Biomarkers and Cardiotoxicity in Breast Cancer

Corresponding Author

Human epidermal growth factor receptor 2 (HER2) is a transmembrane tyrosine kinase receptor that is overexpressed in approximately 15% to 20% of breast cancer patients.¹ The development of HER2-targeted therapies transformed the prognosis of HER2-positive breast cancer patients, leading to significant improvements in disease-free and overall survival in both early-stage and metastatic settings.² There are several classes of HER2-targeting therapies, including monoclonal antibodies (eg, trastuzumab, pertuzumab), which can be used in combination with other cardiotoxic cancer therapy.

However, these life-prolonging therapies carry a well-recognized risk for cardiotoxicity, especially in patients who have been treated with anthracyclines. In the adjuvant setting, asymptomatic left ventricular dysfunction has been observed in approximately 7% of patients receiving trastuzumab compared with 2% of control subjects,³ and rates reach up to 27% when trastuzumab is combined with anthracycline and cyclophosphamide chemotherapy.⁴ Because of this risk, the American Society of Echocardiography, the U.S. Food and Drug Administration, and the European Society of Cardiology recommend echocardiographic surveillance every 3 months during therapy in all patients receiving HER2-targeted therapies to detect cardiotoxicity, defined variably but often as a ≥10% decline in left ventricular ejection fraction (LVEF) from baseline to <55%.5, 6, 7 Societal recommendations for surveillance include periodic cardiac imaging on the basis of patient risk factors and clinical judgement.⁸ This risk for cardiotoxicity underscores the need to better identify patients at highest risk, balancing the benefits of HER2-targeted therapy against the burden of frequent monitoring and potential treatment interruptions.

There is an unmet need to identify reliable biomarkers that predict and detect cardiotoxicity in breast cancer patients receiving cardiotoxic therapies.⁹ These biomarkers could represent early molecular signals before the development of overt cardiac dysfunction and help identify the patients who are at greatest risk for treatment-related cardiovascular complications. Circulating, blood-derived biomarkers are promising because they could be used for both risk stratification prior to therapy and for early detection of subclinical cardiotoxicity. An ideal biomarker should be easily measurable and actionable. This would enable personalized clinical decisions, such as initiating cardioprotective interventions and reducing cancer therapy interruptions. Importantly, biomarkers should be specific to cardiovascular injury, not confounded by cancer itself or its treatment. They should also predict clinically meaningful cardiac outcomes, such as reduction in left ventricular function or symptomatic heart failure.

To date, there have been inconsistent results in the identification of cancer therapy–related cardiac dysfunction (CTRCD) biomarkers in patients treated with trastuzumab. Cardiac troponins and natriuretic peptides are the most studied and used biomarkers in cardio-oncology. American Society of Clinical Oncology, European Society of Cardiology, and American Heart Association/American College of Cardiology expert consensus and guidelines suggest checking these biomarkers in select high-risk patients receiving trastuzumab.^10,11 However, clinical studies report inconsistent results related to the reliability of troponin and B-type natriuretic peptide and their correlation to reduction in LVEF.^12,13 In addition, these biomarkers are not specific to cardiac injury from cancer-directed therapies and could represent acute coronary syndrome, demand ischemia, myocarditis, or heart failure.

Prior studies in relatively small patient cohorts (n = 78) identified potential biomarkers associated with cardiotoxicity from a preselected panel of 8 candidate biomarkers and found that cardiac troponin I, myeloperoxidase (MPO), and growth differentiation factor-15 are associated with cardiac dysfunction in breast cancer patients treated with serial doxorubicin and trastuzumab. Notably, baseline biomarker levels were not significant predictors of cardiotoxicity. These studies underline the need for further validation of these biomarkers and relevance in larger patient cohorts.^14,15

In this issue of JACC: CardioOncology, Gustafson et al¹⁶ provide a comprehensive evaluation of clinical variables, cardiac magnetic resonance (CMR) imaging parameters, protein biomarkers, and plasma microRNA expression profiles to determine the biomarkers that have the greatest predictive power to determine CTRCD in 136 breast cancer patients treated with trastuzumab. These patients had low-risk cardiovascular profiles, with most patients without histories of hypertension, hyperlipidemia, or diabetes. The study included patients with early-stage HER2⁺ breast cancer receiving sequential anthracycline and trastuzumab. The patients underwent very rigorous cardiovascular surveillance, including CMR imaging every 3 months. In the study population, 37 patients (27.2%) developed CMR-defined CTRCD by Cardiac Review and Evaluation Committee definitions. Of these, the majority (89.2%) experienced ≥10% absolute reductions in LVEF from baseline to <55% without accompanying signs or symptoms of heart failure. Only 2 patients (5.4%) had signs or symptoms of heart failure.

