Abstract
Background:
Anthracycline agents can cause cardiotoxicity. We used multivariable risk prediction models to identify a subset of patients with breast cancer at high risk of cardiotoxicity, for whom the harms of anthracycline chemotherapy may balance or exceed the benefits.
Patients and Methods:
A clinical prediction model for anthracycline cardiotoxicity was created in 967 patients with HER2 negative breast cancer treated with doxorubicin in the ECOG-ACRIN study E5103. Cardiotoxicity was defined as LVEF decline of ≥10% to <50% and/or a centrally adjudicated clinical HF diagnosis. Patient-specific incremental absolute benefit of anthracyclines (compared to non-anthracycline taxane chemotherapy) was estimated using the PREDICT model to assess breast cancer mortality risk.
Results:
Of the 967 women who initiated therapy, 51 (5.3%) developed cardiotoxicity (12 with clinical HF). In a multivariate model, increasing age (OR 1.04, 95% CI 1.01-1.08), higher body mass index (BMI) (OR 1.06, 95% CI 1.02-1.10) and lower baseline LVEF (OR 0.93, 95% CI 0.89-0.98) at baseline were significantly associated with cardiotoxicity. The concordance statistic of the risk model was 0.70 (95% CI: 0.63-0.77). In patients with low anticipated treatment benefit (n=176) from the addition of anthracycline (<2% absolute risk difference of breast cancer mortality at 10 years), 16/176 (9%) had a >10% risk of cardiotoxicity and 61/176 (35%) had a 5-10% risk of cardiotoxicity at 1 year.
Conclusion:
Older age, higher BMI and lower baseline LVEF were associated with increased risk of cardiotoxicity. We identified a subgroup with low predicted absolute benefit of anthracyclines but with high predicted risk of cardiotoxicity. Additional studies are needed incorporating long-term cardiac outcomes and cardiotoxicity model external validation prior to implementation in routine clinical practice.
MicroAbstract:
Anthracycline chemotherapy can cause cardiotoxicity. We derived a multivariable risk prediction model to predict anthracycline cardiotoxicity in 967 participants with HER2 negative breast cancer with close cardiac monitoring. We identified a subset of patients at high risk of cardiotoxicity but low predicted benefit from anthracyclines. Multivariable risk models can be used to generate patient-specific estimates of both benefits and harms with specific cancer regimens and with additional model validation and updating, use of these models may improve the shared decision-making process.
Introduction:
As advances in screening and treatment have improved survival for patients with breast cancer(1), cardiovascular disease has emerged as a major cause of long-term non-cancer related morbidity and mortality(2, 3). This likely reflects the role of shared risk factors (in particular age) for breast cancer and cardiovascular disease(4) as well as cancer therapy related cardiac effects(5–7). The increased risk of cardiovascular events in adults treated for cancer was highlighted in recent American Society of Clinical Oncology guideline(8) and American Heart Association scientific statements(9). Anthracycline chemotherapeutic agents, such as doxorubicin and epirubicin, result in increased oxidative and nitrosative stress, mitochondrial dysfunction, and cardiomyocyte apoptosis and can cause clinical heart failure (HF) and asymptomatic reductions in left ventricular ejection fraction (LVEF) (5, 10–13). HF after anthracycline treatment has a poor prognosis with median survival of less than 3 years after clinical diagnosis(14).
The recently published anthracyclines in early breast cancer (ABC) series of trials demonstrated a small benefit of anthracyclines plus taxane over taxane regimens without anthracyclines (15), while another trial failed to show a significant incremental benefit of anthracyclines in addition to taxanes in patients with TOP2A-normal breast cancer(16). With this evidence, we postulated that the harms of anthracycline chemotherapy may outweigh the benefits compared to taxane based chemotherapy for some patients, and that such patients may be identifiable on the basis of pretreatment characteristics. While prediction models for trastuzumab-associated cardiotoxicity have been reported(17, 18), there are currently no multivariable clinical prediction models for cardiac outcomes in HER2 negative breast cancer patients treated with anthracycline-based chemotherapy.
Thus, using a large cohort of patients with HER2 negative early stage breast cancer receiving adjuvant doxorubicin, cyclophosphamide and paclitaxel who were monitored closely for cardiotoxicity with routine LVEF assessment and adjudicated HF events, we sought to develop a clinical prediction model for cardiotoxicity. We used this model to examine the distribution of predicted risk of anthracycline-related cardiotoxicity (harm). Finally, we estimated the distribution of anticipated patient-specific incremental benefits to anthracyclines (i.e. in addition to taxane based adjuvant therapy) in reducing breast cancer mortality by applying recent trial results to each patient using the PREDICT model to estimate mortality. We hypothesized that this approach would identify subgroups of patients with substantially different benefit-harm trade-offs from anthracyclines.
