Risk of hip fracture among older people using antihypertensive drugs: a nationwide cohort study

The nationwide cohort design is suitable to compare people exposed and non-exposed to antihypertensive drugs with regard to the relatively infrequent outcome, hip fracture, without being prone to selection and recall bias. The health registries provided us a unique opportunity to link complete data on all antihypertensive drugs purchased by a large unselected community-dwelling older population with all primary hip fractures registered in Norway, and the 6-year follow-up period yielded a high number of cases.

However, the databases have some limitations. The NorPD lacks individual information on medications dispensed to people staying in nursing homes (predominantly long-term care) and hospitals (mostly short stays), leading to systematic misclassification as drug non-users. Because frail old people in nursing homes are particularly prone to both treatment with antihypertensive drugs and hip fracture [22, 23], bias from immeasurable exposure time probably caused underestimation of associations among exposed people. The Norwegian Hip Fracture Registry comprised about 90 % of all hip fracture operations in Norway [24], with somewhat lower completeness during the first years. Unfortunately, clinical information regarding diagnoses, BMD, functional level, socioeconomic factors and life style was not available from the included or any other registry. This hampered adjustments for potentially confounding factors such as fall-risk-related comorbidities (FRICs), i.e. heart failure, ischemic heart disease, chronic obstructive lung disease, dementia, depression, Parkinson’s disease and stroke. Many older people treated with antihypertensive drugs probably use other drugs concomitantly, e.g. fall-risk-increasing drugs (FRIDs) such as other cardiovascular drugs, psychotropics, opioids and systemic steroids. However, with the time-varying exposure used in the analysis (SIR), matching exposure periods for other drugs was not possible. We consider the time-varying exposure a major strength of our study, as the alternative, fixed exposure, would have led to extensive and unmeasurable misclassification yielding unreliable results. In Norway, antihypertensive drugs are mainly prescribed for hypertension (88–99 % of overall drug volume); however about 40 % of ACE inhibitors (plain products) and beta-blockers are prescribed for conditions such as heart failure and ischemic heart disease [8], and loop diuretics are mainly used for fluid retention. Because many drugs have several different indications, we decided not to use drugs as proxies of comorbidities. The Norwegian Prescription Database includes reimbursement codes for drugs prescribed for chronic conditions, linked to ICD-10 diagnoses. We did not consider this information because reimbursement codes were neither validated nor complete (selected drugs only) and not available before 2009. Previous studies generally revealed protective associations between hip fracture and antihypertensive drugs to remain when adjusting for comorbidity, socioeconomic factors, life style and concomitant drug use [25, 26]; loop diuretics, however, were sensitive for adjustment [26].

Because antihypertensive drugs are predominantly used on a daily basis, we considered the number of days corresponding to the number of DDDs dispensed the best proxy for the number of drug-exposed person-days. A study in Germany observed the actually prescribed mean daily dose (PDD) to be nearly 2.0 DDD for ACE inhibitors and angiotensin II receptor blockers and close to 1.0 DDD for beta-blockers, calcium-channel blockers and thiazides [27]. Consequently, we may have overestimated the exposed person-time, and thus the strength of the associations between ACE inhibitors and angiotensin II receptor blockers, and the risk of hip fractures.

Health care in Norway is public and tax-financed; very low out-of-pocket money and good reimbursement systems may explain the greater use of newer and more expensive antihypertensive drugs in Norway compared to other countries in Europe and North America [9]. One could suspect that the indication and frequency of treatment with antihypertensive drugs differs in a public health care setting and a setting with insurance based health care.

Different mechanisms have been postulated for effects of antihypertensive drugs on fracture risk depending on the duration of drug use; through hypotension and falls in recently started treatment and through changes in bone metabolism and bone strength in long-term treatment [26]. In addition to direct (falls) and indirect (BMD) drug effects, treatment of the underlying cardiovascular disease may possible reverse its negative effects on bone quality and contribute to reduction in fracture risk.

