Dr Black served as a consultant and received grants from Merck & Co., NIAMS, NPS Pharmaceuticals, and Novartis. All other authors have no conflict of interest.
β-Blocker Use, BMD, and Fractures in the Study of Osteoporotic Fractures
Article first published online: 6 DEC 2004
Copyright © 2005 ASBMR
Journal of Bone and Mineral Research
Volume 20, Issue 4, pages 613–618, April 2005
How to Cite
Reid, I. R., Gamble, G. D., Grey, A. B., Black, D. M., Ensrud, K. E., Browner, W. S. and Bauer, D. C. (2005), β-Blocker Use, BMD, and Fractures in the Study of Osteoporotic Fractures. J Bone Miner Res, 20: 613–618. doi: 10.1359/JBMR.041202
- Issue published online: 4 DEC 2009
- Article first published online: 6 DEC 2004
- Manuscript Accepted: 22 OCT 2004
- Manuscript Revised: 7 JUL 2004
- Manuscript Received: 29 APR 2004
- central nervous system;
A role for osteoblastic β-adrenoreceptors in bone regulation is suggested by the finding that β-blockers increase bone mass in mice. We studied the association of β-blocker use with BMD and fractures in the Study of Osteoporotic Fractures. β-blocker use and BMD are unrelated in this cohort, and associations with fracture risk are inconsistent.
Introduction: The central nervous system has been shown to regulate bone mass in mice, possibly by way of the β2-adrenoreceptors on osteoblasts. β-blockers have been shown to increase bone mass in mice. Because these agents are widely used therapeutically, it is possible that they may influence fracture epidemiology in humans, and they are a potential therapy for osteoporosis.
Materials and Methods: We have studied the association of β-blocker use with BMD and fracture rates in the Study of Osteoporotic Fractures. β-blocker use was recorded at the fourth visit, in 8412 women, of whom 1099 were users, and these women were followed for 7 years.
Results: Users had significantly higher weight, more thiazide use, more estrogen use, less glucocorticoid use, more statin use, and more hypertension than nonusers, and they smoked less. Total hip BMD at the fourth visit was higher in the β-blocker users (0.746 versus 0.735 g/cm2, p = 0.02), but adjustment for weight alone, or together with these other variables, eliminated this difference (p = 0.62). There was no effect of β-blocker use on loss of hip BMD over a mean follow-up of 4 years (p = 0.48). Os calcis BMD at visit 4 was also higher in those taking β-blockers (0.385 versus 0.375 g/cm2, p = 0.005), but weight adjustment eliminated this difference (p = 0.14). The frequencies of hip or any fracture (since age 50) were similar in users and nonusers (p = 0.80 and p = 0.51, respectively). Over a mean follow-up of 7 years, there were 2167 total fractures, including 431 at the wrist and 585 at the hip. Among β-blocker users, hazards ratios were 0.92 (0.81, 1.05) for any fracture, 0.74 (0.54, 1.01) for wrist fracture, and 0.76 (0.58, 0.99) for hip fracture. Adjustment for weight and other factors previously shown to influence hip fracture incidence in this cohort made little difference to the outcome. When fracture data were analyzed for nonselective and β1-selective agents separately, trends toward fewer fractures were confined to the users of selective β1-blockers.
Conclusions: β-blocker use and BMD are unrelated in this cohort, and associations with fracture risk are inconsistent. Therefore, a history of use of these drugs is not useful in assessing fracture risk, nor do they have a role in osteoporosis management at this time. The relationship between β-blocker use and hip fracture deserves further study.
UNTIL RECENTLY, IT was believed that bone mass was regulated by local mechanical and cytokine influences and by the effects of systemic hormones. The work of Ducy et al.(1) added a major new dimension to our understanding of this area by showing that bone formation and bone mass in mice are responsive to intracerebroventricular administration of the hormone leptin. This implied that the central nervous system played a role in the regulation of skeletal cell activity. These observations were extended by Takeda et al.,(2) who concluded that the central effects of leptin were mediated by the sympathetic nervous system, ultimately acting through β2-adrenergic receptors on osteoblasts. These studies further showed that the systemic administration of β-agonists resulted in bone loss in mice, whereas the administration of propranolol, a nonselective β-blocker, increased bone mass.
