Use of β-Blockers and Risk of Fractures FREE

Studies have suggested that the sympathetic nervous system has a catabolic effect on bones.^1- 4 In vitro data show that adrenergic agonists stimulate bone resorption in organ culture of mouse calvariae.⁴ Chemical sympathectomy with guanethidine, a sympathetic neurotoxic agent, impairs bone resorption by inhibiting preosteoclast differentiation and disturbing osteoclast activation in adult rats.¹ Additionally, the β-blocker propranolol increased bone formation in ovariectomized female rats.³ These results suggest that β-blockers may overcome loss of bone mass in postmenopausal women.⁵ Indeed, in a recent observational study including 569 fracture case and 775 control patients, use of β-blockers was associated with a higher bone mineral density at the hip and forearm in women aged 50 years or older, and β-blocker use was associated with a 30% decrease in fracture risk.⁶

Thiazide diuretics are thought to be protective against bone loss by reducing urinary calcium excretion. Several epidemiologic studies suggest that exposure to thiazides is associated with a reduced fracture risk.^7- 11

To our knowledge, no studies have evaluated the association of β-blocker use and bone fractures in men and young women or the association of combined use of β-blockers with thiazide diuretics with the risk of fractures. In this large population-based study we explored the association between β-blocker use—with or without concomitant use of thiazides—and fracture risk in men and women aged 30 to 79 years.

Data were derived from the UK-based General Practice Research Database (GPRD), which has been described in detail elsewhere.^12- 13 More than 3 million people in the United Kingdom are enrolled with selected general practitioners who use office computers and have agreed to provide data for research purposes. General practitioners have been trained to record medical information including demographic data, medical diagnoses, details of hospital stays, and deaths using a standard, anonymous form. The physicians generate prescriptions directly with the computer; this information is automatically transcribed into the computer record. A modification of the Oxford Medical Information System classification (similar to the International Classification of Diseases, Eighth Revision) is used to enter medical diagnoses, and a coded drug dictionary based on the UK Prescription Pricing Authority dictionary is used for recording prescriptions. The recorded information on drug exposure and diagnoses has been validated and proven to be of high quality.^14- 15

The GPRD has been the source of several observational studies, including research on antihypertensive drugs^16- 17 as well as fractures.¹⁸ Based on previous record review by ourselves¹⁸ and by others,¹⁹ the validity of fracture diagnoses in the GPRD is high, with a confirmed proportion of at least 90% after comparing computer-recorded diagnoses with hospital discharge letters and/or questionnaire information provided by general practitioners. The study was approved by the Scientific and Ethical Advisory Group of the GPRD.

We identified all patients with an incident diagnosis of a fracture recorded in the computerized medical record between January 1993 and December 1999. More recent data were not included because of coding changes in the database. Cases were identified without any exposure information. Potential cases had to be aged 30 to 79 years at the date of the fracture (ie, index date), and they had to have been recorded in the database for at least 3 years. In a previous study, we reviewed a random sample of 200 records of patients with a fracture diagnosis and found that the proportion of individuals with fractures due to severe polytrauma (eg, car crashes) was relatively low (0.5%-1%).¹⁸ Thus, we decided to include all cases identified on computer in this study. Case patients with a medical disorder known to substantially affect bone metabolism (ie, a history of osteoporosis, osteomalacia, Paget disease, cancer [excluding nonmelanoma skin cancer], or alcoholism) as well as patients who used bisphosphonates (as an indicator of osteoporosis or bone metastases) prior to the index date were excluded from the analysis.

Controls were persons without a fracture selected from the base population. We randomly selected up to 4 controls per case matched on age (±2 years), sex, general practice attended, calendar time (by using the same index date as for cases), and years of prior history in the GPRD (matched on the year when the individual's record in the database began, ±1 year). Furthermore, controls had to be alive on the index date. The same exclusion criteria were applied to controls as to case patients.

