The aim of this study was to investigate the occurrence of heart failure in patients treated with bisphosphonates.
The aim of this study was to investigate the occurrence of heart failure in patients treated with bisphosphonates.
In this nationwide retrospective cohort study from Denmark, all users of bisphosphonates and raloxifene between 1996 and 2006 (n = 102 342) were included in the ‘exposed’ group and three age- and gender-matched subjects (n = 307.026) from the general population comprised the control group. The risk of heart failure was estimated by Cox proportional hazard analyses.
The mean follow-up times were 2.8, 5.5 and 4.9 years for alendronate-, etidronate- and raloxifene-treated patients, respectively. The absolute risk of heart failure was 4.4% in the exposed group and 3.7% in the control group (P < 0.01). The relative risk (RR) of heart failure was significantly increased in users of bisphophonates: crude RR 1.71 [95% confidence interval (CI) 1.63–1.79]; adjusted hazard ratio (HR) 1.41 (95% CI 1.34–1.48). By comparison, raloxifene, which is used for the same indication but has a different mechanism of action, was not associated with an increased risk of heart failure: adjusted HR 1.07 (95% CI 0.76–1.50). When the two most commonly used bisphosphonates, alendronate and etidronate, were analysed separately, significant trends in the risk of heart failure were observed across refill compliance strata. The risk of heart failure increased significantly with increasing refill compliance for etidronate (P for trend <0.01), whereas it decreased for alendronate (P for trend <0.01).
Bisphosphonate users were at increased risk of heart failure compared to age- and gender-matched control subjects. However, users of alendronate showed a dose-dependent reduction in this risk, suggesting that alendronate may reduce the risk of heart failure.
Bisphosphonates inhibit bone resorption and are widely used in clinical practice to prevent and treat osteoporosis. Osteoporotic fractures are associated with increased mortality , and epidemiological studies have shown a clear association between osteoporosis and cardiovascular disease . A randomized, double-blind, placebo-controlled trial investigating the intravenous bisphosphonate zoledronic acid demonstrated a significant reduction in mortality of patients treated with bisphosphonates . Amongst adjudicated cardiac events, myocardial infarction and heart failure were the two most frequent causes of death [3, 4]. Concurrent with these results, a meta-analysis of clinical trials testing the oral bisphosphonate risedronate found a trend towards lower cardiovascular mortality . Several hypotheses have been proposed to explain these findings , but the underlying mechanism of the mortality-reducing effect of bisphosphonates remains unknown. Thus, we investigated the occurrence of heart failure in patients treated with bisphosphonates in this nationwide cohort study.
In this cohort study, patients treated with bisphosphonates were compared to an age- and gender-matched control group of subjects naive to these drugs. The main outcome was occurrence of heart failure.
This study was approved by the National Board of Health and the Danish Data Protection Agency.
All patients who had filed a prescription for bisphosphonates or raloxifene between 1 January 1996 and 31 December 2006 were included in the study. Patients were considered ‘exposed’ if they were treated with any type of bisphosphonate: Anatomical Therapeutic Chemical (ATC) codes M05BA01 (etidronate), M05BA02 (clodronate), M05BA03 (pamidronate), M05BA04 (alendronate), M05BA06 (ibandronate), M05BA07 (risedronate), M05BA08 (zoledronic acid), M05BB01 (etidronate plus calcium) and M05BB03 (alendronate plus vitamin D). In addition, we included patients treated with raloxifene (ATC code G03XC01), which has a similar indication to bisphosphonates, but a different mechanism of action. All drugs were administered orally except zoledronic acid and pamidronate, which were administered intravenously.
For each exposed subject treated with an osteoporosis medication, three age- and gender-matched control subjects were randomly selected from the general population during the time period of treatment. Unexposed subjects were required not to have been treated with strontium or parathyroid hormone or analogues, in addition to bisphosphonates. A dummy baseline date was assigned to each unexposed control subject based on the date of first use of an osteoporosis medication in the matched patient.
