A retrospective, observational, pharmacovigilance study was done on a de-identified publicly available FAERS dataset, which did not require IRB approval. FAERS complies with the international safety reporting guidance issued by the International Conference on Harmonization (ICH E2B). It collects AER submitted by healthcare professionals, consumers, and manufacturers around the world.

The database was searched using the reaction term “Atrial Fibrillation.” All the reported cases of AF for each drug (queried by generic names) were collected from inception to February 10, 2021, and a shortlist created of the top 30 anti-cancer drugs based on reported AF cases. The selected drugs were then individually searched using the generic names and the reported AE were collected individually for each drug. Data were further narrowed down by applying various filters such as age, received year, sex, and outcomes (Figure 1).

First, a descriptive analysis of the data was performed. Then, a comparative analysis was done among the drugs in the group. The proportional reporting ratio (PRR) (5) was calculated for various parameters to determine the disproportionality in reporting. Data on patient characteristics and outcomes were collected and analyzed.

Nilotinib had the highest proportion of CAE, which accounted for 15.6% of the total reported AE of the drug, followed by trastuzumab (14.5%), ibrutinib (12.8%), carfilzomib (12.5%), and doxorubicin (11%). Nilotinib had the highest PRR for CAE (2.05, 95% CI = 1.99–2.11), followed by trastuzumab (1.92, 95% CI = 1.87–1.97), ibrutinib (1.68, 95% CI = 1.64–1.73) and carfilzomib (1.63, 95% CI = 1.56–1.71) (Figure 2).

When analyzed for AF as a percentage of all reported AE for a particular drug, Ibrutinib had the highest percentage (5.3%) followed distantly by venetoclax (1.6%), bortezomib (1.6%), carfilzomib (1.5%), and nilotinib (1.4%). Similarly, the percentage of CAE attributable to AF was also highest for ibrutinib (41.5%), followed by venetoclax (28.4%), pomalidomide (23.9%), bortezomib (18.2%), and lenalidomide (18.2%). Drugs with the highest PRR for AF included ibrutinib (5.96, 95% CI = 5.70–6.23), bortezomib (1.65, 95% CI = 1.52–1.79), venetoclax (1.65, 95% CI = 1.46–1.86), carfilzomib (1.53, 95% CI = 1.33–1.77), and nilotinib (1.46, 95% CI = 1.31–1.63) (Figure 3).

The PRR for AF in men was highest for ibrutinib (6.95, 95% CI = 6.55–7.37), followed by venetoclax (2.17, 95% CI = 1.87–2.51), carfilzomib (1.81, 95% CI = 1.50–2.17), nilotinib (1.70, 95% CI = 1.47–1.96), and bortezomib (1.65, 95% CI = 1.47–1.84) (Supplementary Figure 1). Similarly, PRR for AF in women was also highest for ibrutinib (4.12, 95% CI = 3.78–4.49), followed by trastuzumab (1.66, 95% CI = 1.45–1.91), paclitaxel (1.59, 95% CI = 1.44–1.76), nilotinib (1.37, 95% CI = 1.13 −1.65), and bortezomib (1.32, 95% CI = 1.14–1.53) (Supplementary Figure 2).

The percentage of reported AF cases in patients who died, of the total number of deaths reported for a drug, was highest for cytarabine (35.9%), followed by capecitabine (30.2%), etoposide (26.9%), gemcitabine (26.5%) and cisplatin (25.9%). Paralleling this, the comparative mortality ratio (CMR) was highest for cytarabine (2.7), followed by capecitabine (2.2), etoposide (2.03), gemcitabine (2) and cisplatin (1.9). However, when assessed using disproportional reporting measures, the PRR for deaths in reported AF cases was highest for ibrutinib (4.62, 95% CI = 4.07–5.24), followed by carfilzomib (2.06, 95% CI = 1.51–2.82), bortezomib (1.72, 95% CI = 1.44–2.06), cisplatin (1.44, 95% CI = 1.22–1.70), and cytarabine (1.44, 95% CI = 1.21–1.71) (Figure 4).

Our comparison of anti-cancer drugs using the PRR to assess the signal-to-noise ratio of public reporting to FAERS of cardiac-related AE showed that nilotinib, ibrutinib, trastuzumab, and carfilzomib had the most CAE reported, while ibrutinib, venetoclax, bortezomib, carfilzomib, and nilotinib had the highest AF cases reported when compared to other drugs in the analysis (Figure 5). The drugs with highest reported deaths among AF cases were ibrutinib, bortezomib, carfilzomib, cisplatin, and cytarabine.

