The association of trastuzumab with atrial fibrillation and heart failure in breast cancer patients in routine clinical practice: a population-based propensity score matching and competing risk model analysis

Purpose Trastuzumab, a potent anti-human epidermal growth factor receptor 2 (HER2) monoclonal antibody, is conditionally reimbursed by the Taiwan National Health Insurance (NHI) for HER2-positive breast cancer (BC). Trastuzumab-induced cardiotoxicity studies have well characterized heart failure (HF) but fewer addressed arrhythmia, particularly the association of potential life threatening atrial fibrillation (Af) is poorly characterized. We aimed to study the trastuzumab-related risk of Af and HF using the claimed data of Taiwan NHI. Methods A nationwide retrospective cohort of patients with BC from the Taiwan NHI reimbursement database from January 2007 to December 2016 was analyzed. Propensity score matching and competing risk model analysis were used for adjusting confounding concurrent medication or comorbidities and competing events. The HF study was used to validate the method used. Results For Af, 12,472 trastuzumab users were matched with 12,472 non-trastuzumab users. For HF, 12,241 trastuzumab users and 12,241 non-users were enrolled. We found that trastuzumab users had significantly worse HF-free survival but not Af-free survival than non-trastuzumab users. In the competing risk analysis, the use of trastuzumab did not increase the risk of Af (hazard ratio [HR] 0.76, P = 0.0006) but was associated with HF (HR 1.19, P = 0.0052). The risk trends among stratifications by comorbidities and concurrent medication remained in similar directions for both Af and HF. Conclusion Trastuzumab in real-world practice was associated with an increased risk of HF, but was not associated with an increased risk of Af in BC patients. Trastuzumab-induced arrhythmogenic effects may be masked by concurrent heart-protecting measures, more prominent roles of comorbidities or concurrent medications under real-world settings. Further studies are required. Supplementary Information The online version contains supplementary material available at 10.1007/s10549-022-06753-7.


Introduction
Amplification and overexpression of the human epidermal growth factor receptor 2 (HER2) drive the proliferation, aggression, and metastatic behaviors of the breast epithelial cells [1,2]. Twenty to 25% of patients with invasive breast cancer (BC) have high HER2 expression [2,3]. Trastuzumab is a recombinant humanized monoclonal antibody that binds to HER2, and has been approved in treating HER2-positive BCs [4][5][6]. In Taiwan, HER2 positive BC affected 1 in 120 women [7]. The Taiwan's National Health Insurance (NHI) has reimbursed trastuzumab for HER2 positive BCs under several conditions: (1) One-year adjuvant and or neoadjuvant therapy for HER2-positive nodal positive early BC. (2) HER2-positive metastatic BC.

3
Cardiac toxicities had long been a concern when using HER2 receptor antagonists, especially trastuzumab. As early in 2002, after analyzing seven phase 2 or 3 studies regarding combination use of trastuzumab and other chemo-regimen in advanced disease, Seidman et al. suggested that trastuzumab was associated with congestive heart failure (CHF) [8]. Large randomized trials demonstrated that trastuzumab, combined with chemotherapy regimens, especially anthracycline-base, caused higher cardiac dysfunction and subclinical loss of left ventricular ejection fraction (LVEF) in patients with early BC [4,9,10]. According to the US Food and Drug Administration (FDA) package insert, trastuzumab might cause "left ventricular dysfunction, arrhythmias, hypertension (HTN), disabling cardiac failure, cardiomyopathy, and cardiac death." [11].
Besides CHF, other cardiac insults were less mentioned until recently. Also, most of the trastuzumab-induced cardiotoxicity research has been focused on heart dysfunction rather than the detailed description or rhythm of arrhythmias [12]. Furthermore, the established data on trastuzumab on cardiotoxicity in Asia was still limited.
We evaluated the characteristics, risk factors, and prognosis of heart failure (HF) and atrial fibrillation (Af), the most common arrhythmias predisposed to severe thromboembolism effects, from the Taiwanese National Health Insurance Research Database (NHIRD).