The establishment of a role of the sympathetic nervous system in bone cell regulation is potentially of immediate relevance to the epidemiology and therapeutics of osteoporosis, because β-blockers have been in widespread clinical use for several decades. This awareness led Pasco et al.(3) to study the association of β-blocker use, BMD, and fractures in a case-control study using data from the Geelong Osteoporosis Study. Their study of 569 fracture cases and 775 controls suggested that the odds ratio for fracture associated with β-blocker use was 0.68 (95% CI: 0.49, 0.96) for any fracture, a finding that was little affected by adjustment for age, weight, other medications, and lifestyle factors. In addition, BMD at the total hip and ultradistal forearm were ∼0.2 SD higher in β-blocker users. These findings suggest that a history of β-blocker use or of β-agonist use (e.g., in asthma) should be used in assessing individuals' fracture risk. Furthermore, they point to the possibility of using β-blockers in the therapeutics of osteoporosis. The potentially substantial impact of these observations on both fracture risk assessment and treatment of osteoporosis require that they be confirmed in a large independent study. The Study of Osteoporotic Fractures is a well-characterized prospective cohort study of older women, which was ideally suited to this purpose.
MATERIALS AND METHODS
Study participants were ambulatory, community-dwelling white women ≥65 years of age. They were recruited through population-based listings from four clinical centers in the United States: the Kaiser-Permanente Center for Health Research in Portland, OR; the University of Minnesota in Minneapolis, MN; the University of Maryland in Baltimore, MD; and the University of Pittsburgh in Pittsburgh, PA. Women who had undergone bilateral hip replacements or were unable to walk alone were excluded. The study was approved by the appropriate institutional review boards, and written informed consent was obtained from all participants. The general methods used in this study have been described previously.(4)
These analyses are limited to those who participated in the fourth visit (1992-1994) of the Study of Osteoporotic Fractures, when detailed medication histories were first obtained. Study participants were instructed to bring all prescription and nonprescription medications to the clinic visit for verification by interviewers. Current use of a specific medication was defined as use during the previous 2 weeks. β-blockers were classified as being β1-selective or nonselective and as being taken in high or low doses according to whether the doses were above or below the midpoint of the recommended dose range. BMD at the hip was measured using DXA with Hologic QDR-1000 scanners (Waltham, MA, USA). Calcaneal BMD was measured with single-photon absorptiometry using Osteon Osteoanalyzers (Wahiawa, HI, USA). The mean CV between centers was 0.9% for the total hip and 1.2% for the calcaneus. Hip measurements were repeated at visit 6, ∼4 years later. Physical activity was assessed by asking women if they walked for exercise and by calculating a weekly caloric expenditure based an exercise questionnaire.(4) Participants were weighed with a balance beam scale in indoor clothing with shoes removed. Examiners classified women as frail if the participants appeared weak, unsteady, or fragile, and neuromuscular performance was assessed, including the woman's ability to rise from a chair without using her arms.
During the follow-up period, women were contacted every 4 months, by postcard or by telephone, to determine if they had suffered a fracture. All fractures were confirmed by review of radiographic reports. Fractures that occurred over a mean follow-up period of 7 years after visit 4 were included in this analysis.
Differences between users and nonusers were sought using standard parametric (Student's t-test for unrelated groups) procedures for continuous, normally distributed variables and Fisher's exact test for dichotomous data.
The time to first fracture was modeled using a proportional hazards approach. Among women with more than one fracture, only the first fracture of a specific type was included in the analysis. Multiple linear regression was used to model the relationship between β-blocker use and various BMD endpoints after adjustment for prespecified potential confounders. Model fit was assessed as the overall adjusted R square value, and comparisons between models were made using Akaike's information criterion. In multivariate analyses, no iterative model building procedures were used. Prespecified hypotheses were modeled rather than generated. General linear modeling was used to test for differences in BMD between β-blocker users (ANOVA) and after adjustment for prespecified confounders (ANCOVA). An overall pairwise 5% error rate was maintained using the method of Tukey. Marginal least squares means are presented for adjusted variables.
