Author + information
- Received June 24, 2018
- Revision received September 5, 2018
- Accepted September 19, 2018
- Published online February 25, 2019.
- Ann Banke, MDa,b,∗ (, )
- Emil L. Fosbøl, MD, PhDc,
- Marianne Ewertz, Professor, MD, DMscb,d,
- Lars Videbæk, MD, PhDa,
- Jordi S. Dahl, MD, PhDa,b,
- Mikael Kjær Poulsen, MD, PhDa,
- Søren Cold, MD, PhDd,
- Maj-Britt Jensen, CandSciente,
- Gunnar H. Gislason, MD, PhDf,g,
- Morten Schou, MD, PhDh,∗ and
- Jacob E. Møller, MD, DMScia,b,∗
- aDepartment of Cardiology, Odense University Hospital, Odense, Denmark
- bInstitute of Clinical Research, University of Southern Denmark, Odense, Denmark
- cDepartment of Cardiology, Rigshospitalet, Copenhagen, Denmark
- dDepartment of Oncology, Odense University Hospital, Odense, Denmark
- eDanish Breast Cancer Cooperative Group, Rigshospitalet, Copenhagen, Denmark
- fDepartment of Cardiology, Copenhagen University, Herlev and Gentofte Hospital, Hellerup, Denmark
- gDanish Heart Foundation, Copenhagen, Denmark
- hDepartment of Cardiology, Copenhagen University, Herlev and Gentofte Hospital, Herlev, Denmark
- ↵∗Address for correspondence:
Dr. Ann Banke, Department of Cardiology, Odense University Hospital, J.B. Winsløws Vej 4, 5000 Odense, Denmark.
Objectives This study sought to evaluate the long-term risk of developing heart failure (HF) in patients receiving trastuzumab therapy.
Background Trastuzumab has improved the prognosis in patients with HER2-positive breast cancer, but it can induce left ventricular dysfunction with reduced ejection fraction or HF during treatment. The long-term risk of HF is less well described.
Methods In a nationwide Danish retrospective cohort study, 9,901 patients scheduled for adjuvant treatment for early-stage breast cancer were identified in the Danish Breast Cancer Cooperative Group database. Of these, 8,812 patients (25% HER2-positive; 51.7 ± 8.5 years of age) received chemotherapy including anthracycline; and if they were HER2 positive, trastuzumab was added. The primary endpoint was a diagnosis of HF assessed before and after 18 months in a landmark analysis to distinguish short- and long-term risks.
Results Median follow-up was 5.4 years (interquartile range [IQR]: 4.1 to 6.8 years). In the trastuzumab group, 60 patients had HF by 9 years versus 51 in the group who were treated with chemotherapy alone, corresponding to incidence rates per 1,000 patient years of 5.3 (95% confidence interval [CI]: 4.1 to 6.8) versus 1.4 (95% CI: 1.1 to 1.8), respectively. The cumulative incidence of HF was higher in the trastuzumab group at both the short- and long-term (p < 0.01), yielding adjusted hazard ratios of 8.7 (95% CI: 4.6 to 16.5; p < 0.01) for early HF and 1.9 (95% CI: 1.2 to 3.3; p = 0.01) for late HF associated with trastuzumab treatment.
Conclusions Trastuzumab treatment is associated with a 2-fold increased risk of late HF compared with chemotherapy treatment alone.
Breast cancer is the most frequent malignant disease among women worldwide with estimated 1.67 million new cases diagnosed in 2012, corresponding to 25% of all cancers in women (1). Approximately 15% to 30% of breast cancer patients have a subtype of cancer expressing the human epidermal growth factor receptor 2 (HER2) gene, which is associated with accelerated tumor growth, early metastasis, and thus, a poor prognosis (2–4). Trastuzumab, a humanized monoclonal antibody directed against HER2, has significantly improved survival in HER2-positive breast cancer patients but can induce asymptomatic left ventricular dysfunction with reduced ejection fraction or symptomatic heart failure (HF) (5–7).
