Abstract
Objective:
To compare the risk of prostate cancer mortality among men treated with 5- alpha reductase inhibitors (5-ARIs) with those treated with alpha-adrenergic blockers (ABs) in community practice settings.
Patients and Methods:
A retrospective matched cohort (N=174,895) and nested case-control study (N=18,311) were conducted in 4 regions of an integrated health care system. Men 50 years and older who initiated pharmaceutical treatment for benign prostatic hyperplasia between January 1, 1992, and December 31, 2007, and had at least 3 consecutive prescriptions were followed through December 31, 2010. Adjusted subdistribution hazard ratios, accounting for competing risks of death, and matched odds ratios were used to estimate prostate cancer mortality associated with 5-ARI use (with or without concomitant ABs) as compared with AB use.
Results:
In the cohort study, 1,053 men died of prostate cancer (mean follow-up, 3 years), 15% among 5-ARI users (N= 25,388) and 85% among AB users (N=149,507) (unadjusted mortality rate ratio, 0.80). After accounting for competing risks, it was found that 5-ARI use was not associated with prostate cancer mortality when compared with AB use (adjusted subdistribution hazard ratio, 0.85; 95% CI, 0.72–1.01). Similar results were observed in the case-control study (adjusted matched odds ratio, 0.95; 95% CI, 0.78–1.17).
Conclusion:
Among men being pharmaceutically treated for benign prostatic hyperplasia, 5-ARI use was not associated with an increased risk of prostate cancer–specific mortality when compared with AB use. The increased prevalence of high-grade lesions at the time of diagnosis noted in our study and the chemoprevention trials may not result in increased prostate cancer mortality.
The use of 5-alpha reductase inhibitors (5-ARIs) along with alpha-adrenergic blockers (alpha-blockers [ABs]) to manage lower urinary tract symptoms (LUTSs) associated with benign prostatic hyperplasia (BPH). Because 5-ARIs inhibit the conversion of testosterone to dihydrotestosterone, which reduces prostate size, they have been suggested as potential chemopreventive agents for prostate cancer. To assess this, 2 randomized trials were conducted, the Prostate Cancer Prevention Trial (PCPT)1 and Reduction by Dutasteride of Prostate Cancer Events,2 which found a reduced risk of prostate cancer of 23% to 25% when compared with placebo. There was, however, an increase in high-grade (Gleason 7–10) cancers among the 5-ARI group in the PCPT and a significant increase in years 3 and 4 of the Reduction by Dutasteride of Prostate Cancer Events trial.1,2 Because it was not possible to determine whether the association was a result of bias or an effect of the drugs, uncertainty continues to exist regarding the use of 5-ARIs for the chemoprevention of prostate cancer.3–5
Before the safety of using 5-ARIs for the primary prevention of prostate cancer can be established, studies assessing their long-term use in community practice settings are necessary.6,7 However, the studies to date were limited by the use of total mortality rather than prostate cancer—specific mortality and the rarity of this outcome, limited information on dose and duration, and importantly, the use of highly-selected populations that have limited generalizability to clinical practice.8–10 Therefore, the goal of this study was to specifically assess the risk of prostate cancer—specific mortality associated with 5-ARIs use as compared with AB use for the treatment of BPH/LUTSs in a large population-based cohort of 214,272 men in community practice settings over a 19-year observation period.
METHODS
Study Population
This study was conducted in 4 regions of Kaiser Permanente (Northern and Southern California, Colorado, and Northwest) that are each integrated health care systems that collectively provide comprehensive health care services to more than 8.5 million members. The racial and socioeconomic diversity of the members closely reflects that of the regions each serves.11 Comprehensive electronic health records along with the standardized Virtual Data Warehouse12 allow for the collection of demographic characteristics, health services utilization, disease diagnoses, and death records and includes pharmacy data and inpatient, outpatient, and emergency department encounters, as well as care received outside of the system. This study was approved by the institutional review boards at each site.
A cohort of male members aged 50 years and older who received a new prescription for a BPH/LUTS medication between January 1, 1992, and December 31, 2007, and were members for at least 1 year before first dispensing were eligible for inclusion (N=281,034). Men with a diagnosis of prostate cancer before or within 3 months of their first prescription (n=20,170), those with less than 90 days of consecutive medication filled (n=44,516), and those treated with finasteride 1 mg for alopecia (n=2023) and men with incomplete data (N=82) were excluded, leaving 214,272 men eligible for matching (Figure).
FIGURE.
Analytic samples for cohort and nested case-control studies. Note: Exclusion steps are not mutually exclusive, which resulted in men meeting multiple exclusion criteria. BPH = benign prostatic hyperplasia; CO = Colorado; LUTS = lower urinary tract symptom; NC = Northern California; NW = Northwest; SC = Southern California.
Exposure Assessment
The primary exposure of interest was a new prescription for a 5-ARI or an AB from 1992 to 2007 identified via prescription fills in electronic pharmacy records. Because of variability in the availability of pharmacy records in each region, the start dates were 1992 (Southern California, Northwest, and Colorado) and 1995 (Northern California). Men with exposure to either medication class before baseline were excluded. Combination users (men who were exposed to both an AB and 5-ARI) were defined as 5-ARI users for matching purposes in both studies. In the cohort study, combination users contributed person-time to both exposure groups. However, in the case-control study, men were defined as 5-ARI users if they ever used a 5-ARI. Cumulative exposure and dose were calculated from 5-ARI initiation in men who were 5-ARI users and corresponding matched index date among AB users in the cohort study and from 5-ARI initiation among 5-ARI users and AB initiation among AB users in the nested case-control study.
