Atrial fibrillation (AF) is the most common cardiac arrhythmia. Its overall incidence is less than 1% in individuals younger than 60 years of age, increasing up to 9% at age 801 and affecting approximately 2.3 million individuals in the United States. It may occur in isolation or in association with other clinical conditions, including hypertension, cardiac failure, myocardial ischemia or infarction, mitral stenosis, thyrotoxicosis, and alcohol abuse, as well as following cardiac surgery.2 AF is a major risk factor for stroke, thromboembolism, and cardiac failure,3, 4 resulting in a twofold increase in overall mortality.2
Bisphosphonates are front-line agents in the management of osteoporosis, malignant disease affecting the skeleton, and Paget's disease of bone. There were more than 30 million prescriptions for oral bisphosphonates in the United States in 2006 alone.5 However, a recent study associated zoledronic acid treatment with increased risk of AF, reported as a serious adverse event, in postmenopausal women with osteoporosis,6 and subsequently, pamidronate treatment was reported to be associated with a higher risk of AF than zoledronic acid in a comparator study in patients with myeloma or breast cancer.7 These observations have raised concerns about cardiac complications with the use of bisphosphonates and have led to regulatory action to include AF as a possible side effect in the product information for both zoledronic acid and pamidronate. This article reviews the evidence for and against an association between bisphosphonate therapy and AF and speculates on potential pathophysiologic mechanisms.
Bisphosphonates and Atrial Fibrillation: Clinical Studies
The double-blind, placebo-controlled three-year HORIZON Pivotal Fracture Trial was designed to examine the antifracture efficacy of yearly intravenous administration of the aminobisphosphonate zoledronic acid6 in 7765 postmenopausal women with osteoporosis. After 3 years, significant risk reductions in vertebral, nonvertebral, and hip fracture of 70%, 25%, and 41%, respectively, were shown in women treated with zoledronic acid; however, the number of women with arrhythmia, reported as an adverse effect, was significantly higher in the treatment group than in the placebo group (6.9% versus 5.3%, p = .003). Furthermore, the risk of AF reported as a serious adverse event (defined as fatal, life threatening, or resulting in hospitalization or disability) also was significantly higher in the treatment than in the placebo group [50 women (1.3%) versus 20 women (0.5%); p < .001], a difference that remained substantially unchanged after adjudication. In 47 of the 50 women treated with zoledronic acid in whom AF was reported as a serious adverse event, symptoms developed more than 30 days after the infusion, excluding the possibility that AF was related to early, transient hypocalcemia or another electrolyte disturbance. In a subgroup of 599 patients, 12-lead electrocardiographic (ECG) examination was performed before and 9 to 11 days after the third and final infusion. The selection criteria for the subgroup, other than the exclusion of women who were taking medications that cause prolongation of the QT interval, and their baseline demographic variables were not provided. In this subgroup, the prevalence of AF was balanced between women receiving zoledronic acid and placebo (2.1% versus 2.8%). In the same edition of the New England Journal of Medicine in which the HORIZON study was published, Cummings and colleagues reported a trend toward increased risk of AF in a reanalysis of data from the Fracture Intervention Trial (FIT). In the 6459 postmenopausal women in this study, AF was reported as a serious adverse event in 47 (1.5%) of the women treated with oral alendronate and 31 (1.0%) of the women in the placebo group.8
Subsequent studies have produced conflicting results. The prospective randomized, double-blind, placebo-controlled HORIZON Recurrent Fracture Trial9 examined the effects of once-yearly intravenous zoledronic acid in 1065 patients (817 females and 248 males) and 1062 matched placebo-treated control individuals over a median follow-up period of 1.9 years. Patients recruited into this trial had previously suffered a hip fracture and were older than patients in the HORIZON and FIT studies, indicating greater frailty and a potentially higher risk of drug-induced arrhythmogenic effects. However, no excess of cardiac arrhythmias, including AF, was demonstrated in patients treated with zoledronic acid; AF was reported as an adverse event in 2.8% and 2.6% and as a serious adverse event in 1.1% and 1.3% of patients in the treatment and placebo groups, respectively.