Randomized trials testing trastuzumab in an adjuvant setting have reported a 5-fold increased risk of cardiac toxicity with development of symptomatic HF in 1% to 2% of the patients within the first 2 years after treatment initiation (8). There is more uncertainty about the long-term risk, with evidence from clinical trials revealing no excess risk of HF after the first 2 to 3 years (9–11), whereas some observational studies of clinical populations suggest that the incidence may be higher, up to 20%, and that the duration of the risk period may be longer (12,13). With the current 5-year survival at more than 85% after a diagnosis of breast cancer in developed countries, long-term cardiac adverse effects of therapy become important (14), especially because cardiovascular comorbidity is proposed to be a major contributor to overall long-term survival (15).
After randomization in the HERA (Herceptin Adjuvant) trial (16), trastuzumab was introduced in 2006 as a guideline treatment in Denmark, recommended by the Danish Breast Cancer Cooperative Group (DBCG). This nationwide multidisciplinary group with an associated database was established in 1977 with the aim of improving the prognosis in breast cancer (17). With this unique source of data, we were able to evaluate the long-term risk of clinical HF after adjuvant chemotherapy with and without trastuzumab in an entire population of unselected women with early stage breast cancer.
This was a nationwide, retrospective cohort study based on data linked on an individual level between the complete Danish administrative and clinical registries by a unique civil registration number, which is provided to all Danish citizens at birth or at achievement of permanent residency in Denmark (18).
The primary study endpoint was a diagnosis of HF in the Danish National Patient Registry, defined by the International Classification Diseases-10th Edition (ICD-10) codes DI50-50.9, DI42-42.9, DI11.0, DI13.0, and DI13.2. Diagnosis of HF consists mainly of systolic HF (19). The Danish National Patient Registry contains nationwide data for ICD codes for all hospital admissions since 1977 and since 1995 data for outpatient and emergency patients (20). Diagnoses based only on emergency visits or in outpatients seen on 1 occasion with no further follow-up of chronic HF were not counted as events. Diagnosis of HF in the Danish National Patient Registry has been validated on several occasions with positive predictive values between 81% and 100% (19,21). Vital status was recorded from the Danish Civil Registry (18) and causes of death from the Danish Register of Causes of Death (20).
Covariate data regarding tumor characteristics and cancer treatment were collected from the DBCG database, which contains information on Danish women with diagnoses of breast cancer since 1977 (17). Data for cardiovascular comorbidity were obtained from the Danish National Patient Registry.
Consecutive patients who underwent surgery for unilateral nonmetastatic breast cancer between January 1, 2007, and December 31, 2012, were identified in the DBCG database. Patients allocated to undergo chemotherapy and, if HER2 positive, also to receive trastuzumab but no other biological treatment were included. Patients were excluded if they did not receive chemotherapy or trastuzumab, were ≥70 years of age, or had a diagnosis of HF prior to the diagnosis of breast cancer. Adjuvant chemotherapy consisted of 3 cycles of epirubicin (90 mg/m2) and cyclophosphamide (600 mg/m2), followed by 3 cycles of docetaxel (100 mg/m2) and, if they were HER2 positive, 1 year of trastuzumab therapy distributed in 17 individual cycles. A total of 987 patients (108 HER2-positive and 879 HER2-negative) included in this cohort were randomized to a different chemotherapy regimen consisting of 6 cycles of docetaxel (75 mg/m2) and cyclophosphamide (600 mg/m2) (22). If indicated by guidelines, patients also received radiation and endocrine therapy (23,24). During their trastuzumab therapy, patients underwent repeated isotope multigated acquisition (MUGA) scans to evaluate left ventricular ejection fraction at weeks 0, 9, 18, 30, and 48, after trastuzumab therapy was initiated according to DBCG guidelines. Patients receiving chemotherapy alone did not undergo any routine cardiac evaluations.
Patients were followed from the date of surgery to the first occurrence of the primary endpoint, a competing event of death, a censoring event of emigration, or end of study follow-up period (December 31, 2015).
Baseline characteristics were described using proportions for categorical variables and median and interquartile ranges for continuous variables. The Wilcoxon 2-sample test and chi-squared test were used for comparison of baseline characteristics.
A landmark time point was set at 18 months. Landmark analyses of HF risk were used to assess incidence of HF before and after 18 months from index to distinguish between short- and long-term risks. Hence, those patients followed after 18 months had not died or received a diagnosis of HF from index to 18 months.