Methods:
Model Derivation Cohort:
The prediction model was developed using the 1000 patients in arm A of the ECOG-ACRIN study E5103, a randomized controlled phase III trial that enrolled patients with HER2 negative, node positive or higher-risk node negative breast cancer(19). Reflecting current clinical practice, patients in arm A of this trial received doxorubicin (cumulative dose 240mg/m2) and cyclophosphamide for four cycles [classical (every 3 weeks) or dose dense (every 2 weeks) based on investigator discretion] followed by 12 doses of weekly paclitaxel (arm A). Patients enrolled in other arms of this trial were excluded from this analysis because they received an additional cardiotoxic agent (bevacizumab), which is also not routinely used in the treatment of early breast cancer. Radiation doses and laterality were recorded for each patient. Higher risk node negative disease was defined as either 1) Estrogen Receptor (ER) negative tumor >=1cm, 2) ER pos tumor >=5cm or 3) ER pos tumor >=1 but <5cm with recurrence score of >=11. Cardiovascular exclusion criteria for E5103 included a LVEF less than the lower institutional limit of normal; uncontrolled hypertension or arrhythmias at the time of enrollment; prior myocardial infarction, unstable angina, symptomatic HF or peripheral vascular disease within 12 months prior to enrollment; or any prior history of transient ischemic attack or stroke.
Predictors for Cardiotoxicity Model:
Candidate predictor variables were selected from easily and reliably obtainable pre-treatment clinical characteristics a priori based upon prior studies suggesting an association with anthracycline-related cardiomyopathy or HF(8). The following 4 variables were pre-specified for inclusion in the model: age, body mass index (BMI), hypertension and baseline LVEF. BMI was calculated using measured weight in kilograms divided by the square of the measured height in meters. Hypertension was defined as the use of an antihypertensive agent or SBP>140 or DBP>90 at the baseline visit prior to initiation of cancer therapy. LVEF was measured in all patients prior to initiation of cancer therapy using either echocardiography or multigated acquisition (MUGA) scan. In addition, we considered several additional “exploratory” candidate variables for model inclusion using Akaike’s information criteria (AIC)– radiation therapy, menopausal status, dose dense chemotherapy as well as alternative hypertension definitions including systolic blood pressure alone and different classes of blood pressure medications.
Cardiac Outcomes:
The primary outcome for the prediction model was the composite outcome of a reduction from baseline in LVEF of 10% or greater with a resultant LVEF of less than 50% and/or a clinical diagnosis of HF through the first year of follow up. LVEF was assessed with either echocardiogram or MUGA scan (with protocol specification for the same method to be used throughout) at baseline, Day 1 of cycle 5, within two weeks of completing chemotherapy and at one year from study entry per protocol in all patients. A physician-directed assessment of cardiac symptoms was conducted 2 years from study entry with LVEF assessment performed at the discretion of the treating providers. Qualifying cardiotoxicity events based on LVEF criteria could occur at any of these time points. Clinical HF events were centrally adjudicated by 2 cardiologists and the study chair and required symptoms of one of the following: grade ≥2 dyspnea or grade ≥2 edema with either a reduction in LVEF to below the lower limit of normal or diastolic dysfunction. Patients with grade 1 dyspnea or edema could also be classified as having clinical HF if the LVEF was <40%. Other signs of HF including auscultation of an S3 gallop, bibasilar rales or evidence of cardiomegaly were collected and used in the central adjudication process.
Anthracycline benefit estimation:
To illustrate the heterogeneity in the benefit-harm trade-offs of anthracycline therapy, we estimated the patient-specific benefit of anthracyclines in addition to taxane-based therapy using two approaches. For our main (“model-based”) analysis, we applied version 2 of the PREDICT model, an externally validated multivariable risk prediction tool, to estimate patient-specific 5 and 10 year breast cancer mortality according to baseline characteristics assuming that all patients with hormone receptor positive tumors were treated with hormone therapy and all patients received third generation therapy with anthracycline + taxane (20, 21). To calculate the increased mortality from a non-anthracycline based regimen, we assumed a hazard ratio of 1.23 based on the observed effect seen in the pooled results of the three Anthracycline in Early Breast Cancer (ABC) trials (15) (comparing docetaxel + cyclophosphamide versus anthracycline + taxane). In an additional “trial-based” analysis, we imputed strata-specific benefits based on the observed absolute benefits in 4 year invasive disease-free survival reported in each of 6 “risk” strata from the ABC trial, based on hormone receptor status (negative or positive) and number of positive lymph nodes (none, 1-3 or 4+).
