Abstract
Background
Strategies to estimate risk of cancer therapy‐related cardiac dysfunction (CTRCD) before initiating cardiotoxic cancer treatment are needed. We hypothesized that baseline ECG markers could identify patients at risk for CTRCD.
Methods and Results
In this retrospective cohort study, 1278 female patients with stage I–III HER2 (human epidermal growth factor receptor 2)‐positive breast cancer meeting the following inclusion criteria were included: baseline ECG with QRS <120 milliseconds, baseline echocardiogram, and ≥1 follow‐up echocardiogram. Quantitative measurements of ECG waveform parameters were performed using MUSE (GE Healthcare). The primary outcome of interest was CTRCD at 1 year, defined by left ventricular ejection fraction decline (≥10% to <53% or ≥16% from baseline), or clinical heart failure (New York Heart Association class III/IV). Mean age was 51.7±11.1 years, 990 (77%) received anthracyclines, and all received HER2‐targeted therapy. CTRCD occurred in 160 (13%) patients. In a multivariable Cox proportional hazards model adjusting for our previously published CTRCD risk score (composed of patient and treatment‐specific factors), 4 ECG markers remained independently associated with CTRCD risk: QRS axis, R‐wave duration (lead II), ST segment deviation (lead II), and Sokolow–Lyon voltage (all P<0.05). Compared with a model using only clinical CTRCD risk variables, addition of ECG parameters provided incremental value for predicting CTRCD risk (P<0.001, likelihood ratio test) with continuous net reclassification improvement of 34.9% and integrated discrimination improvement of 3.4%.
Conclusions
Baseline ECG variables are predictive of subsequent CTRCD and provide incremental value to established clinical risk factors for CTRCD risk classification.
Keywords: cardio‐oncology, cardiotoxicity, ECG, risk factors
Subject Categories: Cardio-Oncology
Nonstandard Abbreviations and Acronyms
- CTRCD
cancer therapy‐related cardiac dysfunction
- HER2
human epidermal growth factor receptor 2
- NRI
net reclassification improvement
Clinical Perspective.
What Is New?
In this analysis of a large retrospective cohort of women with HER2 (human epidermal growth factor receptor 2)‐positive breast cancer, several ECG markers assessed before starting cardiotoxic cancer therapy were independently associated with the risk of cancer therapy‐related cardiac dysfunction.
Notably, compared with a model composed of only clinical risk variables, addition of ECG parameters provided incremental value for predicting risk of cancer therapy‐related cardiac dysfunction.
What Are the Clinical Implications?
Findings from this study suggest the potential of baseline ECG data to improve risk stratification for cancer therapy‐related cardiac dysfunction among patients with breast cancer, which may help inform the optimal level of cardiac surveillance during cardiotoxic cancer therapy or identify at‐risk patients who may benefit from cardioprotective strategies, and further investigation is warranted.
Cardiotoxicities are well‐defined, treatment‐limiting, adverse events for patients receiving treatment with several classes of anticancer therapies (eg, anthracyclines, HER2 [human epidermal growth factor receptor 2]‐targeted therapy). Cancer therapy‐related cardiac dysfunction (CTRCD), manifested by clinical heart failure or left ventricular (LV) dysfunction, is a common cardiotoxicity that leads to early anticancer treatment interruption and adversely affects cancer and cardiovascular outcomes. 1 , 2 , 3 Treatment and patient‐specific variables associated with CTRCD‐risk have been well described, including age, anthracycline exposure, cardiovascular risk factors (eg, hypertension, diabetes), and baseline LV ejection fraction (LVEF). Current strategies for CTRCD risk stratification rely on established clinical variables and routine surveillance echocardiograms but are imprecise for predicting which patients are at high (and low) risk for cardiotoxicity.
The 12‐lead ECG has been shown to identify and predict LV systolic dysfunction and heart failure. 4 , 5 , 6 Although not all ECG correlates are known, especially when artificial intelligence (AI) approaches are used, several specific ECG characteristics have been identified to correlate with LV systolic dysfunction or heart failure, including abnormal QRS duration, 7 ST changes, 8 abnormal P‐wave terminal force in V1, 9 and LV hypertrophy. 10 The 2022 European Society of Cardiology guidelines recommend routinely performing an ECG at baseline, before starting any cardiotoxic anticancer therapies. 11 However, the value of this recommendation for CTRCD risk stratification by itself or in combination with other clinical variables remains undefined.
The purpose of this study was to (1) determine if and which baseline ECG parameters are associated with risk of CTRCD in women with breast cancer receiving HER2‐targeted therapy, and (2) ascertain if and which ECG parameters are incremental to established clinical variables for estimating risk of CTRCD.
METHODS
The data that support the findings of this study are available from the corresponding author upon reasonable request.
Study Design
This was a single‐center retrospective cohort study of consecutive patients with HER2‐positive breast cancer treated at an academic cancer institute between January 2004 and December 2013. Inclusion criteria for this study included female sex, stage I to III disease, treatment with trastuzumab‐based therapy, and a baseline 12‐lead ECG and echocardiogram before initiation of trastuzumab. Exclusion criteria included atrial fibrillation, ventricular‐paced rhythm, QRS duration >120 milliseconds, and absence of follow‐up echocardiograms during trastuzumab‐based therapy. A waiver of informed consent was obtained, and the study was approved by the institutional review board.