We ascertained exposure to β-blockers (ie, acebutolol, atenolol, betaxolol, bisoprolol, carvedilol, celiprolol, labetalol, metoprolol, nadolol, nebivolol, oxprenolol, pindolol, propranolol, sotalol, timolol) and thiazide or thiazide-like diuretics (hydrochlorothiazide, chlorthalidone, bendrofluazide, benzthiazide, chlorothiazide, cyclopenthiazide, polythiazide, mefruside, xipamide, metolazone, hydroflumethiazide, clopamide, indapamide) for all cases and controls prior to the index date. Exposure to study drugs as well as for a variety of other drugs was assessed as follows. Patients were classified as current users if their last prescription for a study drug was recorded 1 to 59 days prior to the index date, as recent users if the last prescription for a study drug was recorded 60 to 119 days prior to the index date, and as past users if the last prescription was recorded at least 120 days prior to the index date. All others were considered nonusers.

We further evaluated the number of computer-recorded prescriptions prior to the index date as a proxy for duration of exposure. For the drugs of interest, one prescription corresponded to approximately 1 to 3 months of therapy. For the main analysis, we created mutually exclusive categories of users of β-blockers only, thiazides only, or a combination thereof. We adjusted this analysis for use of other antihypertensive drugs as well as for a variety of other drugs potentially affecting fracture risk. Users of both β-blockers and thiazides were individuals who either used a fixed combination or had separate prescriptions for both β-blockers and thiazide diuretics. They were classified as current users if both β-blockers and thiazides were prescribed within 60 days prior to the index date. If they currently used only a β-blocker and previously used thiazides, or vice versa, they were categorized in a separate group called "other combinations."

We conducted conditional logistic regression analyses using SAS version 8.1 (SAS Institute Inc, Cary, NC). Risk estimates are presented as odds ratios (ORs) with 95% confidence intervals (CIs). P values are 2-sided and were considered statistically significant if less than .05.

Odds ratios were adjusted for the potential confounders age, sex, practice, calendar time, and years of recorded history in the GPRD prior to the index date (by matching), as well as smoking status (none, current, past, unknown), body mass index (BMI; <25, 25-29.9, ≥30 measured as weight in kilograms divided by the square of height in meters), number of practice visits prior to the index date (<5, 5-9, 10-29, 30-59, ≥60), medical conditions associated with an increased fracture risk (including chronic renal failure, hyperthyroidism, hyperparathyroidism, malnutrition, malabsorption, diabetes), drug exposure to other antihypertensive drugs (ie, calcium channel blockers, angiotensin-converting enzyme [ACE] inhibitors, angiotensin II receptor blockers, potassium-sparing diuretics), benzodiazepines, antipsychotics, antidepressants, corticosteroids, antiepileptic drugs, hormone therapy, calcium, fluoride, vitamin D preparations, and statins. Missing values for smoking and BMI were treated as a separate group in the analysis.

We identified 30 601 cases with a fracture and 120 819 matched controls. The most frequent fractures were of the hand/lower arm (n = 12 837 [42.0%]) and of the foot (n = 4627 [15.1%]). The age and sex distribution of cases and controls, as well as the distribution of smoking status, BMI, and number of practice visits, are displayed in Table 1. The association between fracture risk overall and high BMI was close to 1.0; however, stratification by fracture site yielded adjusted ORs of 0.42 (95% CI, 0.34-0.51; P<.001) for femur fractures and 0.65 (95% CI, 0.48-0.89; P = .007) for vertebral fractures for patients with a BMI of at least 30 compared with the reference group of patients with a BMI less than 25. Fractures were evenly distributed across age categories but occurred more often in women than in men (60.3% vs 39.7%).

To explore the association between use of β-blockers and/or thiazides and fracture risk, we first grouped current users into fine exposure strata (1-2, 3-4, 5-9, 10-14 prescriptions, etc). Based on these findings, we pooled them into broader categories of 1 to 2, 3 to 19, or 20 or more prescriptions. Compared with the reference group of nonusers of β-blockers or thiazides, current use of 1 to 2 prescriptions as well as recent or past use of β-blockers only, thiazides only, or a combination thereof were all associated with risk estimates close to 1.0, adjusted for smoking, BMI, number of practice visits prior to the index date, use of calcium channel blockers, ACE inhibitors, antipsychotics, antidepressants, statins, antiepileptics, benzodiazepines, corticosteroids, and estrogens (Table 2).