Incidence rates for HF were calculated per 1,000 patient years. Due to competing risk of death, the incidence of HF was evaluated in a cumulative incidence function with death as the competing endpoint, and the Grey test for equality was applied to examine differences between groups. An additional analysis of cumulative incidence of HF stratified according to age was conducted. Kaplan-Meier estimates based on all-cause mortality were calculated, and the log-rank test was applied.
Risk of HF was examined multivariately in a cause-specific Cox Proportional Hazard Model adjusted for age and baseline comorbidity (ischemic heart disease, acute myocardial infarction, atrial fibrillation, hypertension, type 2 diabetes mellitus, and chronic obstructive pulmonary disease) with the Wald chi-squared test applied. When the proportional hazards assumption was assessed graphically by a log(-log) S plot and by adding a time-dependent variable to the model for the early and late HF assumptions of proportional hazards were met.
For purposes of sensitivity analyses, we modeled the propensity for receiving or not receiving trastuzumab by using all available baseline characteristics (logistic regression analysis with trastuzumab treatment as the outcome variable) (Tables 1 and 2). The propensity score yielded by this model was divided into quintiles, and the cause-specific hazard model was performed again with model stratification for propensity, that is, cases and controls were only compared within strata of propensity scores.
SAS software version 9.4. (SAS Institute, Gary, North Carolina) was used for statistical analysis and data management. A 2-sided p value of <0.05 was considered statistically significant.
The study was approved by the Danish Data Protection Agency (ref. 2007-58-0015/GEH-2014-012; I-suite number 02720). Approval by the Ethics Committee and informed consent are not required for retrospective registry-based studies in Denmark.
From the DBCG database, 9,901 patients scheduled to receive adjuvant medical treatment after surgery for unilateral nonmetastatic breast cancer were identified. Of these, 1,089 patients were excluded because they did not receive treatment (n = 288), they were ≥70 years of age (n = 753), or they had a pre-existing diagnosis of HF before the diagnosis of breast cancer (n = 48). Of the remaining 8,812 patients, 2,117 patients had HER2-positive disease and received both chemotherapy and trastuzumab treatment, whereas 6,695 patients were HER2-negative and received only chemotherapy as part of the adjuvant medical treatment. Details are presented in the consort diagram (Figure 1).
As presented in Table 1, patients in the chemotherapy-plus-trastuzumab treatment group were on average slightly older than patients in the chemotherapy group, with mean ages of 53.5 ± 9.6 years versus 51.2 ± 8.1 years, respectively. Differences in tumor characteristics and other treatment modalities between the chemotherapy-plus-trastuzumab and the chemotherapy group are presented in Table 1. Median accumulated dose of trastuzumab was 6,761 mg (interquartile range [IQR]: 5,700 to 7,800 mg) among the chemotherapy-plus-trastuzumab-treated patients.
Table 2 shows a low prevalence of comorbidity in the cohort at baseline, which was balanced between groups.
The cohort used for the analysis of late HF consisted of 8,611 patients who survived and did not develop HF within the first 18 months after initiation of trastuzumab therapy (the consort diagram is presented in Online Figure 1). Baseline characteristics and comorbidity for the cohort used for analysis of late HF are presented in Online Tables 1 and 2. Comorbidity was updated to 18 months.
Follow-up was complete with a median follow-up period of 5.4 years (IQR: 4.1 to 6.8 years). Mortality was similar in both groups: 14.5% (95% confidence interval [CI]: 12.4% to 17.0%) in the chemotherapy-plus-trastuzumab group versus 15.3% (95% CI: 13.4% to 17.2%) in the chemotherapy group (p = 0.67). Cause of death was primarily cancer, with no differences (86.7% vs. 89.6%; p = 0.27) between the 2 groups. Cardiovascular death constituted 3.3% versus 2.7%, respectively, with no differences between groups (p = 0.70). Divided according to the landmark analysis before and after 18 months, rates of mortality were also similar between the 2 treatment groups, with 1.2% (95% CI: 0.9% to 1.8%) versus 1.9% (95% CI: 1.6% to 2.2%; p = 0.06), respectively, from 0 to 18 months and 13.3% (95% CI: 11.1% to 15.7%) versus 13.6% (95% CI: 11.7% to 15.6%) (p = 0.24), respectively, after 18 months. There were no differences between the groups regarding cancer and cardiovascular mortality before and after 18 months.
