Background
Breast cancer (BC) radiotherapy (RT) reduces BC recurrence and improves survival [
1]. However, increased risk of cardiac death many years after BC RT has been documented [
2]. For long-term radiotherapy-induced cardiac complications, dose-response relationships between the mean heart dose and the rate of major coronary events were observed [
3‐
5]. Long before the onset of clinically relevant cardiac events, evaluation of early myocardial dysfunction was investigated after BC RT [
6]. This was based on two-dimensional speckle tracking echocardiography (2DSTE) that has allowed accurate measurements of global and regional myocardial deformation with strain [
7,
8]. Several studies have showed the higher sensitivity and prognostic value of the global longitudinal strain (GLS), compared with left ventricular ejection fraction (LVEF), for early detection of left ventricular dysfunction, and it has been shown that detecting a decreased LVEF after RT may be too late for treatment [
9,
10].
In most of previous studies on early myocardial dysfunction post BC RT, a statistically significant decrease in GLS was observed among left-sided BC patients, ranging from 5 to 14% at different post-RT time points from few weeks to 14 months [
11‐
16], whereas no measurable alteration of LVEF was observed. However, in these studies, no or few results were specifically presented for patients with a drop in GLS > 10% whereas this threshold is considered to define subclinical left ventricular dysfunction and has been reported to be predictive of subsequent cardiotoxicity [
17,
18].
Knowledge on the relationship between cardiac exposure and the decrease of strain is limited. A dose-related regional myocardial dysfunction in the acute phase after RT was found in left-sided BC patients with the greatest reduction in the apical part of the left ventricle, which received the highest radiation dose [
15], but no significant association was found between the mean heart dose and the GLS reduction [
19]. Further studies are needed to investigate the association between cardiac exposure and the evolution of GLS after RT, considering in particular the doses absorbed to the whole heart as well as to cardiac substructures such as the left ventricle or the coronary arteries. Indeed, these doses could enhance knowledge on the dose-response relationship according to the type of cardiotoxicity and its location [
20].
Based on the BACCARAT prospective cohort of BC patients treated with 3D-CRT (3-Dimensional Conformal Radiotherapy), we aimed to present a 6-month follow-up analysis of the association between radiation exposure to the whole heart and the left ventricle (LV) and the evolution of GLS from baseline to 6 months after RT, in particular for subclinical LV dysfunction defined as GLS reduction > 10%, a secondary endpoint of BACCARAT.
Discussion
In this prospective study, we evaluated the association between cardiac exposure and subclinical LV dysfunction defined as GLS reduction > 10% from baseline before RT to 6 months after RT in BC patients treated with 3D-CRT without chemotherapy. The associations with mean doses to the heart and the LV as well as the heart and LV V20 were significant in univariate analysis. In multivariate analysis, these associations were no longer significant after adjustment for non-radiation factors including BMI, hypercholesterolemia and endocrine therapy. Further exploratory analysis allowed identifying a subgroup of patients (LV V20 > 15%) with a significant association with subclinical LV dysfunction which remained significant after adjustment.
Decrease in longitudinal strain was previously observed in left-sided BC patients with follow-up ranging from few days to 14 months after RT [
11,
14,
15,
26] and mean relative decrease in longitudinal strain ranging from 5% to nearly 15% [
12]. In our study, the mean decrease of GLS was 6% at 6 months after RT, in the range of previously observed decreases. The absence of significant decrease of GLS in right-sided BC patients was also previously observed in these studies even if the limited size of right-sided BC patients group could partly explain these non-significant results. Similarly to most other previous studies [
12‐
15,
26], no significant decrease in LVEF was observed at RT + 6 months in our patients compared with pre-RT.
We considered a subclinical LV dysfunction defined as GLS reduction > 10% which has been considered clinically relevant [
18] and which was also considered in other previous studies [
17,
27]. This early index of cardiotoxicity was observed in 48% of our left-sided BC patients 6 months after RT, which is higher than the 28% observed elsewhere with a shorter follow-up limited to end of RT [
14]. Such subclinical dysfunction may need longer follow-up to develop. No other study evaluated the frequency of this subclinical event as they mainly considered the GLS as a continuous variable which may limit the clinical implication and applications.
Even if all previous studies concluded that longitudinal strain was decreased after RT for left-sided BC patients and not for right-sided BC patients, little was known on the association between cardiac doses and decrease in longitudinal strain. A modest correlation was observed between GLS reduction 6 weeks after RT and mean heart dose or V30 (
R = 0.35,
R = 0.22) [
26,
27], but no difference in radiation dose between the group with or without a > 10% reduction in GLS was observed. That could be explained by the limited number of patients and also the potential contribution of factors apart from radiation dose [
16]. With a very short follow-up (end of RT), another study did not find an association between mean heart dose and the mean GLS reduction [
14]. In our study, we observed significant associations between cardiac doses and subclinical LV dysfunction defined as GLS reduction > 10%. However, these associations were no longer significant after adjustment for non-radiation factor also associated with subclinical LV dysfunction including BMI, hypercholesterolemia and endocrine therapy. It was interesting to note that LV exposure, in particular V20 of the LV, was associated with subclinical LV dysfunction. As exploratory analysis, we identified a subgroup of patients with LV V20 > 15% and there was for this group a 4-fold increase in the adjusted odds ratio for subclinical LV dysfunction, as compared with the 0% category. This may be an indication for future research investigating strain evolution post-RT and subclinical LV dysfunction: instead of whole heart doses, analysis of LV exposure may provide better information to understand the association with RT. In a previous work published by Van Den Bogaard et al. [
5], an association between the volume of the left ventricle receiving 5 Gy (LV-V5) and cumulative incidence of acute coronary event (ACE) was observed. Their analysis showed that LV-V5 was the most important prognostic dose-volume parameter. In our study, LV-V5 was significantly associated with GLS reduction > 10% in univariate analysis (OR = 1.04 [1.00–1.08],
p = 0.04) but was no longer significant in multivariate analysis (OR = 1.03 [0.99–1.07],
p = 0.21).
