Dose to cardiac substructures and cardiovascular events in esophageal cancer patients treated with definitive radiotherapy

Abstract

Introduction

While there is a growing amount of data on the cardiac toxicity of radiotherapy (RT) in relation to its impact on cardiac sub-structures (CSS), there are only few studies addressing this issue in patients followed for esophageal cancer (ESOC). We aimed to evaluate the association between independent parameters of dose received by CSS and major cardiac events (MACEs) in this population.

Materials and methods

We retrospectively analyzed 122 patients treated with exclusive RT or chemo-RT for ESOC. Heart and CSS i.e. right atrium, left atrium (LA), right ventricle, left ventricle and myocardium, have been automatically segmented, and dose volume histogram were extracted. Cardiac events were collected focusing on the occurrence of MACEs of grade 3 or higher (G3+) and grade 4 or higher (G4+) according to the CTCAE v5.0.

Results

With a median follow-up of 21.9 months and in a population of high to very high cardiovascular risk (95.5%), 21 (17.2%) and 9 (7.4%) patients had G3 + and G4 + MACEs with a respective median time to event of 13.05 and 9.8 months. After multivariate analysis and among all heart and CSS-based dosimetric features, only the volume of LA receiving 15 Gy or more (V15LA) remained significantly associated with the G3 + and G4 + MACEs. The use of volumetric modulated arctherapy significantly reduced V15LA compared with 3D conformal RT.

Conclusion

In a cohort of ESOC patients treated with exclusive RT, incidence of MACEs was associated with V15LA, underlining the importance of CSS. These high cardiovascular (CV) risk patients should benefit from standard CV assessment and strict control of their risk factors.

Highlights

While there is a growing amount of data on the cardiac toxicity of radiotherapy (RT) in relation to its impact on cardiac sub-structures (CSS), there are only few studies addressing this issue in patients followed for esophageal cancer (ESOC).

In a cohort of 122 patients, with a median follow-up of 21.9 months, 21 (17.2%) and 9 (7.4%) patients had G3 + and G4 + MACEs with a respective median time to event of 13.05 and 9.8 months.

After multivariate analysis, only the volume of Left Atrium receiving 15 Gy or more (V15LA) remained significantly associated with the G3 + and G4 + MACEs.

The use of volumetric modulated arctherapy significantly reduced V15LA compared with 3D conformal RT.

Heart sub-structures were generated using a free automatic segmentation tool that can be easily distributed around radiation therapy departments.

Introduction

Exclusive chemo-radiotherapy (CT-RT) [1, 2] is the reference treatment for non-operable localized or locally advanced esophageal cancer (ESOC), while exclusive radiotherapy (RT) may be proposed in cases where concomitant chemotherapy (CT) is contraindicated (expert agreement) [3].

The cardiac toxicity of RT has been well documented in the long-term follow-up of patients treated for breast cancer [4,5,6], Hodgkin’s lymphomas [7, 8] or childhood cancer survivors [9].

However, more recently, data have also shown an impact of the dose to the heart on the occurrence of cardiac events but also on overall survival (OS) in non-small cell lung cancer (NSCLC) [10, 11] with, for example, a decrease in OS directly correlated with V5 (Vx: volume that received xGray (Gy) or more) and V30 of the whole heart in the Radiation Thoracic Oncologic Group (RTOG) 0617 dose escalation trial [12], but also in ESOC despite diseases with an often poorer prognosis [13, 14]. Moreover, these patients are often at high risk of cardiovascular (CV) events because of shared risk factors with their disease, such as smoking in the case of squamous cell carcinoma, or obesity in the case of adenocarcinoma. Therefore, particular attention must be paid to cardio protection in this population, whose long-term outcome is improving as a result of diagnostic and therapeutic progress, with 5-year net survival in France doubling from 9 to 18% between 1990 and 2018, according to data from the french national cancer institute (INCA).

