Global longitudinal strain at 3 months after therapy can predict late cardiotoxicity in breast cancer

Abstract Background Cancer therapy‐related cardiovascular toxicity (CTR‐CVT) is a major contributor to poor prognosis in breast cancer (BC) patients undergoing chemotherapy. Left ventricular global longitudinal strain (LV GLS) has predictive value for CTR‐CVT, while few researchers take into account late‐onset CTR‐CVT. This study sought to provide a guide for the prediction of late‐onset CTR‐CVT in primary BC over the 2 years follow‐up via strain and contrast‐enhanced echocardiography. Methods Anthracycline and anthracycline + targeted medication groups were created from 111 patients with stage I–III primary BC who were prospectively included. The left ventricular diastolic function, LV global long‐axis strain (GLS); left ventricular ejection fraction by contrast‐enhanced echocardiography (c‐LVEF), and electrocardiograms were collected at baseline, 3, 6, 12, and 24 months after the start of cancer treatment. The high‐sensitivity troponin‐T and NT‐pro BNP at baseline and 3 months after chemotherapy were measured. Results (1) LV GLS decreased in BC patients over time. (2) After 12 months' follow‐up, the LV GLS in the anthracycline+ targeted group was lower than in the anthracycline group. After 24 months' follow‐up, the GLS and c‐LVEF in the anthracycline + targeted group declined while the E/e’ increased. (3) Decreased LVEF (56%) and arrhythmia (38%) are the common causes of CTR‐CVT. Lower LVEF was a major factor in late‐onset CTR‐CVT. (4) Combination of LV GLS and c‐LVEF at 3 months were used as predictors for CTR‐CVT and exhibited a higher AUC than either one alone (AUC = 0.929, 95% CI: 0.863–0.970). LV GLS at 3 months can predict the late‐onset CTR‐CVT (AUC = 0.745, p < 0.001), and the cut‐off is 20.32%. Conclusions As time went on, the systolic and diastolic dysfunction of BC patients get worsened. The combination of LV GLS and c‐LVEF is better in the prediction of CTR‐CVT. Only the LV GLS at 3 months can predict the late‐onset CTR‐CVT.


| INTRODUCTION
Breast cancer (BC) accounts for 30% of newly diagnosed malignant tumors in women and causes 15% of cancer deaths in women. 1 In BC treatment today, innovations like chemoradiotherapy, immunotherapy, and gene-targeted therapy have been shown to extend overall survival. Nevertheless, Cancer therapy-related cardiovascular toxicity definitions (CTR-CVT) 2,3 emerge as a major contributor to poor prognosis in those undergoing chemotherapy. Furthermore, long-term or "late" adverse effects of cancer treatments are increasingly important. Therefore, early detection of CTR-CVT and late-onset CTR-CVT is necessary for early intervention.
Echocardiography is of helps to detect early impaired cardiorespiratory fitness. LVEF was suggested by several guidelines 4 to detect CTR-CVT; however, the endocardium is not well-displayed for BC patients who underwent left breast surgery, dilator implantation and left thoracic radiotherapy. Contrast-enhanced echocardiography as the use of ultrasound contrast agents based on conventional transthoracic echocardiography (TTE), can clearly depict the endocardial border and reduce interobserver variation in measuring LVEF. 5 Suwatanaviroj et al. 6 found that contrast-enhanced echocardiography had a comparatively greater sensitivity for detecting changes in volume when the observed LVEF varied by no <3.6%.
Left ventricular global longitudinal strain (LV GLS) has predictive value for CTR-CVT and HF and can detect early subclinical myocardial impairment related to cancer treatment, 4,7,8 and can be used in the regular monitoring during and after treatment. 9 According to a study of BC patients with a 6-year follow-up, 13.5% of participants had abnormal GLS, ranging from 11% to 15% among treatment groups. However, there are few studies examining the connection between early GLS alterations and late cardiotoxicity.
The objectives of this prospective cohort study were to evaluate the changes in the LV function of primary BC patients via GLS and contrast-enhanced echocardiography, sought to provide a guide for the prediction of late-onset CTR-CVT (≥1 year) in primary BC over the 2 years follow-up via strain and contrast-enhanced echocardiography.