Analysis was performed using the procedures of SAS (version 8.12; SAS Institute). All tests were two-tailed. A 5% significance level was maintained throughout these analyses.
Clinical data including medication use were available from 8127 subjects at visit 4, of whom 1099 were current users of β-blockers. Table 1 compares the clinical characteristics of the users and nonusers. Users were significantly heavier, they smoked less, they more frequently used estrogen, thiazides, statins, nitrates, and other cardiac drugs, they less frequently used glucocorticoid drugs, and they were frailer and had poorer overall health than nonusers. Death during follow-up was 18% more common in the β-blocker users, so their duration of follow-up after visit 4 was significantly shorter. Among the users of β-blockers, 42% took nonselective agents, and 5% of these were classified as high-dose users. In those taking β1-selective agents, 18% were using high doses.
The percent of subjects with a history of fractures between age 50 and visit 4 was similar between groups, both for hip fractures and total fractures. This comparison was not affected by adjustment for weight (p = 0.37 and 0.57 for hip and total fractures, respectively) or by adjustment for weight, age, estrogen use, thiazide use, glucocorticoid use, and smoking (p = 0.56 and 0.37, respectively).
Total hip BMD at the fourth visit was 1.0% higher in the β-blocker users (p = 0.02), but adjustment for weight alone eliminated this difference (Table 2). Further adjustment for age, estrogen use, thiazide use, glucocorticoid use, and smoking, all of which were significantly different between groups and are known to substantially influence BMD, resulted in virtual identity of the means for hip BMD.
Similar results were found for os calcis BMD, there being a 1% difference between groups before any adjustment, which was eliminated by adjustment for weight alone. Further adjustment for the cluster of factors listed above that were different between groups and impact on BMD confirmed this finding.
In subjects in whom BMD was remeasured at visit 6 (after a mean follow-up of 4 years), bone loss was not different between the groups, whether the data were adjusted for co-variables or not.
These results were essentially unchanged by analyses that separated selective and nonselective β-blockers. For total hip BMD, unadjusted densities were highest in those taking selective β-blockers (0.735, 0.742, and 0.749 g/cm2 for control, nonselective β-blockers, and selective β-blockers, respectively—only the control/selective comparison was significant, p = 0.02). Adjustment for weight or multiple variables (as above) eliminated this difference. For os calcis BMD, the same pattern was seen, the densities in the respective groups being 0.375, 0.378, and 0.391 g/cm2 (pcontrol-selective = 0.0005). After weight adjustment, users of selective β-blockers still had a significantly higher BMD, but not after adjustment for the other variables. There was no effect of β-blocker selectivity on rate of loss of hip BMD with or without adjustment.
Incident fracture status was known in 1097 β-blocker users and 7001 nonusers (Table 3). In unadjusted or fully adjusted models, there were no significant differences in fracture risk associated with β-blocker use for total fractures, wrist fractures, or nonspine fractures. However, in adjusted models, there was a trend toward fewer wrist fractures among β-blocker users (relative hazard [RH] = 0.75), but this relationship did not reach statistical significance (95% CI: 0.54, 1.06).
Hip fractures were less common in β-blocker users, an effect that was no longer significant after adjustment for weight. Further adjustment for variables likely to impact on BMD and that were different between groups made this effect significant again. We then added a wide range of variables, possibly associated with hip fracture risk, one at a time, to this model. These included use of other cardiac medications (statins, other lipid-lowering medications, calcium-channel blockers, angiotensin-converting enzyme [ACE] inhibitors, nitrate use, non-thiazide diuretics), history of “heart attack,” history of hypertension, indices of frailty, alcohol use, and exercise. The reduction in hip fracture risk related to β-blocker use remained significant with each of these adjustments. From this list of variables, we added to the model all those that were significant when added singly. The hazard ratio remained reduced (0.67, p = 0.009). We then deleted variables that were no longer significant from this model, resulting in the group 2 variables listed in Table 3. With these adjustments, the reduction in hip fracture risk still remained significant. This group of variables was also used to adjust the hazard ratios for the other classes of fracture but made little difference to the outcome.