Clinical Data Collection
Demographic and clinical characteristics were collected from the electronic medical record, including age at treatment, race (Black or non‐Black), cancer treatment, baseline cardiovascular risk factors (ie, hypertension, diabetes, hyperlipidemia, current or former smoking) and cardiovascular conditions including coronary artery disease and arrhythmia (ie, atrial fibrillation, atrial flutter, supraventricular tachycardia, sick sinus syndrome, or frequent premature ventricular contractions), and baseline concomitant cardiac medications.
Digital 12‐lead ECGs were recorded at 25 mm/s and 1 mV/cm according to standard protocol at a sampling rate of 500 Hz using a 150‐Hz high‐pass filter. Variables of interest included heart rate and quantitative indices of atrial depolarization and ventricular depolarization/repolarization as represented by axis (P, QRS, and T), QRS‐T angle, duration (P, QRS), interval (PR, QTc), amplitude (Q, R, S, and T), deviation (of ST segment), and T‐wave principal component analysis of waveforms that were measured from median beats for each of 12 leads acquired from a 10‐second surface ECG (see Table S1 for a comprehensive list). Cornell voltage was measured as SV3+RaVL, and criteria for LV hypertrophy were defined as ≥2.0 mV in women. 10 Sokolow–Lyon voltage was measured as SV1+RV5 or RV6 and criteria for LV hypertrophy were defined as ≥3.5 mV. Two‐dimensional and Doppler echocardiograms were performed per standard of care. To minimize variability, all measurements were performed in an automated fashion from digital 12‐lead ECG files using MUSE and Magellan software (GE Healthcare), as previously described. 12 , 13 , 14
Outcomes
The primary outcome was CTRCD, defined as an absolute decline of LVEF ≥10% to below the lower limit of normal (53%) or ≥16% from pre‐reatment baseline (with or without clinical heart failure), as previously recommended by the Task Force for cancer treatments and cardiovascular toxicity of the European Society of Cardiology. 15 , 16 CTRCD events were identified through review of the medical record including clinic visits by an oncologist or cardiologist, review of LVEF measurements from all available cardiac imaging modalities, and review of the medication administration records to identify interruptions or delays in treatment that could be due to possible cardiac‐related side effects of therapy. CTRCD events were further characterized based upon presence or absence of clinical heart failure (New York Heart Association class III or IV). All cases of CTRCD were adjudicated by a board‐certified cardiologist.
Statistical Analysis
Data are presented as mean± SD for continuous measures and percentage for categorical variables. Differences between patients with or without CTRCD were compared using the Wilcoxon rank‐sum test for continuous variables and Fisher's exact test for categorical variables. The primary end point was 1‐year CTRCD‐free survival from start of HER2‐targeted treatment. Patients with no CTRCD at 1 year were censored. No deaths were observed during the 1‐year follow‐up for patients in this cohort. The associations between ECG variables and 1‐year risk of CTRCD was evaluated using univariate Cox proportional hazards regression. The ECG characteristics considered in the analysis were selected with consideration for clinical significance, including variables that are representative of atrial or ventricular structure and function or variables that have previously been associated with cardiovascular outcomes. For variables that were measured individually for the 12 ECG leads (ie, Q amplitude, R amplitude, R duration, ST deviation, and T amplitude), specific ECG leads were selected based upon statistical significance. A multivariable Cox regression model was constructed for ECG variables that showed a significant association in the univariate analysis at a P value threshold of <0.10 using stepwise selection (entry=0.4, stay=0.3). Models were adjusted for a risk score that included 9 clinical predictors of CTRCD (ie, age, race [Black versus non‐Black], body mass index (≥30 versus <30 kg/m2), baseline LVEF (≥60% versus <60%), systolic blood pressure (≥130 versus <130 mm Hg), coronary artery disease, diabetes, arrhythmia, and anthracycline treatment) based on our prior study of women with HER2‐positive breast cancer. 15 Briefly, each clinical covariate in our prior study was assigned a weighted point value based upon its regression coefficient in the multivariable regression model, and a risk score calculated as the total sum of points was used to build a risk nomogram. The risk nomogram demonstrated good discrimination with a concordance index of 0.687.
To assess the incremental value of ECG variables for CTRCD risk assessment, this study used 2 distinct approaches. First, the final parsimonious model of the CTRCD risk score plus ECG variables was compared with the model of CTRCD risk score alone using the likelihood ratio chi‐square test. Second, the net reclassification improvement (NRI) was calculated based upon methods by Pencina et al. 17 The NRI is a metric that quantifies the relative increase in predicted probabilities for subjects who experience an event and decrease for subjects who do not develop events. We defined 3 risk levels for 1‐year CTRCD based upon the lower and upper quartiles: low risk (<6%), intermediate risk (6%–16%), and high risk (>16%), and the categorical NRI was calculated for participants who experienced a CTRCD event and those who remained free of CTRCD within 12 months of follow‐up. Given the lack of established CTRCD risk categories, a continuous NRI that considers any change in predicted risk in the correct direction as appropriate was also calculated. Finally, we calculated the integrated discrimination improvement, which corresponds to the difference in discrimination slopes of the 2 models in comparison. 17 A P value <0.05 was considered statistically significant. All statistical analyses were performed using SAS Version 9.4 (SAS Institute, Inc., Cary, NC) and R version 4.3.2 (R Foundation for Statistical Computing, Vienna, Austria).