Outcome Assessment
The primary outcome was prostate cancer−specific mortality, which was identified via a comprehensive search across multiple sources for vital status. This included local (regional) Kaiser Permanente death databases, cancer registries, state death records, a Social Security Death Index Match, and National Death Index Match. Prostate cancer death was then defined on the basis of coded underlying or primary cause of death.
Matching
Cohort Study.
Men who initiated a 5-ARI were matched using risk-set sampling 1:6 to alpha-blocker users on age at matching (±2 years), race (African American vs Other), timing of BPH medication initiation (within 2 years), history of AB use, and health plan region. Of the 214,272 eligible men, 73% were successfully matched, resulting in an analytic sample of 157,456 men with 174,895 records (18,321 men were matched as both a 5-ARI and AB user). Men were then passively followed via electronic health records through the end of 2010 for death due to prostate cancer (N=1053).
Nested Case-Control Study.
The complicated matching approach led to 27% of men in the cohort study remaining unmatched, including 631 prostate cancer deaths. To include as many of the prostate cancer deaths as possible, a nested case-control study was also performed. In this study, 1671 of the 1684 (99.2%) men who died of prostate cancer in the underlying cohort were defined as case patients and matched 1:10 on age at prostate cancer death (±2 years), race (African American vs Other), timing of BPH medication initiation (within 2 years), and region to cohort members who did not die of prostate cancer and were alive at the time of the cases’ death using risk-set sampling (such that cases were eligible to be previously matched as a control) (N=18,311).
Covariates
Covariates were defined at the time of matching for both studies (cohort entry for cohort and time of death for case-control). Demographic characteristics such as age (at matching, death, and prostate cancer diagnosis), race/ethnicity (non-Hispanic white, African American, Asian, Hawaiian/Pacific Islander/Native Alaskan/American Indian, and Unknown) were collected from electronic demographic files. Aggregate median household income and education were calculated via geocoding using 2000 US census estimates at the block, block-group, tract, and ZIP level. Medical history was identified in the Virtual Data Warehouse and included a modified Charlson comorbidity score,13 history of BPH, cardiovascular disease, high blood pressure, hyperlipidemia, and use of medications to treat erectile dysfunction or overactive bladder in the year before matching. Characteristics related to prostate cancer care, including prostate specific antigen (PSA) levels (ng/mL), prostate biopsies, and bone scans for metastatic disease, were collected throughout the study period. Pathology-confirmed diagnoses of prostate cancer, history of other cancers, primary prostate cancer treatment modality, Gleason score, and American Joint Committee on Cancer stage at diagnosis were derived from cancer registries in each region.
Statistical Analyses
Cohort Study.
The distribution of demographic, medical history, and prostate cancer characteristics (among those diagnosed) was compared across exposure status, using chi-square tests and 2-sided Wilcoxon-Mann-Whitney tests where appropriate. Person-time at risk was calculated from the time of matching (5-ARI medication initiation or index date in AB user) to death due to prostate cancer (event), death due to other causes, loss to follow-up (disenrollment from health plan), or end of study period. To account for competing risk of death due to causes other than prostate cancer, subdistribution proportional hazards regression14 was used to estimate the incidence of prostate cancer death comparing 5-ARI use with or without AB use to AB use, after adjustment for age, race, region, medication initiation year, history of AB use, Charlson score, history of cardiovascular disease, high blood pressure, hyperlipidemia, diabetes, and other cancer, and use of other medications to treat overactive bladder or erectile dysfunction. Adjusted Cox models were also fit to estimate the relative risk of prostate cancer death among 5-ARI users as compared with AB users. Models were then stratified by increasing cumulative exposure, dose, and history of AB-use quartiles. The proportional hazard assumption was tested using Shoenfeld residuals.15 All analyses were performed using SAS 9.3 and R statistical software packages, and a P value of less than .05 was considered statistically significant.
Nested Case-Control Study.
The distribution of demographic, medical history, and prostate cancer characteristics (among those diagnosed) was compared across case and control status, using chi-square tests and 2-sided Wilcoxon-Mann-Whitney tests where appropriate. Matched odds ratios and 95% CIs were estimated using conditional logistic regression, adjusting for the same covariates as the cohort study except history of AB use, and stratified by cumulative dose and exposure quartiles.
Sensitivity Analyses.
We conducted a number of sensitivity analyses to evaluate the robustness of our findings. Because of secular trends in missingness of PSA and Gleason score, we restricted our study period to 1999 and later when more complete data were available. We also stratified by calendar time and adjusted for calendar time in the adjusted models. We then conducted an analysis stratified by exposure lag time (defined as time from medication initiation to prostate cancer death). All the sensitivity results yielded comparable results (see Supplemental Material available online at http://www.mayoclinicproceedings.org).