Reanalysis separating selective and nonselective β-blockers showed no effect of nonselective agents on any fracture outcome, whereas there was a variable association of fracture with use of selective agents. For any fracture, unadjusted hazard ratios were unaffected (0.90 for selective, 0.93 for nonselective), and this pattern was also seen after the various adjustments (all hazard ratios nonsignificant). The same was true for nonspine fractures before and after adjustment (unadjusted hazard ratios, 0.92 for both selective and nonselective agents; all hazard ratios nonsignificant). For hip fractures, unadjusted hazard ratios were 0.70 (95% CI: 0.49, 0.98) for selective and 0.77 (not significant) for nonselective. After weight adjustment, neither class of β-blocker had a significant effect on hip fracture risk (hazard ratios, 0.69, 0.88), but adjustment for group 1 or group 2 variables resulted in significant hazard ratios for selective agents (0.62 and 0.58, respectively) but not for nonselective agents (0.83 and 0.76, respectively). Wrist fractures were unassociated with use of nonselective agents (hazard ratios: unadjusted, 1.02; adjusted for weight, 1.02; adjusted for group 1 variables, 1.09; adjusted for group 2 variables, 1.04) but were consistently significantly associated with selective β-blocker use (hazard ratios, 0.51, 0.52, 0.54, and 0.52, respectively).
Results were similar in analyses that excluded women taking anti-osteoporotic therapies (fluoride, etidronate, calcitonin). Parallel analyses were performed excluding any subject who died during follow-up, with no significant change in the results. No effects of β-blocker dose were found on either BMD or fracture endpoints.
In this large prospective cohort study, we found no independent effect of β-blocker use on bone mass and bone loss. Although we found no significant relationship with nonspine fracture, use of β1-selective blockers was associated with fewer hip and wrist fractures in some analyses. However, we found that postmenopausal women who use β-blockers are different in a number of respects from those who do not. Not surprisingly, hypertension and heart disease are much more common in β-blocker users, resulting in greater use of other anti-hypertensives, nitrates, lipid-lowering therapies, and estrogen (which was thought to be cardio-protective at the time these data were collected). Estrogen,(5) statins,(6) thiazides,(7) and nitrates(4) have all been suggested to impact on either BMD or fracture risk. The β-blocker users are generally in poorer health, as reflected by greater frailty, lower energy expenditure, and greater mortality. Despite this, they are heavier than nonusers, possibly because of the association of hypertension with higher body weight. Many of these variables are likely to impact on BMD and/or fracture risk. Therefore, the interpretation of these data is potentially complex and requires multiple adjustments for these co-variables.
These considerations probably account for the complexity of the association of bone variables with β-blocker use. Thus, BMD at both the hip and heel is about 1% higher in the users compared with the nonusers, but this effect is likely to be attributable to differences in body weight. Further adjustment for other co-variables does not change the finding that, once weight is taken into account, the two groups have comparable BMDs at both sites assessed. Similar results are found with prospective assessment of bone loss over a period of 4 years, the rates of change being near identical with or without adjustment. Therefore, these findings disagree with the recent clinical study of Pasco et al.(3) and the mouse studies of Takeda et al.(2) It is unlikely that our failure to find a difference in BMD between users and nonusers is related to lack of power—this study has a power of 86% to detect the 2% difference in total hip BMD found by Pasco et al.(3)
The interpretation of the effect of β-blocker use on fracture risk in this study is more complex. Consistent with the BMD results, the unadjusted data show trends toward lower fracture risk in β-blocker users. Despite the number of confounding variables that would tend to produce this result, adjustment for them makes very little difference to the hazard ratios. The reason for this is unclear. One potential difficulty with calculating hazard ratios for hip fracture over a long period of follow-up near the end of life is that the risk of fracture climbs steeply during the follow-up period, and this violates one of the assumptions of the proportional hazards model (that of constant risk over time). Formal testing of the hypothesis of constant risk in this cohort does not show this to be a statistically significant problem, but this does not rule out some confounding from this source. The preferential loss of β-blocker users from the period of highest fracture risk (because of their higher mortality) may introduce a bias toward finding a lower risk of hip fracture in this group (although the tests of all the time-dependent variables were not significant). This may contribute to the internal inconsistency of these data, because it is hard to explain the lower fracture risk in the β-blocker users, whose baseline fracture history, baseline BMD, rate of change of BMD, and incidence of falls are the same as those in the nonuser group. It would be necessary to postulate that β-blocker use had resulted in an alteration in bone turnover or other aspects of bone quality that are not reflected in BMD.