Current users of β-blockers only with 3 to 19 recorded prescriptions had an adjusted OR of fracture of 0.63 (95% CI, 0.55-0.73; P<.001), and for those with 20 or more prescriptions the OR was 0.83 (95% CI, 0.76-0.91; P<.001). These 2 risk estimates differed significantly, ie, the risk estimate was lower for users of 3 to 19 prescriptions than for users of 20 or more prescriptions. This observed difference occurred mainly due to a sex effect in patients who had 20 or more prescriptions. While the adjusted ORs for men (0.59 [95% CI, 0.47-0.74]; P<.001) and women (0.67 [95% CI, 0.56-0.81]; P<.001) with 3 to 19 β-blocker prescriptions were similar, the adjusted ORs for men or women who had 20 or more β-blocker prescriptions recorded were 0.69 (95% CI, 0.59-0.81; P<.001) and 0.92 (95% CI, 0.83-1.02; P = .09), respectively, indicating a possible sex-related difference in longer-term use of β-blockers. Further stratification of women with 20 or more β-blocker prescriptions by age yielded ORs of 1.03 (95% CI, 0.70-1.51; P = .89) for women younger than 50 years and 0.90 (95% CI, 0.81-1.00; P = .05) for women 50 years or older. As opposed to use of β-blockers only, the risk estimates for users of thiazides only as well as for β-blockers plus thiazides combined did not differ materially between men and women; nor did they differ substantially between those with 3 to 19 vs 20 or more prescriptions (data not shown).

Since the fracture risk overall (ie, men and women combined) was significantly less than 1.0 for users of both 3 to 19 and 20 or more β-blocker prescriptions, we decided to pool patients with 3 to 19 and those with 20 or more prescriptions into one group of users of 3 or more prescriptions to gain power for stratified analyses. Compared with the reference group of nonusers of β-blockers or thiazides, the adjusted OR for current users of 3 or more prescriptions of β-blockers only was 0.77 (95% CI, 0.72-0.83; P<.001); for current users of 3 or more prescriptions of thiazides only, 0.80 (95% CI, 0.74-0.86; P<.001); and for current users of 3 or more prescriptions of a combination of both β-blockers and thiazides, 0.71 (95% CI, 0.64-0.79; P<.001). These results stratified by sex, age, and fracture site are shown in Table 3.

Among current users of 3 or more β-blocker prescriptions, we also stratified by type of β-blocker used prior to the index date to explore whether receptor selectivity or physico-chemical properties (ie, hydrophilicity vs lipophilicity) may affect bone fracture risk. The risk estimates were similar for atenolol (adjusted OR, 0.78; 95% CI, 0.71-0.86; P<.001), a hydrophilic β₁-selective β-blocker, and for propranolol (adjusted OR, 0.79; 95% CI, 0.68-0.92; P = .002), a lipophilic unselective β-blocker. The adjusted ORs for patients who were current users of 3 or more prescriptions of β-blockers with intrinsic sympathomimetic activity (ISA) (ie, acebutolol, celiprolol, oxprenolol, and pindolol), and for those without ISA were 0.81 (95% CI, 0.58-1.13; P = .21) and 0.78 (95% CI, 0.72-0.84; P<.001), respectively.

Information on smoking history was unavailable for 15% and BMI for 30% of study patients (Table 1). Patients with missing information on BMI and/or smoking differed slightly from those with recorded information (eg, those with missing BMI and smoking information were more likely to be men <40 years or ≥70 years and had fewer practice visits [data available from the author on request]). We conducted an additional analysis in which we included only patients with known BMI and smoking status; the ORs for patients with 3 to 19 or 20 or more β-blocker prescriptions were 0.60 (95% CI, 0.50-0.71; P<.001) and 0.83 (95% CI, 0.75-0.93; P<.001), respectively.

