Jacobsen SJ, Cheetham TC, Haque R, Shi JM, Loo RK. Association Between 5-α Reductase Inhibition and Risk of Hip Fracture. JAMA. 2008;300(14):1660-1664. doi:10.1001/jama.300.14.1660
Author Affiliations: Department of Research and Evaluation, Kaiser Permanente Southern California, Pasadena (Drs Jacobsen, Haque, and Shi); Pharmacy Analytic Services, Kaiser Permanente Southern California, Downey (Dr Cheetham); and Department of Urology, Bellflower Medical Center, Kaiser Permanente Southern California, Bellfower (Dr Loo).
Context For more than 15 years, 5-α reductase inhibitors, which block the conversion of testosterone to dihydrotestosterone, have been used in the treatment of benign prostatic hyperplasia (BPH).Short-term studies show no effects of these agents on bone metabolism,but long-term data are not available.
Objective To assess the association between use of 5-α reductase inhibitors (eg, finasteride) for BPH and occurrence of hip fracture.
Design, Setting, and Patients Population-based case-control study using data from Kaiser Permanente Southern California, a managed care organization with more than 3 million members. Case patients included 7076 men 45 years and older with incident hip fracture from 1997-2006. Control patients were 7076 men without incident hip fracture, optimally matched at a 1:1 ratio to case patients on age and medical center. Electronic information on pharmaceutical use was used to identify use of finasteride from 1991 forward.
Results Overall, 2547 (36%) and 2488 (35%) case and control patients, respectively, had a diagnosis of BPH (P = .30), and 109 (1.5%) and 141 (2.0%) of case and control patients, respectively, had been exposed to finasteride prior to the index date (matched odds ratio, 0.77; 95% confidence interval, 0.59-1.00; P = .04). There was no suggestion of a dose-response relationship between exposure to 5-α reductase inhibitors when the exposure was stratified into tertiles of total exposure (P = .12). By contrast, there was a slightly higher prevalence of α-blocker use in case vs control patients (32% vs 30%, respectively; P = .04).
Conclusions Exposure to 5-α reductase inhibitors was not associated with increased risk of hip fracture. The reduction in risk observed with exposure to 5-α reductase inhibitors and the modest increase in risk associated with exposure to α-blockers require replication and warrant further investigation.
Benign prostatic hyperplasia (BPH) is a common condition in aging men. It has been estimated that more than 8 million US men aged 50 through 79 years will meet current guidelines for discussing treatment options for BPH by 2010.1 Treatments for BPH include surgical procedures, minimally invasive procedures, and pharmacological agents.2 Most often, the first-line therapy is pharmacological, using either α-blockers or 5-α reductase inhibitors (eg, finasteride). The former work through adrenergic pathways in the bladder, prostate, or both, whereas the latter work through hormonal mechanisms. The use of these agents has been increasing, and annual spending on prescriptions for BPH therapy in 1996 to 1998 totaled nearly $200 million3; this is likely much greater now, with the availability of additional pharmacological agents.
The biological effects of 5-α reductase are fairly broad, since it is responsible for the conversion of testosterone to dihydrotestosterone. Dihydrotestosterone is the more powerfully active androgenic agent; with approximately 10 times the binding affinity for the androgen receptor, it is biologically active in pathways leading to secondary sex characteristics, BPH, acne, and male pattern baldness. Several adverse effects have been associated with the use of 5-α reductase inhibitors, including erectile dysfunction, decreased ejaculate volume, breast pain or tenderness, and gynecomastia.4 It is not clear, however, if these are the result of a decrease in dihydrotestosterone levels, a relative increase in testosterone levels, or a relative increase in estradiol levels due to shunting of testosterone through the aromatization pathway.5 Moreover, it is not clear how 5-α reductase inhibition affects long-term bone health, which has multiple androgenic and estrogenic steroid-dependent pathways.6,7
Previous work has suggested that inhibition of 5-α reductase may have little effect on bone metabolism, despite biological plausibility in either direction. The evidence in human cell-line studies is mixed. Testosterone appears to be related to bone mineral density (BMD) only in older men,8 which suggests that 5-α reductase inhibition may lead to increased bone density due to increases in testosterone levels. Simultaneous administration of testosterone and 5-α reductase inhibitors compared with testosterone alone does not appear to have any differential effect on BMD in men with low testosterone levels.9 A substudy of a large clinical trial for BPH showed no difference in BMD after 4 years of finasteride therapy,10 but this was examined in only 117 patients, and substantial dropout during the trial might have introduced bias. A small study of 71 men with BPH randomized to receive 5-α reductase inhibitors vs control found no difference in markers of bone formation, turnover, or BMD.11 Of interest, however, is that human osteoblast-like cells predominantly express 5-α reductase type 1,12 and dihydrotestosterone appears to have a modest regulatory role in the production of 1-α hydroxylase, which in turn increases levels of 1,25 dihydroxyvitamin D.13 Together, these observations suggest that dihydrotestosterone might have a role in bone metabolism, but no clear evidence exists to support this theory.