Analysis of adverse cardiovascular events in approximately 15,000 postmenopausal women in phase 3 clinical trials of risedronate demonstrated a similar risk of AF classified as either an adverse event or a serious adverse event in treatment and placebo groups.10 In a population-based case-control study,11 Danish medical databases were used to compare the use of bisphosphonates in 13,586 women with AF and atrial flutter and 68,054 population-based control individuals. All had complete hospital and prescription histories. Women with AF and atrial flutter had similar frequencies of etidronate and alendronate use as the control individuals, suggesting no association between bisphosphonate use and atrial arrhythmias. However, the power of the study to demonstrate such an association may have been limited by the relatively small number of women who were prescribed bisphosphonates, only 3.2% (435) of the cases and 2.9% (1958) of the population control individuals being current bisphosphonate users. In addition, it is unclear whether all the cases of AF in this study were truly incident. In a larger study that also used Danish registry data, the risk of AF was compared in 15,795 patients starting treatment with an oral bisphosphonate (alendronate, 62.2%; etidronate, 30.8%; other, 7%) following a fracture and 31,590 gender- and age-matched fracture patients not exposed to bisphosphonates.12 Although the risk of AF was significantly greater in bisphosphonate-treated patients [age- and gender-adjusted hazard ratio of 1.29 with a 95% confidence interval (CI) of 1.17–1.41], this effect was reduced by adjustment for comedications and comorbidities. The latter observation, together with the inverse association demonstrated in this study between AF and adherence to bisphosphonate therapy, led the authors to conclude that the higher risk of AF in bisphosphonate-treated patients may be attributable to an increased background risk of cardiovascular events when compared with nonusers.
A smaller population-based, retrospective case-control study in Washington State in the United States investigated the possible association between use of alendronate (ever) and AF in 719 postmenopausal women with confirmed incident AF and 966 women without AF.13 More women with incident AF than control individuals had a history of ever use of alendronate; however, only a small number of women in the study had used alendronate (47 women with AF and 40 in the control group). Furthermore, the control individuals were younger and had a lower prevalence of risk factors for AF, such as diabetes mellitus, angina, myocardial infarction, valvular cardiac disease, and congestive cardiac failure. In addition, the higher risk of AF in past rather than current bisphosphonate users argues against a causal association. Interestingly, higher AF risks were found in alendronate users currently taking statins compared with those who were not (p = .02).
The U.S. Food and Drug Administration (FDA) has collected data on adverse events including AF from the sponsors of placebo-controlled clinical trials of alendronate, ibandronate, risedronate, and zoledronic acid. A press release (November 12, 2008) reported on data submitted from 19,687 bisphosphonate-treated and 18,358 placebo-treated patients followed for between 6 months and 3 years. Bisphosphonate use was not associated with increased rates of AF, and the authors of the report concluded that the incidence of AF was low, most studies containing two or fewer events, with an absolute difference in event rates between the bisphosphonate and placebo arms of 0 to 3 per 1,000.14
Data from the United Kingdom General Practice Research Database (GPRD) also have been used to assess the risks of AF and atrial flutter in women on oral alendronate or risedronate. A self-controlled case-series method was used to minimize confounding, and information on 2,195 patients exposed to either of the preceding bisphosphonates and diagnosed with AF or atrial flutter was analyzed. No evidence of an overall long-term increase in the risk of AF or atrial flutter with continued exposure to alendronate or risedronate was demonstrated, although post hoc analysis suggested that there might be an increased risk during the first few months of alendronate therapy.15
Recently, data on the risk of AF after treatment with pamidronic acid (90 mg) or zoledronic acid (4 or 8 mg) every 3 to 4 weeks for 24 months have been submitted to the Committee for Medical Products for Human Use (CHMP). These data came from an industry-supported international, multicenter, randomized, double-blind, noninferiority comparator trial in patients with multiple myeloma or metastatic breast cancer.16 Patients treated with pamidronic acid had a significantly greater number of AF events (12/556, 2.2%) than the zoledronic acid group (3/563, 0.5%) with a relative risk (RR) of 4.05 (95% CI 1.15–14.28, p = .018). Since there was no placebo group in this study, the background incidence of AF in the study population is unknown, and hence it is uncertain whether the rates of AF in either treatment group were increased, normal, or decreased. These results are not yet published but have been included in the CHMP Pharmacovigilance Working Party (PhVWP) assessment report.7 Interestingly, AF has not been associated previously with bisphosphonate therapy in patients with skeletal malignant disease despite the use of much higher doses than those prescribed for osteoporosis.
