Feasibility and trade-offs of incorporating cardiac substructure dose constraints in left-sided breast volumetric modulated arc therapy planning

Abstract

Purpose

This study aimed to compare left-sided breast volumetric modulated arc therapy plans using heart-only dose constraints with plans that include additional left ventricle (LV) and left anterior descending coronary artery (LAD) dose constraints (heart + LV + LAD).

Materials and Methods

Thirty patients treated with whole-breast irradiation with simultaneous integrated boost were replanned using both strategies. The heart + LV + LAD strategy applied dose constraints based on German Society for Radiation Oncology recommendations and Danish Breast Cancer Group HYPO protocol (LV: D_mean [mean dose] <3 Gy, V_5Gy [percentage of volume receiving ≥5 Gy] <17%, V_23Gy <5%; LAD: D_mean <10 Gy, D_max [maximum dose] <17 Gy). Target and organ-at-risk (OAR) metrics were compared using Wilcoxon signed-rank tests. For exploratory analyses, LAD–planning target volume (PTV) distance was defined as the minimum distance between LAD and whole-breast PTV, and Spearman’s correlations between LAD-PTV distance and LAD or target metrics were evaluated.

Results

Compared with heart-only plans, heart + LV + LAD plans reduced LV D_mean, V_5Gy, and V_23Gy, as well as LAD D_mean and D_max (all p < 0.001), meeting all prespecified dose constraints except LV D_mean and LAD D_max (achieved in 70.0% and 63.3%, respectively). Target metrics showed small compromises (e.g., PTV V_95% [percentage of volume receiving ≥95% of the prescription dose] 98.391% vs. 99.431%; p < 0.001). Ipsilateral lung D_mean and V_5Gy decreased, whereas contralateral breast D_mean and lung V_10Gy increased (all p < 0.05). In heart + LV + LAD plans, greater LAD-PTV distance correlated with lower LAD D_max and D_mean and higher PTV D_98%.

Conclusion

Applying LV and LAD dose constraints improved cardiac substructure sparing with modest trade-offs in target and contralateral OAR metrics. LAD-PTV distance may influence the feasibility of LAD sparing, but this exploratory association warrants further study.

Introduction

In patients with early breast cancer undergoing breast-conserving surgery (BCS), adjuvant whole-breast irradiation (WBI) is a standard of care that reduces local recurrence and breast cancer death [1-3]. As long-term outcomes of breast cancer continue to improve, the potential for treatment-related late sequelae has become an increasingly important consideration [4,5]. Among these late effects, cardiac toxicity is a particular concern in left-sided breast radiation therapy (RT) due to the anatomical proximity of the heart to the irradiated breast. Therefore, a key challenge in treatment planning is minimizing cardiac dose while maintaining adequate target coverage. To reduce cardiac exposure, treatment planning has generally focused on constraining whole-heart dose metrics [6-8].

Although mean heart dose (MHD) has been linked to increased late cardiac toxicity [9] and cardiac mortality [10], growing evidence suggests that doses to specific cardiac substructures, such as the left ventricle (LV) and left anterior descending coronary artery (LAD), may also be associated with adverse cardiac outcomes [11-13]. Furthermore, even when MHD is low, anterior cardiac substructures such as the LAD may still receive substantial doses [14], supporting the consideration of dose constraints beyond whole-heart dose metrics [15]. Accordingly, the German Society for Radiation Oncology (DEGRO) breast cancer expert panel recommended incorporating LV and LAD dose constraints, in addition to MHD, in left-sided breast RT planning [16]. Similarly, the Danish Breast Cancer Group (DBCG) HYPO trial included a LAD-related dose constraint when the LAD was delineated [17].

Selective dose shaping to spare cardiac substructures has become increasingly feasible with modern delivery techniques such as intensity-modulated radiation therapy (IMRT) and volumetric modulated arc therapy (VMAT). In these inverse-planned techniques, the final dose distribution is generated according to the objectives and priorities assigned during optimization [18,19]. In this setting, adding explicit dose constraints for LV and LAD may improve the sparing of these cardiac substructures beyond that achieved with whole-heart dose constraints alone. To our knowledge, several prior dosimetric studies [20,21] compared left-sided breast IMRT/VMAT plans optimized with a heart-only approach to plans optimized using the LV and LAD as additional organs at risk (OARs). However, none of these studies explicitly incorporated a defined set of LV and LAD dose constraints into plan optimization, leaving the achievability and trade-offs of such constraints uncertain.