We found the risk of hip fracture to be decreased with angiotensin II receptor blockers (plain and combination products); use of ACE inhibitors was associated with increased risk among users aged <80 years and decreased among those 80 years and older. The human studies conducted previously have provided conflicting evidence regarding the effects of ACE inhibitors on bone metabolism and fracture risk, whereas angiotensin II receptor blockers appear to be osteoprotective. A population-based case–control study in Denmark reported a 14 % reduction in hip fracture risk (adjusted OR 0.86, 95 % CI 0.80-0.92) in users of ACE inhibitor compared with nonusers; the crude OR was not significant. There were no major differences according to sex, age and drug dosage [25]. Another case–control study in the United Kingdom revealed a decreased risk of any fracture for longer-term current users (≥20 prescriptions) of ACE inhibitors (adjusted OR, 0.81; 95 % CI 0.73–0.89), though, subanalysis for hip fracture was not conducted. No difference in fracture risk was found with use of angiotensin II receptor blockers [28]. In a large cohort study among low-income older people the United States, new use of angiotensin II receptor blockers was associated with a reduction in hip fracture risk (adjusted Hazard Ratio, HR 0.77, 95 % CI 0.64–0.93) compared with calcium channel blockers (reference exposure), whereas no trend was observed for ACE inhibitors (HR 0.95, 95 % CI 0.84–1.07). Of note, nonusers were not considered, and wash out period (30 days) and follow-up period (median 70 days) were short [26]. In a population-based cohort study in Sweden, use of renin-angiotensin-aldosterone active agents was not associated with altered risk of hip fracture (adjusted OR 0.93, 95 % CI 0.79–1.09); however, separate analyses for ACE inhibitors and angiotensin II receptor blockers were not conducted [29]. Another population-based cohort study in Canada failed to demonstrate significant differences in hip fracture risk in hypertensive older people treated with angiotensin II receptor blockers compared to ACE inhibitors (adjusted HR 0.99, 95 % CI 0.78–1.25); nonusers were not considered [30]. The Hypertension in the Very Elderly Trial (HYVET), a randomized trial with hip fracture as a secondary outcome measure, revealed that treatment with a thiazide-like diuretic ± ACE inhibitor may reduce the risk of hip fracture [31]. To the best of our knowledge, previous research has not included combination products. We found protective associations of ACE inhibitor/thiazide and angiotensin II receptor blocker/thiazide with hip fracture to be stronger than for the individual drugs involved. This may possibly be due to several interacting, beneficial effects on underlying cardiovascular disease and bone strength.

A cross-sectional study of hypertensive Chinese men and women reported associations between ACE inhibitor use and higher BMD at the hip, after adjusting for diagnoses, medications and lifestyle [32]. This finding was not confirmed by a more recent prospective cohort study in the United States; ACE inhibitors were associated with hip bone loss whereas angiotensin II receptor blockers were not [33]. An open randomized trial showed beneficial effects of ACE inhibitors ± thiazide in hypertensive people, i.e. BMD loss was avoided and metabolic profile was improved. Subgroup analysis for ACE polymorphism revealed that beneficial BMD response was confined to women with the DD polymorphism [34]. Another open randomized study demonstrated that angiotensin II receptor blockers inhibited bone loss and increased BMD at the hip in older bed-ridden women, as compared to calcium channel blockers [35]. In vitro, angiotensin I stimulates bone resorption in co-cultures of osteoblast and osteoclasts, and this action is inhibited by ACE inhibitors [36]. Angiotensin II accelerates bone resorption by activating osteoclasts, whereas these effects are blocked by angiotensin II receptor blockers [37]. Hence, animal studies suggest that renin-angiotensin-aldosterone active agents inhibit bone resorption, rather than stimulate bone formation [37, 38].

Our finding of a reduced fracture risk among users of calcium channel blockers is partly in line with the above-mentioned epidemiologic studies conducted in Denmark and Sweden. Rejnmark et al. reported a small reduction in hip fracture risk by 7 % (adjusted OR 0.93, 95 % CI 0.87–0.99) [25] while Thorell et al. found a reduction by 18 % (crude OR 0.82, 95 % CI 0.68–0.98); the differences in the latter study were not maintained, however, after adjustment for age, sex and comorbidity [29]. Although in vitro studies suggest that calcium-channel blockers may inhibit osteoclast function [38], previous research has found no association with hip BMD [32].