Methods
This study has been approved by the institutional review board of Taipei Veterans General Hospital and conducted following the Declaration of Helsinki. We implemented a nationwide retrospective cohort study extracting data from the patients having visits or hospitalizations for BC from Taiwan's NHIRD. The NHIRD consists of the medical fee claims; it was contributed by enrollees' medical utilization, including more than 99% of the entire population in Taiwan. All data were extracted via the Applied Health Research Data Integration Service from Taiwan National Health Insurance Administration (NHIA). All applications are reviewed for approval of data release. Data analysis was guided by the monitoring regulation guidelines of the Scientific Data Center of the Ministry of Health and Welfare, Taiwan. The application for data and processing of the data were all conducted via the service and application process were detailed described previously [13,14].
Female patients who were visited or admitted with a diagnosis of BC for the first time between 2007 and 2016 were identified. The patients were excluded if the incidence of HF and Af preceding index date. Meanwhile, the BC diagnosis was confirmed by the catastrophic illness database, another part of NHIRD. A tissue proof with an immunohistochemistry staining for the HER2 protein of 3 + or positive on fluorescence in situ hybridization was required to confirm HER2 positive. Due to the regulation of HER2 reimbursement, this medication was reimbursed for satisfying the previous requirement. The patients were divided into two groups based on their use of trastuzumab, identified with specific drug codes in NHIRD, treated group, and untreated group, respectively. Af and HF were the primary outcomes in this study, and the follow-up for both groups began on the index date, and they followed up until the date of Af/HF hospitalization, death, or end of 2016. Because the chance of primary outcomes can be confounded by competing risks of mortality, a proportional hazards model for the cumulative incidence of a failure cause of primary outcomes, proposed by Fine and Gray [15], was conducted to evaluate the related hazards. Propensity scores, the probability of evaluating a patient to the treated group given control covariates, were implemented using greedy matching. The control covariates included age, comorbidities, and medications.
Demographic data included age, comorbidities and associated medications. The comorbidities listed by ICD 9 or ICD 10 codes (Supplement 1) included HTN, diabetes mellitus (DM), chronic kidney diseases (CKD), and hyperlipidemia or dyslipidemia. The medication used was defined by the Anatomical Therapeutic Chemical Classification system proposed World Health Organization, including aspirin, beta-blockers, renin-angiotensin system (angiotensin-converting enzyme inhibitor, ACEI/angiotensin receptor blocker, ARB), non-steroid anti-inflammatory drugs (NSAIDs), statins, and metformin. The codes for the medication mentioned above were demonstrated in Supplement 2. Medication use was defined as the cumulated prescription days of more than 90 days.
To clarify the association between the cardiotoxic anthracyclines and cardiac outcomes, we further analyzed the frequency of cardiac examinations and the relationship between BC-specific anthracyclines and Af/HF in enrolled patients. The examinations for HF included echocardiography with or without color Doppler, ventricular ejection fraction (EF) or multi-gated acquisition (MUGA) radionucleotide scanning. The examinations to detect Af included 12-lead-electrocardiogram (ECG), Treadmill-exercise ECG, and 24 h Holter ECG monitoring.
Data management and analysis were performed using SAS 9.4 software (SAS Institute Inc). Calculated results were presented as the estimated number and the 95% confidence.

Results
A total of 100,861 women diagnosed with BC without known Af and 99,283 women without HF before the index date from 2007-2016 were enrolled for propensity score matching. Baseline demographic characteristics of patients before and after the matching were illustrated in Table 1. There were a substantial proportion of BC patients with comorbidities in the entire cohort, including 19.7% patients with DM, 17.1% patients with hyperlipidemia, up to 33.9% patients with HTN, and a much smaller proportion with CKD (Table 1a and b).
Kaplan-Meier curves demonstrating HF-free and Affree survival in patients with and without trastuzumab are illustrated in Fig. 1a and b, respectively. After matching, patients who received trastuzumab treatment had worse HFfree survival than non-trastuzumab patients (P < 0.0001). In contrast, a patient who received trastuzumab treatment was not associated with inferior Af-free survival (P = 0.13).
Next, we performed a competing risk analysis to estimate the effects of trastuzumab on HF and Af by considering death as a competing event to HF or Af after adjusting for comorbidities and concurrent medications. Tables 2a and b estimate the effect of trastuzumab on HF and Af from multivariate cox proportional hazards models. As shown in Table 2, after adjusting for age, DM, HTN, hyperlipidemia, CKD, Aspirin, NSAIDs, Statins, metformin, beta-blockers, and renin-angiotensin system (angiotensin-converting enzyme inhibitor, ACEI/angiotensin receptor blocker, ARB) drugs, trastuzumab treatment is associated with increased risk of HF, with an estimated hazard ratio (HR) of 1.193 (95% confidence interval [CI] 1.054-1.351, P = 0.0052). In contrast, trastuzumab treatment is not associated with an increased risk of Af (HR = 0.759 [95 CI 0.649-0.889], P = 0.0006). We also examined the risk of trastuzumab in each subgroup stratifications stratified by comorbidities and concurrent medications ( Fig. 2a and b). As shown in Fig. 2a, the increased risk of HF by trastuzumab treatment remained in similar trends in each subgroups stratified by comorbidities and concurrent medication. In contrast, Fig. 2b demonstrates a decreased risk of Af by trastuzumab with a similar trend in each stratified subgroups. Interestingly, wider ranges of confidence interval of hazards were seen in patients with CKD, in both analyses of Af and HF, probably due to much smaller numbers of patients with CKD.
Further evaluations of the effects of BC-specific anthracyclines were shown in Supplement 3 and Supplement 4. In supplement 3, there were slightly more anthracyclines usage in non-trastuzumab patients, reflecting the real-world practice to avoid concomitant trastuzumab and anthracycline treatment. The proportion of cardiac examinations for cardiac ejection function and rhythms, doxorubicin (including conventional and liposomal doxorubicin), and epirubicin were similar in Af and HF groups (Supplement 3). After adjusted for comorbidities, concurrent medications including anthracyclines (doxorubicin and epirubicin), trastuzumab treatment is still significantly associated with increased risk of HF, but is not associated with an increased risk of Af, the hazard ratios of trastuzumab treatment on outcome were similar compared to Table 2a and b (Table 2a and b, and Supplement 4).