Statistical Analysis:
Baseline characteristics were summarized using mean and standard deviation for normally distributed values or median and interquartile ranges otherwise. The following predictors assessed at baseline were pre-specified - age, BMI, hypertension and baseline LVEF. Logistic regression was used to estimate the effect of the pre-specified covariates on the outcome of cardiotoxicity at one year. Continuous variables were assessed for linearity. Internal validation was performed by creating 200 bootstrap samples from the original dataset, re-estimating the model coefficients in each bootstrap sample and applying the bootstrapped model coefficients to the original dataset. For both the PREDICT (model-based) and ABC stratified (trial-based) analysis, patients were stratified according to their anticipated absolute benefit of anthracyclines+taxanes compared to non-anthracycline taxane-based therapy into low (<2%), intermediate (2-6%), and high (>6%) predicted absolute benefit; patients in each benefit strata were then classified according to cardiotoxicity risk (harm) into low (<2%), low-intermediate (2-5%), intermediate-high (5-10%) or high (>10%) risk of cardiotoxicity and results depicted graphically.
Results:
Model Derivation Cohort:
Of the 1000 individuals randomized to arm A of E5103, men (n=5) and those that never started therapy (n=28) were excluded, leaving 967 women who initiated therapy in arm A for model derivation. The mean age was 52 years, 85% were Caucasian, the median BMI was 29 kg/m2 and 42% of individuals were on antihypertensive medication or had an elevated blood pressure (Table 1). A total of 51 out of 967 (5.3%) developed cardiotoxicity by one year, 12 (1.2%) with clinical HF and the remaining 39 (4.0%) with asymptomatic LVEF reductions of ≥10% to LVEF <50%. Median LVEF values in those with and without cardiotoxicity over are shown in Figure 1.
Table 1:
E5103 Baseline Characteristics of Patients Receiving Anthracyclines
| Entire Cohort (n=967) | Cardiotoxicity (N=51) | No Cardioxicity (N=916) | |
|---|---|---|---|
| Age (years), mean ± SD | 51.5 ± 9.6 | 55 ± 9.8 | 51.3 ± 9.6 |
| Race N(%) | 2 (<1%) | ||
| Caucasian | 820 (85) | 41 (80) | 779 (85) |
| African American | 96 (10) | 5 (10) | 91 (10) |
| Asian | 41 (4) | 1 (2) | 40 (4) |
| Other | 8 (1) | 4 (8) | 4 (0.4) |
| Peri/Post Menopausal, N (%) | 569 (59) | 38 (75) | 531 (58) |
| ECOG Performance Status | |||
| PS 0 (Fully Active) | 833 (86) | 42 (82) | 792 (86) |
| PS 1 (Mild restriction) | 133 (14) | 9 (18) | 124 (14) |
| BMI (kg/m2), median (IQR) | 28.5 (24.5, 33.7) | 32.9 (26.9, 37.5) | 28.3 (24.4, 33.3) |
| Hypertension, N(%) | 403 (42) | 26 (51) | 377 (41) |
| ACE inhibitor, N(%) | 99 (10) | 8 (16) | 91 (10) |
| Angiotensin II receptor blocker, N(%) | 57 (6) | 4 (8) | 53 (6) |
| Beta-blocker, N(%) | 91(9) | 8 (16) | 83 (9) |
| Calcium channel blocker, N(%) | 52 (5) | 2 (4) | 50 (6) |
| Thiazide Diuretics | 120 (12) | 6 (12) | 114 (12) |
| Systolic blood pressure (mmHg), mean ± SD | 126.2 ± 15.1 | 129.6 ± 18.0 | 126.0 ±14.9 |
| Diastolic blood pressure (mmHg), mean ± SD | 75.2 ± 9.0 | 76.4 ± 10.2 | 75.1 ± 8.9 |
| Left ventricular ejection fraction, median (IQR) | 64 (60, 68) | 60 (57, 65) | 64 (60, 68) |
| Disease Site, N (%) | |||
| Left | 509 (53) | 33 (65) | 476 (52) |
| Right | 434 (45) | 18 (35) | 416 (45) |
| Bilateral | 24 (3%) | 0 | 24 (3) |
| Node status, N (%) | |||
| Node negative | 251 (26) | 11 (22) | 240 (26) |
| 1-3 positive nodes | 406 (42) | 18 (35) | 388 (42) |
| 4+ positive nodes | 310 (32) | 22 (43) | 288 (31) |
| Hormone Receptor Status, N (%) | 0 | ||
| ER and PgR negative | 341 (35) | 19 (37) | 322 (35) |
| ER and/or PgR positive | 626 (65) | 32 (63) | 594 (65) |
| Histologic Grade, N(%) | |||
| I | 72 (8) | 3 (6) | 69 (8) |
| II | 342 (36) | 19 (38) | 323 (36) |
| III | 534 (56) | 28 (56) | 506 (56) |
| Tumor Size, N (%) | |||
| ≤2 cm | 367 (38) | 22 (43) | 345 (38) |
| >2 to ≤5 cm | 503 (52) | 22 (43) | 481 (52) |
| >5 cm | 97 (10) | 7 (14) | 90 (10) |