Therefore, guided by the dose constraints suggested by DEGRO and the DBCG HYPO trial, we aimed to investigate the dosimetric feasibility and potential trade-offs of moving from a conventional heart-only constraint strategy to a combined constraint strategy incorporating dose constraints for the heart, LV, and LAD (heart + LV + LAD strategy) in left-sided breast cancer VMAT planning. We hypothesized that these cardiac substructure dose constraints would be achievable while maintaining clinically acceptable target coverage and doses to other OARs. Additionally, as the anatomical proximity between the LAD and whole-breast planning target volume (PTV) may influence the achievable degree of LAD sparing and potentially affect target coverage, we explored the correlation between LAD-PTV distance and key LAD or target metrics.

Materials and Methods

1. Patients

This dosimetric study retrospectively included 30 consecutive patients with early breast cancer who underwent BCS, with or without sentinel lymph node biopsy, followed by left-sided WBI with a simultaneous integrated boost (SIB) using VMAT at our institution between December 2023 and October 2024. Eligible patients had pTis or pT1–2N0 diseases according to the American Joint Committee on Cancer Staging Manual, eighth edition. Patients who received neoadjuvant chemotherapy, regional nodal irradiation, sequential boost, or bilateral breast RT were excluded.

2. RT planning

Computed tomography (CT) simulation was performed in the supine position under free breathing without contrast, with a 5-mm slice thickness on a Brilliance Big Bore CT simulator (Philips Medical Systems, Cleveland, OH, USA). The breast clinical target volume (CTV) was contoured according to the European Society for Radiotherapy and Oncology consensus guideline [22]. The PTV was generated using a uniform 5-mm expansion from the CTV, with its posterior extent toward the chest wall adjusted when clinically appropriate. Boost volume was defined as the volume encompassing surgical clips and seroma in the lumpectomy cavity, with an approximately 1-cm margin. The PTV and boost volumes were cropped 5 mm from the skin surface. The whole heart, lungs, contralateral breast, esophagus, and spinal cord were delineated as OARs in both the heart-only and heart + LV + LAD strategies. In the heart + LV + LAD strategy, the LV and LAD were additionally delineated according to the Danish Multidisciplinary Cancer Groups atlas [23], with reference to prior contrast-enhanced chest CT scans when available. Prescription doses were 42.56 Gy in 16 fractions to the whole-breast PTV and 51.20 Gy in 16 fractions to the SIB volume.

RT planning was performed using the Eclipse treatment planning system (version 15.6) for Halcyon (Varian Medical Systems, Palo Alto, CA, USA). For each patient, two VMAT plans (heart-only and heart + LV + LAD) were generated using two half arcs with a single isocenter. Both plans were created on the same simulation CT with identical arc geometry, beam energy (6-MV flattening-filter-free photon beam), and dose calculation algorithm. For the heart-only strategy, the optimization template from each patient’s clinically treated plan was saved and applied to the corresponding study planning dataset without adding LV or LAD objectives. For the heart + LV + LAD strategy, a predefined optimization template including LV and LAD objectives was developed and applied to all patients.

VMAT optimization was performed with the Photon Optimizer (version 15.06.03, Varian Medical Systems), and dose calculation was performed using the Anisotropic Analytical Algorithm (version 15.06.03, Varian Medical Systems). The predefined optimization template used for the heart + LV + LAD strategy is summarized in Supplementary Table S1. Each plan was generated using a single inverse optimization, without patient-specific iterative modification of optimization objectives or priorities/weights. In the SIB setting, plan normalization was performed with consideration of both target dose levels. Plans were normalized such that each target volume (whole-breast PTV and tumor-bed boost volume) received at least 95% of its respective prescription dose, while hot spots were aimed to remain within 105% of the corresponding prescription dose. The same normalization and target coverage criteria were applied to both the heart-only and heart + LV + LAD strategies.

For OAR sparing in both strategies, MHD was kept below 5 Gy when achievable and otherwise minimized according to the as low as reasonably achievable (ALARA) principle, while also limiting heart V_30Gy to <5%. The following dose constraints were used for non-cardiac OARs in both strategies: ipsilateral lung V_20Gy <20%, whole lung V_20Gy <10%, and ALARA-based minimization for the contralateral lung and contralateral breast. In the heart + LV + LAD strategy, additional cardiac substructure dose constraints were applied according to the DEGRO recommendations [16] and the DBCG HYPO protocol [17] (Table 1): LV D_mean (mean dose) <3 Gy, LV V_5Gy (percentage of volume receiving ≥5 Gy) <17%, LV V_23Gy <5%, LAD D_mean <10 Gy, and LAD D_max (maximum dose) <17 Gy.

Table 1.