In addition, numerous studies have highlighted the importance of considering cardiac-substructures (CSS) when planning RT, since the heart is a heterogeneous structure in terms of tissue, with different pathological response mechanisms to irradiation [15, 16], again in thoracic cancers with long-term follow-up [17, 18], but also in NSCLC. Atkins et al. found a V15 > 10% of the left anterior descending coronary artery to be an independent risk factor for cardiac events or mortality in patients with no history of coronary disease [19], and McWilliam et al. found an association between dose to the base of the heart and OS [20]. Certain dosimetric factors can even predict the risk of cardiac toxicity more reliably than considering the heart as a whole, underlining the importance of such parameters. For example, in a sub-group analysis, Hahn et al. concluded that a model using dose constraints to the coronary arteries was more robust than the dose to the whole heart in predicting the occurrence of cardiac ischemic events when treating patients with Hodgkin’s lymphomas [21].

However, in the case of ESOC, data on the subject remain limited. For this reason, the aim of our study will be to evaluate the association between independent parameters of dose received by CSS and the occurrence of major cardiac events (MACEs) in patients treated with CT-RT or exclusive RT as part of the management of locally advanced non-operated ESOC.

Materials and methods

Population

In this retrospective bicentric study, patients from Brest University Hospital and Morlaix Hospital were selected from a prospectively-maintained database in the RT department of Brest University Hospital. The patients included were men or women aged 18 or over, treated with CT-RT or exclusive RT with a minimum dose of 50 Gy on the high-risk volume, for histologically or cytologically proven, localized or locally advanced, non-operated ESOC from 01/01/2010 to 30/06/2022.

Patients under 18 years of age, under legal protection, with metastatic disease or concomitant malignancy, who received a high-risk volume dose of less than 50 Gy, with missing clinical or dosimetric data, or who refused to participate, were excluded. Patients who received esophageal brachytherapy were excluded due to the lack of dose accumulation possibility between the two treatment modalities. We did not exclude patients with a history of neoplasia if considered to be in remission, or with a history of RT, except if thoracic, but dosimetric data from previous treatments had been taken into account during initial planning in order to comply with dose constraints.

All patients underwent extension work-up by esophagogastroduodenoscopy (EGD), thoraco-abdominal-pelvic computed tomography (TAP-CT) with contrast, positron emission tomography (PET)-CT, and/or endoscopic ultrasound, and were staged according to the 8th version of the American Joint Committee on Cancer TNM classification system [22].

The patients’ histories were gathered, in particular with regard to CV disease, and a score of CV risk, from low to moderate (I), to high (II) or very high (III), was assigned to each according to the recommendations of the European Society of Cardiology (ESC). Patients with a history of diabetes, chronic renal failure, familial hypercholesterolemia or established CV disease were classified separately according to the severity of their pathology or associated co-morbidities, while the other patients were classified according to the SCORE2 and SCORE2-OP (for elderly patients) composite scores adapted to low-risk CV countries [23,24,25].

The study was approved by the Local institutional review board (BTCRET: 29BRC23.0144) and has been registered on the Clinical Trials website (NCT05996276).

Treatment modalities

All patients underwent a free breathing preparatory scan in supine position, followed by exclusive RT or CT-RT using 3-dimensional conformal radiotherapy (3DCRT) or intensity modulated radiotherapy (IMRT) using volumetric modulated arc therapy (VMAT).

Target volumes, including Gross Tumor Volume (GTV), Clinical Target Volume (CTV) and Planning Target Volume (PTV), and organs at risk were defined according to current recommendations [26, 27], and planning was carried out using Pinnacle treatment planning system v9.10 or v16.2 (Philips Medical Systems^®, Fitchburg, WI, USA). 3DCRT was delivered on an ONCOR linear accelerator (Siemens Healthineers^®, Erlangen, Germany) while VMAT plans were delivered on a Truebeam (Varian^®, Palo Alto, California, USA), both equipped with a multileaf collimator.