RESULTS
Cohort Study
The mean age after matching did not differ significantly across exposure groups (P=.16). Men who used 5-ARIs were slightly more likely to be non-Hispanic white compared with AB users (78.3 vs 76.5%) (P<.001). Median serum PSA level at 5-ARI initiation (or index) was higher among 5-ARI users (4.1 ng/mL) than among AB users (2.1 ng/mL) (P<.001). Duration of AB use before matching was slightly lower in the 5-ARI group than in the AB group (3.2 vs 3.4 years) (P<.001) (Table 1). The median dose and duration were greater among 5-ARI users than among AB users (both P<.001). The median follow-up time after matching was also greater among 5-ARI users (3.3 vs 2.3 years) (P<.001).
TABLE 1.
Demographic and Clinical Characteristics in the Matched Cohort and Case-Control Studies by Exposure and Outcome Status (N=174,895)a
| Cohort (N=174,895) |
Case-control (N=18,311) |
|||
|---|---|---|---|---|
| Characteristics at matching | 5-ARI users (n=25,388) | AB users (n=149,507) | Prostate cancer deaths (n=1671) | No prostate cancer death (n=16,640) |
| Exposure | ||||
| α-Blocker | – | – | 1506 (90.1) | 14,894 (89.5) |
| 5-ARI | – | – | 165 (9.9) | 1746 (10.5) |
| Age (y), mean ± SD, median | 72.4±9.3, 72.6 | 72.3±9.2, 72.5 | 81.5±8.1, 82.4 | 81.4±8.0, 82.3 |
| Age (y) (%) | ||||
| <60 | 2546 (10.0) | 15,338 (10.3) | 15 (0.9) | 152 (0.9) |
| 60–69 | 7639 (30.1) | 45,341 (30.3) | 144 (8.6) | 1413 (8.5) |
| 70+ | 15,203 (60.0) | 88,828 (59.4) | 1512 (90.5) | 15,075 (90.6) |
| Race (%)b | ||||
| Non-Hispanic white | 19,889 (78.3) | 114,331 (76.5) | 1269 (75.9) | 12,639 (76.0) |
| African American | 1885 (7.4) | 10,826 (7.2) | 205 (12.3) | 1991 (12.0) |
| Asian | 2017 (7.9) | 12,619 (8.4) | 60 (3.6) | 1004 (6.0) |
| HP, AI, MU, and UN | 1597 (6.3) | 11,731 (7.9) | 137 (8.2) | 1006 (6.1) |
| PSAb | ||||
| Missing (%) | 6359 (25.1) | 58,903 (39.4) | 364 (21.8) | 10,067 (60.5) |
| PSA level, mean ± SD, median | 6.6±30.2, 4.1 | 4.5±44.8, 2.1 | 297.0±746.0, 44.0 | 15.5±141.1, 3.1 |
| Median household income ($1000), mean ± SD, median | 68.3±29.7, 63.5c | 66.1±28.0, 61.8c | 63.8±28.3, 58.4 | 64.5±28.7, 59.6 |
| Geocoded education | ||||
| <9th grade | 0.07 (0.09), 0.04c | 0.07 (0.09), 0.04c | 0.08 (0.10), 0.04 | 0.08 (0.09), 0.04 |
| 9th-12th grade | 0.09 (0.07), 0.08c | 0.10 (0.0.7), 0.08c | 0.11 (0.08), 0.09d | 0.10 (0.07), 0.09d |
| High school graduate | 0.20 (0.08), 0.20 | 0.21 (0.08), 0.21 | 0.21 (0.08), 0.21c | 0.21 (0.08), 0.21c |
| Some college, no degree | 0.24 (0.07), 0.24 | 0.24 (0.07), 0.25 | 0.24 (0.07), 0.25c | 0.24 (0.07), 0.25c |
| Associate degree | 0.08 (0.03), 0.07 | 0.08 (0.03), 0.08 | 0.07 (0.03), 0.07d | 0.07 (0.03), 0.07d |
| Bachelor degree | 0.20 (0.11), 0.20d | 0.19 (0.11), 0.18d | 0.18 (0.10), 0.17c | 0.19 (0.11), 0.18c |
| Graduate or professional degree | 0.12 (0.10), 0.09d | 0.11 (0.09), 0.08d | 0.10 (0.09), 0.08c | 0.11 (0.09), 0.08c |
| Alpha-blocker history (y), mean ± SD, medianb | 4.2±4.0, 3.2 | 4.4±4.1, 3.4 | 5.0±4.1, 4.1 | 5.7±4.3, 5.0 |
| Charlson comorbidity index (%)b | ||||
| 0 | 11,633 (45.8) | 64,176 (42.9) | 159 (9.5) | 6560 (39.4) |
| 1 | 5017 (19.8) | 29,653 (19.8) | 33 (2.0) | 2956 (17.8) |
| 2+ | 8738 (34.4) | 55,678 (37.2) | 1,479 (88.5) | 7124 (42.8) |
| History of cardiovascular disease (yes) | 8992 (35.4%)c | 50,467 (33.8%)c | 915 (54.8) | 7113 (42.8) |
| History of high blood pressure (yes) | 21,289 (83.9%)c | 136,264 (91.1%)c | 1462 (87.5)d | 14,206 (85.4)d |
| History of hyperlipidemia (yes)b | 16,753 (66.0%) | 101,221 (67.7%) | 677 (40.5) | 9312 (56.0) |
| History of diabetes (yes) | 5554 (21.9%)c | 39,719 (26.6%)c | 432 (25.9)d | 3910 (23.5)d |
| History of cancer (other than prostate) (yes)b | 2473 (9.7%) | 13,404 (9.0%) | 309 (18.5) | 1999 (12.0) |
| Use of other medications to treat ED or OAB (yes)b | 3679 (14.5%) | 19,505 (13.1%) | 52 (3.1) | 1295 (7.8) |
AB=alpha-adrenergic blocker; ED=erectile dysfunction; 5-ARI=5-alpha reductase inhibitor; HP=Hawaiian or Pacific Islander; AI=American Indian; MU=multiple; OAB=overactive bladder; PSA=prostate specific antigen; UN=unknown.