The analyses of the effects of selective and nonselective β-blockers separately are of interest. There is a trend for BMD to be higher in users of selective agents, although this is eliminated by adjustment for confounders. The same lack of an effect of selectivity is seen with total and nonspine fractures, but the trends toward lower fracture rates at the hip and wrist are confined to use of selective blockers of the β1-receptors. These may be chance associations and are contrary to the original hypothesis, which was that any effects would be mediated by actions on the β2-receptors on osteoblasts. The absence of any effect of the agents that block these β2-receptors argues strongly against this hypothesis, although it does not rule out a role for other components of the sympathetic nervous system in bone metabolism.
These fracture data do not confirm the findings of Pasco et al.(3) They only assessed total fracture numbers and found this to be significantly decreased in the β-blocker users. There was a substantial difference in weight between the β-blocker users and nonusers, but adjustment for this and for other factors made little difference to the observed difference in fracture risk. Despite a downward trend in total fractures in this study, the odds ratio recorded by Pasco et al.(3) (0.68) does not lie within the 95% CIs of the hazard ratio for “any fracture” in these data. Their study is much smaller than this one, using data obtained from only 569 fracture cases and 775 controls. Also, it used a case-control design. Whereas there may be differences in types of β-blocker used, dose, or duration, there are no obvious reasons for the different findings between the two studies.
This study does have a number of limitations. The principal one is that it is observational, so the indications for β-blocker use may be confounders. The use of other bone-active agents, such as thiazides, statins, and nitrates, which are prescribed for similar indications, are major confounders, although they have been factored into the analysis. The study is also limited by a lack of information regarding the durations of use of β-blockers. Differences between these findings and those in mice may relate to the doses of β-blocker used. Kondo and Togari(8) have shown that the dose of propranolol necessary to block the effects of sympathetic activation on bone in mice are 10-fold higher (25 mg/kg) than those typically used to treat hypertension in humans and that slightly lower doses (15 mg/kg) do not produce the same effects.
A potential source of human data relating to the bone effects of β-blockers, not subject to the biases inherent in an observational study, is the incidence of fracture (captured as adverse events) in randomized controlled trials of β-blockers. We have recently analyzed data pooled from nine randomized controlled trials of a nonselective β-blocker (carvedilol) in the management of congestive heart failure. This patient population was chosen because it was likely to have the highest risk of fracture. From a total subject population of 5865, 2.0% of placebo-treated subjects experienced a fracture compared with 2.3% of those randomized to carvedilol (relative risk, 1.15; 95% CI, 0.81-1.64). Data on falls are not available. This database does not provide any evidence to support the hypothesis that β-blockers reduce fracture numbers.