RESULTS
Patient Characteristics
A total of 1278 patients were included in this analysis. Baseline characteristics of the study population are shown in Table 1. The mean age of the study population was 51.7±11.1 years, 990 (77%) received anthracycline chemotherapy, and 869 (68%) received radiotherapy for stage 1 (37%), 2 (40%), or 3 (23%) invasive breast cancer (52% left sided and 48% right sided). Cardiovascular risk factors were common including hypertension (23%), hyperlipidemia (17%), and diabetes (8%). Prevalence of coronary artery disease (1%) and arrhythmia (1%) was low. Renin angiotensin aldosterone antagonists (15%), statins (13%), and beta blockers (8%) were the most common cardiac medications at baseline. Median baseline LVEF was 66% (interquartile range, 63–70), with 88% having a baseline LVEF ≥60%. Median number of follow‐up LVEF assessments within 1 year of treatment initiation was 3 (interquartile range, 3–4). Overall, 160 (13%) women developed CTRCD: 130 with asymptomatic CTRCD based upon a significant LVEF decline without heart failure and 30 with symptomatic CTRCD based upon New York Heart Association class III/IV heart failure.
Table 1.
Patient Characteristics
| Characteristic | Overall Cohort (n=1278) |
|---|---|
| Age, y | 51.7±11.1 |
| Body mass index, kg/m2 | 26.7±6.0 |
| <25 kg/m2 | 592 (46) |
| 25–29 kg/m2 | 395 (31) |
| ≥30 kg/m2 | 291 (23) |
| Estrogen receptor‐positive | 810 (63) |
| Progesterone receptor‐positive | 614 (48) |
| Stage | |
| I | 475 (37) |
| II | 506 (40) |
| III | 297 (23) |
| Anthracyclines (prior/current) | 990 (77) |
| Median trastuzumab dose, mg/kg (IQR) | 106 (104–110) |
| Radiation therapy | 869 (68) |
| Median baseline left ventricular ejection fraction, % (IQR) | 66 (63–70) |
| <60% | 147 (11.5) |
| ≥60% | 1131 (88.5) |
| Heart rate (beats/min) | 81±10 |
| Baseline systolic blood pressure | |
| <130 mm Hg | 953 (75) |
| ≥130 mm Hg | 325 (25) |
| Race | |
| Black | 151 (12) |
| Non‐Black* | 1110 (87) |
| Hypertension | 297 (23) |
| Diabetes | 96 (8) |
| Hyperlipidemia | 217 (17) |
| Coronary artery disease | 18 (1) |
| Arrhythmia† | 13 (1) |
| Smoking (current or former) | 425 (33) |
| Cardiac medications at baseline | |
| Beta‐blockers | 99 (8) |
| Calcium‐channel blockers | 68 (5) |
| 3‐hydroxy‐3‐methylglutaryl coenzyme A reductase inhibitors | 163 (13) |
| Renin angiotensin aldosterone antagonist | 192 (15) |
Values are presented as mean±SD or n (%), unless otherwise specified. IQR indicates interquartile range.
Non‐Black includes White 969 (76); Asian 114 (9); Other/unknown 44 (3). American Indian (n=1) and Native Hawaiian or other Pacific Islander (n=1) were included in Other (n=25) and Unknown (n=17) categories.
Arrhythmia was defined as a history of atrial fibrillation, atrial flutter, supraventricular tachycardia, sick sinus syndrome, or frequent premature ventricular contractions.
Baseline ECG
Baseline ECG characteristics are summarized in Table 2. The baseline heart rate was higher (74±13 bpm versus 72±12 bpm; P=0.019) and QTc was longer (435±22 versus 428±21; P<0.001) in patients who developed CTRCD compared with those who did not. LV hypertrophy by ECG criteria was more prevalent in patients who developed CTRCD compared with those who did not, based upon either Sokolow–Lyon criteria (6% versus 1%; P<0.001) or Cornell criteria (11% versus 5%; P=0.006).
Table 2.
ECG Characteristics According to the Presence or Absence of Cancer Therapy‐Related Cardiac Dysfunction
| Total (n=1278) | CTRCD (n=160) | No CTRCD (n=1118) | P value | |
|---|---|---|---|---|
| ECG variable | ||||
| Heart rate, bpm | 71.8±11.7 | 74.0±12.9 | 71.5±11.5 | 0.019 |
| PR interval, ms | 151.2±22.7 | 155.5±25.1 | 150.6±22.3 | 0.067 |
| QRS duration, ms | 85.0±8.8 | 85.1±9.1 | 85.0±8.7 | 0.709 |
| QTc interval, ms | 429.2±21.1 | 434.7±22.2 | 428.4±20.8 | <0.001 |
| P axis, degree | 51.5±57.0 | 48.9±19.7 | 51.9±60.5 | 0.968 |
| QRS axis, degree | 36.0±30.2 | 27.6±26.4 | 37.2±30.5 | <0.001 |
| T axis, degree | 39.1±21.9 | 34.7±23.6 | 39.7±21.6 | 0.001 |
| P duration, ms | 104.1±13.6 | 106.0±12.6 | 103.9±13.7 | 0.130 |
| QRST angle, degree | 34.8±19.5 | 35.5±20.5 | 34.7±19.3 | 0.751 |
| PCA ratio | 16.8±9.0 | 17.3±8.1 | 16.7±9.1 | 0.169 |
| Q amplitude aVL, μV | 39.5±54.3 | 50.4±57.1 | 37.9±53.7 | 0.002 |
| R amplitude I, μV | 765.5±345.3 | 933.4±379.4 | 741.4±333.5 | <0.001 |
| R amplitude aVL, μV | 427.1±328.5 | 563.5±362.4 | 407.6±318.8 | <0.001 |
| R duration II, ms | 58.7±16.9 | 63.3±17.3 | 58.1±16.8 | <0.001 |
| Max S amplitude V1, μV | 827.5±324.9 | 926.1±367.6 | 813.4±316.0 | <0.001 |
| STm V5, μV | 19.3±27.3 | 13.6±28.2 | 20.1±27.0 | 0.015 |
| STe II, μV | 45.5±36.3 | 37.0±33.6 | 46.7±36.5 | 0.001 |
| Max STm V5V6, μV | 21.9±25.6 | 16.7±25.0 | 22.6±25.6 | 0.014 |
| T amplitude V5, μV | 291.4±146.6 | 256.1±158.8 | 296.5±144.1 | <0.001 |
| Sokolow–Lyon voltage, μV | 2014.5±596.9 | 2156.4±738.7 | 1994.2±571.3 | 0.067 |
| LVH (by Sokolow–Lyon criteria) | 21 (1.6) | 9 (5.6) | 12 (1.1) | <0.001 |
| Cornell voltage, μV | 1143.9±516.0 | 1273.9±547.6 | 1125.2±508.9 | 0.002 |
| LVH (by Cornell criteria) | 76 (6.0) | 18 (11.3) | 58 (5.2) | 0.006 |
Values are presented as mean±SD or n (%). CTRCD indicates cancer therapy‐related cardiac dysfunction; and LVH, left ventricular hypertrophy.