P<.001 for both studies.
P<.001.
P<.05.
The unadjusted prostate cancer incidence rate among 5-ARI users was lower than among AB users (4.1 vs 9.3 per 1000 per year; P=.06). A greater proportion of men 70 years or older at diagnosis were in the 5-ARI group (70.0%) compared with the AB group (58%) (P<.001). Among men with a prostate cancer diagnosis and Gleason score (52% in cohort study), a greater proportion of men exposed to 5-ARI (12.0%) were diagnosed with a Gleason 8–10 prostate cancer when compared with AB users (9.2%) (P=.005). Similarly, among men with stage at diagnosis available (75%), 14.2% of 5-ARI users were diagnosed with stage III/IV disease compared with 9.9% of AB users (P<.001). The unadjusted prostate cancer–specific mortality rate was slightly lower among 5-ARI users than among AB users (1.6 vs 2.0 per 1000 per year; P=.85) (Table 2).
TABLE 2.
Characteristics During Follow-Up in Matched Cohort (N=174,895) and Case-Control Studies (N=18,311)a
| Cohort (N=174,895) |
Nested case-control (N=18,311) |
|||
|---|---|---|---|---|
| Characteristics | 5-ARI users (n=25,388) | AB users (n=149,507) | Prostate cancer deaths (n=1671) | No prostate cancer death (n=16,640) |
| Exposure and follow-up | ||||
| Cumulative exposure time (y), mean ± SD, medianb | 2.2±2.1, 1.5 | 1.4±1.4, 1.1 | 2.3±2.6, 1.3 | 3.1±3.2, 2.0 |
| Cumulative dose (g)b | 4.1 (4.0), 2.9 | 2.3 (3.8), 1.0 | 4.4 (7.3), 1.8 | 5.6 (9.2), 2.1 |
| Duration of follow-up time (y),b mean ± SD, median | 4.0±3.1, 3.3 | 3.0±2.6, 2.3 | 5.4±3.8, 4.5 | 6.1±4.1, 5.3 |
| Number of PSA tests, mean ± SD, medianb | 3.1±4.2, 2.0 | 2.0±3.3, 1.0 | 10.5±11.2, 7.0 | 4.8±5.0, 3.0 |
| Patients with ≥1 biopsy (%)b | 3190 (12.6) | 11,468 (7.7) | 485 (29.0) | 1,447 (8.7) |
| Prostate cancer characteristicsc | ||||
| Prostate cancer diagnosed (%)b | 837 (3.3) | 4,359 (2.9) | 1,035 (61.9) | 1,427 (8.6) |
| Age at diagnosis (y) | ||||
| <60 | 30 (3.6)d | 274 (6.3)d | 34 (3.3)e | 24 (1.7)e |
| 60–69 | 221 (26.4) | 1,573 (36.1) | 127 (12.3) | 192 (13.5) |
| 70+ | 586 (70.0) | 2,512 (57.6) | 874 (84.4) | 1,211 (84.9) |
| Gleason score | ||||
| Missing | 365 (43.6)e | 1,948 (44.7)e | 728 (70.3)e | 954 (66.9)e |
| ≤6 | 240 (28.7) | 1,194 (27.4) | 88 (8.5) | 206 (14.4) |
| 7 | 132 (15.8) | 818 (18.8) | 51 (4.9) | 163 (11.4) |
| 8–10 | 100 (12.0) | 399 (9.2) | 168 (16.2) | 104 (7.3) |
| Stage at diagnosisb | ||||
| Missing | 171 (20.4) | 1,002 (23.0) | 362 (35.0) | 332 (23.3) |
| I/II | 547 (65.4) | 2,925 (67.1) | 293 (28.3) | 955 (67.0) |
| III/IV | 119 (14.2) | 432 (9.9) | 380 (36.7) | 140 (9.7) |
| Bone scan at diagnosis (%)b | 261 (31.2) | 994 (22.8) | 397 (38.4) | 256 (17.9) |
| Other tests for metastatic disease at diagnosis (%) | 518 (61.9) | 2,756 (63.2) | 592 (57.2)e | 679 (47.6)e |
| Prostate cancer treatment within 6 mo of diagnosis | ||||
| Surgery | 294 (35.1)e | 1637 (37.6)e | 189 (18.3)e | 339 (23.8)e |
| Radiation | 115 (13.7) | 720 (16.5) | 136 (13.2) | 224 (15.7) |
| No treatment | 427 (51.0) | 2,002 (45.9) | 709 (68.6) | 863 (60.5) |
| Prostate cancer incidence rate (per 1,000 person-years)f | 4.1 | 9.3 | – | – |
| Prostate cancer death (%) | 158 (0.6) | 895 (0.6) | ||
| Prostate cancer mortality rate (per 1,000 person-years)g | 1.6 | 2.0 | – | – |
AB = alpha-adrenergic blocker; 5-ARI = 5-alpha reductase inhibitor; PSA = prostate specific antigen.
