Breast cancer (BC) among women represents a public health challenge due to its rising incidence and its life-threatening consequences (1). Prescribed to reduce local recurrence and BC‐related mortality, radiation therapy (RT) has widely demonstrated effectiveness in the treatment of BC (2). However, radiation-induced adverse effects in healthy tissues could occur. Cardiotoxicity resulting from incidental irradiation of the heart in BC patients is now better documented (3). Indeed, BC RT leads to an increased risk of long-term major adverse cardiovascular events (MACEs), mainly coronary heart diseases, as well as excess cardiovascular (CV) mortality rates (3, 4). Up to several decades, the relative risk of clinically significant cardiac events ranged from 1.2 to 3.6 after RT (5). Darby et al. (2013) showed an incidence of acute coronary events increased by 7.4% per Gray (Gy) of mean heart dose already within 5 years following RT, later confirmed by van den Bogaard et al. (2017) who found an incidence of 16.5% per Gy (6, 7) in the first 9 years. Other authors suggested a 0.04 (95% CI: 0.02–0.06) excess relative risk per Gy received at the whole heart (8). However, the asymptomatic phase between acute heart damage occurring early after RT and the longer-term onset of MACEs leads to an underrecognized CV risk during the clinical management of BC patients immediately following RT (9).

Therefore, early screening for subclinical CV changes following RT could prove beneficial for asymptomatic patients who could nevertheless have subclinical left ventricle (LV) dysfunction. According to the American Society of Echocardiography and the European Association of Cardiovascular Imaging, oncological cardiotoxicity is diagnosed when the left ventricular ejection fraction (LVEF) is reduced by ≥10% points to below 53% after RT (10). However, the myocardial deformation [i.e., global longitudinal strain (GLS), measured by two-dimensional (2D) speckle-tracking echocardiography] appeared to be an earlier marker of subclinical LV dysfunction. Specifically, strain imaging characterizes cardiac wall deformation considering speckles. Therefore, a reduction in LVEF reflects late and advanced myocardial injury in relation to substantial cardiac damage (11). Therefore, the measurement of GLS appears to be more sensitive and relevant for detecting early LV dysfunction before the onset of LVEF deterioration, and in identifying a population at greater risk of longer-term CV morbidity and mortality (12, 13).

The ability of GLS to detect cardiotoxicity early has been little investigated among BC patients treated with RT. Some studies have shown that a statistically significant reduction of the GLS can be detected in BC women from a few weeks to 12 months following RT (14–17). However, it remains to be determined whether the observed reduction can be considered clinically relevant. Negishi et al. suggested that a reduction of GLS >15% compared with baseline appears to be clinically meaningful to highlight post‐RT cardiotoxicity, but this GLS cutoff limit was scarcely applied in onco-cardiology research (18). Moreover, few studies investigated the dose‐dependent relationship between RT and changes in GLS. In 2019, Walker et al. investigated the clinical relevance of the reduction of GLS in 79 BC patients included in the BACCARAT study (14) by defining a subclinical LV dysfunction as a relative reduction of GLS >10%. A dose–response relationship was observed, and the risk of subclinical LV dysfunction was increased by 37% per 1 Gy of mean heart dose. Nevertheless, the association was no longer statistically significant after adjustment for age, body mass index (BMI), hypertension, hypercholesterolemia, and endocrine therapy, and the study suffered from its small size and statistical power (19).

Therefore, within the frame of the European MEDIRAD project, the multi-center EARLY-HEART cohort study was designed to investigate early cardiac changes arising after BC RT in the largest population ever studied, using three approaches based on echocardiography, cardiac MRI and heart CT, and computed tomography coronary angiography (20). The present manuscript originally focuses on the specific purpose of evaluating the impact of RT (using individual patient dosimetry) on subclinical LV function changes (using speckle-tracking echocardiography) occurring in the first 6 months after BC RT. This study will open many research possibilities to find markers of early subclinical LV dysfunction potentially predicting long-term MACEs.

Designed to early identify cardiotoxicity in BC women treated with RT, the EARLY-HEART study suggested a strong relationship between cardiac absorbed dose and the occurrence of subclinical LV dysfunction at 6 months following RT based on >15% reduction in GLS estimated by echocardiography.