Heart failure was defined as a diagnosis of International Classification of Diseases, 8th revision (ICD8) 42709, 42710, 42711, 42719 or 42799 or 10th revision (ICD10) I500, I501 or I509, before or after the start of exposure to osteoporosis medication. The diagnosis of heart failure has a specificity of 99% and a sensitivity of 29% in the Danish registers , with positive and negative predictive values of 81% and 90%, respectively .
Information on the occurrence of heart failure and comorbidities such as alcoholism was obtained from the Danish National Patient Register. This register covers all inpatient contacts from 1 January 1977 to 1994, and from 1995 onwards, it also covers all outpatient activities and emergency room contacts [7, 8]. The ICD system is used to classify the contact reason. The register has a nationwide coverage and an almost 100% capture of contacts, making it very useful for research purposes [7-9].
The Danish Health and Medicines Authority (http://laegemiddelstyrelsen.dk/en/) keeps a nationwide register of all drugs sold at all pharmacies from 1994 onwards (the Danish National Prescription Registry. All drugs are registered using the Anatomical Therapeutic Chemical (ATC) classification system, and the dosage sold, form of medication (tablets, injections etc.) and date of sale are also registered. The start date of exposure in this study was the first date of prescription of an osteoporosis medication and a matched dummy date for the unexposed controls, thus minimizing the effects of immortal time bias . Drug exposure was calculated as the average daily dose (number of defined daily dosages [DDD] per day – equal to the average normal dose of a drug per day). This average dose was calculated as the sum of all dispensed prescriptions from first prescription to the date of censoring divided by the time in days from first prescription to the date of censoring. Information on vital status and migration came using the Danish Civil Registration System (http://www.cpr.dk/cpr/site.aspx?p=34). All study participants were followed up until the time of death, migration, occurrence of heart failure or 31 December 2006, whichever came first. Patients treated with drugs against osteoporosis were followed up, despite any treatment gaps. It is possible to link these data using the Civil Registration Number, which is a permanent and unique 10-digit identification number given to every Danish citizen.
Data are presented as mean and standard deviation. A Cox regression analysis was applied to calculate the risk of the outcome after initiation of any osteoporosis medication or the corresponding dummy baseline date amongst the controls. The assumption of proportional hazards was evaluated by inspection of the survival curves and hazards. Adjustments were made for previous diagnoses (dementia, peripheral atherosclerosis, liver disease, kidney disease, acquired immunodeficiency syndrome, connective tissue disease, chronic obstructive pulmonary disease, alcoholism, ulcerations of the duodenum or ventricle, cancer, stroke, diabetes, coronary atherosclerosis and myocardial infarction), prior medication use (inhaled bronchodilators or corticosteroids, diuretics and antihypertensives) and number of hospital bed days, employment status, marital status and income. Separate analyses were performed for sensitivity in patients who only redeemed one prescription for bisphosphonates. Analyses were performed using STATA 9.0 (STATA Corp., College Station, TX, USA) and SPSS 19.0 (SPSS Inc., Chicago, IL, USA).
Baseline characteristics for patients and control subjects are presented in Table 1. The study population included a total of 102 342 patients exposed to bisphosphonates and 307 026 age- and gender-matched control subjects. The mean age at baseline was 71 years, and 85% were women. Mean age was similar in the bisphosphonate groups: 71.6 ± 11.7 and 71.1 ± 12.7 years amongst those treated with alendronate and etidronate, respectively, but lower in the raloxifene-treated group (64.0 ± 10.9 years). Comorbidities and use of comedications, including antihypertensive drugs, platelet function inhibitors and diuretics, were significantly more common at baseline amongst patients exposed to bisphosphonates, compared to control subjects (Table 1).
