Shown is the incidence of bone fracture in the rosuvastatin calcium and placebo groups. The P value was calculated using a log-rank test of the effect of rosuvastatin use, using a proportional hazards model.
Shown are the hazard ratios (squares) for rosuvastatin as compared with placebo. The size of the square is inversely proportional to the confidence interval (horizontal lines). Interaction P values are shown to the right. BMI indicates body mass index (calculated as weight in kilograms divided by height in meters squared); and hs-CRP, high-sensitivity C-reactive protein (to convert to nanomoles per liter, multiply by 9.524).
eFigure. Trial Flow Diagram
eTable. Model Covariates and Risk of Bone Fracture
Customize your JAMA Network experience by selecting one or more topics from the list below.
Peña JM, Aspberg S, MacFadyen J, Glynn RJ, Solomon DH, Ridker PM. Statin Therapy and Risk of Fracture: Results From the JUPITER Randomized Clinical Trial. JAMA Intern Med. 2015;175(2):171–177. doi:10.1001/jamainternmed.2014.6388
Osteoporosis and cardiovascular disease may share common biological pathways, with inflammation playing a role in the development of both. Although observational studies have suggested that statin use is associated with a lower risk of fractures, randomized trial data addressing this issue are scant.
To determine whether statin therapy reduces the risk of fracture and, in a secondary analysis, whether baseline levels of the inflammatory biomarker high-sensitivity C-reactive protein (hs-CRP) are associated with the risk of fracture.
Design, Setting, and Participants
The JUPITER (Justification for the Use of Statins in Prevention: an Intervention Trial Evaluating Rosuvastatin) trial was an international, randomized, double-blind, placebo-controlled study enrolling 17 802 men older than 50 years and women older than 60 years with hs-CRP level of at least 2 mg/L. Participants were screened from 2003 to 2006 and observed prospectively for up to 5 years (median follow-up, 1.9 years).
Rosuvastatin calcium, 20 mg daily, or placebo.
Main Outcomes and Measures
Incident fracture was a prespecified secondary end point of JUPITER. Fractures were confirmed by radiographs, computed tomography, bone scan, or other methods. Cox proportional hazards models were used to calculate hazard ratios (HRs) and associated 95% confidence intervals for the risk of fracture according to randomized treatment assignment, as well as increasing tertiles of hs-CRP, controlling for potential confounders.
During the study, 431 incident fractures were reported and confirmed. Among participants allocated to rosuvastatin, 221 fractures were confirmed, compared with 210 among those allocated to placebo, such that the incidence of fracture in the rosuvastatin and placebo groups was 1.20 and 1.14 per 100 person-years, respectively (adjusted HR, 1.06 [95% CI, 0.88-1.28]; P = .53). Overall, increasing baseline hs-CRP level was not associated with an increased risk of fractures (adjusted HR for each unit increase in hs-CRP tertile, 1.06 [95% CI, 0.94-1.20]; P for trend, .34).
Conclusions and Relevance
Among men and women with elevated hs-CRP level enrolled in a large trial of rosuvastatin therapy for cardiovascular disease, statin therapy did not reduce the risk of fracture. Higher baseline hs-CRP level was not associated with an increased risk of incident fracture.
clinicaltrials.gov Identifier: NCT00239681
Osteoporotic fractures contribute substantially to the burden of disease facing an aging population. Cardiovascular disease (CVD) and osteoporosis are both age-related systemic diseases that may share common biological pathways,1,2 and several epidemiologic studies have linked them together. Inflammation is key to the pathogenesis of atherosclerosis and may also play an important role in the development of osteoporosis. Chronic inflammation promotes bone loss, and extensive reciprocal relationships exist between bone metabolism and the immune system.3-5
There are several mechanisms by which statins may exert positive biologic effects on bone. In an early rodent study, statin injection was shown to stimulate bone formation.6 Statins and nitrogen-containing bisphosphonate drugs both act in the mevalonate pathway of cholesterol synthesis.7 These observations have fueled interest in the role of statins in bone metabolism and the hypothesis that statin use may have clinical benefits beyond CVD prevention.
