Introduction

Bisphosphonates (BPs) are a class of synthetic compounds that reduce bone resorption by inhibiting osteoclast activity, which are widely used for the treatment of various bone and joint diseases, such as osteoporosis, Paget’s disease of bone and malignancy-associated bone disease1,2,3,4,5. Development of BPs has reached their third generation, including clodronate disodium, etidronate disodium, zoledronic acid, risedronate sodium, ibandronate sodium, alendronate sodium and pamidronate disodium.

The proarrhythmic effect of BPs is a significant concern in using them, which showing conflicting results in previous studies6,7,8,9,10,11,12,13,14,15. The definitive confirmation of association between bisphosphonates and cardiac arrhythmia has since been strongly debated because some6,7,8,9,13,14 but not all studies have identified a similar risk10,11,12,15. A large population cohort study showed that zoledronic acid users had a higher risk of arrhythmias (adjusted hazard ratio 1.18 [1.06–1.31]) and atrial fibrillation (adjusted hazard ratio 1.18 [1.05–1.32]) compared with untreated subjects8, while another Danish cohort study revealed a reduced risk of cardiac arrhythmia [0.61; 0.41 to 0.89] and atrial fibrillation [0.67; 0.55 to 0.81] associated with oral BP use10. A meta-analysis including 9 randomized controlled trials (RCTs) with zoledronic acid observed elevated risk of arrhythmia and atrial fibrillation in women with primary osteoporosis6, which contradicted the results of another meta-analysis15. Furthermore, it is unclear if proarrhythmic effects of bisphosphonates are class-specific or related to individual drugs. The two recent pharmacovigilance studies did not comprehensively and systematically describe the effects of different BPs on different arrhythmias: both studies included only two kinds of BP, and one study was limited by grouping six arrhythmias as a single outcome, making it impossible to determine the differences of these BPs on different arrhythmias, while the other study included only three arrhythmias7,16. Moreover, the outlined relationship between BPs and arrhythmias, potential signal spectrum, factors associated with death, and clinical information of BP–related arrhythmias are still unclear.

Given the increasing use of BPs in real world, and the conflicting risk of arrhythmia induced by BPs, we conducted a systematic review to investigate arrhythmias in different BP therapies, leveraging real-world pharmacovigilance data from the FDA Adverse Event Reporting System (FAERS) database.

Methods

Data source

To explore the relationship between BPs and cardiac arrhythmias, we used the FAERS database containing spontaneous adverse event reports from the first quarter of 2016 to the second quarter of 2022 to perform the pharmacovigilance study. The bisphosphonates in the study included zoledronic acid, risedronate sodium, ibandronate sodium, alendronate sodium, pamidronate disodium, clodronate disodium and etidronate disodium. Open Vigil FDA, an application programming interface (API), was a pharmacovigilance tool adapted to extract FAERS data by removing duplicate reports or reports with missing data1714. We used OpenVigil 2.1 to query the FAERS database and retrieve reports on the generic name of BPs (“zoledronic acid"+“risedronate sodium”+“ibandronate sodium”+“alendronate sodium”+“pamidronate disodium”+“clodronate disodium”+"etidronate disodium”) from January 1, 2016 to June 30, 2022.

Definition of AEs

Based on Medical Dictionary for Regulatory Activities (MedDRA), the high-level group term (HLGT) we investigated was “Cardiac arrhythmias (10007521).” The complete list of preferred terms (PTs) within cardiac arrhythmias can be accessed in supplemental table S1. The above PT level adverse events (AEs) were grouped into the following four High Level Terms (HLTs): “Ventricular arrhythmias and cardiac arrest [10047283],” “Cardiac conduction disorders [10000032],” “Rate and rhythm disorders NEC [10037908],” and “Supraventricular arrhythmias [10042600].” Moreover, we collected the following demographic features of AE cases: patient’s characteristics (age, gender, country of origin), drug information (drug indication, receipt date, concurrent medications, treatment start and end date), and arrhythmia characteristics (seriousness, final outcome).