P<.001 for both studies.
In cohort study, those with prostate cancer diagnosed before matching were excluded.
P<.001.
P<.05.
Person-time started at matching and ended at the end of the study or at prostate cancer diagnosis.
Person-time for men who were lost to follow-up and then died later during the study period was calculated up to the date of loss to follow-up.
In the cohort study, 5-ARI use was not associated with an increased risk of prostate cancer−specific mortality when compared with AB use, after adjustment for age, medication initiation year, race, region, AB history, Charlson score, and other comorbidities, and accounting for competing risk of death due to other causes (adjusted subdistribution hazard ratio [SHR], 0.85; 95% CI, 0.72–1.01). The adjusted Cox models yielded similar results (adjusted hazard ratio, 0.72; 95% CI, 0.61–0.85). When stratified by cumulative exposure and dose, there was a trend toward a decreased risk of prostate cancer mortality with increasing cumulative exposure and dose quartiles. Among men exposed for more than 2 years, 5-ARI use was associated with a 34% reduction in the risk of prostate cancer mortality compared with AB users, after accounting for competing risks (adjusted SHR, 0.66; 95% CI, 0.48–0.92). Similarly, men in the highest dose quartile were also less likely to die of prostate cancer if they were exposed to a 5-ARI as compared with an AB (adjusted SHR, 0.64; 95% CI, 0.47–0.86) (Table 3).
TABLE 3.
Multivariable-Adjusted Prostate CancereRelated Mortality Rates, Subdistribution Hazard Ratios, and Matched Odds Ratios Stratified by Duration of Cumulative Exposure, Dose, and History of AB Usea,b
| Number of prostate cancer deaths |
Mortality rate (per 1,000 person-years) |
Adjusted subdistribution hazard ratio (95% CI)c | |||
|---|---|---|---|---|---|
| Cohort study | 5-ARI | α-Blocker | 5-ARI | AB | |
| Overall (n=174,895) | 158 | 895 | 0.12 | 0.15 | 0.85 (0.72–1.01) |
| Cumulative exposured | |||||
| <6 mo (n=45,226) | 37 | 242 | 0.15 | 0.16 | 1.07 (0.75–1.51) |
| 6 mo-1 y (n=35,812) | 41 | 198 | 0.01 | 0.02 | 1.14 (0.82–1.61) |
| 1–2 y (n=45,498) | 36 | 249 | 0.04 | 0.03 | 0.93 (0.65–1.33) |
| 2+ y (n=48,359) | 44 | 206 | 0.009 | 0.02 | 0.66 (0.48–0.92) |
| Cumulative dose (g) | |||||
| Q1 (n=43,705) | 4 | 280 | 0.08 | 0.02 | 1.62 (0.59–4.44) |
| Q2 (n=42,735) | 47 | 230 | 0.12 | 0.06 | 1.16 (0.84–1.60) |
| Q3 (n=44,702) | 51 | 207 | 0.002 | 0.0007 | 1.27 (0.93–1.74) |
| Q4 (n=43,753) | 56 | 178 | 0.001 | 0.005 | 0.64 (0.47–0.86) |
| AB history at matching (y) | |||||
| First quartile (0–0.6, n=43,652) | 80 | 278 | 0.03 | 0.03 | 1.32 (1.02–1.70) |
| Second quartile (0.6–3.4, n=43,825) | 42 | 208 | 0.01 | 0.005 | 1.11 (0.79–1.55) |
| Third quartile (3.4–7.1, n=43,697) | 27 | 213 | 0.0007 | 0.02 | 0.60 (0.40–0.89) |
| Fourth quartile (≥7.1, n=43,721) | 9 | 196 | 8.4 × 10−6 | 0.003 | 0.23 (0.12–0.44) |
| Cases, N (%) (N=1671) |
Controls, N (%) (N=16,640) |
Adjusted MOR (95% CI)e | |||
| Nested case-control study | 5-ARI | AB | 5-ARI | AB | |
| Overall | 165 (9.9) | 1506 (90.1) | 1746 (10.5) | 14,894 (89.5) | 0.95 (0.78–1.17) |
| Cumulative exposured | |||||
| <6 mo (n=3,767) | 46 (11.2) | 366 (88.8) | 370 (11.0) | 2985 (89.0) | 1.03 (0.61–1.73) |
| 6 mo-1 y (n=2,602) | 46 (15.0) | 260 (85.0) | 339 (14.8) | 1957 (85.2) | 0.86 (0.46–1.60) |
| 1–2 y (n=3,013) | 38 (11.4) | 296 (88.6) | 390 (14.6) | 2289 (85.4) | 0.46 (0.21–1.00) |
| 2+ y (n=8,929) | 35 (5.7) | 584 (94.4) | 647 (7.8) | 7663 (92.2) | 0.78 (0.50–1.22) |
| Cumulative dose (g) | |||||
| Q1 (n=4,645) | 22 (4.5) | 469 (95.5) | 213 (5.1) | 3941 (94.9) | 0.71 (0.34–1.49) |
| Q2 (n=4,378) | 58 (12.8) | 397 (87.3) | 455 (11.6) | 3468 (88.4) | 0.94 (0.55–1.60) |
| Q3 (n=4,736) | 65 (15.3) | 361 (84.7) | 661 (15.3) | 3649 (84.7) | 0.94 (0.57–1.56) |
| Q4 (n=4,552) | 20 (6.7) | 279 (93.3) | 417 (9.8) | 3836 (90.2) | 0.62 (0.30–1.29) |
AB = alpha-adrenergic blocker; BPH = benign prostatic hyperplasia; 5-ARI = 5-alpha reductase inhibitor; MOR = matched odds ratio.