Incidence and risk of heart failure
In the chemotherapy-plus-trastuzumab treatment group, 60 new cases of HF (2.8%) were identified versus 51 cases (0.8%) in the group treated with chemotherapy alone, corresponding to incidence rates per 1,000 patient years of 5.3 (95% CI: 4.1 to 6.8) versus 1.4 (95% CI: 1.1 to 1.8), respectively. In the chemotherapy-plus-trastuzumab group, 41.7% of the diagnoses of HF were given during a hospital admission as oppose to in an outpatient clinic, whereas 76.5% (p < 0.001) of the diagnoses in the in chemotherapy group were given during an admission.
The landmark analysis of patients diagnosed with HF within the first 18 months after the diagnosis of breast cancer revealed 36 cases (1.7%) of early HF in the chemotherapy-plus-trastuzumab group versus 13 (0.2%) in the chemotherapy group, corresponding to incidence rates per 1,000 patient years of 11.8 (95% CI: 8.5 to 16.3) versus 1.3 (95% CI: 0.8 to 2.3), respectively, whereas analysis after 18 months demonstrated 24 cases (1.2%) of late HF in the chemotherapy-plus-trastuzumab treatment group vs. 38 (0.6%) in the chemotherapy group, corresponding to incidence rates per 1,000 patient years of 2.9 (95% CI: 1.9 to 4.3) versus 1.4 (95% CI: 1.0 to 1.9), respectively.
The cumulative incidence of HF after 9 years, including all cases of HF, was 3.3% (95% CI: 2.5% to 4.2%) in the chemotherapy-plus-trastuzumab group versus 1.3% (95% CI: 0.9% to 1.8%; p < 0.0001) in the chemotherapy group. Development of both early and late HF was significantly increased among patients treated with chemotherapy-plus-trastuzumab compared with those treated with chemotherapy alone as presented in the landmark analysis (Figure 2).
Increased cumulative incidence in the group treated with both chemotherapy and trastuzumab compared to chemotherapy alone stratified according to age is presented for early HF (p < 0.0001) (Figure 3A) and for late HF (p = 0.03) (Figure 3B).
Including all cases of HF in the Cox proportional hazard model, an age- and comorbidity-adjusted hazard ratio (HR) of 3.61 (95% CI: 2.47 to 5.26; p < 0.0001) was found for HF associated with chemotherapy-plus-trastuzumab treatment compared to treatment with chemotherapy alone. In the landmark analysis including only cases of early HF, the adjusted HR associated with trastuzumab was 8.69 (95% CI: 4.59 to 16.47; p < 0.0001) and 1.93 for cases of late HF (95% CI: 1.15 to 3.25; p = 0.01). In the multivariate analysis, ischemic heart disease, acute myocardial infarction, and chronic obstructive pulmonary disease were associated with development of early HF, whereas ischemic heart disease, atrial fibrillation, and type 2 diabetes were associated with development of late HF. Estimates are illustrated in Online Figure 2.
In a sensitivity analysis, a propensity score based on all baseline variables was added to the model without any major change in the HR associated with trastuzumab, which in this analysis were 3.87 (95% CI: 2.61 to 5.74; p < 0.0001) for all cases, 10.19 (95% CI: 5.27 to 19.70; p < 0.0001) for early cases, and 1.90 (95% CI: 1.10 to 3.26; p = 0.02) for late cases.
This nationwide Danish cohort study evaluated the long-term risk of clinical HF after chemotherapy-plus-trastuzumab treatment compared to treatment with chemotherapy in an unselected cohort of unilateral early-stage breast cancer patients. The main finding was an increased long-term risk of HF after chemotherapy-plus-trastuzumab treatment compared with that after treatment with chemotherapy alone but with an overall low incidence. The risk of HF was weakly associated with baseline comorbidity. Finally, the study confirmed the early risk of HF associated with trastuzumab during treatment.