Statistical analysis

Descriptive analysis was performed for the clinical characteristics of BPs-associated cardiac arrhythmias. We used Chi-square (Chi2) tests to compare categorical variables between the fatal and non-fatal groups. To compare the onset time of BP-associated arrhythmias, we used t-test and nonparametric test (Kruskal Wallis test). Disproportionality analysis was performed to identify potential safety signals for BP-associated arrhythmias by using reporting odds ratio (ROR) and information component (IC)18. The calculation formulas for IC and ROR were as follows:

$${\text{ROR}}=\frac{{{N_{{\text{observed}}}}+0.{\text{5}}}}{{{N_{{\text{expected}}}}+0.{\text{5}}}}$$
$${\text{95}}\% {\text{CI}}={{\text{e}}^{{\text{In}}({\text{ROR}})}}^{{ \pm {\text{1}}.{\text{96}}}}\sqrt {\frac{1}{a}+\frac{1}{b}+\frac{1}{c}+\frac{1}{d}}$$
$${N_{{\text{expected}}}}=\frac{{{N_{{\text{drug}}*}}{N_{{\text{event}}}}}}{{{N_{{\text{total}}}}}}$$
$${\text{IC}}={\text{lo}}{{\text{g}}_{\text{2}}}\frac{{{N_{{\text{observed}}}}+0.{\text{5}}}}{{{N_{{\text{expected}}}}+0.{\text{5}}}}$$
$${\text{I}}{{\text{C}}_{0{\text{25}}}}={\text{IC}}\, - \,{\text{3}}.{\text{3}} \times {\left( {{N_{{\text{observed}}}}+0.{\text{5}}} \right)^{ - 0.{\text{5}}}} - {\text{2}} \times {\left( {{N_{{\text{observed}}}}+0.{\text{5}}} \right)^{ - {\text{1}}.{\text{5}}}}$$
$${\text{I}}{{\text{C}}_{{\text{975}}}}={\text{IC}}\,+\,{\text{2}}.{\text{4}} \times {\left( {{N_{{\text{observed}}}}+0.{\text{5}}} \right)^{ - 0.{\text{5}}}} - 0.{\text{5}} \times {\left( {{N_{{\text{observed}}}}+0.{\text{5}}} \right)^{ - {\text{1}}.{\text{5}}}}$$

Nobserved represented the count of observed records for the target drug AEs, Nexpected was the anticipated count of such records. Ndrug was the total count of records for the target drug irrespective of AEs. Nevent was the total count of records for the specific AEs, regardless of the drug involved. Lastly, Ntotal signified the overall count of records in the entire database. If there were not less than three reports and one algorithm was positive, the arrhythmia AE was considered to be associated with BP treatment. In this study, all statistical analyses were conducted using SPSS 24.0, and p values < 0.05 was considered statistically significant.

Results

Descriptive analysis

From the first quarter of 2016 to the second quarter of 2022, the FAERS database recorded 91,588 bisphosphonate-related AEs and 164,227 cases associated with cardiac arrhythmias. We identified 2377 reports of bisphosphonate-related cardiac arrhythmias and presented the clinical characteristics of cases in Table 1. Among individual areas, the North America reported the highest number of arrhythmia AEs with 52.25% (1242/2377). The median age of patients was 70 years (interquartile range [IQR] 59–78), and 61.88% of BP recipients were female. Zoledronic acid (1020 reports, 42.91%), alendronate sodium (823 reports, 34.62%), and risedronate sodium (352 reports, 14.81%) accounted for a higher proportion of events. As shown in Table 1, there was no significant difference in patient age, sex, reporter, region, BP regimen between fatal and nonfatal reports.

Table 1 Characteristics of patients with BP-associated cardiac arrhythmias sourced from the FAERS database (January 2016 to June 2022).
Full size table

Signal values associated with different bisphosphonate regimens

The signal values and the correlation between bisphosphonates and arrhythmias were shown in Table S1 and Fig. 1. All studied bisphosphonates except ibandronate sodium and clodronate disodium were significantly associated with the reporting frequency of cardiac arrhythmia (HLGT), with ROR ranging from 1.28 with pamidronate disodium to 11.08 with etidronate disodium (Fig. 1). Regarding four specific arrhythmias in HLT level, the bisphosphonates with the highest ROR were etidronate disodium in cardiac conduction disorders (ROR025 = 66.5) and supraventricular arrhythmias (ROR025 = 26.17), risedronate sodium in rate and rhythm disorders (ROR025 = 1.54), and pamidronate disodium in ventricular arrhythmias (ROR025 = 1.42). Ibandronate sodium presented no signal in the above four specific arrhythmias respectively.

Fig. 1
figure 1

The bar chart displays the AEs at HLGT/HLT level. (A) Rank by AE cases (N); (B) Rank by ROR025. HLGT: high-level group term; HLT: high level term; N: number of records; ROR025: the lower end of the 95% confidence interval of ROR.