Several observational studies found a reduced risk of fractures in users of statins,8-12 but others found no association.13,14 Several studies have also shown an association between statin use and greater bone mineral density.15-17 Post hoc analyses of randomized clinical trials of statin therapy have not demonstrated a reduced risk of fracture.18,19 Such analyses have been limited by their post hoc consideration of fractures, use of statins that may be less effective on bone in vitro, and insufficient power.20
In the JUPITER (Justification for the Use of Statins in Prevention: an Intervention Trial Evaluating Rosuvastatin) trial, we sought on an a priori and prespecified basis to determine (1) whether treatment with rosuvastatin calcium is associated with a lower risk of fractures, and (2) in an exploratory analysis, whether higher baseline serum high-sensitivity C-reactive protein (hs-CRP) level is associated with an increased risk of fracture.
The trial protocol was approved by the local institutional review board of each participating study site. Written informed consent was obtained from each participant.
Incident fracture was a prespecified secondary end point of JUPITER. The JUPITER trial was a randomized, double-blind, placebo-controlled, multinational trial performed at 1315 centers in 26 countries. Details of the study design and main results of the trial have been described in detail previously.21 Men and women older than 50 and 60 years, respectively, were eligible for participation if they had no prior history of CVD or diabetes mellitus and if, at the screening visit, serum hs-CRP level was at least 2 mg/L (to convert to nanomoles per liter, multiply by 9.524) and low-density lipoprotein cholesterol level less than 130 mg/dL (to convert to millimoles per liter, multiply by 0.0259). The baseline hs-CRP level was calculated as the mean of the screening and baseline visit levels, and therefore baseline hs-CRP levels less than 2 mg/L were not protocol violations. Blood samples were drawn locally and shipped to a central laboratory where they were analyzed unbatched using a validated high-sensitivity assay with the Behring nephelometer and reagent.22
Exclusion criteria relevant to the development of fractures were recent history of alcohol abuse, cancer within the 5 years prior to enrollment (except basal or squamous cell skin cancers), diabetes mellitus, chronic inflammatory conditions such as lupus, severe arthritis, or inflammatory bowel disease, and use of postmenopausal hormone therapy or long-term use of oral glucocorticoids. A total of 17 802 participants were randomized to receive either rosuvastatin calcium, 20 mg daily, or placebo and were observed for up to 5 years (median, 1.9 years). The primary end point of the study was the first occurrence of a major cardiovascular event.
At baseline, participants had a detailed medical history taken and underwent a screening physical examination prior to randomization. Participants were then observed at 3-month intervals and queried for the occurrence of trial end points including fractures and adverse events. A trial flow diagram is shown in the eFigure in the Supplement.
Each case of fracture was reported on a designated secondary end point case report form. The type of fracture and imaging test used for confirmation were recorded. Fractures were confirmed by the site principal investigator using radiographs, computed tomography, bone scan, or other methods. Only the first confirmed fracture for each patient was included in the primary analysis. In the 13 participants who had multiple confirmed fractures occurring on the same day, only 1 fracture was included in the primary analysis. However, all confirmed fractures were included in the analysis of the effect of rosuvastatin use on each individual fracture type. In a sensitivity analysis, we excluded these participants, hypothesizing that these fractures may have been due to extreme trauma. Comment fields on case report forms were searched manually for pathologic fractures and severe trauma, leading to the exclusion of 7 cases of fracture. We also performed an analysis of all first reported fractures, including those identified from adverse event reports and fractures occurring during safety monitoring after the trial ended. On March 30, 2008, JUPITER was terminated after the data and safety monitoring committee determined that the gathered evidence on efficacy and safety demonstrated that prolonged use of rosuvastatin was clearly indicated for specific patients. Although follow-up for the trial’s primary end point terminated on that day, participants continued to receive the study drug and were observed in a blinded manner for the occurrence of adverse events and safety end points, including fracture, until their close-out study visit. The final study visit occurred on August 30, 2008.