|Variable||Exposed (n = 102 342)||Unexposed (n = 307 026)||P|
|Age (years)a||70.5 ± 11.4||70.5 ± 11.4||–|
|Men||15 602 (15.2)||46 806 (15.2%)||–|
|Women||86 740 (84.8)||260 217 (84.8%)||–|
|Raloxifene (mean treatment time)||4831 (4.9 years)||–||–|
|Etidronate (mean treatment time)||39 724 (5.5 years)||–||–|
|Clodronate (mean treatment time)||566 (2.4 years)||–||–|
|Pamidronate (mean treatment time)||45 (1.3 years)||–||–|
|Alendronate (mean treatment time)||55 090 (2.8 years)||–||–|
|Ibandronate (mean treatment time)||612 (0.4 years)||–||–|
|Risedronate (mean treatment time)||1452 (1.4 years)||–||–|
|Zoledronic acid (mean treatment time)||22 (0.6 years)||–||–|
|Low-dose aspirin/clopidogrel/dipyridamole||27 900 (27.3)||71 826 (23.4)||<0.01|
|Diuretics||46 951 (45.9)||122 923 (40.0)||<0.01|
|Antihypertensive agents||37 993 (37.1)||104 814 (34.1)||<0.01|
|Statins||8307 (8.1)||22 998 (7.5)||<0.01|
|Alcoholism||3637 (3.6)||6 048 (2.0)||<0.01|
|Heart failure||4553 (4.4)||11 585 (3.8)||<0.01|
|Prior coronary atherosclerosis or myocardial infarction||11 577 (11.3)||29 703 (9.7)||<0.01|
|Prior ulcerative disease||10 153 (9.9)||19 023 (6.2)||<0.01|
|Prior stroke||4572 (4.5)||14 744 (4.8)||<0.01|
|Prior diabetes||3638 (3.6)||12 886 (4.2)||<0.01|
|Prior cancer||21 393 (20.9)||57 770 (18.8)||<0.01|
|Prior chronic obstructive pulmonary disease||11 888 (11.6)||13 126 (4.3)||<0.01|
|Dementia||1039 (1.0)||4 594 (1.5)||<0.01|
|Peripheral arterial disease||3976 (3.9)||8 949 (2.9)||<0.01|
|Liver disease||3317 (3.2)||6 615 (2.2)||<0.01|
|Kidney disease||3095 (3.0)||7 720 (2.5)||<0.01|
|Acquired immunodeficiency syndrome||22 (<0.1)||25 (<0.01)||<0.01|
|Connective tissue disease||5238 (5.1)||3893 (1.3)||<0.01|
The absolute risk of heart failure was 1.2% in the alendronate-treated group (versus 0.8% amongst controls), 4.9% in the etidronate-treated group (versus 2.9% amongst controls) and 1.1% in the raloxifene-treated group (versus 0.9% controls). The mean follow-up times were 2.8, 5.5 and 4.9 years for patients treated with alendronate, etidronate and raloxifene, respectively. Approximately 85% of patients exposed to bisphosphonates redeemed two or more prescriptions.
Table 2 shows the crude relative risk (RR) and adjusted hazard ratio (HR) values for heart failure after initiation of osteoporosis medication in patients with no prior diagnosis of heart failure. The risk of heart failure was significantly increased in users of bisphophonates with a crude RR of 1.71 [95% confidence interval (CI) 1.63–1.79], when all users of bisphophonates were combined (all types and doses). The risk of heart failure remained significant after adjusting for previous diagnoses, prior medication use and other factors as described in Methods (adjusted HR 1.41, 95% CI 1.34–1.48). In users of raloxifene, the risk of heart failure was not significantly increased (crude RR 1.23, 95% CI 0.89–1.71; adjusted HR 1.07, 95% CI 0.76–1.50). Considering treatment with bisphosphonates as a time-dependent variable did not affect the results.