Association of Baseline ECG Characteristics and CTRCD
Several routine ECG parameters were associated with CTRCD in univariate Cox analyses, including heart rate (hazard ratio [HR] 1.02 per bpm, P=0.007), PR interval (HR 1.01 per ms, P=0.024), QRS axis (HR 0.99 per degree, P<0.001), and QTc interval (HR 1.01 per ms, P<0.001). Patients in the highest heart rate quartile (≥80 bpm) had a higher risk of CTRCD (HR, 1.74 [95% CI, 1.13–2.69]) compared with the lowest heart rate quartile (≤64 bpm). Patients in the highest QTc quartile (≥444 milliseconds) had a higher risk of CTRCD (HR, 2.16 [95% CI, 1.40–3.34]) compared with the lowest QTc quartile (≤416 ms). Several ECG waveform measurements including amplitude (Q in aVL, R in I and aVL, T in V5), duration (R in II), and segment deviation (ST in II and V5) were significantly associated with CTRCD (Table 3). Using continuous measures, Sokolow–Lyon voltage (HR 1.05 per 100 μV [95% CI, 1.02–1.08]) and Cornell voltage (1.05 per 100 μV [95% CI, 1.02–1.08]) were significantly associated with CTRCD. ECG evidence of LV hypertrophy was associated with risk of CTRCD, based upon both Sokolow–Lyon criteria ≥3500 μV (HR, 5.15 [95% CI, 2.41–11.02]) and Cornell criteria ≥2000 μV (HR, 2.00 [95% CI, 1.21–3.30]). In multivariable adjusted analyses, QRS axis (HR, 0.99; P=0.031), R duration in lead II (HR, 1.01, P<0.007), ST‐segment deviation in lead V5 (HR, 0.99, P=0.039), and Sokolow–Lyon voltage (HR, 1.04, P=0.009) remained significantly associated with CTRCD.
Table 3.
Association Between ECG Parameters and Cancer Therapy‐Related Cardiac Dysfunction Risk
| ECG variable | Univariate | Multivariable* | ||
|---|---|---|---|---|
| HR (95% CI) | P value | HR (95% CI) | P value | |
| Heart rate (per bpm) | 1.018 (1.005–1.031) | 0.0067 | 1.012 (0.997–1.027) | 0.1310 |
| PR interval (per ms) | 1.008 (1.001–1.015) | 0.0244 | 1.004 (0.997–1.012) | 0.2309 |
| QRS duration (per ms) | 0.999 (0.981–1.017) | 0.8901 | … | … |
| QTc interval (per ms) | 1.014 (1.006–1.022) | 0.0003 | 1.007 (0.999–1.015) | 0.1021 |
| P axis (per degree) | 0.999 (0.997–1.002) | 0.5898 | … | … |
| QRS axis (per degree) | 0.991 (0.986–0.995) | <0.0001 | 0.991 (0.984–0.999) | 0.0314 |
| T axis (per degree) | 0.989 (0.980–0.999) | 0.0300 | … | … |
| P duration (per ms) | 1.010 (0.999–1.022) | 0.0819 | 0.990 (0.979–1.001) | 0.0782 |
| QRST angle (per degree) | 1.003 (1.001–1.006) | 0.4589 | … | … |
| PCA ratio (per unit) | 1.008 (0.994–1.021) | 0.2777 | … | … |
| Q amplitude aVL (per μV) | 1.003 (1.001–1.006) | 0.0030 | 1.001 (0.998–1.004) | 0.3434 |
| R amplitude I (per 100 μV) | 1.139 (1.096–1.184) | <0.0001 | … | … |
| R amplitude aVL (per 100 μV) | 1.117 (1.076–1.160) | <0.0001 | … | … |
| R duration II (per ms) | 1.016 (1.007–1.026) | 0.0004 | 1.014 (1.004–1.024) | 0.0069 |
| Max S amplitude V1 (per μV) | 1.001 (1.000–1.001) | <0.0001 | … | … |
| STm V5 (per μV) | 0.991 (0.984–0.997) | 0.0049 | 0.993 (0.987–1.000) | 0.0389 |
| STe II (per μV) | 0.992 (0.988–0.997) | 0.0009 | … | … |
| Max STm V5V6 (per μV) | 0.990 (0.984–0.997) | 0.0051 | … | … |
| T amplitude V5 (per μV) | 0.998 (0.997–0.999) | 0.0046 | … | … |
| Sokolow–Lyon voltage (per 100 μV) | 1.047 (1.016–1.079) | 0.0031 | 1.044 (1.011–1.077) | 0.0090 |
| Cornell voltage (per 100 μV) | 1.048 (1.019–1.077) | 0.0009 | 0.974 (0.939–1.010) | 0.1506 |
CTRCD indicates cancer therapy‐related cardiac dysfunction; and HR, hazard ratio.