To compare the rates of fracture in the placebo and rosuvastatin groups, we fitted Cox proportional hazards models to estimate hazard ratios (HRs) and associated 95% confidence intervals. The multivariate model chosen on the basis of a priori knowledge included age (continuous), sex, blood pressure of at least 140/90 mm Hg or use of antihypertensive medications (yes/no), randomized treatment assignment, current tobacco use (yes/no), body mass index (categories: <25.0, 25.0-29.9, ≥30.0, calculated as weight in kilograms divided by height in meters squared), exercise (categories: rarely or less than once per week, once per week, 2-6 times/wk, daily), race, alcohol use (categories: ≤1-3 times/mo, 1-6 times/wk, 1-3 times/d, ≥4 times/d), baseline hemoglobin A1c level in quartiles (<5.50%, 5.50%-5.69%, 5.70%-5.90%, >5.90%), and history of previous fracture. To evaluate potential effect modification, multiplicative interaction terms between drug assignment and various baseline characteristics were inserted into the unadjusted Cox model. Rates of incident fracture in the 2 groups were compared using the Kaplan-Meier method, with differences between the 2 groups tested using the log-rank test. The proportional hazards assumption was tested using a multiplicative interaction term between each covariate and the mean centered logarithm of study time. We did not detect a violation of the proportional hazards assumption. All analyses were performed according to the intent-to-treat principle, and a 2-sided P < .05 was considered significant.
We also performed an exploratory analysis to determine whether baseline hs-CRP level in tertiles, as well as continuous log-transformed hs-CRP level, predicts incident fracture within the JUPITER study using the same multivariate model as described. Because women typically have higher levels of hs-CRP and different fracture risks than men, we also performed additional analyses using sex-specific tertiles of hs-CRP level.23 Baseline use of thiazide diuretics, bisphosphonates, calcium supplementation, vitamin D, and inhaled or oral steroids did not result in a change in effect estimates, and these were thus omitted from the multivariate model.
Of the 17 802 participants in JUPITER, 38.2% were women and the median age was 66.0 years. History of a previous fracture was reported by 36.6% of participants, with 57.7% of those participants reporting 1 to 3 fractures in the 20 years prior to randomization. The median hs-CRP level was 4.3 mg/L in participants in the placebo group and 4.2 mg/L in the rosuvastatin group (Table 1).
During the study period, there were 431 confirmed fractures, 221 in participants in the rosuvastatin group and 210 in participants assigned to placebo (Table 2). Overall, the incidence of fracture in the rosuvastatin and placebo groups was 1.20 and 1.14 per 100 person-years, respectively (adjusted HR, 1.06 [95% CI, 0.88-1.28]). The risk of fractures during follow-up was increased in women as compared with men: adjusted HR, 2.06 (95% CI, 1.66-2.56) (eTable in the Supplement). Among women, the incidence of fracture among those treated with rosuvastatin was 1.80 per 100 person-years (adjusted HR, 1.16 [95% CI, 0.89-1.50]), and among men, 0.85 per 100 person-years (adjusted HR, 0.97 [95% CI, 0.74-1.28]) (Table 2). The HRs for each of the covariates in the adjusted model are shown in the eTable in the Supplement. The cumulative incidence curves for bone fracture in the placebo and rosuvastatin groups are shown in Figure 1. There were 137 upper-extremity fractures (including 71 wrist fractures) and 135 lower-extremity fractures (including 55 ankle fractures) (Table 2). There were no significant differences in the rate of specific fractures between the 2 study groups (Table 2).
The effect of rosuvastatin use on risk of fracture seemed to be consistent across major subgroups without evidence of treatment benefit in men or women or among those with and without history of previous fracture (Figure 2). There was no evidence of effect modification by number of previous fractures (P = .73).
The inclusion of 10 additional confirmed fractures occurring after trial termination during safety monitoring did not alter our overall results (adjusted HR, 1.06 [95% CI, 0.88-1.28]), nor did the inclusion of all 571 cases of reported (not confirmed) fracture (adjusted HR, 1.04 [95% CI, 0.88-1.22]). The exclusion of fractures less likely to be associated with osteoporosis such as skull, face, finger, and toe fractures led to consistent findings (HR, 1.05 [95% CI, 0.86-1.28]). The exclusion of vertebral fractures, which can be challenging to diagnose without baseline radiographs, yielded similar results (adjusted HR, 1.05 [95% CI, 0.86-1.29]). Because participants with multiple fractures occurring on the same day may have undergone extreme trauma, we also performed an analysis excluding these participants (adjusted HR, 1.03 [95% CI, 0.85-1.25]).
