Background

A key prognostic biomarker for breast cancer is the level of the human epidermal growth factor receptor-2 (HER2) expression. Women with overexpressed HER2-positive breast cancer are more likely to present at an advanced stage and have a significantly shortened disease-free and overall survival [1]. Trastuzumab (Genentech Inc., San Francisco, CA) is a recombinant humanized monoclonal antibody that inhibits tumor cell growth by binding to the extracellular region of the HER2 protein and blocking the signaling cascade [2,3,4]. When used in combination with adjuvant chemotherapy, trastuzumab reduces the relative risk of recurrence and death by 50 and 33%, respectively [3,4,5,6].

While trastuzumab is generally well tolerated, the literature has reported increased risk of cardiotoxic effects [3,4,5, 7, 8]. Many studies have suggested that the blockage of HER2 in cardiomyocytes leads to over-production of reactive oxygen species (ROS) [3, 4, 7,8,9]. Over time, the accumulation of harmful ROS causes cell damage and apoptosis, leading to the development of cardiac dysfunction [9]. As a result, the reduction in cardiomyocytes can induce a drop in left ventricular ejection function (LVEF), potentially causing congestive heart failure (CHF) [3].

In a meta-analysis of ten randomized control trials (RCTs), the incidence of trastuzumab-induced cardiotoxicity (TIC) among breast cancer patients is 1.9% for symptomatic CHF and 7.5% for asymptomatic LVEF drop [10]. Other observational studies suggest that the incidence is much higher [11, 12]. This raises concerns that there may be discrepancies between clinical trial and observational data.

This retrospective cohort study aims to determine the incidence, time to development, and potential risk factors associated with TIC among less selected breast cancer patients treated in a Canadian “real-world” tertiary care center.

Methods

A retrospective cohort study was conducted at St. Michael’s Hospital in Toronto, Ontario, Canada. From January 2006 to June 2014, 160 consecutive HER2+ breast cancer patients who received adjuvant chemotherapy with trastuzumab were identified. Eligible patients were diagnosed with stage I to III HER2+ breast cancer with normal LVEF levels (defined as >50%). Metastatic breast cancer patients were excluded from the study. For this retrospective study, formal consent from patients was not required. Research ethics board approval was obtained from St. Michael’s Hospital.

Patient demographics, cardiovascular risk factors, and treatment details were extracted from the hospital’s electronic medical record. The decision to discontinue or rechallenge trastuzumab therapy was made at the medical oncologist’s discretion.

Trastuzumab-induced cardiotoxicity

Trastuzumab-induced cardiotoxicity (TIC) was defined as symptomatic (e.g., heart failure and/or dyspnea, and/or referral to a cardiologist for confirmation) or asymptomatic (e.g., decline in LVEF ≥ 10% from baseline or LVEF ≤ 50%). Cardiac monitoring to determine LVEF was performed every 3 months using serial multigated acquisition scans (MUGA), cardiac magnetic resonance imaging (CMR), or an echocardiogram (ECHO) at the discretion of the treating medical oncologist.

Statistical analysis

Descriptive analyses were used to summarize patient demographics. Binary analyses included Pearson’s Chi-square test, Fisher’s exact test, Student’s t test, and Satterthwaite method, when appropriate. Statistical significance was defined as a p value of less than 0.05.

Time to TIC (defined as time from first trastuzumab administration to the onset for TIC or at last oncology follow-up, whichever came first) was estimated using the Kaplan–Meier method and compared using the log-rank test. The effect of potential predictors of time to TIC was estimated by the Cox proportional hazards models. We tested the assumptions of proportionality in our Cox models by plotting log[−log(survival)] versus log(time) and assessing parallelism. A hierarchical approach was used for model building; radiotherapy and anthracycline-based chemotherapy were included in the model a priori. A subgroup analysis was conducted and decided a priori for patients with left-sided tumors. SAS version 9.4 (Cary, NC) was used to perform the analyses.