Finally, the possible association between bisphosphonates and AF has been examined recently in two meta-analyses. In one of these, a significantly higher risk of AF reported as a serious adverse event was found in a meta-analysis of four trial data sets [odds ratio (OR) 1.47, 95% CI 1.01–2.14, p = .04], although a meta-analysis of all AF events (serious and nonserious) from the same data sets yielded a pooled OR of 1.14 (95% CI 0.96–1.36, p = .15). No increase in the risk of stroke or cardiovascular mortality was demonstrated.17 In a Bayesian meta-analysis, Mak and colleagues reported a trend toward a higher risk of AF with bisphosphonate use both in randomized, controlled trials and in observational studies. The authors of both studies are cautious in their conclusions and acknowledge the uncertainties associated with their analyses.18
Interpretation of Data From Randomized, Controlled Trials and Observational Studies: What are the Limitations?
The reporting of adverse events in randomized, controlled trials generates huge databases from which inferences are drawn about drug safety. In general, adverse events are not adjudicated independently and therefore rely to some extent on the subjective assessment of the investigator. Furthermore, there is variability in the terminology used to describe adverse events across studies and often some redundancy in the use of terms. When serious adverse events occur, multiple related conditions may be associated with hospitalization, and the reporting of a specific condition as a serious adverse event may not reflect this accurately. In the analysis of adverse-event data, the convention is not to adjust for multiple testing, and this increases the likelihood of false-positive findings, particularly for rare events. This is particularly relevant for adverse events that are not anticipated, as was the case for AF in the HORIZON study, because adjudication was not planned prospectively. There are also specific issues related to the reporting of AF in clinical trials. AF may be asymptomatic or paroxysmal and therefore may be underreported. Furthermore, in most cases where AF is recorded as an adverse event, the diagnosis is made clinically and is not confirmed by ECG.
While observational studies have the advantage of being able to examine larger numbers over longer periods of time, they also have disadvantages. In particular, the potential influence of confounding factors is a significant limitation. There are several shared risk factors for AF and osteoporosis, and osteoporosis itself may be a risk factor for cardiovascular disease, so that patients who are receiving bisphosphonate therapy already may be at increased risk for AF.19 Second, as in clinical trials, the recording of AF in observational studies may not be robust. Third, observational studies generally are limited to examination of the effects of alendronate because of the predominance of its use over other bisphosphonates in the population. Finally, adherence to therapy is usually not documented accurately in observational studies but is likely to be significantly worse than in clinical trials.
The limitations just described also apply to meta-analyses, the results of which additionally may be influenced by publication bias and heterogeneity between studies. In view of these limitations and the conflicting nature of the evidence to date, an association between bisphosphonate therapy and AF cannot be excluded and requires further investigation.
Atrial Fibrillation and Bisphosphonate Action: Cell Physiologic Correlations
The major pathologic change associated with chronic AF is a progressive atrial fibrosis from collagen deposition leading to structural changes including alterations in chamber structure and size. It results from atrial remodeling, likely initiated by the AF itself that further alters atrial electrophysiology. The remodeling is accompanied by decreased action potential duration and effective refractory period.20 The precise causal relationships between these changes are unclear. Nevertheless, alterations arising from each of these processes each could contribute to development of the observed pathophysiology.