Full size image

The signal spectrum of cardiac arrhythmias differs in bisphosphonate strategies

The arrhythmia signal spectra of different bisphosphonate therapies were shown in Fig. 2. Zoledronic acid presented a broadest spectrum of cardiac arrhythmias AEs, with 9 PTs detected as positive signals, ranging from supraventricular tachycardia (IC 025 = 0.14) to atrial fibrillation (IC 025 = 1.17). There were 6 PTs as signals associated with risedronate sodium, with signal values ranging from IC 025 = 0.02 (tachycardia) to IC 025 = 1.68 (supraventricular extrasystoles). However, the drug with the least PTs was ibandronate sodium, with no signal detected, followed by pamidronate disodium, with one signal detected. Atrial fibrillation was the most overlapping PTs, which was found significantly associated with zoledronic acid, risedronate sodium, alendronate sodium and etidronate disodium. Bundle branch block left and atrial fibrillation were detected as the strongest two signals in etidronate disodium (IC 025 = 5.56; IC 025 = 4.36). No BPs showed positive signals in cardiac flutter, bradycardia, extrasystoles, sinus bradycardia, cardiac arrest, ventricular tachycardia and ventricular fibrillation.

Fig. 2
figure 2

Arrhythmia Signal Profiles of Different bisphosphonates Strategies.

Full size image

Time to onset of BP-associated cardiac arrhythmia adverse effects

The onset time of BP-associated arrhythmia was shown in Table S2 and Fig. 3. According to all BPs, the median time to onset (TTO) is 71 days (IQR 2–444 days). There were significant differences in the reported TTO of arrhythmias among BP regimens (p < 0.001). The median TTO was 42 days for zoledronic acid (IQR 1–355), 38 days for risedronate sodium (IQR 10.25–873.75), 251 days for ibandronate sodium (IQR 23.5–649.0), 145 days for alendronate sodium (IQR 13–741), 92 days for pamidronate disodium (IQR 38–356), and 59 days for clodronate disodium (IQR 16–122), respectively.

Fig. 3
figure 3

Onset time of BPs-associated arrhythmias.

Full size image

Discussion

This study is the first pharmacovigilance study to comprehensively evaluate the cardiac arrhythmia AEs associated with BPs based on the FAERS database, and to explore the clinical characteristics of BP-induced arrhythmia AEs.

Our research detected significant signals between cardiac arrhythmia and all BPs except ibandronate sodium and clodronate disodium, with ROR ranging from 1.28 with pamidronate disodium to 11.08 with etidronate disodium. Notably, the risk signals of BP are different, with high-risk signal of arrhythmia detected in etidronate disodium, no risk of signal of arrhythmia detected in ibandronate sodium and clodronate disodium, and modest risk signal of arrhythmia detected in other BPs. The risk of cardiac arrhythmia of BPs varied in different literatures, some of which showed a positive arrhythmia risk of BPs6,8,13,14,16,19,20,21, while others showed the opposite10,11,12,15,22,23. In HORIZON clinical trial, zoledronate versus placebo presented an increased risk of arrhythmia and AF (6.9% vs. 5.3%, p = 0.003; 1.3% vs. 0.5%, p < 0.001)20. Moreover, a case-control study from the United States showed that the ever-users of bisphosphonates had a higher risk of incident atrial fibrillation compared to never-users (OR 1.86, 95% CI 1.09–3.15)21. Interestingly, the risk for AF was inversely correlated with adherence to BPs22, which was supported by another research23. Two large population cohort studies respectively revealed the opposite risk of arrhythmia and atrial fibrillation for BPs8,10. A meta-analysis re-assessing all the RCTs employing alendronate found no increased risk of AF for this drug in any single trial nor in the pooled analysis15. Following the study, the meta-analysis including 9 RCTs with zoledronic acid observed elevated risk of arrhythmias and atrial fibrillation in women with primary osteoporosis6. Given the conflicting results presented by RCTs and observational studies, previous research recommended avoiding the prescription of BPs in patients with a history of atrial fibrillation24,25; however, the varying signal strengths of BPs observed in our study may not only provide more options for such patients, such as choosing ibandronate sodium or clodronate disodium with negative arrhythmia signals, and avoiding etidronate disodium with a strong positive arrhythmia signal, but also favor clinicians to spontaneously report arrhythmia incidents through heightened awareness and enhanced development of pharmacovigilance frameworks. The mechanisms that might explicate such a relation remained unclear but it had been suggested that hypocalcemia and related secondary hyperparathyroidism might be potential factors for triggering arrhythmias26. Supporting the possibility that the risk of arrhythmia may not be uniform across different BPs may be due to difference in predominant use27, affinity for bone and powerful inhibition of resorption28, but may not be attributed to the nitrogen containing moiety23. Both our study and previous research had shown that BPs used primarily for neoplastic indications or rare bone disorders presented a lower risk of arrhythmia compared to those used for non-neoplastic indications27. Although the benefits of BP use may outweigh the modest increased risk of incident arrhythmia for most patients, the findings of this study require further exploration and discovery of these causal relationships in clinical studies to further validate their value.