Results

The patient characteristics of 160 patients who met the inclusion criteria are summarized in Table 1. The median age at diagnosis was 52.4 years (with a baseline LVEF of 62.5%), and 48.1% of the patients were on anthracycline-based chemotherapy. The mean cumulative dose for patients who received epirubicin and doxorubicin was 300 and 240 mg/m2, respectively. Baseline patient characteristics by TIC status are described in Table 2.

Table 1 Patient characteristics at baseline
Full size table
Table 2 Baseline patient characteristics by TIC status (n = 160)
Full size table

In the patient sample (n = 160), 34 patients (21.3%) experienced TIC from January 2006 to June 2014. Of the 34 TIC patients, 7 (20.6%) patients were referred to a cardiologist for dyspnea and 27 (79.4%) asymptomatic patients were diagnosed based on changes in LVEF levels. The median follow-up time from the first dose of trastuzumab to the last oncology follow-up was 55.4 months (IQR 35.1–73.9 months). A total of 23 (14.3%) patients were lost to follow-up, and 9 (5.6%) died during follow-up. Of the 34 patients who experienced TIC, 31 (91.1%) were referred to a cardiologist for further investigation. A total of 22 (70.9%) patients were prescribed cardiovascular medication (angiotensin-converting enzyme inhibitors, beta-blockers, and/or angiotensin II receptor blockers) after referral.

The median time to development to TIC was 28.5 weeks (IQR 14.3–42.5 weeks) during trastuzumab therapy. Of the 160 patients, 15 (9.4%) patients were dose delayed for trastuzumab and 6 (3.8%) patients discontinued trastuzumab early due to LVEF changes.

There was a significant difference between those who undergone a mastectomy compared to lumpectomy in time to TIC development (p = 0.028) (Fig. 1). In the unadjusted analysis, the hazard ratio for individuals who had a mastectomy was 2.10 (95% CI 1.07–4.11) compared to those who had a lumpectomy. However, after adjusting for covariates, the association was weakened and not significant (Table 3).

Fig. 1
figure 1

Kaplan–Meier curve for time to TIC development. Kaplan–Meier curve from trastuzumab initiation to development of TIC for patients who undergone a mastectomy (blue line) compared to lumpectomy (red line). TTIC time to trastuzumab initiation

Full size image
Table 3 Cox proportional hazard ratios for type of surgery
Full size table

Left-sided tumor subgroup

A total of 102 patients had left-sided tumors (including those who had bilateral tumors), and 19 patients (18.6%) developed TIC. There was a statistically significant difference in type of surgery (p = 0.01) but not in radiotherapy, tumor size, or use of anthracycline-based chemotherapy between TIC groups.

Discussion

The incidence of TIC in this “real-world” sample of patients was 21.3% in the adjuvant setting (median follow-up of 55.4 months). The median time to development of TIC from trastuzumab initiation was 28.5 weeks, approximately the 10th treatment of trastuzumab. We found that mastectomy was not an independent risk factor for TIC.

In the four international RCTs, symptomatic TIC was reported in 1.7% in the HERA trial, 2.3% in the NCCTG N9831 trial, 3.8% in the NSABP-31 trial, and 2.0% in the doxorubicin and cyclophosphamide followed by docetaxel and trastuzumab (AC-TH), and 0.4% in the docetaxel, carboplatin, and trastuzumab (TCH) arm of the BCIRG 006 trial [13,14,15,16]. Incidence of asymptomatic TIC in the HERA trial was 7.1 and 18.6% in the AC-TH and 9.4% in the TCH arm of the BCIRG 006 trial [13, 15]. A meta-analysis of the five major adjuvant trastuzumab RCTs reported more grade III or IV cardiac toxicity after trastuzumab use (4.5%) compared to no trastuzumab use (1.8%) [17].

Our study contributes to the growing amount of literature that the incidence of TIC is higher compared to RCTs. Our incidence of 21% at a single urban center is similar to other retrospective studies from various countries that reported 16 [18], 21 [19], and 24% [20]. Our incidence was lower compared to studies that reported 53 [11] and 33% [21]. However, these studies also included metastatic breast cancer patients, potentially increasing the incidence of TIC as metastatic patients may have been exposed to higher doses of prior systemic treatment including anthracyclines and endocrine therapies [11].