Results are from negative binomial regression, with the exception of the subdistributional hazard ratios, which were estimated using competing risk proportional hazard regression, and hazard ratios, which were estimated using proportional hazard regression.
Adjusted for age, BPH initiation year, race, region, AB history, Charlson score, and comorbidities.
Cumulative exposure was calculated from time of matching (5-ARI initiation and matched index date in AB users) in cohort study and from 5-ARI initiation among 5-ARI users and AB initiation in AB users in case-control study.
Adjusted for age, BPH initiation year, race, region, Charlson score, and comorbidities.
Nested Case-Control Study
Mean serum PSA levels were markedly higher among men who died of prostate cancer (44.0) than among men who did not die of prostate cancer (3.1) (P<.001). History of years of AB use was lower among men who died of prostate cancer than among those who did not (4.1 vs 5.0 years) (P<.001) (Table 1). Men who died of prostate cancer had shorter mean follow-up times than did men who did not die of prostate cancer (5.4 vs 6.1 years) (P<.001) (Table 2).
Similar to the cohort study results, 5-ARI use was not associated with an increased risk of prostate cancer−specific mortality when compared with AB use, after adjustment in the case-control study (adjusted matched odds ratio, 0.95; 95% CI, 0.78– 1.17). The case-control study yielded results similar to those of the cohort study, which suggested a decreased risk with increasing dose and duration, but did not reach statistical significance (Table 3).
DISCUSSION
The results of this large, population-based study in community practice settings suggest that the use of 5-ARIs to treat BPH/LUTSs is not associated with an increased risk of prostate cancer mortality when compared with AB use. Although there was an increase in Gleason 8–10 prostate cancers diagnosed among men treated with 5-ARIs (among those for whom we had Gleason scores) in this population, this did not translate into an increased risk of prostate cancer death.
Our findings are in line with those reported by Azoulay et al8 in a study of men with prostate cancer in the United Kingdom.8 Similarly, Preston et al9 found that 5-ARI use was not associated with lethal prostate cancer in the Health Professional’s Follow-up Study, albeit with limited statistical power. Follow-up results from the PCPT suggested that men treated with finasteride were not more likely to die of all causes compared with men randomized to placebo, and the Finnish Prostate Cancer Screening Trial found no increase in the risk of prostate cancer or all-cause mortality associated with 5-ARI use in men diagnosed with prostate cancer.10,16 However, these results are difficult to interpret because prostate cancer is an infrequent cause of death in the overall population, the generalizability of the results is limited by the highly-selected trial populations,10,16 and the populations had low exposure rates.8,9,16
In contrast to these studies that assessed the association between 5-ARI use and prostate cancer mortality,8,9,16 the present study specifically addresses the safety of use of 5-ARIs relative to a comparator class of drugs (AB), in men treated for BPH/LUTSs in real-world community practice settings. In addition, our findings support the suggestion that the increase in Gleason 7–10 prostate cancers diagnosed among 5-ARI users in the chemoprevention trials1,2 may be related more to bias than to biology as previously suggested.17–19 The increase in high-grade tumors seen in these trials and other observational studies, including ours, may result from a detection bias,18–21 through the effect of 5-ARIs on improving the sensitivity of prostate-specific antigen testing22 and decreasing prostate volumes,23 resulting in the increased detection of higher-grade tumors.
Also, among men who were diagnosed with prostate cancer in our sample, those who used 5-ARIs were more likely to be diagnosed with stage III/IV disease compared with men who used ABs. This stage difference did not translate into an increased risk of prostate cancer death for 5-ARI users during the study period and may be more related to detection and surveillance patterns that are different across exposure groups. Men who used 5-ARIs were older and were more likely to have bone scans at diagnosis compared with men who used ABs. Men with larger prostate glands, more severe symptoms, and higher PSA levels are also more often treated with 5-ARIs in clinical practice and thus followed more closely.18,19,22 The 5-ARIs also reduce PSA levels, which may have delayed diagnosis via PSA screening in these men.18–21 These detection and surveillance biases may therefore result in differences in the proportion of metastatic disease at diagnosis across exposure groups in this study.