Atrial fibrillation in cancer patients has been shown to be associated with poor outcomes (7). The incidence of thromboembolism in cancer patients with new onset AF increases two-fold (8). In addition, various anti-cancer drugs increase the risk of thrombogenesis (9). Finally, the risk of bleeding in cancer patients on anticoagulants is higher than that in patients who do not have cancer (10). All these factors make the management of AF in cancer patients a unique challenge, underscoring the relevance of our findings identifying newer anti-cancer drugs that are disproportionately reported to FAERS for AF as an AE.

A variety of modalities of cancer treatment, including chemotherapy, immunotherapy, hormone and radiation therapy, and surgical resection, have each been separately associated with AF. Various anti-cancer drugs, including ibrutinib, doxorubicin, and cisplatin, amongst others, have been associated with AF (11). We have previously established that radiation therapy may be an independent risk factor for AF development (12). Postoperative AF has been associated with surgical resection of lung, colorectal, and esophageal tumors (13–15). Patients with cancer often receive multiple treatment modalities and multiple chemotherapeutic agents. The amalgamation of conventional risk factors, cancer itself, and cancer treatment creates a unique scenario in which the assessment of the specific AF risk with a certain drug can be challenging. Below, we briefly discuss some of the drugs from our analysis that had significantly disproportional reporting for AF and what is known in terms of their association with AF.

Ibrutinib is an irreversible Bruton's tyrosine kinase inhibitor that prevents downstream activation of B-cells. It is currently used in the treatment of chronic lymphocytic leukemia (CLL), mantle cell lymphoma, marginal zone lymphoma, small lymphocytic lymphoma, and the plasma cell disorder Waldenstrom's macroglobulinemia. An AF incidence of 6% was noted for ibrutinib in the landmark RESONATE and RESONATE-2 trials that led to the FDA approval of the drug (16). A meta-analysis by Yun et al. (17) of AF in patients treated with ibrutinib vs. other chemo-immunotherapeutic agents for B-cell malignancies showed an approximately nine-fold increased risk of AF following treatment with ibrutinib. Our results showing the highest AF PRR for ibrutinib (5.96, 95% CI = 5.70–6.23) are in line with the known evidence for AF and the drug. Based on a risk prediction model, the risk of AF in patients treated with ibrutinib may be modified by co-existent conditions, including age >65 years, male sex, valvular heart disease, cardiomyopathy, thyroid abnormality, chronic lung disease, diabetes mellitus, and grade-3 infections (18). The precise mechanism behind the pathogenesis of AF in connection with ibrutinib treatment is unknown.

Nilotinib is another member of the TKI family used in the treatment of CML (19), that is known to be highly vasculotoxic and cardiotoxic. No prior evidence of an association between AF and nilotinib exists in the literature. In a Phase II trial in which 73 patients treated with nilotinib were followed for 6 years, only 1 developed AF. Similarly, another study reported that AF occurred in 1 of 81 patients treated with nilotinib (20). Another study of the cardiotoxicity of TKI in the FAERS database reported a significantly increased adjusted Reporting Odds Ratio (aROR) for cardiac arrhythmias following treatment with nilotinib (2.7), but did not specify the type of arrhythmia (21). While our study highlights a potential novel association, further prospective studies are warranted to confirm the association and determine the causation.

Bortezomib and carfilzomib are both proteasome inhibitors (PI) that act by inhibiting the NF-κB pathway on multiple myeloma (MM) cells leading to cell death. The available data on the incidence of AF in connection with these drugs are scarce and inconsistent. A study of four Phase II clinical trials of carfilzomib showed that cardiac arrhythmias developed in 13.3% of the patients (n = 526); however, 73.6% of the patients enrolled had a prior history of cardiovascular events (22). Another meta-analysis of cardiovascular adverse events related to treatment with carfilzomib reported cardiac arrhythmias in 2.4% of patients from a total of 13 studies (23). A smaller prospective study (n = 95) compared the CAE that resulted from treatment with carfilzomib to those of bortezomib. Results showed that the CAE were significantly higher in patients treated with carfilzomib than in those treated with bortezomib (51% vs. 17% respectively), consistent with our results (PRR for CAE: carfilzomib, 1.63 (95% CI = 1.56–1.71) vs. bortezomib, 1.15 (95% CI = 1.11–1.19). The incidence of AF was comparable for the two drugs, with 2 of 65 patients treated with carfilzomib and 1 of 30 patients treated with bortezomib developing AF (24). Similar to these results, we found a comparable increase in the reporting of AF for both bortezomib and carfilzomib [1.65 (95% CI = 1.52–1.79) and 1.53 (95% CI = 1.33–1.77), respectively].