We hypothesize that these differences between our “real world” and RCT incidence lie in the strict inclusion and wide exclusion criteria. Contradictions to study entry in RCTs include: patients with previously documented cardiac disease, and adequate baseline hepatic, renal, and bone marrow function. Additionally, it is possible that the heterogeneity in the definition of asymptomatic TIC contributes to the differences between RCT and observational studies. In the HERA trial, decrease in LVEF was defined as a decline of ≥10% from baseline to an LVEF < 50% at any time [13], while in the BCIRG 006 trial, LVEF was defined as a relative reduction from baseline of more than 10% at the last evaluation [15]. Observational studies also vary in the definition of asymptomatic TIC from LVEF absolute reduction ≥16–10% from baseline and drops in LVEF < 50–55% [11, 12, 18, 22]. In our study, we used a definition of asymptomatic TIC as a decline in LVEF ≥ 10% from baseline or LVEF ≤ 50%.

The median time to development of TIC from trastuzumab initiation was 28.5 weeks, approximately the 10th treatment of trastuzumab, independent of prior anthracycline-based therapy. Our data are similar to other retrospective studies that reported cardiac events or cardiomyopathy between 3 and 6 months in the adjuvant setting [11, 23].

In our retrospective study, known risk factors such as anthracycline-based chemotherapy and history of cardiovascular disease were not associated with development of TIC (Table 2). Mastectomy was not independently associated with developing TIC (Table 3). Previous studies have identified higher body mass index, higher cumulative dose of anthracycline, heavy alcohol use, and Ile655Val HER2 polymorphism [8]. Interestingly, two prospective studies conducted in Poland and Brazil reported no risk factor associated for TIC [11, 24]. Clinical trial data reported other risk factors such as older age (≥60 years) and use of hypertensive medications in the NSAPBP B-31 and NCTTG N9831 joint analysis [14].

Due to the proximity of the heart in the chest wall, we hypothesized an association between left-sided mastectomy with radiation and the development of TIC. Data from the NCCTG N9831 trial investigated concomitant trastuzumab and radiotherapy in 908 patients and found no significant differences between radiated and non-radiated groups [25]. A study at MD Anderson found higher rates of TIC in patients with left-sided breast radiation compared to right (20 vs. 7 cardiac events) [26]. A study by Cao et al. [27] reported that left-sided irradiation with increased low-dose volume and mean heart dose was associated with reversible low-grade cardiac toxicity. In our study, there were no differences between groups in radiotherapy, regimen, or dose. In our left-sided tumor subgroup, no differences were found between groups for radiotherapy. However, type of surgery was significant. Given the small sample size of the subgroup, we were limited in statistical power to perform further exploratory analyses.

This study has several limitations associated with its retrospective nature. We were limited by the quality of information available from the electronic medical records and other potential variables such as physical inactivity and alcohol consumption. We did not collect cardiac biomarkers such as troponin-I and B-type natriuretic peptide, as they are not part of standard of care. The small sample size and wide confidence intervals also limited the statistical power to further detect independent associations between risk factors and TIC. LVEF assessments were also conducted with MUGA, CMR, or ECHO, adding inter-study and inter-operator variability across different tests. Additionally, this study was limited by the short follow-up time and therefore cannot determine the long-term incidence of TIC.

Results from this study identifying a higher risk of cardiotoxicity in the real world have led to the creation of a rapid cardio-oncology clinic at our center. This clinic provides rapid assessment by a cardiologist within 1–2 weeks for further investigation and appropriate therapeutic management if cardiotoxicity is identified or if concerning cardiac risk factors are present in the setting of active cancer therapy.

Conclusions

The incidence of TIC is higher in our “real-world” population compared to clinical trial data. No cardiovascular risk factors were independently associated with TIC. Early identification of women who are at risk of cardiac complications is critical to optimize cardiac care, ensuring safe delivery of curative oncologic treatments. Future prospective studies are warranted to further assess the incidence and risk factors for complications of curative adjuvant trastuzumab therapies in the real world.