| Surgery type, N (%) | |||
| Breast conserving surgery | 443 (46) | 23 (45) | 420 (46) |
| Mastectomy | 524 (54) | 28 (55) | 496 (54) |
| Radiation, N (%) | 740 (78) | 43 (84) | 697 (77) |
| Dose Dense chemotherapy, N (%) | 783 (81) | 36 (71) | 747 (82) |
ECOG refers to Eastern Cooperative Oncology Group; PS Performance Status; BMI body mass index; ACE Angiotensin Converting Enzyme; ER Estrogen Receptor; PgR Progesterone Receptor. Hypertension was defined as either on antihypertensive medication at baseline or systolic blood pressure greater than or equal to 140mmHg or diastolic blood pressure greater than or equal to 90mmHg. Dose Dense chemotherapy refers to every 2 week doxorubicin-cyclophosphamide.
Figure 1:
Median LVEF at baseline and protocol-directed time points in the first year. LVEF assessments at 2 years were at the discretion of the treating physician and were only available in a subset of patients. Median LVEF is shown stratified by whether the participant met criteria for cardiotoxicity at any time point. Blue=no cardiotoxicity, Red=Cardiotoxicity
Cardiotoxicity Clinical Prediction Model:
In the multivariate model that included only prespecified covariates, increasing age, higher BMI and lower baseline LVEF were significantly associated with cardiotoxicity at one year (Table 2). Hypertension was included in the prespecified model given associations seen in other cohorts but was not significantly associated with the outcome in multivariate analysis. There was no model improvement (according to AIC) with inclusion of any of the “exploratory” variables (menopausal status, ECOG performance status, radiation exposure or dose dense versus classical). The risk model discrimination, based on c-statistic, was 0.701 (95% CI: 0.627 to 0.774). Bootstrap internal validation yielded an optimism-corrected c-statistic of 0.68 (95% PI 0.62-0.75). Cardiotoxicity rates varied from 2% in the lowest risk quartile to 11% in the highest risk quartile, an extreme quartile risk ratio of 5.5.
Table 2:
Multivariate Clinical Prediction Model for Cardiotoxicity
| Variable | Beta coefficient | OR (95% CI) | p-value |
|---|---|---|---|
| Intercept | −2.6421 | ||
| Age | 0.0427 | 1.04 (1.01 to 1.08) | 0.0119 |
| Body Mass Index | 0.0592 | 1.06 (1.02 to 1.10) | 0.0010 |
| Hypertension | −0.2637 | 0.77 (0.39 to 1.50) | 0.4416 |
| Baseline Left Ventricular Ejection Fraction | −0.0684 | 0.93 (0.89 to 0.98) | 0.0044 |
| Predicted benefit in breast cancer mortality absolute risk reduction at 5 years | Predicted benefit in breast cancer mortality absolute risk reduction at 10 years | ||
Anthracycline benefit versus cardiotoxicity risk:
Of the 967 patients included in the cardiotoxicity analysis, 19 had missing data on breast cancer grade and were excluded from the main (“model-based”) benefit analysis. For our main (“model-based”) analysis, we found substantial heterogeneity of absolute benefit in the E5103 cohort, which was independent of cardiotoxicity risk . In those 19% (176/948) of patients with low anticipated anthracycline-related benefit at 10 years (<2% absolute risk difference in breast cancer mortality at 10 years), 16/176 (9%) had a >10% risk of cardiotoxicity and 61/176 (35%) had a 5-10% risk of cardiotoxicity at 1 year (Figures 2 and 3). In the 67% (631/948) with an intermediate anticipated anthracycline-related benefit (2-6% absolute risk difference in breast cancer mortality at 10 years), 69/631 (11%) had a >10% risk of cardiotoxicity and 182/631 (29%) had a 5-10% risk of cardiotoxicity at 1 year. In a sensitivity analysis, results were similar in our “trial-based” analysis using the ABC trial results subgrouped based on nodal status and hormone receptor status (Supplemental Figure 1), including all 967 individuals from the cardiotoxicity analysis. In subgroups (405/967) with no or low benefit of anthracyclines (<3% risk difference in 4 year invasive disease free survival), 35/405 (9%) had a >10% risk of cardiotoxicity and 118/405 (29%) had a 5-10% risk of cardiotoxicity at one year.