Dose constraints for cardiac substructures

Structure	Dose constraints	Reference	Associated findings
LV	Dmean <3 Gy	DEGRO [16]	LV segments receiving >3 Gy showed reduced regional strain on Doppler echocardiography immediately after RT and at 2-month f/u [24]
	V5Gy <17%	DEGRO [16]	Mean LV V5Gy of patients who experienced an acute coronary event vs. those who did not: 29.3% vs. 16.9% (median f/u 7.6 years) [11]
	V23Gy <5%	DEGRO [16]	New LV perfusion defects on SPECT when <5% vs. ≥5% of LV were included within the RT fields (i.e., ≥23–25 Gy): 10%–20% vs. 50%–60% (f/u 6–24 months) [25]
LAD	Dmean <10 Gya)	DEGRO [16]	Left-sided breast RT was associated with increased stenosis in LAD “hotspot” segments compared with right-sided breast RT, and RT with greater expected hotspot exposure (“high-risk RT”) was associated with increased stenosis in predefined hotspot areas (median f/u 10.3 years) [12]
			After mediastinal Hodgkin lymphoma RT, coronary segments that were stenotic at median 6-year f/u had received higher RT doses than segments without stenosis (median segment dose 30.3 Gy vs. 26.3 Gy) [26]
	Dmax <17 Gya)	DBCG HYPO trial [17]	N/A

LV, left ventricle; D_mean, mean dose; DEGRO, German Society for Radiation Oncology; RT, radiation therapy; f/u, follow-up; V_xGy, percentage of volume receiving ≥x Gy; SPECT, single-photon emission computed tomography; LAD, left anterior descending coronary artery; D_max, maximum dose; DBCG, Danish Breast Cancer Group; N/A, not available.

^a)Expert panel-recommended or protocol-derived dose constraints, rather than endpoint-derived thresholds.

3. Dosimetric evaluation

For each patient, heart-only and heart + LV + LAD VMAT plans were compared. Plan quality for the PTV and boost volumes was evaluated using Radiation Therapy Oncology Group (RTOG)-based indices, homogeneity index (HI), quality of coverage (QOC), and conformity index (CI). These indices were categorized according to the RTOG protocol compliance/deviation criteria as follows: HI ≤2.0, >2.0 to 2.5, and >2.5; CI 1.0 to 2.0, 0.9 to <1.0 or >2.0 to 2.5, and <0.9 or >2.5; and QOC ≥0.9, 0.8 to <0.9, and <0.8, corresponding to per-protocol, minor deviation, and major deviation ranges, respectively [27,28]. Dose distributions for the PTV and boost volumes were also compared using D_2% (minimum dose received by the hottest 2% of the structure), D_50%, D_98%, V_95% (percentage of volume receiving ≥95% of the prescription dose), and V_107%. For OARs, D_mean and relevant V_x metrics were evaluated for the heart, lungs, and contralateral breast, and LV and LAD doses were assessed using D_mean and/or D_max, as well as selected V_x metrics.

4. Statistical analysis

Paired comparisons of dose-volume histogram (DVH) metrics between the heart-only and heart + LV + LAD plans were performed using two-sided Wilcoxon signed-rank tests. For exploratory analyses, the LAD-PTV distance was defined as the minimum distance between the LAD and whole-breast PTV contours on the planning CT. This distance was measured manually for each patient using the distance measurement tool in the treatment planning system. Correlations of LAD-PTV distance with LAD dose metrics (D_max and D_mean) as well as PTV D_98% were evaluated within each planning strategy, along with correlations between LAD-PTV distance and between-plan differences (Δ = heart + LV + LAD − heart-only) in these metrics, using Spearman’s rank correlation coefficient. All statistical analyses were performed using IBM SPSS Statistics for Windows (version 29.0, IBM Corp., Armonk, NY, USA), and p < 0.05 was considered statistically significant.

Results

Thirty patients with pathologically confirmed left-sided early breast cancer were analyzed. Patient, tumor, and planning characteristics are summarized in Table 2. Representative dose distributions for the heart-only and heart + LV + LAD plans are shown in Fig. 1, and the cohort-averaged cumulative DVHs are presented in Fig. 2.

Table 2.

Patient, tumor, and planning characteristics (n = 30)

Characteristic	Value
Age (year)	50.5 (31–84)
Histology
Ductal carcinoma in situ	7 (23.3)
Invasive carcinoma of no special type	18 (60.0)
Mucinous carcinoma	3 (10.0)
Tubular carcinoma	1 (3.3)
Invasive micropapillary carcinoma	1 (3.3)
T classification
pTis	7 (23.3)
pT1mi	1 (3.3)
pT1	20 (66.7)
pT2	2 (6.7)
Hormone receptor status
ER- or PR-positive	28 (93.3)
ER- and PR-negative	2 (6.7)
HER2 status
HER2-positive	6 (20.0)
HER2-negative	24 (80.0)
Tumor location
UIQ	12 (40.0)
LIQ	3 (10.0)
UOQ	6 (20.0)
LOQ	5 (16.7)
Central	4 (13.3)
PTV volume (cm3)	705.1 (326.2–1351.5)
Boost volume (cm3)	163.0 (40.4–386.9)
LAD-PTV distance (cm)	0.68 (0.13–1.38)

Values are presented as median (range) or number (%).