Atrial fibrillation (AF) induced by zoledronic acid, risedronate sodium, alendronate sodium and etidronate disodium was over-reported, but only the signal intensity of etidronate disodium was strong; ibandronate sodium and pamidronate disodium did not present a significant signal value. This was the first comprehensive study to compare the effects of different BPs on atrial fibrillation, and to find that the risk of atrial fibrillation varied among different BPs. The previous studies showed that either all BPs in the report were associated with atrial fibrillation13,14,19, or none of the BPs are associated10,11,12. Our study shows that zoledronic acid, risedronate, and alendronate indicated a similar risk of atrial fibrillation, which is consistent with previous clinical data for the three BPs indicating no significant differences from placebo in overall incidence of AF AEs29. The no signal value of ibandronate sodium and pamidronate disodium in AF found in our study, which was supported by a meta-analysis of the pivotal RCTs for ibandronate30 and a pharmaco-epidemiological study detecting a lower risk of severe AF following pamidronate acid compared to zoledronic acid use27. Atrial fibrillation and serious atrial fibrillation events with ibandronate were 0.8% and 0.4% in previous study30, which is lower than that reported in the study on zoledronic acid (2.4% and 1.3%, respectively)20. Ibandronate and pamidronate were deemed to be safe overall in AF, which required further research and certification. Several mechanisms were postulated for the increased risk of AF with BPs, including calcium shifts after the infusion, cytokine release, and deposition of bisphosphonates in myocardial tissue31,32 which needed further validation in future studies.

Our study demonstrated that BP-related arrhythmias occurred early after BP initiation, with significant differences among different BP therapies (p < 0.001). The median TTO of BP-related arrhythmias was 71 (IQR 2- 444) days, and most known cases (406/969, 41.49%) occurred within the first thirty days after the initiation of BP therapy, which indicated the significance of early cardiac monitoring within the higher-risk time window of 30 days. It is important to recognize the difference in onset time of arrhythmias among different BP regimens, providing evidence for effective interventions in affected population.

This study involves several limitations that should be acknowledged. Firstly, the FAERS database has some inherent limitations, including data duplication, missing data, and reporting bias (e.g., selective reporting, inevitable underreporting, and false or inaccurate reporting), which can lead to disproportionate analysis inaccuracies, confuse signal detection work, and result in incomplete description of drug safety. Additionally, as a spontaneous reporting system, the FAERS database does not provide detailed clinical information. Secondly, it is difficult to control for confounding factors, such as medication history for patients, concomitant diseases status, potential drug interactions, the use of multiple drug combinations and other health conditions, all of which may influence the risk of arrhythmias. Thirdly, data mining using Bayesian analysis and non-proportional analysis can only indicate a statistical association between adverse reactions and drugs rather than a causal relationship. The slight increase in ROR does not indicate a risk of AEs in clinical practice, which may be relevant. Therefore, these values only provide safety signals, not real risks, and require further confirmation. Moreover, future pharmacovigilance efforts need to explore the use of advanced artificial intelligence methods, such as employing machine learning33,34,35 and deep learning models35,36 to identify patterns in adverse event data, while also utilizing natural language processing to extract insights from unstructured text fields in the FAERS database.

Conclusion

We reviewed BP-associated arrhythmia AEs from the FAERS database to explore the safety profile of BPs for arrhythmia and optimize their use among individual patients. Our study showed that BPs presented different spectrum and risk of positive arrhythmia signals, with high-risk signal of arrhythmia detected in etidronate disodium, no risk of signal of arrhythmia detected in ibandronate sodium and clodronate disodium, and modest risk signal of arrhythmia detected in other BPs. Our study is practical for clinicians to enhance the management of BP-related arrhythmia AE signals and improve BPs treatment safety. Due to the spontaneous nature of the FAERS reporting system, which includes omissions, misreporting, and incomplete reports, this study could not definitively infer a causal relationship between BPs and arrhythmias. Therefore, further clinical research is needed in the future to validate this association.