Some of this lack of clarity is owing to several limitations in the existing literature. Most of the studies have included a fairly limited number of individuals and therefore had limited power to detect differences. Most have included fairly short-term follow-up, and most have focused on BMD and not fracture. The latter outcome is of great public health import, given the morbidity and costs associated with fracture. If 5-α reductase inhibition has an effect on fracture risk, it could be important from a societal perspective. Thus, to help clarify the potential adverse role of 5-α reductase inhibitors in bone metabolism, we conducted a case-control study of hip fracture among men enrolled in a large managed care organization.
Kaiser Permanente Southern California (KPSC) is a large managed care organization that covers the region from Bakersfield to San Diego. In 2006, KPSC had a membership of more than 3.2 million, with a racial/ethnic composition similar to that of the source population. The majority of health care for members is delivered in 1 of 12 medical centers or in more than 100 affiliated outpatient facilities. A small fraction of emergent and specialty care is obtained from other institutions through contractual arrangements or through a claims reimbursement system. Health plan members are assigned a primary medical center based on geographic proximity. All health care encounters are tracked through electronic data systems, including detailed information on diagnoses applied and procedures performed during those encounters, regardless of setting. Most members have pharmacy benefits as well. All dispensed prescriptions are electronically tracked, with information on drug, dose, frequency, amount, and date dispensed.
Case patients included men identified from KPSC membership rolls from 1997 to 2006 who had a new diagnosis of hip fracture (International Classification of Diseases, Ninth Revision, Clinical Modification code 820.x) coded in the encounter information and who were 45 years or older at the time of fracture. Control patients were selected from the same population and who were optimally matched14 to case patients at a 1:1 ratio on age, membership in the health plan on the index (fracture) date, and medical center; because of the known association of age and race/ethnicity with risk of fracture, patients also were matched on these factors. Race/ethnicity was determined by an administrative code primarily assigned to members at the time of enrollment or during health care encounters.
The primary exposure measure was the dispensing of a 5-α reductase inhibitor (dutasteride or finasteride) from 1991 (the year the electronic pharmacy files were established) forward. Information was recorded on the start of the prescription (date), dose, total doses, amount dispensed, and the total dose dispensed. Prescriptions for α-blockers (terazosin, tamsulosin, doxazocin, alfuzosin,phentolamine, phenoxybenzamine, and prazosin) were also assessed as a second comparison exposure.
Information on additional variables was collected for consideration of these variables as potential confounders or effect modifiers. Among these were comorbid conditions, determined by the Deyo modification15 of the Charlson Comorbidity Index16 and based on diagnoses made during inpatient and outpatient encounters from 1981 until the index date. A diagnosis of BPH or surgical treatment of BPH also were considered potential confounders. The matching factor of race/ethnicity was considered a potential effect modifier.
The primary analysis was the comparison of the off-diagonal frequencies of exposure to 5-α reductase inhibitors for the matched case-control pairs. Differences were tested with the McNemar test. The association between 5-α reductase inhibition and hip fracture was estimated with a matched-pairs odds ratio (OR) and its respective 95% confidence interval (CI). Conditional logistic regression was used to estimate the OR, taking into account potential confounders. Potential effect modifiers were tested by introducing appropriate interaction terms to the conditional logistic regression models.