The mean value of GLS in the whole population did not significantly decrease from baseline to 6 months post-RT. Other authors previously showed significant GLS changes after BC RT, with a mean reduction of the GLS following RT of 5% in Erven et al., 6% in Walker et al., and 7.9% in Trivedi et al. (follow-up from 3 to 12 months) (14, 16, 17). In these studies, significant changes were highlighted in left-sided BC only.

GLS damage was further studied as a relative change in each individual and from a clinical perspective (10). A binary clinical endpoint of subclinical LV dysfunction was set by categorizing BC patients with or without reduction of GLS >15% as previously suggested in order to be largely beyond the possible errors related to the accuracy and reproducibility of measurements (18). Among the 186 women, 22 presented a subclinical LV dysfunction (11.8%). The prevalence of subclinical LV dysfunction was slightly higher when applying the cutoff of 10% (19.3%). Although not negligible, these two rates were lower than those obtained in other studies (applying the 10% cutoff) where they ranged from 27.5% to 46.8% (14, 16, 17). The high proportion of right-sided BC, cardiac dose differences, and the chemotherapy-naïve status of BC women in our study may explain this lower rate. Fourati et al., using similar study criteria, also obtained a lower prevalence rate of cardiac dysfunction (6.8%) (30) (i.e., 1.76 ± 1.12 Gy versus in our study versus 2.8 Gy of mean dose in the study of Fourati et al.; 22% of left-sided BC versus 42% of right-sided BC).

The EARLY-HEART study robustly showed a relationship between dose-volume parameters and an increased risk of subclinical LV dysfunction (adjusted ORs ranging from 1.13 [1.05–1.23] (V₅) to 1.74 [1.20–2.61] (D_mean) for the WH structure and from 1.10 [1.05–1.17] (V₅) to 1.46 [1.17–1.87] (D_mean) for the LV structure). The magnitude of the association was consistent with previous studies (or even stronger): OR = 1.37 [1.01–1.86] in Walker et al. and OR = 1.04 [1.01–1.06] in Fourati et al. (both analyzing relationship between D_mean and a reduction in GLS >10%) (30, 31). Furthermore, three parameters were able to properly distinguish BC women at risk or not of a reduction of GLS>15%, 6 months after RT: D_mean, D_max, and V_5. The lowest AUROC was 0.765 for D_max (WH) and the highest AUROC was 0.815 for V₅ (LV). Then, V₅, a dose-volume parameter, seemed highly relevant, as previously shown by other studies (7, 14). Indeed, van den Bogaard et al. showed that V₅ (LV) was the best predictor of acute coronary events (HR = 1.016 [1.002–1.030], p-value = 0.016). However, the mean heart dose remains currently the most widely used predictor of cardiotoxicity (31). Furthermore, intrinsic to our sample, the threshold of 3 Gy for D_mean, previously identified by Erven et al. (17), was also highlighted. Indeed, a D_mean of 2.74 Gy (WH) or a V₅ >5.2% (WH) should not be exceeded to prevent the CV risk. The threshold of 3 Gy for D_mean, previously identified by Erven et al. (17), was also highlighted. Some impactful studies (e.g., Darby et al., showing a dose–response relationship between acute coronary events and mean heart dose) incited RT protocols to evolve to limit the risk of MACE and cardiac doses (e.g., breath‐hold, VMAT, and hypofractionation can reduce cardiotoxicity) (6). Proton therapy may also be applied for patients still at increased risk. However, our study combining different techniques of RT showed that some patients remained in dose ranges that should be considered with caution (e.g., D_mean of 3 Gy to LV). Vigilance must be brought to this specific point, especially in randomized controlled trials where a systemic assessment of radiation‐induced cardiotoxicity should be investigated as a clinical endpoint.

Our study was the first to demonstrate, with sufficient statistical power, a dose-dependent relationship between early cardiotoxicity defined using the stringent and recommended criterion of a reduction of GLS >15% and a wide range of doses absorbed (inclusion of both right- and left-sided BC patients). Its prospective design allowed us to include only BC women without baseline overt CV diseases and chemotherapy and to control CV risk factors, making the results on the observation of an early subclinical LV dysfunction induced by RT more robust.