Dosimetric assessment

In our study, we used a publicly available deep learning segmentation model called Totalsegmentator [28, 29] to delineate the following cardiac substructures on the original planning CT: heart, right atrium, left atrium (LA), right ventricle, left ventricle and myocardium. All contours were reviewed by two trained radiotherapists and no modifications were needed.

Dosimetric parameters were then extracted using MIM Maestro (MIM software^® Inc., Cleveland, OH, USA) and dose volume histogram (DVH) were generated, including Dx_y (dose in Gy received by x% of the organ y) ranging from D5 to D100 in increments of 5, and Vx_y (volume of the organ y in cc receiving at least xGy) ranging from V5 to V70 in increments of 5, as well as volumes of the structures studied. Vx_y can also be expressed in the percentage of the y volume receiving at least xGy. Mean dose to the heart (MHD) was also recorded.

Follow-up

Patients were followed up every 3 months for 2 years and then monitored every 6 months for 5 years by TAP-CT (PET-CT could be used) and EGD according to French recommendations [3]. Local recurrence was defined as recurrence in the esophagus, regional recurrence as recurrence in nearby lymph nodes and distant recurrence as the occurrence of metastases.

End points

The primary aim of this study was to determine clinical or dosimetric parameters, particularly related to CSS that were predictive of the primary endpoint, i.e. the occurrence of G3 + MACEs.

Cardiac events including pericarditis, acute coronary disease, heart failure, valvulopathy and heart rhythm disorders, were defined by certified cardiologists, and we obtained the details after a thorough reading of the patients’ clinical records. They were classified according to the Common Terminology Criteria for Adverse Events version 5.0 (CTCAE 5.0), without knowledge of the treatment plan. They were labeled as MACEs if their grade was ≥ 3 (G3+). The time-to-event (TTE) was calculated from the end of RT to the occurrence of the event.

Secondary endpoints were the identification of dosimetric or clinical parameters that were predictive of grade ≥ 4 (G4+) MACEs, as well as OS, defined as the time from end of radiotherapy to death from any cause, patterns of relapse and progression-free survival (PFS), defined as the time from end of radiotherapy to tumor progression as assessed by an expert panel, or death from any cause. We also looked at the impact of the radiotherapy technique (3DCRT vs. VMAT) on the dosimetric parameters of interest.

Statistical analysis

In order to limit the amount of data to be included in the univariate and multivariate analyses, we first used the non-parametric Mann Whitney U test to determine which dosimetric parameters could be of interest. Receiver operating characteristics (ROC) curves were then plotted and parameters with an area under the curve (AUC) > 0.65 were retained. Then, we used Spearman’s correlation test to eliminate highly correlated variables (with a correlation coefficient > 0.7). When two features appeared as correlated, only the one harboring the highest AUC remained. Finally, an optimal cut-off was defined by the Youden index of accuracy (sensitivity + specificity – 1) under a ROC analysis.

These dosimetric parameters along with MHD were then included in the univariate analyses using the Cox proportional hazard model, along with clinicopathological factors chosen by the investigators according to their relevance to the subject i.e. MACEs: patient-related parameters such as age at diagnosis, performance status (PS), CV risk, obesity, LVEF (left ventricular ejection fraction), history of heart failure, history of valvulopathy, history of heart rhythm disorders, the presence of a pulmonary disease (chronic obstructive pulmonary disease, obstructive sleep apnea-hypopnea syndrome, pulmonary arterial hypertension) potentially responsible for pulmonary hypertension (PHT) and hence heart failure; treatment-related parameters such as treatment modality (CT-RT vs. RT), RT technique (3DCRT vs. VMAT), low-risk CTV (CTV-LR) prescribed dose, high-risk CTV (CTV-HR) prescribed dose, the type of CT if used, with a focus on the use of 5FU/Herceptin during induction CT or CT-RT, or at any time during treatment; and tumor-related parameters such as location. Significant parameters, or close to significance (p < 0,2), in univariate analysis were then tested in multivariate analysis.