Adjusted for CTRCD score comprised of conventional risk factors including left ventricular ejection fraction (≥60% vs <60%), body mass index (≥30 vs <30 kg/m2), baseline systolic blood pressure (≥130 vs <130 mm Hg), coronary artery disease (yes vs no), anthracycline (yes vs no), arrhythmia (yes vs no), diabetes (yes vs no), race (Black vs non‐Black), and age, as previously described.
Incremental Value of ECG Characteristics for Risk Prediction of CTRCD
Compared with a model using only clinical CTRCD risk variables, addition of ECG parameters provided incremental value for predicting risk of CTRCD (P<0.001 using likelihood ratio test; Figure). Of the 160 patients who developed CTRCD, 28 (17.5%) were correctly reclassified to a higher risk category and 17 (10.6%) were reclassified to a lower risk category, accounting for a net improvement of 6.9% patients better reclassified using ECG. Of the 1118 patients who remained free of CTRCD, 272 (24.3%) were correctly reclassified to a lower risk category and 98 (8.8%) were reclassified to a higher risk category, accounting for a net improvement of 15.5%. The overall NRI was 22.4% (95% CI, 13.7–31.2%; P<0.001). The continuous NRI and integrated discrimination improvement for predicting CTRCD using clinical variables plus ECG was 34.9% (95% CI, 18.5–51.2%; P<0.001) and 3.4% (95% CI, 1.9–4.8%; P<0.001).
Figure 1. Incremental value of baseline ECG characteristics for prediction of cancer therapy‐related cardiac dysfunction.
The addition of ECG variables added incremental value for prediction of CTRCD over a baseline model of clinical CTRCD risk variables (P<0.001) in patients with breast cancer receiving human epidermal growth factor receptor 2‐targeted therapy. Created with BioRender.com. CTRCD indicates cancer therapy‐related cardiac dysfunction.
DISCUSSION
In this study, we investigated the utility of baseline ECG parameters to estimate risk of CTRCD during HER2‐targeted therapy among patients with breast cancer. Several ECG markers were significantly associated with CTRCD risk, independent of established CTRCD risk factors. Furthermore, building upon previous work by our group, we demonstrate that baseline ECG parameters provide incremental value to established CTRCD risk factors for prediction of subsequent CTRCD during HER2‐targeted therapy. These findings offer promising evidence to support the role of ECGs for risk stratification and surveillance of CTRCD during cardiotoxic cancer therapy that warrants further investigation.
CTRCD represents an important treatment‐limiting adverse effect of cardiotoxic cancer therapies that affects both oncologic and cardiovascular outcomes. Strategies for risk stratification and early detection of CTRCD are therefore critical so that preventative treatment and appropriate cardiac surveillance can be implemented. We and others have previously developed risk models for prediction of CTRCD using clinical variables. 18 , 19 , 20 Our 9‐variable risk nomogram is a simple‐to‐use tool based upon readily accessible clinical variables that estimates 1‐year risk of CTRCD during HER2‐targeted therapy with good discrimination (c‐index=0.687). Results from the current study suggest that such CTRCD risk prediction can be improved by incorporating parameters from a baseline ECG, which is a routine cardiac exam recommended for baseline risk assessment before initiation of cancer therapy. In analyses adjusted for established CTRCD risk variables, Sokolow–Lyon voltage, QRS axis, R duration (lead II), and median ST deviation (lead V5) were significantly associated with CTRCD risk.
The association between ECG changes and cardiovascular outcomes has been reported in several prior studies. In the Copenhagen Heart Study, multiple ECG findings (ie, Q waves, ST and T wave changes, conduction defects, and LV hypertrophy) were associated with cardiovascular disease events, and adding ECG findings to conventional risk factors improved cardiovascular disease risk prediction. 21 A similar analysis of the Women's Health Initiative Study also demonstrated that minor and major ECG abnormalities were associated with risk for cardiovascular disease, independent of established risk factors. 22 Current evidence on the association of ECG with incident cardiovascular toxicity in patients with cancer is more limited. Potential ECG predictors that have been identified include T wave changes, QT prolongation, QRS duration, fragmented QRS morphology, and repolarization indices. 23 , 24 , 25 , 26 , 27 Limitations of these prior studies include qualitative or semiquantitative visual methodology for ECG analysis, small sample size, and lack of assessment for incremental value when ECG is added to established clinical risk variables. We hypothesize that baseline ECG variables identified in this study may reflect subclinical cardiomyopathic (ie, increased LV voltage or abnormal LV axis), arrhythmogenic (ie, P or R wave duration), or ischemic (ie, Q wave or ST deviation) changes that may increase susceptibility to further cardiovascular injury upon exposure to sequential cardiovascular insults, such as cardiotoxic cancer therapy.