However, our study had some limitations. The interpretation of the present results must be made with knowledge of these. First, our sample of BC women was limited by strict inclusion criteria. Further studies should include a larger representation of BC patients treated with RT only (e.g., risk in younger and older BC patients, risk in patients with or without previous CV diseases, and risk according to regional specificities). Likewise, the lower proportion of left-sided BC patients in our EARLY-HEART population compared to other studies could impact the observed change in mean GLS, which was not significant. Moreover, inter-observer (i.e., different cardiologists) and inter-operator (i.e., different vendors) imprecisions cannot be excluded to explain the absence of a statistically significant decrease in mean GLS in our study even if the good reproducibility of the strain measure using echocardiography was established (32). Indeed, inter-operator relative mean errors ranged from 5.4% to 11.0% when inter-observer relative mean errors varied from 1.9% to 11.3%. These values of errors remained lower than that observed for LVEF and other conventional echocardiography parameters (32). Our one-way sensitivity analysis omitting one center at a time reduced this potential bias and showed the robustness of our findings. Finally, although the current results were adjusted for baseline CV risk factors, it cannot be ruled out that other confounding factors could impact the observed association (e.g., parental history of CV diseases, sedentary habits, and nutritional habits).

In the future, it remains to be investigated whether the occurrence of subclinical LV dysfunction observed at RT+6 months is declining, maintaining, or improving in the longer term. The specific interest in echocardiography data imaging from the EARLY‐HEART study will be further studied to determine which specific segments of the longitudinal strain (i.e., basal, mid, or apical) could be the most affected by dose-volume parameters. Indeed, Tuohinen et al. recently showed that the dose absorbed at the level of the apical region of the anterior wall of the LV was linked to a significantly higher deterioration of the GLS than in other locations (15). Furthermore, the same research team recently showed that diastolic strain rate was an earlier predictor of dysfunction than systolic LV strain rate (23), which could be of interest knowing that diastolic function is involved in diffuse fibrosis following RT.

The present analysis of BC women from the EARLY-HEART study showed that the cardiac doses absorbed during RT were strongly associated with the occurrence of a subclinical LV dysfunction at 6 months after RT. Therefore, primary and secondary CV health prevention could be beneficial at this early asymptomatic phase to reduce long-term CV complications. These findings already suggest the potential relevance of an early screening of BC patients treated with RT to eventually early implement cardioprotective actions during RT by limiting the dose absorbed by the heart as much as possible.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The study protocol and related amendments received approval from the competent ethics authority of each center involved (France: Comité de Protection des Personnes Sud-Ouest IV, ID: CPP2015/66/2015-A00990-69–R1, and Agence Nationale de Sécurité des Médicaments, ID: 150873B-12; the Netherlands: Medisch Ethische Toetsingscommissie van het Universitair Medisch Centrum Groningen [METc UMCG], ID: METc 2017/379, NL62360.042.17; Germany: Ethikkomission der Technischen Universität München, ID: 235/17 S; Spain: Comitè d’Etica d’Investigatio CEAi GIRONA, ID: EARLY HEART v1.1 05/07/2017 i FIP v1.3; Portugal: Comissao de Ética do Centro Hospitalar Lisboa Norte e do Centro Académico de Medicina de Lisboa [CHLN e CAML], ID: 257/2017). The patients/participants provided their written informed consent to participate in this study.

Conceptualization: SJ, AC, HL, AE, SS, KB, GF, and EC. Methodology: SJ, HL, GF, and EC. Software: ML. Validation: ML and SJ. Formal analysis, ML and SJ. Investigation: SJ, AC, ML, DS, FG, MF, SS, SC, KB, EM, and UG. Resources: GF and EC. Data curation: ML, EM, UG, SJ, and DS. Writing—original draft preparation: ML and SJ. Editing: all authors. Supervision: EC, GF, and HL. All authors contributed to the article and approved the submitted version.

The European Community’s Horizon 2020 Programme supported the EARLY-HEART study conducted in the frame of the MEDIRAD - Implications of Medical Low Dose Radiation Exposure - project spanning from 2017 to 2021 granted by the Euratom Research and Training Programme 2014-2014 under agreement No. 755523.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.