Figure 2:
Risk of cardiotoxicity stratified by anthracycline benefit of breast cancer mortality at 5 (A) and 10 years (B). Patients were divided into strata based on expected reductions in breast cancer mortality of <2% (low), 2 to ≤6% (intermediate) or high (>6%) and the predicted cardiotoxicity is shown in each column n(%). Absolute breast cancer mortality benefit was calculated using the PREDICT model with a modification to compare anthracycline+taxane versus non-anthracycline regimens. Cardiotoxicity was estimated using the cardiotoxicity clinical prediction model. Green shading suggest situations of clear net benefit while red shading suggests situations where the risk of cardiotoxicity may outweigh the benefits. Gray shading denotes situations of uncertainty of risk-benefit.
Figure 3:
Risk of cardiotoxicity stratified by anthracycline benefit of breast cancer mortality at 5 (A, B) and 10 years (C, D). Patients were divided into strata based on expected reductions in breast cancer mortality of <2% (low), 2 to ≤6% (intermediate) or high (>6%) and the predicted cardiotoxicity is shown as a stacked bar graph depicted as numbers of patients (A, C) or relative proportions (B, D). Absolute breast cancer mortality benefit was calculated using the PREDICT model with a modification to compare anthracycline+taxane versus non-anthracycline regimens. Cardiotoxicity was estimated using the cardiotoxicity clinical prediction model. BC refers to breast cancer; Mort to mortality; BEN to benefit.
Discussion:
A clinical prediction model for anthracycline cardiotoxicity can assist patients and providers in estimating personalized benefits and harms. Based on our newly derived cardiotoxicity prediction model, we found clinically significant variation in the risk of cardiotoxicity based on 4 established risk factors, with a 5-fold higher risk of heart failure or reduced LVEF in the highest risk compared to the lowest risk quartiles. The heterogeneity in the risk of cardiotoxicity appeared to be consistent regardless of breast cancer mortality or recurrence risk. Thus, particularly for the many patients falling into subgroups with marginal benefit from the addition of anthracycline therapy (i.e. patients in the low benefit groups, comprising 19% of the trial population in our main analysis) patient-specific information regarding cardiotoxicity risk is highly likely to be decisionally relevant. For example, among women with low anticipated benefit of the addition of anthracycline therapy (<2% absolute difference in breast cancer mortality at 10 years), 9% were at a high predicted risk of cardiotoxicity (>10% risk at 1 year). Given that HF after anthracycline therapy has a poor prognosis(14), some individuals may choose a non-anthracycline regimen in the setting of a high predicted risk of cardiotoxicity and low anticipated absolute benefit of anthracyclines. This analysis also demonstrates that summary statistics for the overall study population obscures clinically significant heterogeneity of both benefit and risk, underscoring the need for risk stratified trial analysis of RCTs using multivariable prediction models. Finally, our study highlights the importance of more routine collection of baseline cardiac risk factors, cardiac disease history and prospective collection of long-term cardiovascular events in cancer trials. Additional studies are needed incorporating long-term cardiac outcomes and external validation prior to implementation of this cardiotoxicity prediction model in routine clinical practice. This information is critical to further refine cardiotoxicity clinical prediction models and provide patients with accurate, personalized information of both risks and benefits of cancer therapy.