In addition to refining our previously published approach to CTRCD risk stratification, this study's findings support the potential that ECG may have potential clinical implications. More accurate CTRCD risk stratification can help clinicians to identify high‐risk patients who may benefit from more frequent or advanced surveillance strategies (ie, myocardial strain imaging or circulating cardiovascular biomarkers). Conversely, costly cardiac testing of limited utility could be reduced for patients at low risk. Overall, better risk stratification could help to mitigate risk‐imaging mismatch that is a consequence of the one‐size‐fits‐all approach for CTRCD surveillance. More accurate risk stratification with ECG could also help to identify the most appropriate at‐risk patient populations in whom to conduct efficacy trials of cardioprotective interventions to prevent CTRCD.
AI technology further increases the value of ECGs, enabling the detection of ECG characteristics or patterns that otherwise would be undetectable by experienced clinicians using standard diagnostic criteria. AI‐based ECG algorithms have been developed that can identify LV dysfunction (EF ≤35%) with a sensitivity and specificity of 86% using ECG data alone. 4 Among patients without ventricular dysfunction, this algorithm can also identify patients at risk for developing future ventricular dysfunction. 28 This algorithm was recently validated in a prospective cohort of 989 women with breast cancer and detected an LVEF <50% and ≤35% after anthracycline therapy with an area under the curve of 0.93 and 0.94, respectively. The ability of AI‐based ECG algorithms to estimate risk of future LV systolic dysfunction in patients with cancer receiving cardiotoxic cancer therapy has not previously been studied and warrants further investigation.
Limitations
This study was performed using retrospective data. The limitations in sample size and CTRCD event rate warrant validation of our findings in a larger external data set among patients with breast cancer (or other solid tumors) receiving trastuzumab or other HER2‐targeted therapies. The study population included patients receiving standard treatment regimens, although a larger proportion received anthracycline‐based regimens. Surveillance echocardiograms were performed at variable intervals at the discretion of clinicians, possibly leading to underestimation of CTRCD events. The consensus definition for CTRCD from the 2022 European Society of Cardiology cardio‐oncology guidelines was not used due to absence of data elements in this retrospective study (ie, global longitudinal strain, cardiac biomarkers); however, we used similar LVEF criteria that are consistent with several prior studies. 29 , 30 , 31 The number of symptomatic CTRCD cases was limited in this study population, thus assessment of the association between ECG parameters and CTRCD severity could not be performed. Only patients with a normal QRS duration (<120 ms) were included in this study due to technical challenges of measuring certain ECG waveform parameters (ie, ST‐segment deviation) in the presence of a bundle branch block. This study analyzed only a limited number of candidate ECG parameters, and we cannot exclude the presence of other ECG characteristics or patterns (which may not have been previously described) that are associated with CTRCD risk and could further refine CTRCD risk stratification. We plan to address this limitation in a future study that will employ an artificial intelligence‐enabled ECG algorithm to estimate CTRCD risk during HER2‐targeted therapy. Current guidelines do not recommend repeat ECGs after the baseline assessment during anthracycline or HER2‐targeted therapy and thus were not available for patients in this cohort, however the predictive value of follow‐up ECGs may warrant future study.
CONCLUSIONS
A baseline ECG may be a powerful tool to identify individuals at increased risk for CTRCD during cardiotoxic cancer therapy. Importantly, this study demonstrates that ECG data provide independent and incremental value to established clinical variables of CTRCD risk and could be used to develop a personalized approach to CTRCD surveillance and prevention that is focused on patients most likely to benefit. Additional studies are needed to validate these findings in independent cohorts of patients with breast cancer or other tumor types and further investigate the predictive value of ECG using AI‐based approaches.
Sources of Funding
Research at Memorial Sloan Kettering Cancer Center is supported in part by a National Institutes of Health/National Cancer Institute Cancer Center Support Grant (P30 CA008748). Anthony F. Yu is supported in part by a National Institutes of Health/National Cancer Institute grant (R37CA273923).
Disclosures
Dr Kosmidou reports personal fees from Abbott, Edwards, Laminar, Pfizer, and Sanofi. Dr Yu reports personal fees from Genentech and Avacta.
Supporting information
Table S1
Acknowledgments
Editorial support at Memorial Sloan Kettering Cancer Center was provided by Katharine Olla Inoue, MA, ELS.
This article was sent to Tochukwu M. Okwuosa, DO, Associate Editor, for review by expert referees, editorial decision, and final disposition.
This work was presented in part at the American College of Cardiology Scientific Sessions on April 8, 2024.
Supplemental Material is available at https://www.ahajournals.org/doi/suppl/10.1161/JAHA.124.039203
For Sources of Funding and Disclosures, see page 8.