Although this study used a large, diverse population in community practice settings, included 19 years of observational data and a large number of men who were exposed to 5-ARIs, searched multiple sources for vital status to identify prostate cancer deaths, and used both a cohort and case-control design and multiple sensitivity analyses to confirm the robustness of our findings, there are some potential limitations to consider. Missing data in the earlier years of this study, particularly for PSA and Gleason score, limits our ability to conduct stratified analyses. However, the sensitivity analysis restricting our study to dates when more complete data were available found similar results. In addition, adjustment in the models for these variables may not have been appropriate because they are potentially in the causal pathway. The number of prostate cancer diagnoses in our study may be underestimated because only 62% of men who died of prostate cancer had a pathology-confirmed diagnosis in the cancer registries. However, most of those who died of prostate cancer had an International Classification of Diseases, Ninth Revision diagnosis code in their records. This most likely reflects prevalent disease, which was diagnosed before health plan enrollment and, therefore, by definition, not captured by the cancer registries. Also, some early cases may have been clinical diagnoses and, therefore, excluded from the registries as well. As a result, there are a small proportion of men in our sample (<2%) who initiated their medications after their prevalent or clinical prostate cancer diagnoses. Although cohort studies comparing 2 treatment modalities are subject to immortal time bias, we addressed this bias by matching on exposure in the cohort study so that the follow-up time in both treatment groups began at the same time point (5-ARI initiation or index date) rather than date of diagnosis or first dispense in controls, using risk-set sampling of AB users in the cohort study and performing a nested case-control analysis of the cohort.24 The average follow-up time after medication initiation in this population and cumulative duration of use was short, mostly due to the advanced age of the cohort and 5-ARIs not becoming widely used until later in the study period. However, our study included 543,523 person-years of follow-up, more than in previous studies.8,9
CONCLUSION
Treatment with 5-ARIs was not associated with an increased risk of prostate cancer mortality when compared with treatment with AB for BPH/LUTSs. Thus, the increased prevalence of high-grade lesions at the time of prostate cancer diagnosis noted in this study and the chemoprevention trials of 5-ARIs may not result in increased prostate cancer−specific death.
Supplementary Material
Grant Support:
The work was supported by a research grant (no. 116059) from GlaxoSmithKline. Although the sponsor had an opportunity to comment on the manuscript, it did not have access to the data and all analyses were conducted at Kaiser Permanente Southern California. Drs Wallner and Jacobsen had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Dr Wallner made the final decision to submit this manuscript for publication and was not paid by any other organization or agency to write this article.
Abbreviations and Acronyms:
- AB
alpha-adrenergic blocker
- BPH
benign prostatic hyperplasia
- 5-ARI
5-alpha reductase inhibitor
- LUTS
lower urinary tract symptom
- PCPT
Prostate Cancer Prevention Trial
- PSA
prostate specific antigen
- SHR
subdistribution hazard ratio
Footnotes
SUPPLEMENTAL ONLINE MATERIAL
Supplemental material can be found online at http://www.mayoclinicproceedings.org. Supplemental material attached to journal articles has not been edited, and the authors take responsibility for the accuracy of all data.
Data Previously Presented: Data was previously presented at the American Urological Association Annual Meeting on May 6, 2016 in San Diego, CA.
Potential Competing Interests: Drs Wallner, DiBello, Van Den Eeden, Abell, Jacobsen, D’Agsotino Jr, Loo, and Horwitz report funding from GlaxoSmithKline for this study. Dr Van Den Eeden report funding from Takeda not related to the topic of this study. Drs Li, Weinmann, Ritzwoller, Aaronson, and Richert-Boe have no additional financial disclosures to report.
Contributor Information
Lauren P. Wallner, Department of Medicine, University of Michigan, Ann Arbor, MI; Department of Research and Evaluation, Kaiser Permanente Southern California, Pasadena, CA.
Julia R. DiBello, Worldwide Epidemiology, GlaxoSmithKline, Collegeville, PA.
Bonnie H. Li, Department of Research and Evaluation, Kaiser Permanente Southern California, Pasadena, CA.
Stephen K. Van Den Eeden, Division of Research, Kaiser Permanente Northern California, Oakland, CA.
Sheila Weinmann, Center for Health Research, Kaiser Permanente Northwest, Portland, ORG.
Debra P. Ritzwoller, Institute for Health Research, Kaiser Permanente Colorado, Denver, CO.
Jill E. Abell, Real World Evidence, Janssen, Philadelphia, PA.
Ralph D’Agostino, Jr, Department of Biostatistical Science, Wake Forest University, Winston-Salem, NC.
Ronald K. Loo, Department of Research and Evaluation, Kaiser Permanente Southern California, Pasadena, CA.
David S. Aaronson, Division of Research, Kaiser Permanente Northern California, Oakland, CA.
Kathryn Richert-Boe, Center for Health Research, Kaiser Permanente Northwest, Portland, ORG.
Ralph I. Horwitz, Temple University, Philadelphia, PA; Institute of Medicine, Washington, DC.
Steven J. Jacobsen, Department of Research and Evaluation, Kaiser Permanente Southern California, Pasadena, CA.