Use of angiotensin II receptor blockers, loop diuretics, or potassium-sparing diuretics was not associated with an altered fracture risk, and adjusting the multivariate model for these drugs did not change the risk estimates of the association between β-blocker use and fracture risk (data available on request). On the other hand, in the final multivariate analysis we found a decreased fracture risk for longer-term current users (≥20 prescriptions) of ACE inhibitors (adjusted OR, 0.81; 95% CI, 0.73-0.89; P<.001).

Additionally, we found altered fracture risks associated with drug classes other than antihypertensives, including oral corticosteroids (adjusted OR in current long-term users [≥20 prescriptions], 1.15 [95% CI, 1.04-1.29]; P = .006), benzodiazepines (adjusted OR in current short-term users [1-2 prescriptions], 1.62 [95% CI, 1.35-1.93]; P<.001), antipsychotic drugs (adjusted OR in current short-term users [1-2 prescriptions], 1.15 [95% CI, 0.95-1.38]; P = .16), antiepileptic drugs (adjusted OR in current long-term users [≥20 prescriptions], 1.66 [95% CI, 1.49-1.85]; P<.001), estrogens (adjusted OR in long-term users [≥20 prescriptions], 0.62 [95% CI, 0.55-0.70]; P<.001), and statins (adjusted OR in current long-term users [≥20 prescriptions], 0.84 [95% CI, 0.67-1.04]; P = .11), with varying risk estimates for statin use depending on the fracture site (femur: adjusted OR, 0.39 [95% CI, 0.16-0.94]; P = .04).

The findings of this large population-based case-control analysis provide further evidence that current use of β-blockers is associated with a statistically significantly decreased fracture risk in both men and women. While short exposure (ie, 1-2 prescriptions, corresponding to at least 1 month and at most 6 months of therapy) did not materially alter the fracture risk, the risk reduction became apparent in users of 3 or more prescriptions. As stated in the results, the relative risk was lower for current users of 3 to 19 β-blocker prescriptions (OR, 0.63) than for users of 20 or more prescriptions (OR, 0.83). This is not necessarily consistent with a linear duration effect, but it may also be a chance finding. To explore whether the OR tended to approach 1.0 with increasing exposure duration, we stratified the 20 or more prescription group into 20 to 29, 30 to 39, and 40 or more prescriptions; the risks remained stable and fluctuated around 0.8 for all longer-term use strata.

The association between fracture risk and longer-term use (ie, ≥20 prescriptions) of β-blockers was weaker in women (OR, 0.92) than in men (OR, 0.69) in our study population. Women tended to have more practice visits in our study population than men did (19.7% of men and 31.3% of women saw a general practitioner 30-59 times, and 6.9% of men and 15.7% of women saw the general practitioner ≥60 times prior to the index date). Women on average also took more medications than men did. However, the final analysis was adjusted for number of practice visits, which serves as a proxy for general health status, medical attention, and number of drugs used. Thus, the differences in drug use between men and women may not explain the differences in effect size with longer-term β-blocker use. This difference should be further explored in future studies.

There is substantial evidence that the nervous system participates in the regulation of bone remodeling. Functional adrenergic receptors are present in osteoblasts as well as osteoclasts,²⁰ and sympathetic nerve fibers have been demonstrated in bone tissue.^2- 3 These findings support the concept that the nervous system is one of the regulators of bone metabolism. The action may be exerted centrally through the hypothalamus and peripherally by the release of neurotransmitters.³ It appears that β-adrenergic receptor stimulation may have anabolic as well as catabolic actions on bone cells mediated by α- and β-adrenergic receptors, respectively.² Effects on osteoblast proliferation, osteoblast maturation, and osteoclast development have been demonstrated.^2,21 In contrast, sympathetic system inactivation in rats results in a significant decrease in osteoclast number and osteoclast activity.²² Propranolol has been shown to increase bone mass in mice, secondary to an increase in both the bone formation rate and the osteoblast number.³ Furthermore, there is also evidence that β-adrenergic antagonists increase bone strength in rat models.²³ The adrenergic regulatory pathway may also operate in humans, although to date there are no investigations directly addressing this issue.