A series of secondary analyses were conducted as well. Additional analyses used α-blockers, a diagnosis of BPH, or surgical treatment of BPH as alternative exposures. The association between hip fracture and α-blocker use was explored further, separating selective vs nonselective α-blocker, recency of initiating use, and current vs past use. A dose-response relationship between cumulative drug exposure and hip fracture was evaluated by Cochran-Armitage test for trend; P = .05 was considered statistically significant. With type I and type II error rates fixed at .05 and .20, respectively, 7076 case patients, and an exposure prevalence of 0.20, the minimal detectable OR was 1.37. All analyses were performed by one investigator (J.M.S.) using SAS version 9.1.3 (SAS Institute Inc, Cary, North Carolina). The protocol for this study was reviewed and approved by the KPSC institutional review board.
Overall, 7076 men had a diagnosis of hip fracture over the 10-year period. The median age at fracture was 77 (interquartile range, 67-84)years (Table 1). Approximately 70% of men were white, with 11% identified as Hispanic and 7% as black. Compared with men without hip fracture, those with hip fracture were more likely to have multiple comorbid conditions (P < .001), despite being matched by age. A modest direct association was observed between comorbidity and use of 5-α reductase inhibitors (Spearman correlation coefficient, 0.057; P < .001).
Use of 5-α reductase inhibitors was fairly low in this population. During this period (1991 to 2006), finasteride was the only 5-α reductase inhibitor dispensed to study patients, and only 109 case patients (1.5%) and 141 control patients (2%) had a history of any exposure to these compounds (Table 2). The matched OR was 0.77 (95% CI, 0.59-1.00; P = .04). There was no suggestion of a dose-response relationship between exposure to 5-α reductase inhibitors when the exposure was stratified into tertiles of total exposure (P = .12). In multivariate models, there was little effect of adjustment for comorbidity.
There was no evidence for an association between a diagnosis of BPH or surgical treatment of BPH and hip fracture (Table 3). Of the 7076 case and control patients, 2547 (36%) and 2488 (35%), respectively, had a prior diagnosis of BPH (P = .30). The use of α-blockers was slightly greater in men with hip fracture (2250/7076 [32%]) compared with those without hip fracture (2139/7076 [30%]) (P = .04). There also was no evidence of a dose response in the association with α-blocker use. The risk associated with α-blocker use was increased only among those with the most recent prescriptions (within 30 days vs none: OR, 2.04; 95% CI, 1.19-3.49). In comparing the selective vs nonselective α-blockers, the effect was slightly stronger in men using selective α-blockers (OR, 1.23; 95% CI, 0.97-1.56) vs nonselective α-blockers (OR, 1.07; 95% CI, 1.00-1.16), although neither attained statistical significance.
These data demonstrate no evidence of a direct association between 5-α reductase inhibitors and hip fracture. In fact, there appears to be an inverse association. This does not appear to be due to the men having BPH, which in and of itself may be the manifestation of an altered hormonal milieu.4 Similarly, the inverse association does not appear to be due to confounding by indication, because there was no suggestion of an inverse association between α-blocker use, a diagnosis of BPH, or surgical treatment for BPH and hip fracture. Thus, the inverse association between 5-α reductase inhibitors and risk of hip fracture may be real and deserves additional consideration.
This observed association is biologically plausible. The role of sex steroids in bone health has been long recognized.17 Testosterone levels in men have been shown to be directly correlated with bone density in older men, although serum levels of bioavailable estradiol were more strongly correlated.8 However, the pathway through which 5-α reductase inhibitors may work is not clear, and the association may not necessarily be affected by bone metabolic pathways. The predominant form of 5-α reductase in osteoblasts is type I, whereas finasteride inhibits the type II isoenzyme, suggesting alternative pathways. It has been noted that finasteride is associated with higher serum levels of estradiol,5 which may be due to shunting of testosterone through the aromatization pathway. In addition, increased levels of bioavailable testosterone have been shown to be indirectly related to the risk of falling.18 Even if the inverse association between finasteride and fracture risk is due to a decreased risk of falls, it is not clear if decreased risk of fracture represents a direct effect of androgen or an increase in estrogen levels due to increased levels of substrate for aromatase. And while previous studies of 5-α reductase inhibitors have not shown an association with bone density or with markers of bone metabolism, many of these studies may have been underpowered, may have had limited follow-up, or both.