At the time of trial termination, 75% of participants continued to take the study drug.21 Adherence was assessed at each clinic visit by means of pill count. In an analysis of the 14 594 participants who continued to take the study drug or did not initiate off-label statin therapy at 1 year, rosuvastatin was not associated with a reduced risk of fracture (adjusted HR, 1.06 [95% CI, 0.86-1.29]).
Increasing tertiles of baseline hs-CRP level were not associated with the risk of fracture (adjusted HR for each unit increase in hs-CRP tertile, 1.06 [95% CI, 0.94-1.20]; P for trend, .34) (Table 3). Results were similar in the continuous hs-CRP analysis (adjusted HR for 1-unit increase in log-transformed hs-CRP level, 1.09 [95% CI, 0.95-1.24]; P = .23), as well as in sex-specific analyses (Table 3).
In this randomized, double-blind placebo-controlled trial of rosuvastatin therapy for primary CVD prevention, allocation to rosuvastatin did not reduce the risk of fracture. Furthermore, higher baseline hs-CRP level was not associated with an increased risk of fracture. Our results were consistent in sensitivity analyses and various subgroups.
Our findings with regard to statin therapy are similar to prior analyses of cardiovascular trials18,19,24 and some observational and case-control studies.13,14 In exploratory analyses of the Scandinavian Simvastatin Survival Study (4S) and Long-Term Intervention with Pravastatin in Ischaemic Disease (LIPID) trials, statin therapy did not reduce the rate of adverse event reports of fracture.18,19 In a prespecified analysis of the Heart Protection Study, which enrolled participants with vascular disease or diabetes mellitus, use of simvastatin, 40 mg daily, was not associated with a decreased risk of hospitalization for fracture.24 Among postmenopausal women participating in the Women’s Health Initiative, statin use was not associated with a lower fracture risk or improved bone density.14
However, multiple observational studies have observed a salutary effect of statin use on fracture risk. In a meta-analysis of 4 large observational studies, statin use was associated with a lower risk of hip fracture (odds ratio, 0.43 [95% CI, 0.25-0.75]).20 Furthermore, case-control studies among varied populations in the United Kingdom,11 elderly Medicare and Medicaid beneficiaries,10 and American veterans9 have suggested a reduced risk of fracture in those exposed to statins, with odds ratios ranging from 0.29 to 0.64. Whereas the effect estimates in these observational reports are compelling, the limitations of observational data are noteworthy. These studies may be limited by selection bias, confounding by indication, or a “healthy user” phenomenon, as well as residual confounding by use of other cardiovascular medications, frailty, and obesity.25-27 The discordance between observational studies and randomized trials may also be explained by other factors. For example, it is possible that long-term statin therapy, intermittent use of statins over a long period, or statin therapy in selected populations reduces the risk of fracture.
Elevation of levels of the inflammatory biomarker hs-CRP has been associated with an increased risk of fractures in several observational cohort studies.28-30 In 1 cohort of elderly women, increasing hs-CRP level was associated with a higher risk of fractures (adjusted HR for highest hs-CRP tertile, 2.40 [95% CI, 1.10-5.24]).28 Because JUPITER enrolled participants with elevated hs-CRP levels, our analysis of baseline hs-CRP levels and fracture risk is limited to those at the upper end of the range of hs-CRP values. The truncated range of hs-CRP level may have obscured our ability to detect a relationship between hs-CRP level and fracture risk. In addition, the relatively short follow-up and younger age of participants in our study in contrast to those in observational studies may have contributed to the observed lack of association.
To our knowledge, our study is the first randomized clinical trial examining the relationship between statin therapy and incident fracture in a prespecified manner. This design is a strength of our analysis because it minimizes the risk of bias and residual confounding inherent in observational investigations. Furthermore, the use of rosuvastatin, a potent statin, extends prior investigations in the LIPID, 4S, and Heart Protection Study trials using pravastatin sodium and simvastatin, respectively.