| Data source and study population
This study is a single-center prospective cohort study from a registered study with the identifier ChiCTR1900022108 (http://www.chictr.org.cn). Consecutive patients with stage I-III primary breast invasive ductal carcinoma receiving anthracycline chemotherapy or followed by sequential targeted therapy between May 2019 and December 2019 were enrolled. The patients were diagnosed with BC by Version 3. 2018, National Comprehensive Cancer Network (NCCN) Clinical Practice Guidelines. 10 Inclusion criteria: (1) ≥18 years old; (2) LVEF ≥53% measured by biplane Simpson's rule by TTE before chemotherapy. 11 Exclusion criteria include: ventricular bigeminy, atrial flutter, atrial fibrillation, and other severe arrhythmias; severe valvular disease, including moderate-severe aortic stenosis, moderate-severe mitral stenosis, and moderate-severe mitral regurgitation; myocardial infarction or symptomatic HF; history of cardiac surgery; use of cardiotoxic chemotherapy drugs in the past; and previous left thoracic radiotherapy. According to treatment procedures, the patients were assigned into two groups: anthracyclines group and anthracyclines + targeted therapy group. This study was approved by the Biomedical Ethics Committee of West China Hospital, Sichuan University (Approval Number: 20180517), and all patients provided informed consent.

| Clinical evaluation
The following information was collected: age, height, weight, systolic and diastolic blood pressure, history of chronic illness (hypertension and diabetes), tumor node metastasis (TNM) stage and endocrine therapy, and thoracic radiation. Holter was carried out if the electrocardiogram (ECG) revealed any new arrhythmias. At baseline (T0), follow-up at 3 months (T1), 6 months (T2), 12 months (T3), and 24 months (T4) post-chemotherapy, TTE and ECG were conducted. At T0 and T1, the myocardial injury biomarkers including NT pro-BNP, troponin-T, and creatine kinase isoenzyme (CK-MB) were collected. Follow-up terminated by January 31, 2022. Figure 1 prediction of CTR-CVT. Only the LV GLS at 3 months can predict the late-onset CTR-CVT.

K E Y W O R D S
breast cancer, cardio-oncology, echocardiography, heart failure, late-onset cardiotoxicity displayed the follow-up chart. The control group was comprised of 77 individuals who had completed normal ECG, TTE, and laboratory tests during the same period. The controls were used to compare the cardiac structure and function of patients before anthracycline-based therapy, in order to exclude the possibility that surgery and the tumor itself had affected their cardiac function. In addition, we matched the prevalence of hypertension and diabetes between the case group and the control group since these diseases can also impact the LV GLS.

| Echocardiography
The ultrasound system (Philip Epiq7C and Philip CVx) with a 1-5 MHz phased-array transducer (M5S) was used. Standard echocardiography, including twodimensional (2D), M-mode, and Doppler echocardiography, was performed according to the guidelines of the American Society of Echocardiography. 12 LVEF was measured based on the modified biplane Simpson's rule. Mitral inflow velocity was measured at early (defined as E wave) and late (defined as A wave) diastole. The velocity of the mitral annulus at early diastolic (defined as e') and late diastolic (defined as a') was recorded by pulsed tissue Doppler imaging. The E/e' ratio was used as an index of LV diastolic function. All images were captured by a senior operator.
Apical four-, two-, and three-chamber images of at least five cardiac cycles were captured and used to calculate the LV GLS. The strain analytical software (AutoStrain, Philips) automatically detects, segments, and monitors LV myocardium after selecting the region of interest for examination, with strain curves being computed simultaneously. To provide a good portrayal of ventricular motion, every track was examined. Poorly tracked segments were manually corrected until a higher tracking quality was attained. When more than two portions of a video could not be tracked satisfactorily, the video was deemed to have "failed". 13 All cases were analyzed using the Auto Strain software (Philips Healthcare, Figure 2E-H). The analysts were blinded to the measurements performed and the clinical data. Bland-Altman analysis showed good intraand inter-observer agreement with a small nonsignificant bias for GLS ( Figure 3C,D).

| Contrast-enhanced echocardiography
All patients received an intravenous injection of Sono vue (Bracco SpA). Apical four-, three-, and two-chamber images of at least five cardiac cycles were digitally recorded and archived for analysis. LVEF (designated as c-LVEF) was estimated from the contrast recordings in the Philips system using the biplane Simpson's rule (Figure 2A-D). Bland-Altman analysis showed good intra-and interobserver agreement with a small nonsignificant bias for c-LVEF ( Figure 3A,B).