|Drug||DDD/day||Users||Crude RR||Adjusted HR||P for trend|
|Alendronate||Never user||1200/16 5521||Reference||Reference|
|≤0.25||270/9984||3.66 (3.24–4.14)a||1.98 (1.73–2.27)a||<0.01|
|0.26–0.80||187/10 731||2.38 (2.05–2.76)a||1.69 (1.44–1.97)a|
|>0.80||254/32 213||1.09 (0.95–1.24)||1.03 (0.89–1.18)|
|Combined||1.84 (1.68–2.02)a||1.42 (1.29–1.57)a|
|Etidronate||Never user||3480/115 689||Reference||Reference|
|≤0.25||447/9285||1.57 (1.43–1.73)a||1.17 (1.06–1.29)a||<0.01|
|0.26–0.80||789/14 093||1.82 (1.68–1.96)a||1.45 (1.34–1.58)a|
|>0.80||718/14 392||1.63 (1.50–1.76)a||1.54 (1.42–1.67)a|
|Combined||1.68 (1.60–1.78)a||1.40 (1.32–1.49)a|
|Any bisphosphonate||Never user||4739/289 403||Reference||Reference|
|≤0.25||730/19 758||2.26 (2.09–2.55)a||1.42 (1.30–1.54)a||0.03|
|0.26–0.80||985/25 349||2.37 (2.22–2.54)a||1.59 (1.49–1.70)a|
|>0.80||984/48 182||1.24 (1.16–1.34)a||1.26 (1.17–1.35)a|
|Combined||1.77 (1.69–1.85a)||1.41 (1.34–1.48)a|
|Raloxifene||Never user||124/14 369||Reference||Reference|
|≤0.25||13/1420||1.06 (0.60–1.87)||0.99 (0.63–1.54)||0.48|
|0.26–0.80||14/1079||1.50 (0.87–2.59)||0.94 (0.48–1.85)|
|>0.80||24/2281||1.22 (0.79–1.88)||1.32 (0.68–2.57)|
|Combined||1.23 (0.89–1.71)||1.07 (0.76–1.50)|
Table 3 shows gender-stratified adjusted risks of heart failure after initiation of treatment with bisphosphonates or raloxifene in patients with no prior diagnosis of heart failure. The risk of heart failure did not differ between men and women. When the two most commonly used bisphosphonates (alendronate and etidronate) were analysed separately, significant trends in the risk of heart failure were observed across refill compliance strata (DDD per day, Medication Possession Ratio, MPR). This dose–effect relationship differed between the first-generation bisphosphonate etidronate and the newer nitrogen-containing bisphosphonate alendronate. The risk of heart failure increased significantly with increasing refill compliance for etidronate (men/women: adjusted HR 1.13/1.19 with MPR ≤0.25 and 2.09/1.45 with MPR >0.80, P for trend <0.01), whereas it decreased for alendronate (men/women: 1.84/2.02 with MPR ≤0.25 and 1.20/0.99 with MPR >0.80, P for trend <0.01). By contrast, there was no MPR-associated trend in raloxifene users (Table 3). The dose–response relationships for etidronate and alendronate are clearly shown in Figs 1 and 2; Figures 1-4 shows the cumulated hazard of heart failure after initiation of bisphosphonates and raloxifene by time after first use.
|Men||Grouped dose (DDD/day)|
|≤0.25||1.84 (1.27–2.69)a||1.13 (0.85–1.50)||–|
|0.26–0.80||1.97 (1.33–2.92)a||1.70 (1.37–2.11)a||–|
|>0.80||1.20 (0.85–1.68)||2.09 (1.71–2.56)a||–|
|Combined (vs. never user)||1.51 (1.16–1.96)a||1.68 (1.43–1.98)a||–|
|Combined; only filed one prescription (vs. never user)||1.81 (1.03–3.16)a||2.35 (1.76–3.14)a||–|
|Women||Grouped dose (DDD/day)|
|≤0.25||2.02 (1.74–2.33)a||1.19 (1.06–1.32)a||0.95 (0.54–1.70)|
|0.26–0.80||1.64 (1.38–1.95)a||1.43 (1.31–1.56)a||1.34 (0.77–2.35)|
|>0.80||0.99 (0.85–1.16)||1.45 (1.32–1.59)a||1.01 (0.65–1.58)|
|Combined (vs. never user)||1.41 (1.27–1.57)a||1.37 (1.28–1.46)a||1.07 (0.76–1.50)|
|Combined; only filed one prescription (vs. never user)||2.46 (2.02–3.01)a||1.73 (1.51–1.98)a||1.23 (0.57–2.66)|
Amongst patients who only redeemed one prescription for alendronate or etidronate, the unadjusted RRs for heart failure were 3.18 (95% CI 2.60–3.74) and 2.58 (95% CI 2.29–2.90), respectively. After adjustment, the HRs amongst users of alendronate and etidronate were 2.35 (95% CI 1.95–2.84) and 1.85 (95% CI: 1.63–2.08), respectively.