A specific multivariate analysis was also performed including the retained CSS dosimetric features and the following dose constraints: V40_Heart < 30% (33), V45_Heart < 66% (34), MHD < 15 Gy, V5_Heart < 80%, V30_Heart < 30%, V50_Heart < 7% (14) as well as V30_Heart < 21% (35).

Finally, for the remaining parameters, survival curves were plotted using the Kaplan Meier method, then compared using the risk stratified log-rank test. A risk-stratified Cox regression model was used to estimate hazard ratios (HRs) and corresponding confidence intervals (CIs).

The same process was used for cardiac events of G4+. For all tests, a p-value < 0.05 was considered statistically significant.

All statistical analyses were performed using SPSS Statistics v24.0 and MedCalc Statistical Software version 15.8 (MedCalc Software, Ostend, Belgium, https://www.medcalc.org, 2015). Quantitative variables will be presented with their median (interquartile range) or mean +/- standard deviation.

Results

Patient and treatment characteristics

Fig. 1

Patient selection flowchart

122 patients met the inclusion criteria (Fig. 1), with a median age of 70.5 years (63 to 78), of whom 106 (86.88%) were men. Most of them, 88/122 (72.13%), had squamous cell carcinoma, and 74 (60.66%) were at very high CV risk, while 70 (57.38%) had a pre-existing CV pathology. The majority, 106/122 (86.88%), received CT-RT, mostly Folfox (5FU–oxaliplatin) for 48/122 patients (39.24%), or Cisplatin-5FU for 24/122 patients (19.67%). All patients eligible for 5 FU had a dihydropyrimidine dehydrogenase assay prior to chemotherapy. Almost all, 115/122 (94.26%), received lymph node irradiation (CTV-LR) with a median dose of 41.4 Gy (41.4 to 48.4), while the median dose for CTV-HR was 59.4 Gy (50.4 to 60) with a median duration of 46 days (41 to 50). Some, 7/122 (5.74%), had an intermediate volume. All irradiations were performed with conventional fractionation of 1.8–2 Gy. Baseline characteristics are summarized in Table 1.

Table 1 Patient, tumor and treatment characteristics

Cardiac events

With a mean follow-up of 25.8 months (IC95% 20.7–31.0), 36/122 patients (29.51%) experienced cardiac events with a mean of 1.8 +/- 1.08 events per person and a median time to event (TTE) of 13.8 months (7.9 to 31.9). Among them, 21/36 (58.33%) had G3 + MACEs with a median TTE of 13.1 months (5.3 to 25.8) and an average of 2.3 +/- 1.52 events per person, and 9/36 (25%) had G4 + MACEs with a median TTE of 9.8 months (5.3 to 19.9) and an average of 3.1 +/- 1.9 events per person. A total of 22 G3 events, 5 G4 events and 6 G5 events occurred during follow-up. Among these, the most frequent event was heart failure (n = 15), followed by heart rhythm disorders (n = 5), acute coronary disease (n = 5), valvulopathy (n = 5) and pericarditis (n = 3). The 6 grade 5 MACE were 6 cases of heart failure due to ischemic coronaropathy or cardiomyopathy.

Predictors for G3 + events

Of the numerous significant dosimetric parameters in the Mann Whitney U test, only one feature was retained after feature set reduction using the AUC and the inter-correlation analyses: left atrium V15 (V15LA) (p = 0.011; AUC = 0.661). The optimal cut-off was defined at 76.5 cc using the ROC curve. This parameter, together with MHD, was then included in univariate and multivariate analyses along with the clinicopathological parameters previously described.