ER, estrogen receptor; PR, progesterone receptor; HER2, human epidermal growth factor receptor 2; UIQ, upper inner quadrant; LIQ, lower inner quadrant; UOQ, upper outer quadrant; LOQ, lower outer quadrant; PTV, planning target volume; LAD, left anterior descending coronary artery.

Fig. 1.

Representative dose distributions for a typical patient: (A) heart-only and (B) heart + LV + LAD plans. LV, left ventricle; LAD, left anterior descending coronary artery.

Fig. 2.

Cohort-averaged cumulative DVHs for all patients. (A) DVHs for target volumes and cardiac structures. (B) DVHs for the remaining OARs. Solid lines indicate the heart-only plans, and dotted lines indicate the heart + LV + LAD plans. DVH, dose-volume histogram; OAR, organ at risk; PTV, planning target volume; LV, left ventricle; LAD, left anterior descending coronary artery.

1. Target coverage and dose distribution

Target coverage and dose distribution metrics are summarized in Table 3. Because the whole-breast PTV included the SIB region, PTV HI, D_2%, and V_107% reflect intentional dose escalation within the boost volume; therefore, boost metrics were reported separately. For the whole-breast PTV, the heart + LV + LAD plans showed a higher HI, with lower QOC and CI, than the heart-only plans (all p < 0.001). For the boost volume, HI and QOC also differed between strategies (both p < 0.05), whereas CI did not differ significantly. The heart + LV + LAD plans had higher PTV D_2% (53.879 ± 0.559 Gy vs. 53.539 ± 0.527 Gy) and lower PTV D_98% (40.819 ± 0.821 Gy vs. 41.751 ± 0.887 Gy) and V_95% (98.391 ± 1.118% vs. 99.431 ± 0.841%) than the heart-only plans (all p < 0.001). For the boost volume, D_2%, D_50%, D_98%, and V_107% differed between strategies (all p < 0.05).

Table 3.

Dosimetric comparison between heart-only and heart + LV + LAD plans: target coverage and dose distribution

Structure	Metric	Heart-only	Heart + LV + LAD	Difference (heart + LV + LAD – heart-only)	p-value
PTV	HI (Imax/RI)	1.258 ± 0.012	1.266 ± 0.013	0.008 ± 0.013	<0.001
	QOC (Imin/RI)	0.920 ± 0.041	0.879 ± 0.038	–0.041 ± 0.036	<0.001
	CI (VRI/TV)	0.958 ± 0.017	0.934 ± 0.022	–0.024 ± 0.027	<0.001
	D2% (Gy)	53.539 ± 0.527	53.879 ± 0.559	0.340 ± 0.556	<0.001
	D50% (Gy)	46.682 ± 1.012	46.787 ± 1.052	0.105 ± 0.487	0.329
	D98% (Gy)	41.751 ± 0.887	40.819 ± 0.821	–0.932 ± 1.054	<0.001
	V95% (%)	99.431 ± 0.841	98.391 ± 1.118	–1.040 ± 1.138	<0.001
	V107% (%)	61.900 ± 7.880	62.942 ± 8.030	1.041 ± 7.094	0.530
Boost	HI (Imax/RI)	1.055 ± 0.012	1.061 ± 0.011	0.006 ± 0.012	0.002
	QOC (Imin/RI)	0.965 ± 0.011	0.961 ± 0.012	–0.004 ± 0.009	0.022
	CI (VRI/TV)	0.943 ± 0.051	0.952 ± 0.001	0.009 ± 0.051	0.780
	D2% (Gy)	54.011 ± 0.612	54.312 ± 0.565	0.301 ± 0.639	0.001
	D50% (Gy)	52.603 ± 0.396	52.953 ± 0.404	0.350 ± 0.496	<0.001
	D98% (Gy)	50.703 ± 0.230	50.642 ± 0.177	–0.061 ± 0.188	0.023
	V95% (%)	99.964 ± 0.102	99.946 ± 0.092	–0.018 ± 0.101	0.176
	V107% (%)	1.137 ± 4.805	1.874 ± 3.808	0.738 ± 5.715	0.045

Values are presented as mean ± standard deviation.