At the cellular level, acute changes in electrophysiologic function of otherwise normal myocardium with intact electrical conducting pathways typically are associated with altered cytosolic and mitochondrial electrolyte homeostasis, particularly involving Ca2+, Mg2+, and Cl–. Alterations in ion currents then lead to electrical remodeling. L-type Ca2+ channels (LTCCs) normally are involved in cardiac action potential generation. This leads to ryanodine receptor (RyR)–mediated Ca2+-induced release of intracellularly stored sarcoplasmic reticular (SR) Ca2+, in turn, initiating excitation-contraction coupling.21 Transient intracellular perturbations in cytosolic [Ca2+], caused either by LTCC-mediated extracellular Ca2+ influxes or SR Ca2+ release, could contribute to or initiate arrhythmogenic situations.22–25 Intracellular Ca2+ overload of myocardial cells could cause AF. The ion channel remodeling known to accompany chronic AF then appears to reduce LTCC densities.26
Levels of ionized Mg (Mg2+) similarly may influence cardiovascular function. Severe Mg2+ deficiency could cause prooxidative and proinflammatory changes. Mg depletion in humans may lead to cardiac arrhythmias that include AF. Diabetes mellitus is the most commonly recorded clinical condition associated with Mg2+ deficiency that may result from the osmotic diuresis caused by glycosuria.27 Mg2+, by reducing ventricular response, has been used prophylactically to reduce the occurrence of AF following cardiac surgery.28 It also can be considered as a safe adjunct to digoxin in controlling the ventricular response in AF.29 Reductions in L-type channel activity during electrical remodeling induced by rapid pacing may involve the putative zinc transporter 1 (ZnT-1) that transports Zn2+ out of cells.21
Finally, gap junctions mediate cell-to-cell electrical coupling through their component connexins, including the extensively studied connexin 40, restricted predominantly to atrial tissue; connexin 43, highly expressed in both atrial and ventricular tissue; and the more ubiquitous connexin 45.30 Animal models deficient in cardiac connexins are vulnerable to reentrance arrhythmias.31 Altered distributions, densities, or types of expressed connexins could well influence arrhythmogenic tendency. Thus AF is associated with alterations in connexin 40 expression and localization.32–34 [Ca2+]i also could influence expression of cardiac connexins.35
Bisphosphonates (Fig. 1) are stable analogues of naturally occurring inorganic pyrophosphates and show significant interactions with electrolytes similar to those involved in cardiac excitability. They bind strongly to Ca2+, Mg2+, and Zn2+. Individual bisphosphonates vary in their mineral-binding properties and biochemical actions at the cellular level.36 Bisphosphonates with larger (e.g., alendronate, ibandronate, and zoledronate) or smaller (e.g., risedronate) electrical charges on their nitrogen atom in the R2 group may differ in their cytosolic penetration and interactions with charged cations.36 The aminobisphosphonate alendronate increases [Ca2+]i in human osteoclast-like cells.37 The nonaminobisphosphonate clodronate reduces release of intracellularly stored Ca2+ and LTCC-mediated Ca2+ influx in vascular smooth muscle.38 Sequestration of intracellular Ca2+ has been reported with the aminobisphosphonate pamidronate and with clodronate in splenic and hepatic macrophages,39 although not in cardiac myocytes. Nevertheless, a recent report describes oscillatory Ca2+ dynamics in isolated atrial myocytes following long-term alendronate treatment.40
However, there is no evidence either confirming or excluding direct causal links between bisphosphonate use and AF attributable to ion chelation or to direct effects on ion channels that might alter immediate atrial electrophysiologic function or electrical remodeling. Acute AF events did not occur around the time of zoledronic acid infusion in the HORIZON Pivotal Fracture Trial.6 Bisphosphonates have high selective affinities for bone, are selectively taken up by osteoclasts, and therefore are unlikely to reach high concentrations in other cells or tissues to thereby act directly on ionic, particularly Ca2+ homeostasis in cardiac myocytes or other tissues that may affect atrial function. Clinically significant hypocalcemia is rare following intravenous bisphosphonate administration, and there is no direct evidence that the AF observed in the HORIZON Pivotal Fracture Trial6 resulted from altered serum electrolyte levels.