| Cancer therapy-related cardiotoxicity
CTR-CVT includes cancer treatment-related cardiac dysfunction (CTRCD), which is defined by the European Society of Cardiology guidelines as having an LVEF<50% or a decrease of >10% post-chemotherapy, or a new GLS decrease of 15% relative to baseline, immune checkpoint inhibitor-induced myocarditis, cancer treatment-related vascular toxicity, cancer treatment-related arterial hypertension, and cancer treatment-related arrhythmia. 14 Early-onset CTR-CVT includes acute cardiotoxicity which can occur during or soon after initiation of therapy and occurs within the first year of completion of chemotherapy, whereas Late-onset CTR-CVT occurs beyond the first year and up to 2 years in our study. 15

| Statistical analysis
For normally distributed data, continuous variables were displayed as mean standard deviation (SD) or median (25th percentile and 75th percentile) for non-normally distributed data. Frequencies were used to present categorical variables. For comparing two and three independent groups of normally distributed data, the t-test, and one-way analysis of variance (ANOVA) were used, respectively. The nonnormal distribution was tested using the Wilcoxon signedrank test. Binary variables were compared using Fisher's exact test and the Chi-square test. To evaluate the risk factors of CTR-CVT in more detail, multivariable logistic regression models were applied. The multivariate regression analysis model included contained univariate regression analysis of positive factors. After accounting for echocardiographic indices, a receiver operating characteristic (ROC) curve analysis was carried out. Bland-Altman plots were used to display the difference in intra-observer and interobserver of c-LVEF and GLS, along with the average bias and 95% limits of agreement. p < 0.05 were taken into consideration. IBM SPSS for Windows, version 19.0 (IBM Corp) was used for all analyses.
Age, body mass index (BMI), chronic diseases like hypertension, and diabetes, tumor node metastasis (TNM) stage, and other treatment indicators for BC did not significantly alter between baseline and follow-up (Table 1).

| Cardiac structure and function
At baseline, there was no discernible difference between controls and patients in terms of structure and function (Table 1). No difference in cardiac chamber's size at follow-up either. LV GLS decreased at T1, T2, T3, and T4 compared to baseline (p < 0.05). When compared to LV GLS at 3 months (T1) follow-up, LV GLS at 1 year (T3) and 2 years (T4) decreased. E/e' increased at 3 months (T1), 1 year (T3), and 2 years (T4) compared to baseline (T0). Additionally, when compared to baseline (T0) and 3 months (T1), serum myocardial injury biomarkers, such as CK-MB, increased at 6 months (T2) (p < 0.05) ( Table 2). These findings implied that myocardial injury may occur after a 6-month pharmaceutical regimen, and subclinical systolic dysfunction may occur at 3 months.
After a year of follow-up, GLS in the anthracycline + targeted group decreased in comparison to the anthracycline group (p < 0.05), but the c-LVEF showed no difference ( Table 3). The GLS and c-LVEF both decreased after 2 years of follow-up, while the E/e' rose (p < 0.05) in the anthracycline + targeted group compared to the anthracycline group.

| Cancer therapy-related cardiotoxicity in BC patients
The occurrence of CTR-CVT was regarded as the endpoint event. A total of 34 patients had CTR-CVT during the follow-up, among them, 20 with CTRCD (5 patients with LVEF<50%, 7 patients with decreased LVEF by >10%, 8 patients with declined GLS by >15%), 13 arrhythmias, and 1 with newly developed hypertension. Twelve patients had late-onset CTR-CVT, including nine with decreased LVEF, two with arrhythmia, and one with newly developed hypertension. Figure 4 displayed the CTR-CVT's proportion.