LV, left ventricle; LAD, left anterior descending coronary artery; PTV, planning target volume; HI, homogeneity index; I_max, maximum isodose in the target; RI, reference isodose; QOC, quality of coverage; I_min, minimum isodose in the target; CI, conformity index; V_RI, reference isodose volume; TV, target volume; D_x%, minimum dose received by the hottest x% of the structure; V_x%, percentage of volume receiving ≥x% of the prescription dose.

2. OAR sparing

OAR dosimetric parameters are presented in Table 4, and box plots comparing cardiac and cardiac substructure metrics between the two plans are shown in Fig. 3. For the whole heart, V_30Gy was lower in the heart + LV + LAD plans than in the heart-only plans (0.038% ± 0.099% vs. 0.333% ± 0.734%, p = 0.001), whereas D_mean did not differ significantly (4.639 ± 1.357 Gy vs. 5.848 ± 3.183 Gy, p = 0.086). LV dose parameters were reduced with the heart + LV + LAD strategy, including D_mean (2.843 ± 0.433 Gy vs. 5.472 ± 2.236 Gy, p < 0.001), V_5Gy (3.025% ± 2.230% vs. 39.584% ± 32.579%, p < 0.001), and V_23Gy (0.013% ± 0.053% vs. 0.546% ± 1.743%, p < 0.001). LAD D_mean and D_max were also lower in the heart + LV + LAD plans (D_mean: 6.264 ± 1.102 Gy vs. 10.507 ± 3.709 Gy; D_max: 15.888 ± 4.124 Gy vs. 25.412 ± 7.284 Gy; both p < 0.001).

Table 4.

Dosimetric comparison between heart-only and heart + LV + LAD plans: organs at risk

Structure	Metric	Heart-only	Heart + LV + LAD	Difference (heart + LV + LAD – heart-only)	p-value
Heart	Dmean (Gy)	5.848 ± 3.183	4.639 ± 1.357	–1.208 ± 2.293	0.086
	V30Gy (%)	0.333 ± 0.734	0.038 ± 0.099	–0.294 ± 0.675	0.001
LV	Dmean (Gy)	5.472 ± 2.236	2.843 ± 0.433	–2.629 ± 2.234	<0.001
	V5Gy (%)	39.584 ± 32.579	3.025 ± 2.230	–36.559 ± 32.842	<0.001
	V23Gy (%)	0.546 ± 1.743	0.013 ± 0.053	–0.534 ± 1.738	<0.001
LAD	Dmean (Gy)	10.507 ± 3.709	6.264 ± 1.102	–4.243 ± 3.665	<0.001
	Dmax (Gy)	25.412 ± 7.284	15.888 ± 4.124	–9.524 ± 7.754	<0.001
Ipsilateral lung	Dmean (Gy)	9.376 ± 2.101	8.651 ± 1.242	–0.725 ± 1.704	0.035
	V20Gy (%)	6.457 ± 5.737	5.161 ± 2.365	–1.296 ± 5.138	0.360
	V5Gy (%)	79.372 ± 13.161	75.547 ± 12.531	–3.825 ± 9.137	0.035
Contralateral lung	Dmean (Gy)	4.090 ± 0.993	4.117 ± 0.849	0.027 ± 0.616	0.168
	V10Gy (%)	3.162 ± 2.112	3.886 ± 2.831	0.724 ± 1.832	0.010
	V5Gy (%)	28.986 ± 14.858	27.108 ± 11.584	–1.878 ± 10.061	0.877
Whole lung	Dmean (Gy)	6.379 ± 1.112	6.088 ± 0.759	–0.291 ± 0.773	0.049
	V20Gy (%)	2.805 ± 2.361	2.296 ± 1.023	–0.509 ± 2.160	0.441
	V5Gy (%)	50.923 ± 11.903	48.142 ± 9.128	–2.782 ± 6.924	0.116
Contralateral breast	Dmean (Gy)	4.179 ± 0.894	4.510 ± 0.817	0.331 ± 0.706	0.021
	V5Gy (%)	24.911 ± 15.079	29.836 ± 14.967	4.925 ± 12.277	0.052

Values are presented as mean ± standard deviation.

LV, left ventricle; LAD, left anterior descending coronary artery; D_mean, mean dose; V_xGy, percent volume receiving ≥x Gy; D_max, maximum dose.

Fig. 3.