Contribution of factors apart from radiation dose on the risk of long term cardiac disease, such as age, hypertension, diabetes or preexisting cardiac diseases was previously observed [
3] as they had an additive effect on the risk of cardiac disease. At the scale of subclinical LV dysfunction, quantified by longitudinal strain, it was also important to consider their contribution on the associations. Among the different factors that we considered, endocrine therapy was associated with subclinical LV dysfunction (OR = 3.20, 95% CI (1.02–10.10)), in particular for aromatase inhibitors (OR = 4.25, 95% CI (1.32–15.53)) which are known risk factors for cardiovascular disease [
28]. An independent association between a reduction in GLS and the use of aromatase inhibitors was also previously observed [
14]. The BMI was also important to consider in multivariate analysis, as it was a confounding factor: associated with both the cardiac event and the dose. We observed that patients with higher BMI had higher cardiac doses (Heart Dmean = 2.1 Gy for patient with BMI < 25 kg/m
2 vs. 3.3 Gy for patients with BMI > 25 kg/m
2,
p < 0.01), as previously observed [
29]. Last, the cardiotoxicity of chemotherapy, such as anthracyclines or trastuzumab, are known to alter the longitudinal strain [
30,
31]. A strength of our study was to include chemotherapy-naïve patients, which allowed a precise evaluation of the association with radiation exposure without confounding due to chemotherapy.
Some studies investigated more precisely the longitudinal strain changes based on the segmental evaluation within the LV [
32]. Lo et al. detected dose-related regional myocardial dysfunction in the acute phase after RT with the greatest reduction in the apical part of the LV, which received the highest radiation dose [
15]. In the Erven’s study [
11], changes were more pronounced in the LV wall receiving the highest RT dose (anterior wall). Heterogeneity of cardiac exposure [
20] may be considered for precise evaluation of cardiotoxicity with the assessment of doses to cardiac substructures such as LV or LAD. The LV segmentations and assignment of these segments to coronary arterial territories are still not sharply defined and confusing [
32]. However, we are now collecting strain segmental values for each echocardiography of our patients and will further analyze this data according to the precise individual dose evaluation of each coronary artery [
20]. On the other side, the strain is not uniform over the LV. The LV wall is not homogenous and comprises an endocardial, a mid-myocardial, and an epicardial layer [
25]. Recent 2DSTE software allows separate evaluation of myocardial layers deformation [
33]. Our measurements of longitudinal strain focused on individual evaluation of midwall deformations. Our results at baseline (GLS = − 16.1 ± 2.7%) is consistent with the GLS observed in previous study [
34]. Other studies averaged the three layers of GLS [
11,
26], but we chose to focus here on the middle layer considering that this GLS would be a good indicator of the association with radiation exposure in the LV wall in addition to the fact that it has been shown in several studies to be robust and reproducible [
24]. Further analysis detailing the three layers are in progress and may allow refining our results as previously observed for chemotherapy [
35].
Research on radiation-induced changes in LV-function and association with cardiac exposure in patients treated with RT is not only relevant for BC patients. Patients treated for other organ with relevant heart doses like esophagus cancer, lung cancer or Hodgkin’s lymphoma are also of concerns. Some studies showed that, similarly to BC patients, these patients could receive high doses to the heart and its substructures [
36] and GLS could be decreased after lung RT [
37]. It is thus important to further develop studies, not only for BC patients, but also for other patients with thoracic RT.
Limitations
Even if we could observe associations between cardiac doses and the risk of subclinical LV dysfunction, our study showed the importance to consider also non-radiation factors such as BMI or endocrine therapy in the investigation of these association with doses as the results didn’t remain significant after adjustment. Our study is one of the largest ever published in this research area. However, with 79 patients, the size of the population involved a limited statistical power for odds ratios sub-analyses which may partly explain the absence of significant results in multivariate analysis. A large multicenter European study (MEDIRAD EARLY-HEART study) is ongoing and plan to include 250 patients which should provide results without this limitation [
38]. Some alternatives to cardiac echocardiography for assessment of early cardiac damage exist, such as myocardial scintigraphy or cardiac magnetic resonance imaging. Myocardial scintigraphy is an important noninvasive method in the evaluation of patients with suspected coronary artery disease due to its high diagnostic accuracy, as well as being able to define the extent, severity and location of myocardial perfusion abnormalities. It is used for detecting ischemia in symptomatic patients, but its use in asymptomatic ones, like our BC patients, is less clear, which conducted us not to consider this examination in our cohort. However cardiac magnetic resonance imaging for assessment of the function and structure of the cardiovascular system is very promising and will be further investigated in MEDIRAD EARLY-HEART study [
38]. The subclinical LV dysfunction defined as a GLS reduction > 10% is an indication of a beginning effect of radiation on LV function which corresponded to a secondary endpoint of BACCARAT study [
21].
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.