The recent pooled analysis of the Anthracyclines in Early Breast Cancer (ABC) Trials – USOR 06-090, NSABP B-46-I/USOR 07132 and NSABP B-49 - showed a small but statistically significant improvement in invasive disease free survival with adjuvant doxorubicin, cyclophosphamide and a taxane compared with the anthracycline-free regimen of docetaxel and cyclophosphamide in women with HER2 negative early breast cancer (15). In contrast, the DBCG 07-READ trial showed no benefit for the addition of epirubicin to cyclophosphamide and docetaxel in TOP2A normal early breast cancer (16). These studies suggest that many patients would experience no or low absolute benefit of anthracyclines to docetaxel and cyclophosphamide treatment alone. In this setting, personalized risk prediction for breast cancer recurrence and mortality (benefit of anthracyclines), harms of anthracycline treatment (cardiotoxicity and secondary malignancies) and patient preferences may help guide adjuvant treatment regimen selection. We recognize that the outcomes compared in our benefit-harm analysis are not symmetric in terms of severity –we include both symptomatic and asymptomatic cardiotoxicity and compare this with breast cancer mortality. However, due to the shorter duration of follow up for cardiotoxicity events in most cancer clinical trials, including E5103, we are not able to develop a model for clinical HF events alone. Future studies collecting long-term cardiac events and using clinical prediction models for breast cancer recurrence instead of cancer mortality alone are needed in order to compare more analogous outcomes. HF endpoints in E5103 were centrally adjudicated by two cardiologists and the study chair using standard criteria requiring signs and symptoms of HF and evidence of either systolic or diastolic dysfunction. HF events occurred in 1.2% of participants and asymptomatic reductions in LVEF in 4% of participants, these rates are roughly similar to other clinical trials(22) and lower than observational studies(5, 12). While some clinical trials have required severe symptoms (New York Heart Association Functional Class III/IV) or CV death, clinical HF events in E5103 included patients with mild symptoms and objective evidence of structural heart disease. Because any diagnosis of cardiomyopathy including asymptomatic reductions in LVEF (ACC/AHA Stage B HF) or symptomatic heart failure with reduced or preserved ejection fraction (ACC/AHA Stage C HF) is associated with increased risk of mortality from cardiovascular causes and morbidity from symptoms and hospitalizations, we believe that the cardiotoxicity definitions included in this study are relevant to patient outcomes(14, 23, 24).
Our cardiotoxicity model prespecified 4 routinely available variables – age, BMI, hypertension and baseline LVEF. Similar to prior studies we found that older age and higher BMI were associated with increased risk of cardiotoxicity(5, 25, 26). Lower baseline LVEF has previously been described as a risk factor for the development of cardiotoxicity with trastuzumab(17), however, our study is the largest to report this finding for anthracycline cardiotoxicity. While hypertension has been found to be a risk factor for the development of anthracycline cardiotoxicity in other cohorts(5, 27), it was not significantly associated with cardiotoxicity in our multivariable risk model. Modifications in the definition of hypertension did not alter the result. One potential explanation for the difference with prior studies is that patients with uncontrolled hypertension were excluded from E5103. While radiation therapy has been described as increasing the risk of HF in patients receiving anthracycline agents(28, 29), inclusion of radiation did not improve cardiotoxicity model performance. This may be due to the timing of cardiotoxicity outcome assessment (i.e. at 1 year from initiation of chemotherapy, before the effects of radiation may be detectable) or to the exposure assessment, as we did not have access to detailed radiation planning scans to calculate mean heart dose or other cardiac specific exposure estimates.
Strengths of this study include the longitudinal assessment of LVEF at multiple time points, central adjudication of clinical HF events and the protocol-driven collection of several important cardiac risk factors. However, this study has several limitations that should be considered. First, in the derivation cohort, protocol-driven assessment of cardiotoxicity occurred only in the first year and thus the outcomes of cardiotoxicity at one year and breast cancer mortality at 5 and 10 years are not balanced. Additional studies that include longer term clinical cardiac outcomes are needed. Second, the PREDICT estimates were derived by using the PREDICT model and adding the effect estimate from a clinical trial to generate absolute benefit estimates of breast cancer mortality, which assumes a constant relative risk increase. Based on the results of the ABC trials, this assumption appears reasonable. Third, anthracycline-related cardiotoxicity cannot be directly estimated without a non-anthracycline treatment group, which was not available in this cohort; although cardiac events in the first year after treatment are most likely due to cancer therapy. Fourth, some important cardiac risk factors such as diabetes, tobacco use and prior cardiac diagnoses were not collected and, if available, the inclusion of these risk factors may have improved model performance. Fifth, LVEF assessment was performed locally by each site and there was no central core lab review. Lastly, further external validation in large cohorts is necessary prior to applying this model in clinical practice.
In conclusion, results from multivariable risk prediction models for both benefit (reduction in breast cancer mortality) and harm (anthracycline cardiotoxicity) suggest that there are patients with breast cancer requiring adjuvant chemotherapy who are predicted to experience low absolute incremental benefit from the use of anthracycline therapy but are at moderate or high risk of cardiotoxicity. Further refinement of multivariable risk prediction tools for cardiotoxicity using longer-term follow up and routine collection of cardiac risk factors is warranted.