Table 2 Univariate and multivariate analysis for G3 + MACEs

As shown in Table 2, significant parameters in univariate analysis were heart failure (p = 0.009), heart rhythm disorders (p = 0.04), LVEF (p = 0.03) and V15LA (p = 0.001) while parameters close to significance were obesity (p = 0.07), valvulopathy (p = 0.06), PHT (p = 0.08) and location (p = 0.14). Whether or not CT was used, and the type of CT used were not significant predictors of G3 + MACEs. After multivariate analysis, only V15LA emerged as a significant parameter (≤ or > 76.5 cc; 95%CI: 1.66–25.19; HR: 6.974; p = 0.013). Compared with patients for whom V15LA was less than ≤ 76.5 cc, patients with V15LA > 76.5 cc had a significantly higher incidence of G3 + MACEs (2 years rate: 9.7% vs. 35%; 5 years rate: 24.4% vs. 65.9%; HR: 4.31; p = 0.0008) as shown in Fig. 2.

Fig. 2

Kaplan Meier curves stratified on V15LA for G3+ cardiovascular events

When compared to usual heart-based dose constraints, V15LA remained statistically different (p = 0.003) along with MHD as shown in Supplementary Table 1.

Predictors for G4 + events

In the analysis for G4 + MACEs, only V15LA (p = 0.07; AUC = 0.773) remained after feature set reduction using the AUC and the inter-correlation analyses and was therefore included in the univariate and multivariate analyses with the clinicopathological parameters and MHD. The optimal cut-off was defined as 97.6 cc using the ROC curve.

Table 3 Univariate and multivariate analysis for G4 + MACEs

As shown in Table 3, significant parameters in univariate analysis were heart failure (p = 0.01), PHT (p = 0.03) and V15LA (p = 4.1E-05) while parameters close to significance were heart rhythm disorders (p = 0.08), age (p = 0.14), LVEF (p = 0.09) and MHD (p = 0.09). As for G3 + MACEs, the use or type of CT were not significant predictors of G4 + MACEs. This time, in multivariate analysis not only did V15LA emerged again (≤ or > 97.6.cc; HR: 55.448; 95%CI: 3.71-828.11; p = 0.004), but also the history of pathology favoring PHT as previously defined (Yes vs. No; HR: 10.345; 95%CI: 1.22–87.67; p = 0.032). The incidence of G4 + MACEs in patients with V15LA > 97.6 cc was significantly higher (2 years rate: 3.8% vs. 37%; 5 years rate: 6.9% vs. 52.7%; HR: 238.8; p < 0.0001) as shown in Fig. 3.

Fig.3

Kaplan Meier curves stratified on V15LA for G4+ cardiovascular events

When compared to other usual heart-based dose constraints, V15LA was the only statistically different dose constraint (p = 0.001) as shown in Supplementary Table 2.

Survival endpoints

Median OS was 19.0 months (IC95% 12.2 to 25.4) and there was no significant difference between patients with or without MACEs (p = 0.37), as shown in Supplementary Fig. 1.

We did not find any dosimetric parameters of interest in relation to OS in the preliminary analyses, so we did not perform univariate or multivariate analyses subsequently. We did, however, test the V15LA with the previously defined cut-offs at 76.5 cc and 97.6 cc, with a log rank test, but no significant difference in terms of OS emerged between patients with a V15LA > 76.5 cc and those without (p = 0.85) and those with V15LA > 97.6 cc and those without (p = 0.65), as shown in Supplementary Figs. 2 and 3.

Two-year OS was 45.9% (n = 56/122) and 5-year OS was 18.85% (n = 23/122) and 29/122 patients (23.77%) were still alive at the end of the study. The causes of death were distributed as follows: 56/93 (60.21% of all deaths) related to disease progression, including 6 from esomediastinal fistula and 4 from esovascular fistula, 11 from infection (11.83%), 6 (6.45%) from cardiac causes, 3 (3.23%) from other cancers, 3 (3.23%) from natural causes, 2 (2.15%) from chronic obstructive pulmonary disease exacerbation, 2 (2.15%) from pulmonary embolism, 2 (2.15%) from cirrhosis complications, 2 (2.15%) from iatrogenic causes (1 morphine overdose and 1 esophageal prosthesis) and 6 (6.45%) from unknown causes. Of those who died of cardiac causes, 4 died of heart failure, 1 of which was related to PHT, 1 of acute coronary syndrome and 1 of tamponade.