While the size of the effect is modest, it could have important public health implications if it represents a cause-and-effect relationship. It has been projected that approximately 8.1 million US men aged 50 through 79 years will have BPH by 2010.1 If 10% of these men are taking 5-α reductase inhibitors,19 approximately 2000 would be expected to sustain hip fractures in the following year.20,21 Based on the inverse association observed in this study, these men may avoid 500 of these hip fractures, owing to the effects of 5-α reductase inhibition. In this study population, however, the use of these agents was fairly low, and the use of dutasteride almost nonexistent. Thus, it is not clear if these results would be sustained with wider use of finasteride or with the use of 5-α reductase inhibitors that affect both forms of the enzyme, such as dutasteride. In addition, the duration of exposure in this study was somewhat limited, given the availability and uptake of finasteride within the KPSC health plan. It will be important to assess men who start taking 5-α reductase inhibitors at earlier ages for BPH or other indications.
The finding of a modest direct association with α-blockers, although not an a priori hypothesis, is also of interest. These agents are often used as a first-line medication for the treatment of lower urinary tract symptoms often associated with BPH.2 A well-known adverse effect of these agents, particularly when the dose is first being titrated, is orthostatic hypotension.22 A consequent increase in falls could therefore represent the cause of the increase in hip fracture observed in this study. This hypothesis is supported by the observation that the increase in risk was primarily in men who started taking an α-blocker more recently. It was surprising, however, that the effect was somewhat stronger in men receiving selective α-blockers, because the systemic adverse effects are supposed to be less common with these agents. This may represent confounding by indication (men at higher risk of falling being given the selective agents) or a false sense of security and less careful titration of dose.
While these results provide no evidence of a detrimental effect of 5-α reductase inhibitors on the risk of hip fracture, this study does have some limitations. While it does not appear that the observed association is due to the underlying disease—ie, BPH—or to confounding by indication or comorbidity, an unmeasured confounder may be responsible for the observed association. The replication of these results with other fracture sites may provide additional evidence as to whether this is a true association. Also, it is important to remain cognizant that dispensed medications are not necessarily ingested. However, any misclassification of exposure status due to medication received but not ingested is likely to bias the results to the null and thereby lead to an underestimate of the true association. The KPSC population represents an insured population; therefore, the results may be limited in their generalizability. However, the characteristics of the KPSC member population are very similar to those of the California population, with the exception of a modest underrepresentation of the lowest and highest socioeconomic brackets. Finally, this study relied on the diagnostic codes for fracture and on pharmacy codes for exposure. While some modest misclassification might exist with these codes, it seems unlikely that misclassification would introduce a systematic bias on the basis of either exposure or outcome.
These data suggest that 5-α reductase inhibitors do not confer a negative risk for bone health and in fact may lower the risk of hip fracture. While presumably this lower risk is related to hormonal mechanisms, further understanding of the biological mechanisms underlying this phenomenon may lead to new insights that can be exploited for preventive measures. The increased risk of fracture associated with recent receipt of an α-blocker highlights the need for careful titration of these agents.
Corresponding Author: Steven J.Jacobsen, MD, PhD, Department of Research and Evaluation, Kaiser Permanente Southern California, 100 S Los Robles Ave, Second Floor, Pasadena,CA 91101 (email@example.com).
Author Contributions: Dr Jacobsen had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Jacobsen.
Acquisition of data: Jacobsen, Shi.
Analysis and interpretation of data: Jacobsen, Cheetham, Haque, Shi, Loo.
Drafting of the manuscript: Jacobsen,Haque, Shi.
Critical revision of the manuscript for important intellectual content: Cheetham, Haque, Loo.
Statistical analysis: Jacobsen, Haque,Shi.
Administrative, technical, or material support: Jacobsen, Loo.
Study supervision: Jacobsen.
Financial Disclosures: Dr Jacobsen reported receiving research funding from and serving as an unpaid consultant for Merck Research Laboratories. Dr Cheetham reported affiliations with Allergan through family. No other disclosures were reported.
Funding/Support: This study was supported through the Kaiser Foundation Community Benefit fund and the Southern California Permanente Medical Group.
Role of the Sponsors: The Kaiser Foundation Community Benefit fund and Southern California Permanente Medical Group had no role in the design and conduct of the study; the collection,management, analysis, and interpretation of the data; or the preparation,review, or approval of the manuscript.
Additional Contributions: We thank Bianca Cheung, BS, Kaiser Permanente Southern California, for her assistance in the preparation of the manuscript. Ms Cheung received no compensation for her contributions.