| Factors of prediction for CTR-CVT
Plots illustrating the cumulative incidence of CTR-CVT for the three groups-without CTR-CVT, with CTR-CVT, and with late-onset CTR-CVT-were made (Table 4). No difference in Age, systolic, and diastolic blood pressure, the proportion of patients receiving radiation, left thoracic radiotherapy, cumulative dose of anthracyclines (400/300 mg/m), and endocrine therapy, TNM stages,  hypertension, or diabetes among all three groups, but targeted therapy was significantly different. While there was no difference in the 2-year incidence of CTR-CVT between the anthracycline and anthracycline + targeted groups, Kaplan-Meier analysis ( Figure 5) revealed that the late-onset CTR-CVT was more common in the anthracycline + targeted group than in the anthracycline group (p = 0.034). Furthermore, we found that left thoracic radiotherapy can affect the early-onset CTR-CVT (p = 0.009).
Out of a total of 22 patients with early-onset CTR-CVT, 12 patients had received left chest radiotherapy. Univariate logistic regression analysis was used to examine potential risk variables for CTR-CVT. The multivariate regression analysis model took the data with discrepancies into account ( Table 5). The model demonstrated that in BC patients, 3-month GLS, and c-LVEF were independent risk factors for CTR-CVT. Three separate risk factors for late-onset CTR-CVT included the 3-month GLS, E/e', and the proportion of the targeted medication.

T A B L E 3
The left ventricular function in the anthracycline group and anthracycline + targeted therapy group at follow-up.

F I G U R E 5 (A)
The probability of CTR-CVT during the 2 years' followup between the anthracycline and anthracycline + targeted therapy; (B) The probability of late-onset CTR-CVT between the anthracycline and anthracycline + targeted therapy.

T A B L E 5
The risk factors for CTR-CVT in patients with primary breast cancer.