Incidence and risk of heart failure
The increased incidence of late HF observed in the present study differs from data recently published for cardiac side effects from some of the first randomized trials testing trastuzumab treatment in an adjuvant setting with up to 8 years of follow-up. Both the HERA and the NSABP (National Surgical Adjuvant Breast and Bowel Project protokol B-31) trials reported an increased risk of HF during and shortly after trastuzumab treatment but with very few cases of suspected cardiac toxicity on long-term follow-up (9,11). This difference is likely explained by the favorable cardiovascular risk profile of women in the randomized trials compared with a clinical population with a presumably greater burden of comorbidity. Interestingly, Goldhar et al. (25) recently conducted a population-based study of Canadian breast cancer patients and found a similar result as the randomized studies with an HR of 5.77 of developing HF within the first 18 months, but thereafter no excess risk of HF could be detected. Although the Canadian cohort shared age and prevalence of myocardial infarction similar to those of our cohort, diabetes and hypertension were more prevalent in the Canadian cohort; therefore, comorbidity is unlikely to explain the difference alone (25). Other observational studies report increased risk of HF compared to the randomized trials, especially among elderly patients, but follow-up duration was shorter, and there was no differentiation between early and late HF development (13,26).
In the present study, the cumulative incidence of just below 3% of patients developing HF during the first 5 years of follow-up is comparable with that of the large Canadian observational study with approximately 5% (25) and the NSABP (B-31) trial with <4% (27). In the HERA study, the cumulative incidence was even lower at approximately 1% (11). These low incidences are, as mentioned earlier, likely explained by a lower cardiovascular risk profile in the randomized trials but may also suggest that patients referred to chemotherapy in a clinical setting could in some degree be selected. It has been suggested previously in a large epidemiological study, that breast cancer patients have a favorable cardiovascular risk profile compared with that in the background population (28). However, a recent nationwide register study on the general incidence of HF in the Danish background population found incidence rates of 6.4 and 17 per 10,000 person-years among individuals between 45 to 54 and 55 to 64 years of age, respectively, which are lower than the findings in the present study (29). Incidence of HF among women in the Framingham and Hillingdon Heart studies are also below those of the present study, even though direct comparisons are difficult due to differences in age and methods (30).
Age and comorbidity
The cumulative incidence rates stratified according to age in our study indicate, with reservation for the low number of cases in each group, that early HF development during trastuzumab treatment is less associated with age than in patients developing late HF. It has previously been shown that the risk of HF associated with trastuzumab is higher among elderly patients but without the differentiation between early and late HF (13). The association with cardiovascular comorbidity, which has also been described in other studies (13,25), was also present in this study. Furthermore, the isolated association with age in this study was erased when comorbidity was added to the multivariate model.
With many long-term survivors, after modern breast cancer treatment (14), awareness of HF as a possible long-term consequence of trastuzumab treatment is important. Even though the overall risk of HF after trastuzumab treatment is low, it may be relevant to take this risk into consideration when planning the course of follow-up after HER2-positive breast cancer, especially in patients with other comorbidities as this seems associated with a later diagnosis of HF. When HF or symptoms of HF are diagnosed, it is important for the patient to provide detailed information about previous cancer treatment when determining the cause, and keeping in mind that trastuzumab even several years after treatment could be a contributing factor.
Given the observational design of the study, the results should be interpreted with caution. However, to the best of our knowledge, there are very few studies of clinical populations of comparable sizes in terms of number of patients treated with trastuzumab. The complete follow-up in this study is also a strength as is the quality of Danish nationwide registries, where the completion of the registries is ensured by an obligation by legislation to report to these registries (20). Furthermore, the certainty of the HF diagnosis in the Danish National Patient Registry is reasonably high, with positive predictive values between 81% and 100% in validation studies (19,21). A limitation of the National Patient Registry, however, is that only patients who present with symptoms severe enough to require an outpatient hospital visit or hospitalization are recorded. The incidence is likely underestimated because a number of patients with early stages of HF and asymptomatic patients with reduced ejection fraction or very limited symptoms are not diagnosed and consequently not recorded in the Danish National Patient Registry.
There were differences in ages between the 2 treatment groups in this study: patients treated solely with chemotherapy were slightly younger than patients treated with both chemotherapy and trastuzumab, as HER2-negative patients older than 60 years of age, according to treatment guidelines, were scheduled only for adjuvant medical treatment if specific tumor characteristics placed them at high risk for recurrence. The risk analysis, however, was adjusted for age, and the sensitivity analysis with addition of propensity score did not change the risk estimate.