One difficulty in differentiating between these two drugs and their association with AF is that carfilzomib is mostly reserved for relapsed or refractory MM and most of the patients receiving carfilzomib had already been treated with bortezomib (25). Bortezomib is one of the newer anti-cancer drugs identified for its association with AF by Alexandre et al. in their pharmacovigilance study based on the World Health Organization's Vigibase AE dataset; however, any association of AF with carfilzomib was not reported (26). Even though our study highlights a potential association between proteasome inhibitors and AF, the results should be looked upon with caution. Due to lack of comorbidity data and temporal association, causation and definitive conclusions cannot be drawn. An important confounding factor in these drugs is the high prevalence of cardiac disease in MM patients due to increased age, hyper-viscosity, arteriovenous shunts, anemia, and amyloidosis (27). It cannot be determined whether the AF was caused by the drugs alone, or by the combination of all other factors, but the over-reporting in studies from two different databases merits further investigation.

Venetoclax is a B-cell lymphoma (BCL-2) homology domain-3 (BH-3) mimetic, selective B-cell lymphoma (BCL-2) inhibitor used in the treatment of lymphomas and leukemias (28). We found increased reporting of AF with venetoclax (PRR 1.65, 95% CI = 1.46–1.85). It is difficult to positively identify the association of venetoclax with AF, because it is used either as first line for the treatment of CLL or used subsequently after progression on ibrutinib, that is known for its association with AF. A Phase II RCT investigating the combination of ibrutinib and venetoclax used to treat high-risk elderly CLL patients showed an incidence of AF in 15% of the patients (29) that is nearly the same as that reported for ibrutinib (30), providing strong evidence for ibrutinib as the cause of AF rather than venetoclax (29). No studies have reported on the incidence of AF during treatment with venetoclax alone.

Pomalidomide is a novel immunomodulatory agent used in the treatment of MM. Prior to our study, pharmacovigilance analysis of the Vigibase dataset showed that AF was reported to occur following treatment with pomalidomide (26). It should be noted that this drug is approved for use in treating MM that is refractory to at least two prior therapies, including lenalidomide and bortezomib, both of which have been reported to be associated with AF. It is important to note that patients with MM are at increased risk of AF in the first place due to age and high cardiovascular disease burden. While there is increased reporting of AF with the above-mentioned drugs, the results should be interpreted with caution as the association does not imply causation.

AF increases the risk of heart failure, bleeding complications, myocardial infarction, and all-cause mortality (7, 31). Although a large number of AF patients reported to be on treatment with cytarabine (35.9%), etoposide (26.9%), and gemcitabine (26.5%) died compared to AF patients being treated with other chemotherapeutic agents, this could reflect the severity of the cancer being treated with these drugs rather than an association with the AF reported for the drug. These traditional chemotherapeutic drugs are commonly used to treat cancers with unfavorable survival rates such as AML, pancreatic cancer, bladder cancer, metastatic lung cancer, and ovarian cancer. Therefore, to refine the signal-to-noise ratio, we calculated the PRR for deaths in patients with reported AF. Ibrutinib had a significantly higher PRR for deaths in patients with reported AF, almost three-fold higher than the PRR for the next highest anti-cancer drug, carfilzomib. This suggests that the adverse events related to AF with Ibrutinib could be associated with increased mortality and merits further prospective studies.

Recently, a pharmacovigilance study by Alexandre et al. (26) based on the Vigibase dataset identified 19 anti-cancer drugs associated with AF. Of the 19 drugs identified, new associations with AF were reported for nine drugs including lenalidomide, pomalidomide, nilotinib, ponatinib, midostaurin, azacytidine, clofarabine, docetaxel, and obinutuzumab. Our results confirm the over-reporting of AF with lenalidomide, pomalidomide, and nilotinib, but our PRR for docetaxel was <1. In addition, ponatinib, midostaurin, azacytidine, clofarabine, and obinutuzumab were not included in our analysis due to the small number of cases reported to FAERS. For example, midostaurin has had no AF cases reported to FAERS, and the study by Alexandre et al. had only 20 AF cases. On the other hand, we found significant over-reporting of cases for AF following use of carfilzomib and venetoclax.

In general, FAERS had a higher number of total AF cases reported for most of the chemotherapeutic drugs compared to reports in Vigibase, which is a global database. In addition, we used a different quantitative signal detection tool than Alexandre et al. used; in their study, the Reporting Odds Ratio (ROR) was used as a measure of disproportionality analysis, but on the other hand we used PRR. While PRR and ROR are two different measures and have their own advantages and drawbacks, both are authentic qualitative signal detection tools (5). Another difference is that Alexandre et al. adjusted the ROR for age, sex, geographical region, and comorbidities, whereas inconsistent reporting in FAERS prevented us from performing such adjustments.