Box plots comparing cardiac and cardiac substructure dosimetric parameters between the heart-only and heart + LV + LAD plans: (A) D_mean to the heart, LV, and LAD, and D_max to the LAD, (B) heart V_30Gy and LV V_23Gy, and (C) LV V_5Gy. LV, left ventricle; LAD, left anterior descending coronary artery; D_mean, mean dose; D_max, maximum dose; V_xGy, percent volume receiving ≥x Gy. Asterisks (*) indicate statistically significant differences (p < 0.05).

Across the prespecified LV and LAD dose constraints, more plans met each dose constraint with the heart + LV + LAD strategy than with the heart-only strategy (LV D_mean <3 Gy: 21 [70.0%] vs. 0 [0.0%]; LV V_5Gy <17%: 30 [100.0%] vs. 13 [43.3%]; LV V_23Gy <5%: 30 [100.0%] vs. 29 [96.7%]; LAD D_mean <10 Gy: 30 [100.0%] vs. 17 [56.7%]; LAD D_max <17 Gy: 19 [63.3%] vs. 4 [13.3%]).

Ipsilateral lung D_mean and V_5Gy were lower with the heart + LV + LAD strategy compared with the heart-only strategy (8.651 ± 1.242 Gy vs. 9.376 ± 2.101 Gy; 75.547 ± 12.531% vs. 79.372 ± 13.161%; both p = 0.035). For the contralateral lung, V_10Gy was higher in the heart + LV + LAD plans (3.886% ± 2.831% vs. 3.162% ± 2.112%, p = 0.010), whereas D_mean and V_5Gy were not significantly different. Contralateral breast D_mean significantly increased in the heart + LV + LAD plans (4.510 ± 0.817 Gy vs. 4.179 ± 0.894 Gy, p = 0.021), whereas the V_5Gy numerically increased but did not reach statistical significance (29.836% ± 14.967% vs. 24.911% ± 15.079%, p = 0.052).

3. Association of LAD–PTV distance with LAD sparing and target coverage

The median LAD-PTV distance was 0.68 cm (range, 0.13 to 1.38) (Table 2). LAD-PTV distance was not correlated with LAD D_max in the heart-only plans (Spearman ρ = 0.086, p = 0.650) (Fig. 4A), but was inversely correlated with LAD D_max in the heart + LV + LAD plans (ρ = −0.638, p <0.001) (Fig. 4B). The between-plan difference in LAD D_max (ΔLAD D_max = LAD D_{max (heart + LV + LAD)} − LAD D_{max (heart-only)}) (Fig. 4C) was not significantly correlated with LAD-PTV distance (ρ = −0.252, p = 0.178). Similar associations were observed between LAD D_mean and LAD-PTV distance, with significant inverse correlations in the heart + LV + LAD plans (ρ = −0.454, p = 0.012), but not in the heart-only plans (ρ = 0.194, p = 0.303). ΔLAD D_mean was not significantly correlated with LAD-PTV distance (ρ = −0.251, p = 0.181). LAD-PTV distance showed a significant positive correlation with PTV D_98% in the heart + LV + LAD plans (ρ = 0.487, p = 0.006), but not in the heart-only plans (ρ = 0.258, p = 0.169). No significant correlation was observed between LAD-PTV distance and ΔPTV D_98% (ρ = 0.109, p = 0.565).

Fig. 4.

Scatterplots of the association of LAD-PTV distance with LAD D_max in the heart-only plans (A), LAD D_max in the heart + LV + LAD plans (B), and ΔLAD D_max (LAD D_{max (heart + LV + LAD)} − LAD D_{max (heart-only)}) (C). LAD, left anterior descending coronary artery; PTV, planning target volume; D_max, maximum dose; LV, left ventricle.

Discussion and Conclusion

By comparing heart-only and heart + LV + LAD strategies, this study evaluated the dosimetric feasibility and trade-offs of incorporating expert panel-recommended and protocol-derived LV and LAD dose constraints in left-sided breast cancer VMAT planning. Incorporation of these dose constraints substantially reduced LV and LAD doses while largely maintaining target coverage and plan quality, with small but statistically significant trade-offs observed in some target metrics and contralateral OAR parameters.

Depending on the affected cardiac structure, radiation-induced cardiac toxicity can manifest as a spectrum of clinical entities, including coronary artery disease, cardiomyopathy, pericardial disease, valvular heart disease, and rhythm disturbances [29]. Although the pathophysiology of radiation-induced cardiovascular toxicity is not fully understood, macrovascular and microvascular injuries are thought to be driven in part by endothelial damage, leading to inflammatory cell infiltration, cytokine release, and subsequent atherosclerotic changes [30]. Radiation may also directly affect cardiomyocytes by increasing reactive oxygen species, which can impair mitochondrial function and contribute to myocardial fibrosis and dysfunction [30].