| DISCUSSION
Our major findings can be summarized as follows ( The 2-year follow-up of this study showed that the incidence of CTR-CVT including early-onset and late-onset was 30.6% in primary BC patients, which was commensurate with the incidence of cardiotoxicity events (3%-35%) described in earlier studies. 2,16,17 The difference in the definition of the incidence of cardiotoxicity events across studies is also not negligible. As stated in the literature, early-stage CTR-CVT was characterized by a high rate of cardiac arrhythmias, but late-onset CTR-CVT was characterized by a reduced LVEF. 18 Previous studies also showed among patients who developed late severe cardiovascular events, 66 (48.5%) had HF, 40 (29.4%) cardiomyopathy, 19 (14.0%) myocardial injury, and 11 (8.1%) cardiovascular death. 19 So it is still clinically necessary to identify individuals who are at a high risk of developing CTR-CVT. If CTR-CVT is discovered early, cancer treatments may be altered to reduce additional myocardial injury. In this context, imaging methods and biomarkers are required to identify at-risk patients prior to a CTR-CVT. Contrast-enhanced echocardiography and GLS was adopted to assess left ventricular function in this study. For LVEF measures, a stronger association between contrastenhanced echocardiography and CMR was discovered. 20 In the BC patients in our study, c-LVEF decreased after a two-year follow-up, and c-LVEF at 3 months (T1) can predict development of CTR-CVT. Over the past 10 years, GLS has become a reliable indicator of subclinical LV dysfunction and is more sensitive than LVEF at detecting early changes in LV function. 21 The published study with a median follow-up of 24.7 months has shown that left ventricular ejection fraction, strain, and strain remain decreased at follow-up, which was consistent with our study. 22 This might be due to the persistent cardiac impairment of the chemotherapy drugs. In our study, GLS allowed an early detection of subclinical LV dysfunction and had predictive value for CTR-CVT and late-onset CTR-CVT. Few studies also investigated the clinical value of 3-month GLS, the cutoff values of LV GLS for predicting CTR-CVT for 3 months range from −18.0% to −16.6%. 23,24 It was different with our 20.96% in CTR-CVT and 20.32% in late-onset CTR-CVT, which might be related to different machines or our patients taking dexrazoxane. Overall, we found that a combination of GLS and c-LVEF predicted CTR-CVT best. To our knowledge, this is the first study to show the combination of GLS and c-LVEF in the prediction of BC patients' outcomes.
In the 3-month follow-up, the E/e' increased concurrently, implying diastolic dysfunction. Additionally, 3-months E/e' predicted BC patients' late-onset CTR-CVT. Unfortunately, because of the limited sample size, we did not get a positive result by ROC. Anthracyclines can affect diastolic function. 16,25 Abnormal diastolic function was associated with an increased risk of cancer-treated reduced cardiac dysfunction. 16 This was in line with our study. However further investigation is required in a largesample study to confirm the prediction value of E/e'.
We assessed serum cardiac markers including myoglobin, troponin T, and CK-MB at baseline, 3 months, and 6 months after anthracycline-based therapy. Previous research has shown that CK-MB levels also increase during chemotherapy but with a different time course. In the anthracycline + targeted group, CK-MB levels peaked at 3 months and then stabilized at 6 months and 1 year, while in HER2−patients, CK-MB levels behaved similarly to the other biomarkers, with a slight increase at 3 months, a peak at 6 months, and a stabilization at 1 year after treatment. 26 Therefore, we analyzed both troponin T and CK-MB levels to assess the extent of myocardial damage. Based on previous studies, NT-proBNP has been shown to have predictive value in the early diagnosis of cardiotoxicity. However, there are certain biological variations and factors such as age, body mass index, and renal function that need to be considered when interpreting the biomarker results. It is also important to note that earlier studies were limited by small sample sizes and the validity of cardiac assessments. Ponde et al. 27 conducted an analysis using data from 280 patients with early breast cancer treated in the Neo-ALTTO trial to assess the predictive power of NT-proBNP. They found that elevations in biomarker levels were rare and failed to establish an association between these markers and cardiac events. Another study showed that when patients were dichotomized based on meeting CMR CTRCD criteria, there were no increased diagnostic values via hsTnI and BNP. 28 It is important to consider that NT-proBNP is an indicator of cardiac strain or volume overload and may not detect early cell injury. In our study, we found no difference in the size of the cardiac chamber and the cardiac systolic function among T0-T2, which may explain the lack of change in NT-pro BNP until 6 months.
The cumulative dose of doxorubicin was not significantly linked with cardiotoxic events in our investigation, which is consistent with Slamon et al.'s findings [12], despite earlier studies reporting that anthracycline-induced myocardial injury was dose-dependent. 29 This may be because the risk of developing cardiotoxicity is still low at doses below the higher recommended range of 900-1000 mg/m 2 . 30 Anthracycline + targeted therapy can lead to changes in cardiac function in primary BC patients. 31 In this study, anthracycline + targeted treatment was associated with higher cardiotoxicity events than anthracycline chemotherapy alone. 20%-45% of patients and 3% of patients, respectively, have heart failure or cardiotoxicity as a side effect of their targeted therapy. 32 There is a distinct element of long-term, irreversible damage to the myocardium, 33 and cardiotoxicity is closely related to treatment F I G U R E 8 Central illustration LV-GLS at 3 months can predict late-onset CTR-CVT; LV-GLS and c-LVEF in combination: a more reliable predictor of CTR-CVT. stoppage, underscoring the seriousness of the problem, even though some of this damage is thought to be reversible following trastuzumab cessation. 22 Previous studies indicated that cardiac radiation doses were associated with subclinical LV dysfunction evaluated by LV GLS. Some studies also showed through 40 patients treated for left breast cancer, that for a mean heart dose of 1.3-2.5Gy, there was no correlation between GLS and mean heart dose. 34 In our study, there no difference in radiation history among the no CTR-CVT, CTR-CVT, and lateonset CTR-CVT group. However, there was a significant difference in the history of left chest radiation exposure between the early-onset CTR-CVT group and the nonearly-onset CTR-CVT group. Previous studies have also indicated that in patients with left-sided breast cancer, GLS was reduced from −18.3% ± 3.1% to −17.2% ± 3.3% after undergoing radiation therapy for 6 months. 35 Thus, compared to late-onset CTR-CVT, left chest radiation may more easily induce CTR-CVT within the first year after therapy.
There are several limitations of our study. The early stages of the COVID-19 outbreak in 2020 (January to June) had an impact on follow-up rates. As a result, only a small number of BC patients returned to the hospital for follow-up at T2, which is a major limitation of this study. A larger sample size would have allowed us to draw more conclusive results, which could present a statistical limit in multivariate analysis and ROC curve.

| CONCLUSIONS
As time went on, the systolic and diastolic dysfunction of BC patients get worsened. Reduced LVEF was the predominant manifestation of both CTR-CVT and late-onset CTR-CVT, which was then followed by an arrhythmia. The combination of LV GLS and c-LVEF is better in the prediction of CTR-CVT. Only the LV GLS at 3 months can predict the late-onset CTR-CVT.