The investigators used machine learning tools along with their ambitious multiomics approach to identify angiopoietin-2 (ANGPT2), MPO, and endoglin (ENG) as the top-ranked biomarkers associated with CTRCD risk, as defined by changes in LVEF on CMR. These 3 protein biomarkers performed strongly, with an area under the receiver-operating characteristic curve of 0.972 and specificity of 0.87. None of the >2,000 plasma microRNAs was found to strongly predict CTRCD risk. Notably, the previously identified candidate biomarkers, growth differentiation factor-15, cardiac troponin I, and B-type natriuretic peptide, did not predict cardiac dysfunction despite reductions in LVEF, suggesting that these are unreliable biomarkers for the development of CTRCD in this patient population.

The 3 circulating biomarkers (ANGPT2, MPO, and ENG) that were the strongest predictors of CTRCD were higher at baseline (prior to chemotherapy) in patients who subsequently developed cardiotoxicity. However, there was substantial overlap in baseline values between those who did and did not develop CTRCD, highlighting that the cutoff values for risk discrimination are unclear. In addition, their temporal patterns during treatment suggest that their clinical utility is limited to the pretreatment setting. ANGPT2 rose significantly at time point 2 (TP2) in both patients with and without CTRCD, and interestingly, the TP2 values in those without CTRCD exceeded the baseline values in patients who eventually developed CTRCD. MPO levels remained stable throughout therapy, showing no meaningful difference between groups. ENG demonstrated a significant increase at TP2 in both groups. These patterns suggest no added value in serial biomarker measurements during cancer treatment.

Furthermore, the biologic source of these proteins remains uncertain. Their trajectories do not align with expected cardiac-specific injury biomarkers. Instead, they may reflect systemic effects of cancer therapy, such as endothelial dysfunction, vascular injury, inflammation, or even tumor-derived signals. This distinction is clinically important. Although these biomarkers may be valuable for baseline risk stratification, their lack of dynamic discriminatory behavior during therapy limits their use for real-time cardiotoxicity surveillance.

The integration of machine learning approaches in biomarker discovery is an exciting advance. However, the model’s identification of soluble CD62 antigen-like family member E and endothelin-1 as predictive biomarkers despite their similar baseline levels in those who did and did not develop CTRCD raises important questions about the robustness of its biomarker selections. Although the investigators are to be strongly commended for validating the biomarkers in another cohort of patients, the attenuation in performance compared with the discovery cohort highlights the variability and limitations of smaller, heterogeneous clinical samples.

Although these findings suggest potential value for predictive biomarkers in identifying patients at highest risk for cardiotoxicity, their clinical utility remains unclear. Several guidelines recommend LVEF surveillance every 3 months for all patients receiving HER2-targeted therapy, regardless of individual risk. However, the absolute risk for symptomatic heart failure remains low, and this “one-size-fits-all” approach can lead to oversurveillance, unnecessary treatment interruptions, and the initiation of cardioprotective medications without clear evidence of benefit. For example, the SUCCOUR (Strain Surveillance During Chemotherapy for Improving Cardiovascular Outcomes) trial demonstrated that more aggressive global longitudinal strain–guided management increased the initiation of guideline-directed medical therapy but did not improve LVEF at 1 year, highlighting the limitations of intensive monitoring strategies in lower risk patients.¹⁷ Furthermore, the recommendations for frequent surveillance do not differentiate between monoclonal antibodies (such as trastuzumab) and other HER2-targeted therapies, including antibody-drug conjugates and tyrosine kinase receptors, that are much less likely to cause declines in LVEF.¹⁸

Predictive biomarkers could help stratify patients prior to the initiation of cancer-directed therapy, identifying those at highest risk who warrant closer surveillance and early cardioprotective strategies, while allowing lower risk patients to undergo less frequent imaging without compromising outcomes. Such risk assessment, incorporating baseline cardiovascular risk factors, cancer therapy regimens, and baseline biomarkers, could reduce unnecessary surveillance imaging and focus attention on higher risk patients. To realize this potential and application in clinical practice, prospective studies in larger and more diverse populations are needed to determine whether predictive biomarkers can move beyond statistical association to ultimately guide personalized cardiac monitoring strategies in breast cancer patients.

Funding Support and Author Disclosures

Dr Baik is a consultant and an advisory board member for Kiniksa Pharmaceuticals; and is a consultant for Lexeo Therapeutics.