Atrial Fibrillation and Bisphosphonate Action: Effects of Fibrotic and Angiogenic Changes
A number of longer-term changes also might influence atrial arrhythmogenic tendency. Thus the zinc-dependent endopeptidase matrix metalloproteinases, particularly MMP-2 and MMP-9, could be synthesized and released by cardiomyocytes and be involved in fibrotic change.41 Their action is directly inhibited by endogenous tissue inhibitors of metalloproteinases (TIMPs), but AF is often associated with the combination of increased MMP activity and their inadequate inhibition by TIMPs.41, 42 AF is also associated with cardiac ischemia. Coronary heart disease constitutes a significant risk factor for AF, which is a known frequent complication of acute myocardial infarction. Acute atrial ischemia in canine hearts produces an arrhythmic substrate,43 and in patients with lone recurrent AF, abnormal atrial microvascular perfusion could be demonstrated.44 Capillary densities in the posterior wall of the left atrial myocardium are often considerably lower than in other atrial regions in patients with AF associated with mitral valve disease.45, 46
One accordingly may consider whether bisphosphonates are involved in the structural remodeling or the ischemic changes associated with AF. Cell culture studies, in vivo angiogenesis bioassays, and tumour models suggest that bisphosphonates have antiangiogenic properties.47–52 Zoledronic acid is a known metalloproteinase inhibitor, suppressing MMP-9 expression and inhibiting its proteolytic activity by Zn chelation.51 MMP-9 mobilizes the angiogenic vascular endothelial growth factor (VEGF) and increases its association with VEGF-R2, thus promoting microvessel growth.53 Bisphosphonates then potentially may further reduce an already limited atrial blood supply in patients with abnormal myocardial perfusion, triggering, contributing to, or maintaining AF.
Atrial Fibrillation, Bisphosphonate Action, and Inflammatory Changes
Inflammatory events are also associated with AF. There is a high postoperative incidence of acute AF. This correlates with elevated plasma concentrations of inflammatory markers following cardiac surgery.54, 55 Two-thirds of atrial biopsies from patients with lone AF show chronic inflammatory infiltrates and myocyte necrosis.56 Further studies report increased circulating C-reactive protein (CRP), interleukin 6 (IL-6), and tumor necrosis factor α (TNF-α) levels in patients with paroxysmal or persistent AF in nonsurgical situations.57 Studies in experimental animal models suggest that inflammation can alter atrial conduction but have not yet clarified whether inflammation causes or results from AF.58, 59
Bisphosphonate treatment is also associated with inflammatory changes.57 Aminobisphosphonates, particularly when administered intravenously, stimulate release of inflammatory cytokines,60 an acute phase response that involves rapid and copious proinflammatory cytokine production by peripheral blood gammadelta (γδ) T cells and is inhibited by statins.60 Increased proinflammatory cytokine levels and transient flulike symptoms develop in significant numbers of patients less than 72 hours following a first exposure to intravenous aminobisphosphonates. However, there is no published work on direct acute or cumulative effects of the acute-phase reaction owing to bisphosphonates on structural atrial changes or atrial conduction. Such effects also could result from bisphosphonate actions on cardiac macrophages involved in chronic cardiac remodeling. Apart from osteoclasts, macrophages and monocytes are the only major cells that take up bisphosphonates at low concentrations and then potentially become apoptotic. Experimental macrophage depletion in rats after intravenous administration of liposomal clodronate leads to areas of myocyte loss and inflammatory cell infiltration, predominantly by CD4+ T-lymphocytes.61 Thus the findings in the HORIZON Pivotal Fracture Trial6 could be consistent with delayed inflammatory effects of bisphosphonates (>30 days) on the atrial conduction system.