References
- 1. Yu AF, Yadav NU, Lung BY, Eaton AA, Thaler HT, Hudis CA, Dang CT, Steingart RM. Trastuzumab interruption and treatment‐induced cardiotoxicity in early HER2‐positive breast cancer. Breast Cancer Res Treat. 2015;149:489–495. doi: 10.1007/s10549-014-3253-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Copeland‐Halperin RS, Al‐Sadawi M, Patil S, Liu JE, Steingart RM, Dang CT, Yu AF. Early trastuzumab interruption and recurrence‐free survival in ERBB2‐positive breast cancer. JAMA Oncol. 2020;6:1971–1972. doi: 10.1001/jamaoncol.2020.4749 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Yu AF, Flynn JR, Moskowitz CS, Scott JM, Oeffinger KC, Dang CT, Liu JE, Jones LW, Steingart RM. Long‐term cardiopulmonary consequences of treatment‐induced cardiotoxicity in survivors of ERBB2‐positive breast cancer. JAMA Cardiol. 2020;5:309–317. doi: 10.1001/jamacardio.2019.5586 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Attia ZI, Kapa S, Lopez‐Jimenez F, McKie PM, Ladewig DJ, Satam G, Pellikka PA, Enriquez‐Sarano M, Noseworthy PA, Munger TM, et al. Screening for cardiac contractile dysfunction using an artificial intelligence‐enabled electrocardiogram. Nat Med. 2019;25:70–74. doi: 10.1038/s41591-018-0240-2 [DOI] [PubMed] [Google Scholar]
- 5. Attia ZI, Kapa S, Yao X, Lopez‐Jimenez F, Mohan TL, Pellikka PA, Carter RE, Shah ND, Friedman PA, Noseworthy PA. Prospective validation of a deep learning electrocardiogram algorithm for the detection of left ventricular systolic dysfunction. J Cardiovasc Electrophysiol. 2019;30:668–674. doi: 10.1111/jce.13889 [DOI] [PubMed] [Google Scholar]
- 6. Cho J, Lee B, Kwon JM, Lee Y, Park H, Oh BH, Jeon KH, Park J, Kim KH. Artificial intelligence algorithm for screening heart failure with reduced ejection fraction using electrocardiography. ASAIO J. 2021;67:314–321. doi: 10.1097/MAT.0000000000001218 [DOI] [PubMed] [Google Scholar]
- 7. Dhingra R, Pencina MJ, Wang TJ, Nam BH, Benjamin EJ, Levy D, Larson MG, Kannel WB, D'Agostino RB Sr, Vasan RS. Electrocardiographic QRS duration and the risk of congestive heart failure: the Framingham heart study. Hypertension. 2006;47:861–867. doi: 10.1161/01.HYP.0000217141.20163.23 [DOI] [PubMed] [Google Scholar]
- 8. Okin PM, Roman MJ, Lee ET, Galloway JM, Best LG, Howard BV, Devereux RB. Usefulness of quantitative assessment of electrocardiographic ST depression for predicting new‐onset heart failure in American Indians (from the strong heart study). Am J Cardiol. 2007;100:94–98. doi: 10.1016/j.amjcard.2007.02.059 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Okin PM, Kamel H, Kjeldsen SE, Devereux RB. Electrocardiographic left atrial abnormalities and risk of incident stroke in hypertensive patients with electrocardiographic left ventricular hypertrophy. J Hypertens. 2016;34:1831–1837. doi: 10.1097/HJH.0000000000000989 [DOI] [PubMed] [Google Scholar]
- 10. Okin PM, Wachtell K, Gerdts E, Dahlof B, Devereux RB. Relationship of left ventricular systolic function to persistence or development of electrocardiographic left ventricular hypertrophy in hypertensive patients: implications for the development of new heart failure. J Hypertens. 2014;32:2472–2478. doi: 10.1097/HJH.0000000000000432 [DOI] [PubMed] [Google Scholar]
- 11. Lyon AR, Lopez‐Fernandez T, Couch LS, Asteggiano R, Aznar MC, Bergler‐Klein J, Boriani G, Cardinale D, Cordoba R, Cosyns B, et al. 2022 ESC guidelines on cardio‐oncology developed in collaboration with the European Hematology Association (EHA), the European Society for Therapeutic Radiology and Oncology (ESTRO) and the International Cardio‐Oncology Society (IC‐OS). Eur Heart J Cardiovasc Imaging. 2022;23:e333–e465. doi: 10.1093/ehjci/jeac106 [DOI] [PubMed] [Google Scholar]
- 12. Yang H, Marwick TH, Wang Y, Nolan M, Negishi K, Khan F, Okin PM. Association between electrocardiographic and echocardiographic markers of stage B heart failure and cardiovascular outcome. ESC Heart Fail. 2017;4:417–431. doi: 10.1002/ehf2.12151 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Okin PM, Devereux RB, Fabsitz RR, Lee ET, Galloway JM, Howard BV; Strong Heart S . Quantitative assessment of electrocardiographic strain predicts increased left ventricular mass: the Strong Heart Study. J Am Coll Cardiol. 2002;40:1395–1400. doi: 10.1016/s0735-1097(02)02171-x [DOI] [PubMed] [Google Scholar]
- 14. Okin PM, Devereux RB, Fabsitz RR, Lee ET, Galloway JM, Howard BV. Principal component analysis of the T wave and prediction of cardiovascular mortality in American Indians: the Strong Heart Study. Circulation. 2002;105:714–719. doi: 10.1161/hc0602.103585 [DOI] [PubMed] [Google Scholar]
- 15. Yu AF, Lin IH, Jorgensen J, Copeland‐Halperin R, Feldman S, Ibtida I, Assefa A, Johnson MN, Dang CT, Liu JE, et al. Nomogram for predicting risk of cancer therapy‐related cardiac dysfunction in patients with human epidermal growth factor receptor 2‐positive breast cancer. J Am Heart Assoc. 2023;12:e029465. doi: 10.1161/JAHA.123.029465 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Zamorano JL, Lancellotti P, Rodriguez Munoz D, Aboyans V, Asteggiano R, Galderisi M, Habib G, Lenihan DJ, Lip GYH, Lyon AR, et al. 2016 ESC position paper on cancer treatments and cardiovascular toxicity developed under the auspices of the ESC Committee for practice guidelines: the task force for cancer treatments and cardiovascular toxicity of the European Society of Cardiology (ESC). Eur Heart J. 2016;37:2768–2801. doi: 10.1093/eurheartj/ehw211 [DOI] [PubMed] [Google Scholar]