Clinical Practice Points:
It is well recognized that anthracyclines can lead to heart failure. Heart failure due to anthracyclines has a poor prognosis and eligible patients may require advanced therapies such as heart transplantation. Prior to this work, individual risk factors for anthracycline cardiotoxicity were known but there were no multivariable risk prediction models to help estimate patient specific risk of cardiotoxicity. In this study we report a multivariable risk prediction model with four routinely available predictors – age, hypertension, body mass index and baseline left ventricular ejection fraction. In addition, recent randomized clinical trials suggest that the overall benefit of additional anthracycline chemotherapy to taxane based therapy in early breast cancer is small and varies based on risk of breast cancer recurrence or mortality. This study adds to the current understanding of benefits and harms of adjuvant anthracycline therapy by using multivariable risk prediction models to quantify the percentage of patients who have a relatively low benefit from anthracyclines but are at high predicted risk of cardiotoxicity. This work highlights the role for multivariable risk prediction models as tools to generate personalized risk estimates for use in shared decision-making discussions. It also highlights the need for routine collection of cardiac risk factors and long-term cardiac outcome data in breast cancer research so that models such as this one can be created, validated and updated. While further external validation of this model in large cohorts with longer follow up for cardiac events is necessary prior to applying this model in clinical practice, we have demonstrated that developing tools for patients and providers to estimate the benefits and harms of treatments may help with personalized shared-decision making.
Supplementary Material
Acknowledgments
Funding: This study was coordinated by the ECOG-ACRIN Cancer Research Group (Peter J. O’Dwyer, MD and Mitchell D. Schnall, MD, PhD, Group Co-Chairs) and supported by the National Cancer Institute of the National Institutes of Health under the following award numbers: CA180820, CA180794, CA189828, CA180795, CA180816. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. government.
Research reported in this publication was partially funded through a Patient-Centered Outcomes Research Institute (PCORI) Award (RR-1705-0001). The views, statements, opinions in this publication are solely the responsibility of the authors and do not necessarily represent the views of the Patient-Centered Outcomes Research Institute (PCORI), its Board of Governors or Methodology Committee.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Conflict of Interest: The authors declare that they have no conflict of interest
References:
- 1.Siegel RL, Miller KD, Jemal A. Cancer Statistics, 2017. CA: a cancer journal for clinicians. 2017;67(1):7–30. [DOI] [PubMed] [Google Scholar]
- 2.Bradshaw PT, Stevens J, Khankari N, et al. Cardiovascular Disease Mortality Among Breast Cancer Survivors. Epidemiology. 2016;27(1):6–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Hanrahan EO, Gonzalez-Angulo AM, Giordano SH, et al. Overall survival and cause-specific mortality of patients with stage T1a,bN0M0 breast carcinoma. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2007;25(31):4952–60. [DOI] [PubMed] [Google Scholar]
- 4.Johnson CB, Davis MK, Law A, et al. Shared Risk Factors for Cardiovascular Disease and Cancer: Implications for Preventive Health and Clinical Care in Oncology Patients. The Canadian journal of cardiology. 2016;32(7):900–7. [DOI] [PubMed] [Google Scholar]
- 5.Pinder MC, Duan Z, Goodwin JS, et al. Congestive heart failure in older women treated with adjuvant anthracycline chemotherapy for breast cancer. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2007;25(25):3808–15. [DOI] [PubMed] [Google Scholar]
- 6.Darby SC, Ewertz M, Hall P. Ischemic heart disease after breast cancer radiotherapy. The New England journal of medicine. 2013;368(26):2527. [DOI] [PubMed] [Google Scholar]
- 7.Vaz-Luis I, Keating NL, Lin NU, et al. Duration and toxicity of adjuvant trastuzumab in older patients with early-stage breast cancer: a population-based study. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2014;32(9):927–34. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Armenian SH, Lacchetti C, Lenihan D. Prevention and Monitoring of Cardiac Dysfunction in Survivors of Adult Cancers: American Society of Clinical Oncology Clinical Practice Guideline Summary. Journal of oncology practice. 2017;13(4):270–5. [DOI] [PubMed] [Google Scholar]