Our study design has important limitations. First, the relatively short treatment duration may not have been sufficient to achieve a beneficial effect of statin use on bone. We note, however, that our Kaplan-Meier curves offered no suggestion of an emerging benefit in those participants treated for 3 to 5 years. Second, we administered a fixed dose of statin and are unable to exclude the possibility that higher doses of statin may prevent fracture. Orally administered statins have low bioavailability, and the dose used in animal studies of statins and bone metabolism is approximately 10-fold higher than that used in human studies.25 However, the 20-mg dose of rosuvastatin was sufficient to achieve cardiovascular benefits in this trial and higher doses may convey other risks. In addition, because our study end point was fracture, we are unable to evaluate subclinical effects on bone or elucidate further on the mechanism by which statins may mediate effects on bone.
Whether the study was designed with adequate statistical power is a concern in light of the null finding. However, with the accrual of 431 confirmed, incident fractures in JUPITER, the log-rank test with 2-sided α set at .05 has an estimated 96% power to detect a 30% reduction in the hazard of first fracture associated with statin assignment (HR, 0.70), 84% power to detect a 25% reduction (HR, 0.75), and 64% power to detect a 20% reduction (HR, 0.80).
In this large randomized trial of rosuvastatin therapy among men and women with evidence of inflammation, randomization to rosuvastatin did not reduce the risk of fracture and higher baseline hs-CRP level was not associated with an increased risk of fracture. Our study does not support the use of statins in doses used for cardiovascular disease prevention to reduce the risk of fracture.
Accepted for Publication: August 9, 2014.
Corresponding Author: Jessica M. Peña, MD, MPH, Division of Cardiology, Montefiore Medical Center, Albert Einstein College of Medicine, 1825 Eastchester Rd, Bronx, NY 10461 (email@example.com).
Published Online: December 1, 2014. doi:10.1001/jamainternmed.2014.6388.
Author Contributions: Dr Peña 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: Peña, Glynn, Solomon, Ridker.
Acquisition, analysis, or interpretation of data: Peña, Aspberg, MacFadyen, Glynn.
Drafting of the manuscript: Peña, Aspberg.
Critical revision of the manuscript for important intellectual content: Peña, Aspberg, MacFadyen, Glynn, Solomon, Ridker.
Statistical analysis: Peña, Aspberg, Glynn.
Obtained funding: Glynn, Ridker.
Administrative, technical, or material support: Peña, MacFadyen, Ridker.
Study supervision: Ridker.
Conflict of Interest Disclosures: Dr Glynn has received grant support from AstraZeneca through research grants to Brigham and Women’s Hospital for analyses of JUPITER data, as well as investigator-initiated grant support from Novartis. Dr Solomon receives salary support through research grants to Brigham and Women’s Hospital from Amgen, Lilly, Pfizer, and the CORRONA Foundation. He also serves in unpaid roles on trials sponsored by Bristol-Myers Squibb and Pfizer. He receives royalties from UpToDate on chapters regarding nonsteroidal anti-inflammatory drugs and coxibs. He receives an honorarium from the American Orthopaedic Association for serving on a multispecialty board related to their Own the Bone program and is on the Governance Committee of the National Bone Health Alliance. Dr Ridker has received investigator-initiated grant support from the National Heart, Lung, and Blood Institute, Pfizer, AstraZeneca, Novartis, and Amgen; is listed as a coinventor on patents held by the Brigham and Women’s Hospital that relate to inflammatory biomarkers in cardiovascular disease that have been licensed to AstraZeneca and Siemens; and has served as a consultant to Genzyme, Vascular Biogenics, Boston Heart, Amgen, and Isis Pharmaceuticals. No other disclosures are reported.
Funding/Support: The JUPITER trial was supported by AstraZeneca. Dr Peña was supported by grant T32 HL07575 from the National Heart, Lung, and Blood Institute.
Role of the Sponsor: AstraZeneca collected the trial data and monitored study sites but had no role in design of the study protocol; analysis and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.