The results of this study show that the risk of heart failure is significantly higher in users of bisphosphonates than in age- and sex-matched control subjects, even when adjusted for differences in comorbidities and comedication. This finding is important as bisphosphonates are widely used, and heart failure constitutes a major health issue. Moreover, our results may seem surprising, given that patients treated with bisphosphonates have previously been reported to have improved survival both in observational studies  and in randomized clinical trials [3, 12]. However, increased survival was shown in a comparison not with the background population but with other patients with osteoporosis.
Our results may be partly explained by allocation bias as indicated by the finding that the increased risk of heart failure in users of bisphosphonates was also observed in patients who had only redeemed one prescription for bisphosphonates. It is well known that confounding by indication is an important problem in observational studies that is difficult to eliminate , and this fact should be acknowledged when interpreting our results. However, the risk of heart failure was not increased in users of raloxifene, which is used for treatment of the same indication as bisphosphonates, but has a different mechanism of action. The comparison between raloxifene and bisphosphonates is only possible in women, because raloxifene is not indicated for treatment of osteoporosis in men, and prescribers may have sought to avoid using raloxifene in women diagnosed with cardiovascular disease.
We adjusted for a large number of possible confounders that could account for an increased risk of heart failure in patients with osteoporosis, and we also examined whether the risk of incident heart failure was dose-dependent as this would increase the possibility of a causal relationship. Accordingly, the association between heart failure risk and dose was examined using refill compliance, as drugs to treat osteoporosis are used at the same dose for all patients in Denmark. The 5-mg daily dose of alendronate, which is available in the USA, is not used in Denmark. It is interesting that the risk of heart failure in alendronate users declined significantly with increasing exposure and reached unity at a MPR of 80% or higher. By contrast, the HR for incident heart failure in female users of etidronate was significantly increased, whereas no significant differences in the risk of heart failure were observed in patients treated with raloxifene.
These findings thus suggest an increased risk of heart failure in patients with osteoporosis and a dose-dependent reduction in this risk in patients treated with alendronate, but not etidronate or raloxifene. It is interesting that use of alendronate was linked to an inverse dose–risk relationship. Although both alendronate and etidronate are bisphosphonates, they differ considerably in their mechanism of action, with nitrogen-containing bisphosphonates such as alendronate and zoledronic acid inhibiting the enzyme farnesyl pyrophosphate synthase (FPPS) . It has recently been shown that FPPS has a role in the pathophysiology of hypertensive cardiac hypertrophy, and knockout of the gene encoding FPPS reduces cardiac hypertrophy in rats . Moreover, nitrogen-containing bisphosphonates inhibit proliferation, adhesion and migration of vascular smooth muscle cells and may have a statin-like effect on the mevalonate pathway [16, 17]. Furthermore, bisphosphonates have immunomodulatory effects that may influence the development of vascular disease, which has a key role in the development of most heart failure events. Thus, it was recently suggested that bisphosphonates may reduce the risk of myocardial infarction amongst patients with rheumatoid arthritis . It has also been reported that atrial fibrillation might be the link between bisphosphonates and heart failure [19-21], perhaps through changes in calcium dynamics including altered expression of proteins controlling sarcoendoplasmic reticulum calcium . However, recent findings suggest that the previously reported excess risk may be the result of confounding due to allocation bias and chronic obstructive pulmonary disease [23, 24]. Finally, a pathophysiological pathway linking heart failure and catabolic bone remodelling was recently identified .