Conversely, anti-inflammatory factors might modify the risk of arrhythmias. When gluocorticoids are prescribed in combination with bisphosphonates, for example, in skeletal malignancy or gluocorticoid-induced osteoporosis, their anti-inflammatory effects might prevent or reduce the development of arrhythmogenic substrate by bisphosphonates. Glucocorticoids prevent inducible cardiac flutter and slow down atrial remodeling in canine hearts62, 63 and can delay the recurrence of arrhythmic episodes in patients with the first symptomatic and persistent episode of AF.64 Two prospective, randomized, double-blind, placebo-controlled clinical trials demonstrated that dexamethasone reduces release of several inflammatory and acute-phase response mediators associated with adverse outcomes65 and that intravenous hydrocortisone reduced the incidence of AF after cardiac surgery by 33%.66 However, glucocorticoid therapy has been reported as a risk factor for AF. In a large population-based case-control Danish study, Christiansen and colleagues showed that current use of systemic glucocorticoids is associated with an almost twofold increased risk of AF or flutter.67
Statins, another group of anti-inflammatory agents, might mask the link between AF and bisphosphonates in clinical practice. Both statins and bisphosphonates affect the mevalonate pathway that is responsible for the synthesis of sterol isoprenoids, including cholesterol (Fig. 2). Statins inhibit HMG-CoA reductase, which catalyzes mevalonate synthesis in hepatocytes68 and bone cells such as osteoblasts and osteoclasts.69, 70 Bisphosphonates inhibit one or more enzymes downstream of HMG-CoA reductase in osteoclasts. Both agents inhibit prenylation and function of small GTPases.71 Yet several animal studies report that statin use may reduce the risk of AF. High-dose atorvastatin appears to reduce AF by inhibiting inflammation in a canine sterile pericarditis model.72 Daily simvastatin (80 mg) attenuated AF following atrial tachycardia in canine hearts.73
Available clinical evidence concerning statin action in preventing AF is confined to small observational studies and retrospective analyses.2, 74 Thus (1) statin use reduced postoperative AF risk by 48% following coronary artery bypass grafting, an effect attributed to altered matrix remodeling.75 (2) A “nested” study was conducted on 555 patients, combined from threerandomized and controlled Atrial Fibrillation Suppression Trials (AFIST-I, -II, and -III), who had undergone cardiothoracic surgery. There was a substantial use of both prophylactic amiodarone and postoperative beta blocker administration in these patients. Nevertheless, adjunctive use of preoperative statin was associated with a 40% reduction in likelihood of AF following cardiothoracic surgery76. (3) A secondary analysis of the Heart and Estrogen-Progestin Replacement Study (HERS) in 2673 (976 on a statin) postmenopausal women younger than 80 years of age with known coronary heart disease and treated with a statin showed a reduced prevalence and incidence of AF77. However, (4) a large retrospective study in patients without a history of AF, where 2096 patients (52%) received preoperative statins, failed to show protective effects of statins.78 One retrospective report examined the effects of statin therapy in patients treated with bisphosphonates on the risk of AF.13 Statin use was similar in the bisphosphonate-treated and control groups (14.9% and 13.8%, respectively); alendronate users currently taking statins showed an increased risk of incident AF (p = .02).
In conclusion, some of the pharmacologic characteristics of bisphosphonates, such as their ion-binding capacity and proinflammatory effects, theoretically could cause acute or chronic changes similar to those found in the arrhythmogenic myocardium (Fig. 3). However, available clinical or epidemiologic data are inadequate to substantiate such claims, and experimental evidence is lacking completely. Therefore, further translational investigations are required.
Following the report in the HORIZON study of an increased risk of AF as a serious adverse event in women treated with zoledronic acid, of an association between bisphosphonate therapy and AF has been investigated in a number of studies. These have produced conflicting results and have not excluded the possibility that such an association exists. Nevertheless, the risk-benefit balance of bisphosphonate therapy in patients with osteoporosis and other bone diseases such as skeletal malignancy and Paget's disease remains strongly positive, and changes in clinical practice are not indicated on the basis of current evidence.
Potential mechanisms by which bisphosphonates might increase the risk of atrial arrhythmias include changes in intracellular ion concentrations and proinflammatory, profibrotic, and antiangiogenic effects. Evidence that these operate to affect atrial conduction in vivo is currently lacking but is an important area for further research.
MP served as a consultant for Procter & Gamble. JEC is a member of the Data Safety Monitoring Board of the HORIZON study and has received honoraria for speaking and/or advisory commitments from Procter & Gamble, Sanofi-Aventis, Merck, Sharp, and Dohme. CLH is not a recipient of any fees, honoraria, grants, or consultancies that would constitute a conflict of interest with the current review article.
The authors are grateful to Helene Mellor for the illustration services. JEC acknowledges support from the Cambridge Biomedical Research Centre and the National Institutes for Health Research (NHIR). CLH acknowledges support from the Wellcome Trust, the Medical Research Council, and the British Heart Foundation.