Median PFS was 8.1 months (IC95% 6.5 to 12.0). Of the 77/122 (63.11%) patients who recurred, 38 (31.15%) recurred locally, 32 (26.23%) regionally and 53 (43.44%) metastatically. 23/77 patients (29.87% of all recurrences) showed metastatic recurrence only, 13 (16.88%) regional and metastatic recurrence, 11 (14.29%) local recurrence only, 10 (12.99%) local and metastatic recurrence, 10 (12.99%) local and regional recurrence, 7 (9.09%) local, regional and metastatic recurrence and 3 (3.9%) regional recurrence only. The most frequent sites of distant recurrence were lung (n = 25), liver (n = 17), bone (n = 12), distant lymph nodes (n = 6) and peritoneal carcinosis (n = 5). Less frequently, secondary lesions were observed in the adrenal glands (n = 4), pleura (n = 3), soft tissue (n = 3), brain (n = 3) or pancreas (n = 1), mostly in multi-metastatic patients.

Impact of the radiotherapy technique on V15LA

59/122 (48.36%) patients were treated using 3DCRT, mostly before 2016, and 63 (51.64%) were treated using VMAT almost exclusively after 2016, as recommended (32). Median V15LA in the 3DCRT group was 63.4 cc (46.4 to 76.9), while it was 52.6 cc (14.7 to 70.9) in the VMAT group. After performing a Mann Whitney U test to compare V15LA between these two groups, a significant reduction in V15LA in favor of VMAT was confirmed (p = 0.0135). As can be seen in Supplementary Fig. 4, the difference was most marked for locations close to the heart, i.e. middle and lower thirds of thoracic esophagus and esogastric junction.

Discussion

While data on cardiac toxicities in long-term survivors of Hodgkin’s lymphomas, breast cancer, childhood cancer or lung cancer [4,5,6,7,8,9,10,11,12] are accumulating, reports focusing on patients with ESOC and especially on CSS remain rare.

In a 2015 meta-analysis, Beukema et al. [13] analyzed 13 studies of RT toxicity in ESOC published from 1998 to 2012, and already noted an early onset of cardiac events classically defined as late-onset with, for example, a median TTE of G3 + MACEs of 10 months in the study by Ishikura et al. [30], which still appears to be the case today with a median TTE in our study at 13 months (5.3 to 25.8). These emerging data, combined with an increase in OS of ESOC patients in line with therapeutic advances, have raised new concerns about the prediction and management of cardiac toxicities in this population.

Such an evaluation is especially needed considering the negative results of several dose-escalation trials where the potential benefit of dose-escalation may have been counter-balanced by the increase in treatment-related toxicities [12, 31], even in a selected population that may be quite different from “real-life” patients like in our study.

Dosimetry planning for a patient with ESOC often only takes into account whole-heart based dosimetric constraints, for instance MHD. Wang et al. [14] found an occurrence of G3 + MACEs in 18% of their patients, with a significantly more frequent onset in patients with MHD ≥ 15 Gy and a median OS of 54 months. In our study, we found a similar proportion of G3 + MACEs which occurred in 17.21% (n = 21/122) of our patients, but with a median OS of 19.0 months. These results are probably due to the poorer prognosis of our population, consisting only of non-operated patients while only 40.7% had definitive CT-RT in Wang’s study. Most of our patients, 70.49% (n = 86/122), were not eligible for surgery because of their poor general condition or previous medical history, particularly CV history with 57.38% (n = 70/122) of our patients presenting with pre-existing CV pathologies versus 22.8% in the study by Wang et al., underlining the importance of cardiac toxicity data in our population.