A landmark study by Darby et al. [9] reported that MHD was associated with an increased rate of major coronary events (7.4% per Gy) without an apparent threshold, whereas mean LAD dose was not after accounting for MHD. However, because the cohort predated routine three-dimensional CT-based planning, cardiac and LAD doses were estimated by reconstructing treatment plans on a reference CT dataset using archived treatment records. Given the potential uncertainty of this approach for small structures such as the LAD, the interpretation of LAD-specific findings from that study may be limited.

Accumulating evidence suggests that doses to specific cardiac substructures may be associated with adverse cardiac outcomes. Although prospective dose-response data remain lacking, based on the available literature [11,12,24-26], the DEGRO breast cancer expert panel [16] proposed dose constraints for the LV and LAD in left-sided breast RT, as summarized in Table 1. The DEGRO dose constraints of LAD V_30Gy <2% and LAD V_40Gy <1% were not included in the present study because the protocol-derived LAD D_max <17 Gy constraint from the DBCG HYPO trial [17] was applied.

The evidence underlying these dose constraints spans early functional and imaging measures as well as later clinical outcomes following RT for breast cancer. For example, using strain rate imaging, which allows the detection of subtle early changes in regional myocardial function, a Doppler echocardiographic study reported that LV segments exposed to >3 Gy showed reductions in regional LV strain at short-term follow-up after RT [24]. A single-photon emission computed tomography–based study also reported higher rates of new LV perfusion defects when ≥5% of the LV received ≥23–25 Gy [25]. With respect to later clinical coronary outcomes, a CT-based dose-effect modeling study identified low-dose LV exposure (LV V_5Gy) as a key predictor of acute coronary events (ACEs) within 9 years after RT, supporting attention to LV low-dose metrics in addition to MHD [11].

Regarding LAD, evidence indicating specific numerical cut-offs is less direct. Coronary imaging studies have shown that stenotic lesions after breast irradiation tend to cluster in LAD “hotspot” segments (i.e., areas within or near the tangential field) [12]. Segment-level dose-stenosis associations have also been reported after RT for mediastinal Hodgkin lymphoma [26]. In addition, among women who underwent coronary angiography for suspected coronary events after three-dimensional conformal radiation therapy (3DCRT) for breast cancer, mid LAD D_mean >20 Gy was associated with higher odds of coronary intervention at that specific location compared with a D_mean of 0–1 Gy (adjusted odds ratio, 5.23) [31].

Taken together, these findings support the association between radiation exposure and location of subsequent coronary stenoses. Furthermore, among breast cancer patients with pre-existing atherosclerotic plaques in the LAD, the mean dose to the plaques was reported to be the strongest predictor of ACEs after 3DCRT, even after adjusting for cardiovascular risk factors (hazard ratio, 1.269 per Gy) [32]. Therefore, plaque-containing coronary segments may warrant particular attention in treatment planning.

In the present study, the heart + LV + LAD strategy achieved significantly more sparing of LV and LAD than the heart-only strategy (Table 4, Fig. 3A–3C). Among the adopted dose constraints (Table 1), LV V_5Gy, LV V_23Gy, and LAD D_mean constraints were met in all heart + LV + LAD plans, while LV D_mean and LAD D_max constraints were achieved in 70.0% and 63.3% of heart + LV + LAD plans, respectively. Although direct comparisons are limited by differences in techniques, cohorts, and reported parameters, LV and LAD dose parameters reported in prior planning studies incorporating cardiac substructures [20,21] were generally higher than those achieved in the current study and frequently did not meet the DEGRO or DBCG HYPO dose constraints.

The cohort-average LV D_mean decreased from 5.472 Gy to 2.843 Gy, falling below the DEGRO-recommended constraint of <3 Gy, which was informed in part by reports of early regional LV strain impairment in LV segments receiving >3 Gy [16,24]. The cohort-average LV V_5Gy decreased markedly from 39.584% to 3.025%, well below the DEGRO-recommended constraint of <17% [16]. This change is notable because van den Bogaard et al. [11] reported a mean LV V_5Gy of 29.3% in patients who experienced ACEs, compared with 16.9% in those who did not. Their normal tissue complication probability model for ACEs included LV V_5Gy, age, and a weighted ACE risk score based on diabetes, hypertension, and ischemic cardiac events [11]. For the LAD, the cohort-average D_mean and D_max also decreased, falling below the DEGRO- and DBCG HYPO-derived constraints of <10 Gy and <17 Gy, respectively [16,17].