Our findings of a reduced hip fracture risk among thiazide-users align with a Cochrane review including 6 cohort and 15 case–control studies; compared with nonusers, the fracture risk among thiazide users was decreased by 24 %, relative risk (RR) 0.76 (95 % CI 0.64–0.78) [39]. Randomized controlled trials have revealed preservation of cortical bone in older women and men using thiazides, however, fracture was not an outcome measure [40, 41]. Beneficial effects on BMD are probably caused by increased renal calcium absorption, suppressed parathyroid hormone secretion and enhanced calcium uptake in the intestine, leading to a positive calcium balance. Further, thiazides directly affect bone by stimulating osteoblast differentiation [15].

The results of our study suggest that risk of hip fracture with loop diuretics changes with the users’ age; fracture risk being increased among users aged <80 years, and decreased among those 80 years and older. A meta-analysis of 4 cohort and 9 case–control studies reported a 14 % increase in overall risk of hip fracture among loop diuretic users compared with nonusers, RR 1.14 (95 % CI 1.08–1.19) [42]. However, most included studies were adjusted for age, not allowing for comparison between different age groups. Some studies have reported sensitivity of risk estimates to adjustment for other confounders (e.g. prior fracture, comorbidity or use of other drugs), revealing smaller RRs after adjustment [26, 43]. Cohort studies and a randomized controlled trial have reported loss of total hip BMD among older women and men using loop diuretics compared with nonusers [44‐46]. Interestingly, a recent study indicated a clinical association of hyponatremia during loop diuretic use and an increased risk of osteoporotic fractures, possible due to exertion of loss of bone sodium and calcium [47].

We found the risk of hip fracture to be decreased with use of beta-blockers, in accordance with two meta-analyses. Yang et al. (6 cohort and 7 case–control studies) reported reduction in risk by 17 % (RR 0.83, 95 % CI 0.70–0.92) [48], and Toulis et al. (7 cohort and 9 case–control studies) reduction by 14 % in women (pooled effect size (ES) 0.86, 95 % CI 0.80–0.91) and 20 % in men (ES 0.80, 95 % CI 0.71–0.90) [49]. Although observational studies have associated use of beta-blockers with a higher BMD at the hip [50‐52], randomized trials are lacking. Several studies suggest the reduction in hip fracture risk to be associated with β1-selective agents rather than nonselective beta-blockers [49, 53]. The underlying mechanisms are not yet clear. Increased sympathetic nervous activity causes an increase in bone resorption and a decrease in bone formation through stimulation of osteoclasts and inhibition of osteoblasts. In vitro studies suggest that beta-blockers enhance bone formation through blocking input from the sympathetic nervous system [54].

We found increased risk of hip fracture during the first 2 weeks of treatment with loop diuretics; similar findings have been reported by Berry et al. for the initial 3 weeks of treatment with thiazides and loop diuretics [12], and by Butt et al. for the 45 days immediate after a first prescription of various antihypertensive drugs (incidence rate ratio 1.43, 95 % CI 1.19–1.72) [11]. Although the numbers of hip fractures in these studies were relatively small and the findings must be interpreted with caution, the results are interesting because they are opposite to those with long-term drug use. Whereas initiating antihypertensive drugs is known as risk factor for falls in older people [11], two meta-analyses revealed no significant associations of falls with current use of antihypertensive drugs [13, 14]. On the other hand, a recent population-based study found a dose–response relationship between drug dose and falls [55].

We found a lower risk of hip fracture among exposed men than among exposed women. Rejnmark et al. reported similar trends for sex with regard to beta-blockers, ACE inhibitors and calcium channel blockers [25]. Sex hormones interfere with bone metabolism, posing postmenopausal women at a much higher risk of hip fracture than men; most antihypertensive drugs appear to be beneficial for bone health in both sexes, however, risk differences are not eliminated.

Generally, we found the protective associations to be most evident within the youngest cohorts. We probably underestimated the risk of hip fracture among the oldest old because the lack of clinical information (confounding factors could not be adjusted for) and the systematic misclassification of the highly exposed [22] nursing home patients as drug non-users. Amplification of the effects of these drugs with advancing age may suggest the maintenance of protective drug effects despite other risk factors such as frailty, osteoporosis, comorbidity and concomitant drug use. On the other hand, cessation of drug treatment in the oldest age groups could be associated with general clinical deterioration (the “healthy user effect”), explaining the SIRs below 1. The increased hip fracture risk in people younger than 80 years using loop diuretics and plain ACE inhibitors, however, may possibly be due to the “more-ill-user effect” as these drugs are frequently issued to relatively frail people with heart failure.