Oral bisphosphonates and atrial fibrillation

Abstract

The risk is negligible, and does not offset the benefits of reducing fractures

The benefits of bisphosphonates for the prevention of fracture in patients with osteoporosis are not disputed,^{1 2} and one trial has also reported reduced mortality in these patients.³ Nevertheless, bisphosphonates are underused by those most likely to benefit—elderly patients with fractures.² This shortfall in care was difficult to rectify even while bisphosphonates were considered effective and safe,² but matters will probably get worse after recent reports that these drugs might increase the risk of atrial fibrillation.^{3 4 5 6} Even fewer eligible patients are now likely to start taking bisphosphonates, and more treated patients are likely to stop taking them. This would be justified if the risk of atrial fibrillation were real and large enough to offset the known benefits for fracture reduction. However, this potentially increased risk is challenged by an accompanying population based observational study by Sørensen and colleagues, which finds no significant increase in atrial fibrillation or flutter in women taking bisphosphonates for osteoporosis.⁷

It was only a matter of time before osteoporosis, its treatments, and cardiovascular events were linked.^{8 9} Osteoporosis is common, it affects people with an increased risk of cardiovascular disease—such as postmenopausal women and elderly people—and it has been associated with adverse cardiovascular events.⁹ Classic risk factors for atrial fibrillation such as older age, hyperthyroidism, and smoking increase the risk of osteoporosis. Also, both osteoporosis and atrial fibrillation can be relatively clinically silent. For example, implantable event monitor studies of patients with treated atrial fibrillation find almost 40% of those with extended (>48 hours) recurrences are completely asymptomatic.¹⁰ With this scenario of at risk patients sharing similar risk factors, any secondary or post hoc analyses of osteoporosis treatments and adverse cardiovascular events will be difficult to interpret and susceptible to serious confounding and problems related to multiple comparisons.⁸

So how strong is the evidence that bisphosphonates cause atrial fibrillation? To date, four studies have reported the link between these drugs and atrial fibrillation (table).^{3 4 5 6} Of most concern is the horizon pivotal fracture trial,³ which compared annual intravenous zoledronic acid with placebo in almost 7800 postmenopausal women. Over three years, new atrial fibrillation did not significantly increase (2.4% of zoledronic acid users v 1.9% of controls; P=0.12); however, the ill defined entity of “serious” atrial fibrillation did significantly increase (1.3% of zoledronic acid users v 0.5% of controls; P<0.001). The evidence available so far indicates that the absolute risk of atrial fibrillation is very small, and if present at all, it is probably associated with intravenous zoledronic acid and not the more commonly used oral agents (table). A meta-analysis of individual patient data might help to define the risk of atrial fibrillation with bisphosphonates.

Randomised trials of bisphosphonates that have reported data on atrial fibrillation

Trial	Sample size and population	Follow-up	Comparison	Atrial fibrillation (%)*		Absolute % difference (NNH)	P value
				With bisphosphonate	Controls
HORIZON pivotal fracture trial3	~7800 postmenopausal women with fractures or low bone mass (or both)	Up to 3 years	Intravenous zoledronic acid (5 mg annually) v placebo	Overall 2.4	1.9	0.5 (200)	0.12
				Serious 1.3	0.5	0.8 (125)	<0.001
HORIZON recurrent fracture trial4	~2100 survivors of hip fracture (76% postmenopausal women)	Up to 3 years	Intravenous zoledronic acid (5 mg annually) v placebo	Overall 2.8	2.6	0.2 (500)	0.79
				Serious 1.1	1.3	−0.2 (NA)	0.84
FIT programme of trials5	~6500 postmenopausal women with fractures or low bone mass (or both	Up to 5 years	Oral alendronate (5-10 mg daily) v placebo	Overall 2.5	2.2	0.3 (333)	0.42
				Serious 1.5	1.0	0.5 (200)	0.07
Risedronate phase III trials6	~10 000 postmenopausal women with fractures or low bone mass (or both)	Up to 3 years	Oral risedronate (5 mg daily) v placebo	Overall 1.4	1.4	0.0 (NA)	1.0
				Serious 0.6	0.5	0.1 (1000)	0.49

*Serious atrial fibrillation as defined by authors.

NA=not applicable; NNH=number needed to harm derived by taking reciprocal of absolute per cent difference of treatment minus control.

That said, secondary analyses from trials and meta-analyses may not constitute best evidence when dealing with rare drug related adverse event rates (<1-2%) and two common conditions that are found in the same older population. Unlike the use of non-randomised studies to estimate treatment effects without a trial (as used for oestrogen and heart attack) or to find wholly unexpected benefits (as used for statins and prevention of sepsis or cancer), a large and well conducted observational study is perhaps better suited to determining unexpected harm from commonly used drugs.¹¹

Sørenson and colleagues report such a study.⁷ They investigated disodium etidronate and alendronate in a Danish population based nested case-control study (around 14 000 cases of atrial fibrillation and around 68 000 controls) conducted from 1999 to 2005. The timing of their study is important because prescribers would have had no concerns about any risk of atrial fibrillation, which otherwise might have introduced confounding by prognosis.¹¹ In terms of databases and methods used, the study is rigorous and adequately powered. The investigators report almost no difference in the use of bisphosphonates in people with atrial fibrillation and those without (3.2% of current users had atrial fibrillation v 2.9% of non-users). No association was seen in appropriately adjusted analyses that examined new users versus not new users (the least biased comparison), former users versus not former users, and long duration versus short duration of use. To replicate “serious atrial fibrillation” as reported in trials (table), Sørenson and colleagues also restricted analyses to events needing hospital admission or cardioversion, and again they found no clinically important associations.

The main limitation of their work is lack of information about zoledronic acid. This is important. Previous reports^{5 6} and Sørenson and colleagues’ work⁷ suggest that oral bisphosphonates are safe with respect to atrial fibrillation. But atrial fibrillation might possibly be triggered soon after an infusion with zoledronic acid (or other intravenous bisphosphonates) as a result of the release of proinflammatory cytokines or transient hypocalcaemia and secondary hyperparathyroidism.^{5 7} However, in the HORIZON pivotal fracture trial, most cases of atrial fibrillation occurred months after infusion, and electrocardiograms done in 559 patients before and nine to 11 days after infusion did not differ with respect to the presence of arrhythmias, which suggests that neither of these possible mechanisms was responsible.³ Nevertheless, the makers of zoledronic acid should do their utmost to confirm or allay concern, by pooling individual patient data and conducting postmarketing surveillance studies.

What are the implications for clinicians who commonly and appropriately prescribe oral bisphosphonates for their older patients with fractures and osteoporosis? For now, beyond taking the patient’s pulse and ordering an electrocardiogram when it is irregular,¹² available evidence suggests that business should carry on as usual—the risk of atrial fibrillation associated with oral bisphosphonates seems to be vanishingly small if it exists at all, and it is unlikely to ever offset the confirmed benefits of these drugs in the prevention of fractures.