- 9.Mehta LS, Watson KE, Barac A, et al. Cardiovascular Disease and Breast Cancer: Where These Entities Intersect: A Scientific Statement From the American Heart Association. Circulation. 2018;137(8):e30–e66. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Yeh ET, Chang HM. Oncocardiology-Past, Present, and Future: A Review. JAMA cardiology. 2016;1(9):1066–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Narayan HK, Finkelman B, French B, et al. Detailed Echocardiographic Phenotyping in Breast Cancer Patients: Associations With Ejection Fraction Decline, Recovery, and Heart Failure Symptoms Over 3 Years of Follow-Up. Circulation. 2017;135(15):1397–412. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Cardinale D, Colombo A, Bacchiani G, et al. Early detection of anthracycline cardiotoxicity and improvement with heart failure therapy. Circulation. 2015;131 (22): 1981–8. [DOI] [PubMed] [Google Scholar]
- 13.Finkelman BS, Putt M, Wang T, et al. Arginine-Nitric Oxide Metabolites and Cardiac Dysfunction in Patients With Breast Cancer. Journal of the American College of Cardiology. 2017;70(2): 152–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Felker GM, Thompson RE, Hare JM, et al. Underlying causes and long-term survival in patients with initially unexplained cardiomyopathy. The New England journal of medicine. 2000;342(15):1077–84. [DOI] [PubMed] [Google Scholar]
- 15.Blum JL, Flynn PJ, Yothers G, et al. Anthracyclines in Early Breast Cancer: The ABC Trials-USOR 06-090, NSABP B-46-I/USOR 07132, and NSABP B-49 (NRG Oncology). Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2017:JCO2016714147. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Ejlertsen B, Tuxen MK, Jakobsen EH, et al. Adjuvant Cyclophosphamide and Docetaxel With or Without Epirubicin for Early TOP2A-Normal Breast Cancer: DBCG 07-READ, an Open-Label, Phase III, Randomized Trial. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2017;35(23):2639–46. [DOI] [PubMed] [Google Scholar]
- 17.Romond EH, Jeong JH, Rastogi P, et al. Seven-year follow-up assessment of cardiac function in NSABP B-31, a randomized trial comparing doxorubicin and cyclophosphamide followed by paclitaxel (ACP) with ACP plus trastuzumab as adjuvant therapy for patients with node-positive, human epidermal growth factor receptor 2-positive breast cancer. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2012;30(31):3792–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Ezaz G, Long JB, Gross CP, et al. Risk prediction model for heart failure and cardiomyopathy after adjuvant trastuzumab therapy for breast cancer. Journal of the American Heart Association. 2014;3(1):e000472. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Miller KD, O’Neill A, Gradishar W, et al. Double-Blind Phase III Trial of Adjuvant Chemotherapy With and Without Bevacizumab in Patients With Lymph Node-Positive and High-Risk Lymph Node-Negative Breast Cancer (E5103). Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2018;36(25):2621–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Candido Dos Reis FJ, Wishart GC, Dicks EM, et al. An updated PREDICT breast cancer prognostication and treatment benefit prediction model with independent validation. Breast cancer research : BCR. 2017;19(1):58. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Public Health England CU. Predict [cited 2017 1 December]. Available from: http://www.predict.nhs.uk/predict_v2.0.html.
- 22.Mackey JR, Martin M, Pienkowski T, et al. Adjuvant docetaxel, doxorubicin, and cyclophosphamide in node-positive breast cancer: 10-year follow-up of the phase 3 randomised BCIRG 001 trial. The Lancet Oncology. 2013;14(1):72–80. [DOI] [PubMed] [Google Scholar]
- 23.Wang TJ, Evans JC, Benjamin EJ, et al. Natural history of asymptomatic left ventricular systolic dysfunction in the community. Circulation. 2003;108(8):977–82. [DOI] [PubMed] [Google Scholar]
- 24.Yancy CW, Jessup M, Bozkurt B, et al. 2013 ACCF/AHA guideline for the management of heart failure: executive summary: a report of the American College of Cardiology Foundation/American Heart Association Task Force on practice guidelines. Circulation. 2013;128(16):1810–52. [DOI] [PubMed] [Google Scholar]
- 25.Kotwinski P, Smith G, Cooper J, et al. Body Surface Area and Baseline Blood Pressure Predict Subclinical Anthracycline Cardiotoxicity in Women Treated for Early Breast Cancer. PloS one. 2016;11(12):e0165262. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Hequet O, Le QH, Moullet I, et al. Subclinical late cardiomyopathy after doxorubicin therapy for lymphoma in adults. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2004;22(10):1864–71. [DOI] [PubMed] [Google Scholar]
- 27.Armenian SH, Xu L, Ky B, et al. Cardiovascular Disease Among Survivors of Adult-Onset Cancer: A Community-Based Retrospective Cohort Study. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2016;34(10):1122–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.van Nimwegen FA, Schaapveld M, Janus CP, et al. Cardiovascular disease after Hodgkin lymphoma treatment: 40-year disease risk. JAMA internal medicine. 2015;175(6):1007–17. [DOI] [PubMed] [Google Scholar]
- 29.Saiki H, Petersen IA, Scott CG, et al. Risk of Heart Failure With Preserved Ejection Fraction in Older Women After Contemporary Radiotherapy for Breast Cancer. Circulation. 2017;135(15):1388–96. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.