Notably, the incorporation of LV and LAD dose constraints resulted in statistically significant compromises in some target metrics (Table 3, Fig. 2A). Although the heart + LV + LAD plans showed slightly higher HI and lower QOC and CI than the heart-only plans, no major deviations were observed in any of the indices according to the RTOG protocol compliance/deviation criteria [27,28]. The American Society for Radiation Oncology guideline for WBI recommends covering at least 95% of the whole-breast volume with 95% of the prescription dose while minimizing high-dose regions (e.g., >105% of the prescription dose) [6]. In the present cohort, despite statistically significant shifts in some DVH parameters for the PTV and boost volumes, the absolute differences were small, and both strategies achieved high target coverage (e.g., PTV V_95% [heart + LV + LAD vs. heart-only]: 98.391 ± 1.118% vs. 99.431 ± 0.841%; p < 0.001) with limited hot spots (e.g., boost D_2%: 54.312 ± 0.565 Gy [106.1%] vs. 54.011 ± 0.612 Gy [105.5%]; p = 0.001). Overall, the target trade-offs observed with the heart + LV + LAD strategy were statistically detectable but modest in magnitude and remained within clinically acceptable ranges.

In terms of non-cardiac OARs, the heart + LV + LAD strategy resulted in reduced ipsilateral lung D_mean and V_5Gy but increased contralateral lung V_10Gy and contralateral breast D_mean (Table 4, Fig. 2B). A similar pattern of decreased ipsilateral lung V_20Gy and increased contralateral breast D_mean in LAD- and LV-sparing plans was reported in a previous study [21]. Although the absolute increases were small, contralateral exposure warrants attention under the ALARA principle, particularly because radiation-induced cancer risk is commonly classified as a stochastic effect and is modeled using a linear no-threshold assumption [33]. Observational studies have reported dose-dependent risks of second primary cancers after breast irradiation, including contralateral breast cancer [34] and second primary lung cancer [35], although these risk estimates should be interpreted cautiously because they vary across studies and appear to be modified by patient- and treatment-related factors such as age [34], hormone therapy use [36], and smoking history [10,35]. Accordingly, when applying LV and LAD dose constraints, planning strategies to mitigate contralateral low-dose spills should be considered, and the balance between cardiac substructure sparing and contralateral low-dose exposure should be individualized with consideration of factors such as baseline cardiovascular risk and age.

In exploratory analyses, LAD-PTV distance appeared to influence the achievable degree of LAD sparing with the heart + LV + LAD strategy and may partly explain the observed trade-offs in target coverage. In the heart + LV + LAD plans, a greater LAD-PTV distance was correlated with lower LAD D_max and D_mean and higher PTV D_98%. These findings suggest that when the LAD lies close to the target volume, attempts to limit LAD hotspots may be constrained and may entail some compromise in target coverage. Therefore, patients with short LAD-PTV distances may warrant heightened planning attention and consideration for additional heart-sparing approaches. For example, deep inspiration breath-hold (DIBH), which has been shown to reduce cardiac and LAD exposure in breast RT [37], may be particularly beneficial for patients with shorter LAD-PTV distances by creating a greater separation between the LAD and PTV during breath-holding. However, the potential role of DIBH in this setting could not be directly evaluated in the current study because all plans were generated on free-breathing CT.

This study has some limitations. First, respiratory motion and setup variability may affect the position of the heart and cardiac substructures during treatment. These uncertainties were not explicitly accounted for in the planning process. Second, the potential inter- and intra-observer variability in LAD delineation should be acknowledged. Prior studies have shown substantial inter-observer variation in LAD delineation and estimated LAD dose [38]. In the present study, the LV and LAD were delineated by a single radiation oncologist using consensus guidelines [23] to maintain internal consistency. However, observer-related variability could not be excluded. Third, this was a single-institution study with a relatively small sample size, which may limit the generalizability of the findings. Finally, the lack of clinical cardiovascular outcome data is an inherent limitation of this dosimetric planning study.

In conclusion, applying LV and LAD dose constraints in left-sided breast VMAT substantially reduced LV and LAD doses, and most cardiac substructure dose constraints based on the DEGRO recommendations and the DBCG HYPO trial were achievable with small but statistically significant trade-offs in some target metrics and contralateral OAR parameters. In addition, a shorter LAD-PTV distance may limit LAD sparing and make target coverage more difficult to maintain, underscoring the importance of considering LAD-PTV distance during planning. Ultimately, the balance between cardiac substructure protection, target coverage, and non-cardiac OAR sparing should be individualized, weighing recurrence risk against baseline cardiovascular risk and the potential for radiation-induced cardiotoxicity.

Supplementary Materials

Supplementary materials can be found via https://doi.org/10.3857/roj.2026.00437.

Supplementary Table S1.

Optimization objectives used in the heart + LV + LAD planning template