We may have overestimated the effect of trastuzumab on the risk of HF, as part of the risk may be explained by different exposure to an anthracycline between the 2 groups, because we lacked exact information for doses. However, 96% of the patients in the chemotherapy-plus-trastuzumab treatment group and 88% in the group treated solely with chemotherapy received an anthracycline. Therefore, a sensitivity analysis was conducted with exclusion of the number of patients who did not receive an anthracycline. This analysis did not change the risk estimate.
Because the risk of early HF during trastuzumab treatment is well known, patients in the group treated with both chemotherapy and trastuzumab might have had more HF symptoms, and some degree of observational bias cannot be excluded in relation to late HF. This assumption is supported by the larger proportion of patients whose condition was diagnosed in an outpatient clinic (presumably with less critical symptoms) in the chemotherapy-plus-trastuzumab group as opposed to the larger proportion who received diagnoses during a hospital admission in the chemotherapy group.
In this nationwide cohort study based on real-life data, trastuzumab treatment after anthracycline-based chemotherapy was associated with a 2-fold increased risk of late clinical HF compared to that after anthracycline-based chemotherapy alone and associated with baseline cardiovascular comorbidity.
This increased risk should be taken into account when planning the course of follow-up after cancer treatment and should be kept in mind when determining the cause of HF or cardiomyopathy in long- term cancer survivors.
COMPETENCY IN MEDICAL KNOWLEDGE: With many long-term cancer survivors, the finding of increased long-term risk of HF after trastuzumab treatment should be kept in mind when patients present with HF many years after a diagnosis of cancer. Furthermore, previous use of trastuzumab should be kept in mind when the etiology of the cardiomyopathy is determined.
TRANSLATIONAL OUTLOOK: The risk of cardiotoxicity during trastuzumab treatment is well known, but the present study finds indications of comorbidity being more a contributing factor in long-term than in short-term risk of HF in relation to trastuzumab. Furthermore, research is needed to establish the significance of this.
The authors acknowledge the Danish Breast Cancer Cooperative Group for access to the breast cancer database.
↵∗ Drs. Schou and Møller contributed equally to this work and are joint senior authors.
Supported by Danish Heart Foundation grants 14-R97-A5188-22839 and 15-R99-A5940; and the Research Fond of the Region of Southern Denmark, Vejle, to Dr. Banke during her PhD study. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- human epidermal growth factor receptor 2
- heart failure
- isotope multigated acquisition scan
- Received June 24, 2018.
- Revision received September 5, 2018.
- Accepted September 19, 2018.
- 2019 American College of Cardiology Foundation
- Slamon D.J.,
- Clark G.M.,
- Wong S.G.,
- Levin W.J.,
- Ullrich A.,
- McGuire W.L.
- Gusterson B.A.,
- Gelber R.D.,
- Goldhirsch A.,
- et al.
- Perez E.A.,
- Romond E.H.,
- Suman V.J.,
- et al.
- Cameron D.,
- Piccart-Gebhart M.J.,
- Gelber R.D.,
- et al.
- Moja L.,
- Tagliabue L.,
- Balduzzi S.,
- et al.
- Ganz P.A.,
- Romond E.H.,
- Cecchini R.S.,
- et al.
- Perez E.A.,
- Suman V.J.,
- Davidson N.E.,
- et al.
- de Azambuja E.,
- Procter M.J.,
- van Veldhuisen D.J.,
- et al.
- Chen J.,
- Long J.B.,
- Hurria A.,
- Owusu C.,
- Steingart R.M.,
- Gross C.P.
- Christiansen P.,
- Ejlertsen B.,
- Jensen M.B.,
- Mouridsen H.
- Ejlertsen B.,
- Tuxen M.K.,
- Jakobsen E.H.,
- et al.
- Nielsen M.H.,
- Berg M.,
- Pedersen A.N.,
- et al.
- Goldhar H.A.,
- Yan A.T.,
- Ko D.T.,
- et al.
- Thavendiranathan P.,
- Abdel-Qadir H.,
- Fischer H.D.,
- et al.
- Romond E.H.,
- Jeong J.H.,
- Rastogi P.,
- et al.
- Christiansen M.N.,
- Kober L.,
- Weeke P.,
- et al.
- Mehta P.A.,
- Cowie M.R.