MHD is not a ubiquitous parameter, and such a volume-based parameter does not take into account the dose heterogeneity nor the differential response of different heart tissues to RT. This is why dosimetric parameters related to CSS have begun to be investigated, with numerous studies focusing on the coronary arteries in particular, for example, Wang et al. [32] also showed that left anterior descending coronary artery V30 ≥ 10% was significantly associated with the risk of major coronary events, and a better predictive factor than MHD or the classically used heart V30. For our part, we chose not to study these structures, due to their highly variable delineation between observers, as already shown, notably in the study by Duane et al. [33], which found an average contour overlap of 6 practitioners of less than 50%, even with a precise atlas detailing the segments of each coronary. It is also justified by the fact that we used a segmentation software which, even if it does not take coronary arteries into account, has the advantage of being free and public, allowing our results to be used in other centers with this tool. Moreover, studies reporting cardiac toxicities in ESOC patients treated with exclusive RT or CT-RT suggest that coronary events may not predominate, as in the study by Ogino et al. [34] who found none among 9 G3 + cardiac events in 58 patients, which is consistent with our study where we found only 5/33 (15.15%) of G3 + coronary toxicity.

Our study is therefore the first, to our knowledge, to find dosimetric parameters linked to CSS as a predictor of MACEs in non-operated ESOC patients only, treated with exclusive RT or CT-RT.

It is also important to note that a history of factors predisposing to PHT was also found to be significant in multivariate analysis for G4 + MACEs, underlining the potential importance of a PHT mechanism in RT related heart failure phenomena. A study by Ma et al. [35] showed that pulmonary artery V40 > 80%, V45 > 68% and V55 > 32% were independent predictors of OS in patients treated with normofractionated RT for NSCLC, without however finding any obvious clinico-pathological explanations. Even if pulmonary artery extracted dosimetric features were not included in our study, it should be further explored.

Our study also highlights the importance of the RT technique, with a significant reduction in V15LA with VMAT. These results are consistent with the literature, which reports a significant improvement of cardiac sparing with VMAT particularly for low-situated ESOC i.e. close to the heart. Munch et al. [36] described a significant improvement in heart V30 from 50.4 to 17.7% (p = 0.015) for patients with ESOC of middle or lower third, and Hawkins et al. [37] reported a reduction in MHD from 28.4 Gy to 25.9 Gy (p = 0.04) but also in heart V30-40 for ESOC of the esogastric junction or lower third.

In addition, our study is one of the only ones, to our knowledge, to have followed the CV risk SCORE2 and SCORE2-OP published by the ESC in 2021 (25–27), and although we found no significant association with the occurrence of G3 + MACEs, 97.54% (n = 119/122) of our population was found to be in the high or very high-risk category. Therefore, it is important to remember that these patients should benefit from an overall assessment of their CV status and strict control of risk factors.

Our study has a number of limitations: a retrospective design, with a probable information bias concerning the collection of cardiac events, a lack of initial CV evaluation and standardized follow-up as recommended, and therefore a potential underestimate of less symptomatic events, a limited number of individuals, a lack of data on treatment of risk factors and CV diseases, a lack of study of certain cardiac parameters of interest such as coronary arteries, an impossibility to take brachytherapy boost into account in dosimetry reconstruction, and an extended data collection period with heterogeneity of techniques, RT regimens and CT protocols, even though we did not find any impact of these different factors on the primary endpoint.

Conclusion

Our study was able to demonstrate a positive association between V15LA and the occurrence of G3 + MACEs, with a better prediction than conventional parameters such as MHD. Our results highlight the importance of defining and validating new dose constraints on CSS in patients treated for non-operated ESOC.

Particular attention should be paid to modern radiotherapy techniques for cardiac sparing, notably proton therapy, currently under investigation with two phase III studies comparing IMRT and proton therapy (NRG-GI006 in USA and PROTECT in Europe). Moreover, other constraints remain to be explored in different structures such as pulmonary arteries, and it is important to emphasize the need for standardized CV assessment and follow-up in patients treated for ESOC or other thoracic irradiation.