find little support for one possible noncausal explanation for our findings: detection bias. This bias could result from men with more contact with the healthcare system being more likely to receive a PSA-based prostate cancer screening, and if elevated, a follow-up biopsy, as well as screening for cholesterol and, if elevated, a prescription for a lipid-lowering medication. A simulation study performed previously by members of our team showed that such detection bias is unlikely to fully explain the inverse association with advanced disease, but could explain a modest positive association with incident prostate cancer (15). Furthermore, the inverse association between lipid-lowering medication use and risk of more aggressive prostate cancer has been observed in European populations where use of PSA-based prostate cancer screening was low (22). Although we did not have information on prostate cancer screening or PSA concentration for ARIC participants, and therefore could not fully rule out this source of bias, we conducted analyses for incident prostate cancer among men who had health insurance and among men who had private insurance and/or Medicare as well as stratified by frequency of a routine physical examination, the usual time when cholesterol and prostate cancer screenings are done, and found no evidence that the association for current lipid-lowering drug use and incident prostate cancer differed notably by these factors. We had too few fatal cases to conduct these sensitivity analyses for fatal prostate cancer, although this type of bias would only be apparent in the short term for fatal disease.
Second, the overall small size of ARIC for the analysis of fatal prostate cancer, especially in subgroups, is a limitation of this study. Our small sample size limited our ability to detect differences by race and precluded analyzing the association between duration of use and fatal prostate cancer in black men. In addition, we did not have sufficient power to comprehensively examine associations by types of lipid-lowering medication, or the type of statin drug. All but one (3) of the previously conducted studies discussed above (1, 4, 11–14) examined only statin medications rather than all lipid-lowering medications, as was done in our study. Although, as mentioned above, the majority of lipid-lowering drug use in this cohort was statin use, and we did conduct sensitivity analyses restricted only to statins, which produced similar results to our overall findings. Additional studies in larger populations will be necessary to determine whether the observed associations differ by type of lipid-lowering medication or type of statin drug.
Third, we did not adjust for family history of prostate cancer because it was found not to be a confounder in sensitivity analyses.
Fourth, we did not perform an analysis to eliminate the prevalent medication-use bias as described by Hernan and colleagues (23). However, few men in the analytic cohort were using lipid-lowering medications at the start of follow-up for this analysis (3%) and thus, the majority of users were new users, not prevalent users. In a previous analysis of data from another cohort on statin use and the incidence of lower urinary tract symptoms in which follow-up was also begun at the start of the statin drug era when the prevalence of use was low, results from the method of Hernan and colleagues (23) and the time-dependent Cox modeling of duration of use yielded virtually identical estimates (24). Thus, we expect that the HRs we estimated using time-dependent Cox modeling of duration of use in this study in ARIC should approximate what would be observed in a trial on new use of lipid-lowering medication (barring other sources of bias and confounding).
Fifth, we previously documented that ARIC participants have similar prevalences of major cancer risk factors and cancer incidence and mortality rates to the U.S. population within the same age, race, and sex groups (8), and given the consistency of our findings and those of others for lipid-lowering medications, including statins, and with various definitions of advanced prostate cancer, we expect that our suggestive findings for lipid-lowering medications and fatal prostate cancer are likely generalizable to similar groups in the United States as a whole.
In conclusion, longer-term use of lipid-lowering medications may decrease risk of incident prostate cancer, and current use may reduce risk of death from prostate cancer. Further work will need to be conducted to evaluate the full impact of the lower prevalence of statin use in black versus white men on prostate cancer incidence and mortality in the U.S. population as whole. Further work is also needed to determine which lipid-lowering medications used for what duration may provide the most benefit against the development of prostate cancer and when in the natural history of prostate cancer, interception should occur to prevent the fatal phenotype from manifesting. Although these questions must be addressed to support the testing of lipid-lowering medications for primary and tertiary prevention of prostate cancer, the U.S. Preventive Services Task Force has issued a recommendation on the use of statin drugs for the primary prevention of cardiovascular disease (25), which if followed by clinicians may increase the number of older men taking a statin and as a consequence provide a possible ancillary benefit of preventing prostate cancer incidence and mortality.
Disclosure of Potential Conflicts of Interest
E.A. Platz is a senior editor at AACR. No potential conflicts of interest were disclosed by the other authors.
Disclaimer
The content of this work is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.
Authors' Contributions
Conception and design: A.M. Mondul, E.A. Platz
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.E. Joshu, A.R. Folsom, E.A. Platz
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.M. Mondul, C.E. Joshu, J.R. Barber, A.E. Prizment, N.A. Bhavsar, E.A. Platz
Writing, review, and/or revision of the manuscript: A.M. Mondul, C.E. Joshu, A.E. Prizment, N.A. Bhavsar, E. Selvin, A.R. Folsom, E.A. Platz
Study supervision: E.A. Platz
Acknowledgments
The ARIC study has been funded in whole or in part with Federal funds from the National Heart, Lung, and Blood Institute, NIH, Department of Health and Human Services, under contract nos. HHSN268201700001I, HHSN268201700003I, HHSN268201700005I, HHSN268201700004I, and HHSN268201700002I. Studies on cancer in ARIC are also supported by the NCI (U01 CA164975). This research was additionally supported by NCI Cancer Center Support Grant (P30 CA006973). C.E. Joshu was supported by the Prostate Cancer Foundation. E. Selvin was supported by NIH/NIDDK grants K24DK106414 and R01DK089174. The authors thank the staff and participants of the ARIC study for their important contributions. Cancer incidence data have been provided by the Maryland Cancer Registry, Center for Cancer Surveillance and Control, Maryland Department of Health (201 W. Preston Street, Room 400, Baltimore, MD 21201). We acknowledge the State of Maryland, the Maryland Cigarette Restitution Fund, and the National Program of Cancer Registries (NPCR) of the Centers for Disease Control and Prevention (CDC) for the funds that helped support the availability of the cancer registry data.
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.