REFERENCES
- 1.Thompson IM, Goodman PJ, Tangen CM, et al. The influence of finasteride on the development of prostate cancer. N Engl J Med. 2003;349(3):215–224. [DOI] [PubMed] [Google Scholar]
- 2.Andriole GL, Bostwick DG, Brawley OW, et al. ; REDUCE Study Group. Effect of dutasteride on the risk of prostate cancer. N Engl J Med. 2010;362(13):1192–1202. [DOI] [PubMed] [Google Scholar]
- 3.Rubin MA, Kantoff PW. Prevention of prostate cancer with finasteride. N Engl J Med. 2003;349(16):1569–1572. [DOI] [PubMed] [Google Scholar]
- 4.Scardino PT. The prevention of prostate cancer — the dilemma continues. N Engl J Med. 2003;349(3):297–299. [DOI] [PubMed] [Google Scholar]
- 5.Zuger A A big study yields big questions. N Engl J Med. 2003; 349(3):213–214. [DOI] [PubMed] [Google Scholar]
- 6.Figg WD, Thompson IM. Effect of 5α-reductase inhibitor use on mortality from prostate cancer. JAMA Oncol. 2015;1(3):321–322. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.LeFevre M A role for finasteride in the prevention of prostate cancer? N Engl J Med. 2013;369(7):670–671. [DOI] [PubMed] [Google Scholar]
- 8.Azoulay L, Eberg M, Benayoun S, Pollak M. 5A-reductase inhibitors and the risk of cancer-related mortality in men with prostate cancer. JAMA Oncol. 2015;1(3):314–320. [DOI] [PubMed] [Google Scholar]
- 9.Preston MA, Wilson KM, Markt SC, et al. 5alpha-Reductase inhibitors and risk of high-grade or lethal prostate cancer. JAMA Intern Med. 2014;174(8):1301–1307. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Thompson IM Jr, Goodman PJ, Tangen CM, et al. Long-term survival of participants in the prostate cancer prevention trial. N Engl J Med. 2013;369(7):603–610. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Koebnick C, Langer-Gould AM, Gould MK, et al. Sociodemographic characteristics of members of a large, integrated health care system: comparison with US Census Bureau data. Perm J. 2012;16(3):37–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Ross TR, Ng D, Brown JS, et al. The HMO Research Network Virtual Data Warehouse: a public data model to support collaboration. EGEMS (Wash DC). 2014;2(1):1049. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40(5):373–383. [DOI] [PubMed] [Google Scholar]
- 14.Fine JP, Gray RJ. A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc. 1999;94(446):496–509. [Google Scholar]
- 15.Grambsch PM, Therneau TM. Proportional hazards tests and diagnostics based on weighted residuals. Biometrika. 1994; 81(3):515–526. [Google Scholar]
- 16.Murtola TJ, Karppa EK, Taari K, Talala K, Tammela TL, Auvinen A. 5-Alpha reductase inhibitor use and prostate cancer survival in the Finnish Prostate Cancer Screening Trial. Int J Cancer. 2016;138(12):2820–2828. [DOI] [PubMed] [Google Scholar]
- 17.FDA Drug Safety Communication: 5-alpha reductase inhibitors (5-ARIs) may increase the risk of a more serious form of prostate cancer. http://www.fda.gov/Drugs/DrugSafety/ucm258314.htm. Accessed June 30, 2015.
- 18.Andriole GL, Humphrey PA, Serfling RJ, Grubb RL. High-grade prostate cancer in the Prostate Cancer Prevention Trial: fact or artifact? J Natl Cancer Inst. 2007;99(18): 1355–1356. [DOI] [PubMed] [Google Scholar]
- 19.Goodman PJ, Thompson IM Jr, Tangen CM, Crowley JJ, Ford LG, Coltman CA Jr. The prostate cancer prevention trial: design, biases and interpretation of study results. J Urol. 2006; 175(6):2234–2242. [DOI] [PubMed] [Google Scholar]
- 20.Redman MW, Tangen CM, Goodman PJ, Lucia MS, Coltman CA Jr, Thompson IM. Finasteride does not increase the risk of high-grade prostate cancer: a bias-adjusted modeling approach. Cancer Prev Res (Phila). 2008;1(3): 174–181. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Lucia MS, Darke AK, Goodman PJ, et al. Pathologic characteristics of cancers detected in the Prostate Cancer Prevention Trial: implications for prostate cancer detection and chemoprevention. Cancer Prev Res (Phila). 2008;1(3): 167–173. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Thompson IM, Chi C, Ankerst DP, et al. Effect of finasteride on the sensitivity of PSA for detecting prostate cancer. J Natl Cancer Inst. 2006;98(16):1128–1133. [DOI] [PubMed] [Google Scholar]
- 23.Cohen YC, Liu KS, Heyden NL, et al. Detection bias due to the effect of finasteride on prostate volume: a modeling approach for analysis of the Prostate Cancer Prevention Trial. J Natl Cancer Inst. 2007;99(18):1366–1374. [DOI] [PubMed] [Google Scholar]
- 24.Suissa S Immortal time bias in observational studies of drug effects. Pharmacoepidemiol Drug Saf. 2007;16(3):241–249. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.