In addition, fracture risk estimates for users of thiazide diuretics were similar to those found in previous observational studies.^7- 11 Thiazide diuretics are widely used as antihypertensive agents, they are inexpensive, and they have few serious adverse effects.²⁴ In addition to their effect on the cardiovascular system they also affect calcium homeostasis and bone metabolism. The mechanisms by which thiazides influence bone metabolism are not fully elucidated. Thiazide treatment results in an increase in tubular calcium reabsorption and thereby a decrease in urinary calcium excretion, leading to a positive calcium balance.²⁵ Additional mechanisms by which thiazides may decrease bone resorption are changes in acid-base balance by inducing a metabolic alkalosis, or a direct action on osteoclasts by inhibiting carbonic anhydrase, one of the enzymes necessary for acid production by osteoclasts.^26- 27 Furthermore, they may also act directly on osteoblasts.^28- 29 In several controlled studies on the effect of thiazides on bone metabolism, a decrease in bone turnover has been reported.^30- 32 Randomized clinical trials have shown an increase in bone mineral density or a decrease in bone loss in patients treated with thiazides.^10,30- 32

Various drugs with effects on the central nervous system are known to increase the risk of falls and thereby the fracture risk (eg, hypnotics, antidepressants, or antipsychotics). The risk is particularly high in the early phase of a therapy when patients have not yet adapted to centrally suppressing effects that may alter coordination and perception. The findings in our analysis are consistent with previous reports on fracture risks associated with use of centrally acting drugs.^33- 34 We also found a slightly decreased fracture risk associated with current use of statins, but not with other lipid-lowering drugs, as shown in a recent meta-analysis.³⁵ We found no evidence for an altered fracture risk for users of calcium channel blockers, while there was some suggestion of a possibly decreased fracture risk associated with longer-term use of ACE inhibitors. Of interest, in an open prospective study including 134 patients with low to moderate hypertension and stable bone mineral density, calcium and 25-hydroxyvitamin D plasma levels were both increased in patients treated with the ACE inhibitor quinapril, and calcium levels were increased in a group treated with the combination of quinapril and hydrochlorothiazide.³⁶

We were unable to evaluate the association of physical activity, diet, or socioeconomic status with fracture risk because this information is not available in the GPRD on an individual patient level. Additionally, although fracture diagnoses have been validated in previous studies and more than 96% of diagnosed vertebral fractures were confirmed radiographically,¹⁹ likely not all vertebral fractures are captured in the GPRD because a certain proportion of vertebral fractures are asymptomatic and remain undetected. Moreover, the number of vertebral fractures in this study was too low for a meaningful interpretation.

Furthermore, we cannot exclude with certainty that unknown risk factors may have biased or confounded our analysis. However, we controlled the analysis for a variety of medical conditions such as renal failure, hyperthyroidism, hyperparathyroidism, malnutrition, malabsorption, or diabetes mellitus. Adjusting the analysis for these diseases left the results virtually unchanged, and we decided to not include them into the final model.

We also adjusted for drugs known to be associated with altered fracture risks such as benzodiazepines, antipsychotics, antidepressants, antiepileptics, statins, corticosteroids, and estrogens. Even though these covariates did not confound the main association of interest, we included these drugs in the final model because they were independently associated with altered fracture risks. We also adjusted the final model for number of practice visits prior to the index date to have a crude marker for general health status in the model, and we controlled for smoking status and BMI. Smoking history was unavailable for 15% and BMI for 30% of study patients; however, the association between use of β-blockers and fracture risk was virtually identical in an analysis restricted to patients with known values for both BMI and smoking status and in the entire data set.

In summary, the present large case-control analysis provides evidence that use of β-blockers—alone or in combination with thiazide diuretics—is associated with a significantly decreased fracture risk. The association in long-term users was weaker in women than in men in the current study population. Although additional observational studies and controlled trials are needed to confirm these potentially important findings, many elderly patients with hypertension are at risk of developing osteoporosis, and they may potentially profit from positive effects of the relatively inexpensive β-blockers and thiazide diuretics on fracture risk.