Discussions
Cardiotoxicities induced by trastuzumab have long been established. In this study, we utilized the Taiwan NHIRD data to evaluate the risk of Af and HF in patients treated with trastuzumab vs. matched-non-trastuzumab controls. We found trastuzumab users were associated with worse HF-free survival and increased association with HF in a competing risk analysis, but this was not seen with Af.
Considering the impacts on cardiac outcomes on concurrent cardiovascular agents, we analyzed how aspirin, NSAIDs, statins, metformin, beta-blockers, and RAAS inhibitors affect the heart AEs. Aspirin, beta-blockers, and RAAS inhibitors were associated with a higher incidence of heart failure and atrial fibrillations. It was impressive because beta-blockers and RAAS agents were standard treatments for CHF. One possible reason was that patients taking those drugs were prone to have AEs. Aspirin, NSAIDs, statins, metformin, beta-blockers, and RAAS drugs might represent the severity of the underlying diseases. In the subgroup analysis, we found that metformin and statins therapy was associated with a lower incidence of Af. The results were partially consistent with previous studies that statin therapy [16] and metformin [17] were significantly associated with a reduced risk of incidence or recurrence of Af. Aspirin was not recommended to prevent stroke in patients with Af as monotherapy [18] and might result in the association of aspirin and lower Af rate.
Unexpectedly, the results revealed that the trastuzumab group had a lower risk of developing Af than the control arm. We believed the result was valid since a similar population was analyzed for HF. The result was consistent with the previous studies that trastuzumab was associated with increased heart failure incidence. Prior Taiwan NHIRDbased studies have also examined the potential confounding effects of anthracyclines on the cardiotoxicities [19,20]. Chang et al. [19] found BC patients receiving trastuzumab had an increased risk of major adverse cardiac and cerebrovascular event (MACCE), especially HF, a similar but some different outcome endpoint compared to our current study.   is independent of anthracyclines. Similarly, Chieh et al. [20] studied an earlier NHIRD cohort and disclosed that both anthracyclines and trastuzumab contribute to increased risk of HF and cardiomyopathy. Accordingly, our additional analysis (Supplement 3 and 4) revealed that there were slightly more anthracyclines usage in non-trastuzumab patients, reflecting the real-world practice to avoid concomitant trastuzumab and anthracycline treatment. After being adjusted for comorbidities, concurrent medication including anthracyclines, trastuzumab treatment is associated with increased risk of HF, but is not associated with an increased risk of Af. This additional analysis also supported prior NHIRDbased studies which showed trastuzumab-induced HF risk is independent to anthracyclines. Moreover, it is possible that transient arrhythmia during trastuzumab might have been missed and not coded in the NHIRD claims. This is apparently a limitation of this NHIRD-based study [13,14]. Nevertheless, we additionally examined the correlated cardiac exams among trastuzumab and matching non-trastuzumab users (Supplement 3), we found similar frequency patterns of cardiac examinations among Af and HF cohorts. Under these similar frequency patterns of cardiac examinations, the trastuzumab-associated risks for HF and Af still showed a distinct difference in hazard ratios (Supplement 3 and 4).