Our findings in raloxifene users suggest that this drug is prescribed to women who are not at increased risk of developing heart failure, and that there is no clear intrinsic effect on this risk associated with the drug itself. Accordingly, the Raloxifene Use for The Heart (RUTH) trial of raloxifene did not show a reduced risk of cardiovascular events . However, raloxifene users may have a lower cardiovascular disease risk than the average patient with osteoporosis because physicians are likely to avoid the drug in women at increased risk of stroke or venous thrombotic events . By contrast, there are no recommendations regarding caution with bisphosphonate treatment in osteoporosis patients with cardiovascular or cerebrovascular disease risk factors. The increased risk observed in patients with extremely limited exposure to bisphosphonates (defined as less than two prescriptions dispensed or a refill compliance of less than 25%) suggests an increased baseline risk due to patient-related factors and not a drug effect. Alternatively, one would need to assume that an acute effect on myocardial function would cause patients to stop treatment, thus creating an association that was inversely dose related. In general, patients who stop taking bisphosphonates do so not because of typical heart failure symptoms such as oedema or shortness of breath but mainly as a result of upper gastrointestinal discomfort .
This study is based on complete and nationwide data covering the entire Danish population (of approximately 5.5 million). The main limitation is inherent to the observational nature of the study; demonstration of clear cause–effect relationships is not possible, and selection bias cannot be entirely excluded. However, the Danish National Patient Register includes all hospital admissions in Denmark, and data collection was independent of race, age, socio-economic status or participation in health insurance programmes; the risk of selection bias was thus minimized. Moreover, pharmacies are required to register all dispensed drug prescriptions, thus ensuring complete registration. Although prescription data had to be used as a proxy for actual use of bisphosphonates, we did not base bisphosphonate exposure on written prescriptions, but on actual pharmacy dispensing. All types of bisphosphonates can only be purchased through prescription. The agreement between drug dispensing and drug use is likely to be high, because only partial reimbursement of drug expenses is offered. Another limitation is the lack of detailed information about prognostic factors such as left ventricular ejection fraction, New York Heart Association classification, blood pressure, renal function, smoking and body weight. However, lifestyle-related diseases such as coronary artery disease, cancer, diabetes and chronic obstructive lung disease were included in our regression model. Analyses were carefully adjusted for comorbidities and concomitant pharmacotherapy, yet the possible effect of unmeasured confounders should be acknowledged.
This study demonstrates that patients treated with bisphosphonates have an increased risk of heart failure compared with the general population, probably partly due to the higher prevalence of hypertension, hyperlipidaemia, renal disease and peripheral artery disease in these patients. Increased exposure to the second-generation nitrogen-containing bisphosphonate alendronate was associated with a decreased risk of heart failure, whereas this was not the case for the first-generation bisphosphonate etidronate or the selective oestrogen receptor modulator raloxifene. Physicians should be aware that patients with osteoporosis are generally at risk of developing cardiovascular disease, including heart failure. The potential for nitrogen-containing bisphosphonates to reduce the risk of heart failure is interesting and provides a possible explanation for the observation of increased survival after hip fracture treated with the intravenously administered nitrogen-containing bisphosphonate zoledronic acid .
This study was funded by an unrestricted grant from the A.P. Møller Foundation (Fonden til Lægevidenskabens Fremme), Servier Denmark and the Dandy Foundation. None of the funding sources had any role in: the design and conduct of the study; collection, management, analysis and interpretation of the data; and preparation, review or approval of the manuscript. PV had full access to all the data, and takes responsibility for the integrity of the data and the accuracy of the analysis.
ELG has received lecture fees from AstraZeneca, Bayer, Boehringer Ingelheim and Pfizer and serves on advisory boards for AstraZeneca and Bristol-Myers Squibb. BA has received research grants or consulting fees from Amgen, Nycomed and Novartis, and speaker's fees from Amgen Eli Lilly, Merck, Novartis and Nycomed. PV has received travel grants from Novartis, Eli Lilly, Servier and Amgen.