Whereas the immediate stimulus to study β-blocker effects on bone has come from the work of Karsenty's group, which was focused on determining the mechanism of action of centrally administered leptin on bone, there is a significant body of earlier work that supports a role for the sympathetic nervous system in bone metabolism. Thus, others have shown the presence of β-adrenoreceptors on osteoblasts,(9,10) and Moore et al.(11) showed these to be of the β2 subtype. β-agonists have been shown to stimulate production of bone-active cytokines (e.g., interleukin-6, interleukin-11, prostaglandin E2),(12,13) PTH,(14) and RANKL(13) and to increase osteoclastogenesis.(13) Propranolol has been shown to diminish and norepinephrine to increase bone resorption in bone organ culture.(11,15) In vivo, sympathectomy produces effects on bone turnover, although the direction of this effect is inconsistent.(16–18) Furthermore, there is some evidence that propranolol increases cross-linking of type 1 collagen in other tissues, increasing its tensile strength.(19)
Despite this fairly consistent body of experimental evidence, the inconsistency of the available clinical data means that the effects of β-blockers on bone in humans remains an open question. These data do not provide support for concluding that β-blocker use is an important factor to be considered when assessing an individual's future fracture risk nor that they should be deployed, at this time, as therapeutic agents. However, the substantial body of evidence implicating the sympathetic nervous system in bone metabolism suggests that randomized controlled trials could be undertaken to test this possibility further.
This study was supported by U.S. Public Health Service Grants AG05407, AR35582, AG05394, AM35584, and AR35583 and the Health Research Council of New Zealand.
- 12000 Leptin inhibits bone formation through a hypothalamic relay: A central control of bone mass. Cell 100: 197–207., , , , , , , , ,
- 22002 Leptin regulates bone formation via the sympathetic nervous system. Cell 111: 305–317., , , , , , , ,
- 32004 Beta-adrenergic blockers reduce the risk of fracture partly by increasing bone mineral density: Geelong Osteoporosis Study. J Bone Miner Res 19: 19–24., , , , ,
- 41998 Intermittent use of nitrates increases bone mineral density—the study of osteoporotic fractures. J Bone Miner Res 13: 1755–1759., , ,
- 52003 Effects of estrogen plus progestin on risk of fracture and bone mineral density—The Women's Health Initiative randomized trial. JAMA 290: 1729–1738., , , , , , , , , , , , ,
- 62003 HMG CoA reductase inhibitors and the skeleton: A comprehensive review. Osteoporos Int 14: 273–282.
- 72000 Hydrochlorothiazide reduces loss of cortical bone in normal postmenopausal women: A randomized controlled trial. Am J Med 109: 362–370., , , , , , , ,
- 82003 In vivo stimulation of sympathetic nervous system modulates osteoblastic activity in mouse calvaria. Am J Physiol 285: E661–E667.,
- 91988 Effects of retinoic acid on alkaline phosphatase messenger ribonucleic acid, catecholamine receptors, and G proteins in ROS 17/2.8 cells. Endocrinology 122: 456–463., ,
- 101986 Dexamethasone effects on beta-adrenergic receptors and adenylate cyclase regulatory proteins Gs and Gi in ROS 17/2.8 cells. Endocrinology 118: 2510–2518.,
- 111993 Characterization of beta-adrenergic receptors on rat and human osteoblast-like cells and demonstration that beta-receptor agonists can stimulate bone resorption in organ culture. Bone Miner 23: 301–315., , , ,
- 122001 Signal transduction system for interleukin-6 and interleukin-11 synthesis stimulated by epinephrine in human osteoblasts and human osteogenic sarcoma cells. Biochem Pharmacol 61: 319–326., , ,
- 132001 Adrenergic stimulation of osteoclastogenesis mediated by expression of osteoclast differentiation factor in MC3T3-E1 osteoblast-like cells. Biochem Pharmacol 61: 579–586., , ,
- 142003 Beta 1-adrenergic blockade augments pulsatile PTH secretion in humans. J Am Soc Nephrol 14: 3245., , , , , ,
- 151979 Inhibition of bone resorption in tissue culture by membrane-stabilizing drugs. Endocrinology 104: 1644–1648., ,
- 161999 Chemical sympathectomy impairs bone resorption in rats: A role for the sympathetic system on bone metabolism. Bone 25: 545–551., , ,
- 171991 Effects of neonatal sympathectomy and capsaicin treatment on bone remodeling in rats. Neuroscience 44: 747–755., ,
- 181987 Effect of surgical sympathectomy on bone remodeling at rat incisor and molar root sockets. Anat Rec 219: 32–38., ,
- 191991 Effects of propranolol on bone metabolism in the rat. J Orthoped Res 9: 869–875., , , ,