Chang et al. revealed trastuzumab-induced risk of MACCE
Recently, a meta-analysis from 15 studies involving 8124 patients has reported an overall 1.2% incidence of Af in BC patients receiving trastuzumab. The incidence seemed not influenced by the formulation of trastuzumab, the additional use of neoplastic agents, anthracycline exposure status, or concurrent radiotherapy [21]. There have been several Table 2 Multivariate cox proportional hazards with competing risk analysis models for breast cancer patients developing (a) heart failure (b) atrial fibrillation C.I confidential interval, HR hazard ratio, DM diabetes mellitus, HTN hypertension, CKD chronic kidney disease, NSAID non-steroid antiinflammatory drug, ACEI angiotensin converting enzyme inhibitor, ARB angiotensin receptor blockers  in vitro mechanistic studies showing arrhythmogenic effects of trastuzumab under various doses [22][23][24][25]. In contrast, a minority reported trastuzumab at a concentration of 10 mg/L did not trigger more arrhythmias in rat hearts [26]. Unspecified arrhythmias were reported in the relatively little literature regarding arrhythmias in trastuzumab use in HER2 high expression BC patients. As early as 2001, Behr et al. tried to approach trastuzumab-induced cardiotoxicity by applying 111In-DTPA-trastuzumab to 20 patients. Only one developed arrhythmia among the seven patients showing isotope uptake in myocardium on image [27]. In one small cohort including 27 metastatic BC patients, trastuzumab was believed to cause arrhythmias in 6 patients, one of which developed sinus bradycardia and others were unspecified [28]. In a larger study including 68,113 BC patients, the chemotherapy group was shown to have higher Af incidence, while in the subgroup analysis, there were no significant differences regarding trastuzumab use [29]. Other abnormal electrocardiogram findings, including branch bundle blocks, QT prolonged, ventricular bigeminy, and non-sustained ventricular tachycardia in HER2-expressed BC patients, were shown as case reports [30][31][32]. A small cohort of 20 metastatic HER2 BC patients reported no arrhythmias after trastuzumab infusion [33]. The evidence regarding arrhythmias associated with trastuzumab in BC patients was still vague, controversial, and not well-established. According to the above pre-clinical and clinical evidence, we suggested that the AE of arrhythmias of trastuzumab might refer to ventricular tachycardia and ventricular fibrillation rather than Af. There were some limitations to our study. First, the misclassification of false-positive diagnosis of administrative codes for HF and Af might occur as limitations of population-based, including administrative data. Second, more frequent cardiac screening in patients on trastuzumab led to detection bias. Third, education in patients receiving trastuzumab might induce health behavior bias. No information regarding the value of LVEF and the severity of the cardiac outcomes were available using the NHIRD. Also, this study only detected patients who used NHI-supported trastuzumab; those who paid for the regimen themselves were omitted. EF is not a routine examination in realworld clinical practice in non-HER2 BC patients. Better comparisons across exposure groups could be achieved if the evaluation of heart function was performed in non-HER2 patients. Moreover, transient arrhythmia during trastuzumab infusion might have been missed and not coded in the NHIRD claims.
Furthermore, although we attempted to match both groups with age, comorbidities including DM, HTN, hyperlipidemia, CKD, medication usage including aspirin, NSAIDs, statins, metformin, beta-blockers, and drugs against RAAS system, residual confounding still existed. Our study reflects the real-world clinical practice of trastuzumab treatment in Taiwan under the NHI coverage. Our data showed that around 12% of all BC patients had received trastuzumab. HER2-positivity accounts for 20-25% of all BC patients. Patients with HER2-positive node-negative early BC may have never received trastuzumab according to the reimbursement criteria. Nevertheless, these patients would still receive cardiotoxic chemotherapy (such as anthracyclines) and thus may have risks of Af. Moreover, other well-characterized risk factors for Af, such as HTN and DM, and concurrent medications may have masked the less prominent role of trastuzumab in developing Af.
Collectively, the opposite result of trastuzumab on the risk of Af despite its known arrhythmogenic effect may be explained by several reasons below: (1) In general, HF can increase the risk of developing Af subsequently due to structural atrial stretch and interstitial fibrosis [34]. In contrast, most trastuzumab-induced heart dysfunction can be improved or reversible [35,36] and therefore Af would not precede HF. (2) In the real-world setting, clinicians would try to avoid cardiotoxic agents when prescribing trastuzumab, theoretically lowering the incidence of cardiotoxicity. (3) Taiwanese patients with HER2-positive node-negative early BC may have never received trastuzumab according to the reimbursement criteria. Nevertheless, these patients (non-trastuzumab users) would still receive cardiotoxic chemotherapy (such as anthracyclines) or radiotherapies and thus may have risks of Af. Moreover, other well-characterized risk factors for Af, such as HTN and DM, and concurrent medications may have masked the less prominent role of trastuzumab in developing Af.
Indeed, clinical reports have been focused on the more prominent cardiomyopathic role of trastuzumab; the incidence of Af (1.2%) is lower than that of HF (1-4%) [37,38]. Nevertheless, given the possible arrhythmogenic effects of trastuzumab [12], further studies are necessary to characterize the potential electrophysiological side effects.

Conclusions
Trastuzumab in real-world patients was associated with an increased risk of HF, but was not associated with an increased risk of Af in BC patients. Trastuzumab-induced arrhythmogenic effects may be masked by heart-protecting measures, more prominent roles of comorbidities, or concurrent medications under real-world settings. Further studies are required.