- 17. Pencina MJ, D'Agostino RB Sr, D'Agostino RB Jr, Vasan RS. Evaluating the added predictive ability of a new marker: from area under the ROC curve to reclassification and beyond. Stat Med. 2008;27:157–172. doi: 10.1002/sim.2929 [DOI] [PubMed] [Google Scholar]
- 18. Ezaz G, Long JB, Gross CP, Chen J. Risk prediction model for heart failure and cardiomyopathy after adjuvant trastuzumab therapy for breast cancer. J Am Heart Assoc. 2014;3:e000472. doi: 10.1161/JAHA.113.000472 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Bowles EJ, Wellman R, Feigelson HS, Onitilo AA, Freedman AN, Delate T, Allen LA, Nekhlyudov L, Goddard KA, Davis RL, et al. Risk of heart failure in breast cancer patients after anthracycline and trastuzumab treatment: a retrospective cohort study. J Natl Cancer Inst. 2012;104:1293–1305. doi: 10.1093/jnci/djs317 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Lyon AR, Dent S, Stanway S, Earl H, Brezden‐Masley C, Cohen‐Solal A, Tocchetti CG, Moslehi JJ, Groarke JD, Bergler‐Klein J, et al. Baseline cardiovascular risk assessment in cancer patients scheduled to receive cardiotoxic cancer therapies: a position statement and new risk assessment tools from the cardio‐oncology study Group of the Heart Failure Association of the European Society of Cardiology in collaboration with the international cardio‐oncology society. Eur J Heart Fail. 2020;22:1945–1960. doi: 10.1002/ejhf.1920 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Jorgensen PG, Jensen JS, Marott JL, Jensen GB, Appleyard M, Mogelvang R. Electrocardiographic changes improve risk prediction in asymptomatic persons age 65 years or above without cardiovascular disease. J Am Coll Cardiol. 2014;64:898–906. doi: 10.1016/j.jacc.2014.05.050 [DOI] [PubMed] [Google Scholar]
- 22. Denes P, Larson JC, Lloyd‐Jones DM, Prineas RJ, Greenland P. Major and minor ECG abnormalities in asymptomatic women and risk of cardiovascular events and mortality. JAMA. 2007;297:978–985. doi: 10.1001/jama.297.9.978 [DOI] [PubMed] [Google Scholar]
- 23. Kinoshita T, Yuzawa H, Natori K, Wada R, Yao S, Yano K, Akitsu K, Koike H, Shinohara M, Fujino T, et al. Early electrocardiographic indices for predicting chronic doxorubicin‐induced cardiotoxicity. J Cardiol. 2021;77:388–394. doi: 10.1016/j.jjcc.2020.10.007 [DOI] [PubMed] [Google Scholar]
- 24. JAN S, Hewawasam RP, Fernando M, Jayatilaka K. Electrocardiographic and biochemical analysis of anthracycline induced cardiotoxicity in breast cancer patients from southern Sri Lanka. BMC Cancer. 2023;23:210. doi: 10.1186/s12885-023-10673-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Gurdogan M, Ozkan U. A novel predictor of chemotherapeutic cardiotoxicity in patients with non‐Hodgkin lymphoma. Oncol Res Treat. 2019;42:375–381. doi: 10.1159/000500439 [DOI] [PubMed] [Google Scholar]
- 26. Pohl J, Mincu RI, Mrotzek SM, Wakili R, Mahabadi AA, Potthoff SK, Siveke JT, Keller U, Landmesser U, Rassaf T, et al. ECG scoring for the evaluation of therapy‐Naïve cancer patients to predict cardiotoxicity. Cancer. 2021;13:1197. doi: 10.3390/cancers13061197 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Kinoshita T, Onda N, Ohno R, Ikeda T, Sugizaki Y, Ohara H, Nakagami T, Yuzawa H, Shimada H, Shimizu K, et al. Activation recovery interval as an electrocardiographic repolarization index to detect doxorubicin‐induced cardiotoxicity. J Cardiol. 2023;82:473–480. doi: 10.1016/j.jjcc.2023.07.006 [DOI] [PubMed] [Google Scholar]
- 28. Jacobs JEJ, Greason G, Mangold KE, Wildiers H, Willems R, Janssens S, Noseworthy P, Lopez‐Jimenez F, Voigt JU, Friedman P, et al. Artificial intelligence electrocardiogram as a novel screening tool to detect a newly abnormal left ventricular ejection fraction after anthracycline‐based cancer therapy. Eur J Prev Cardiol. 2023;31:560–566. doi: 10.1093/eurjpc/zwad348 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Dang C, Guo H, Najita J, Yardley D, Marcom K, Albain K, Rugo H, Miller K, Ellis M, Shapira I, et al. Cardiac outcomes of patients receiving adjuvant weekly paclitaxel and trastuzumab for node‐negative, ERBB2‐positive breast cancer. JAMA Oncol. 2016;2:1–8. doi: 10.1001/jamaoncol.2015.3709 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Leong D, Cosman T, Alhussein M, Tyagi N, Karampatos S, Barron C, Wright D, Tandon V, Magloire P, Joseph P, et al. Safety of continuing trastuzumab despite mild cardiotoxicity, a phase I trial. JACC: Cardiooncol. 2019;1:1–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Boekhout AH, Gietema JA, Milojkovic Kerklaan B, van Werkhoven ED, Altena R, Honkoop A, Los M, Smit WM, Nieboer P, Smorenburg CH, et al. Angiotensin II‐receptor inhibition with candesartan to prevent trastuzumab‐related cardiotoxic effects in patients with early breast cancer: a randomized clinical trial. JAMA Oncol. 2016;2:1030–1037. doi: 10.1001/jamaoncol.2016.1726 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Table S1