Clinical impact of novel cardiovascular and immune system dosimetric parameters in stage III NSCLC patients treated with radiotherapy

Abstract

Purpose

To evaluate the clinical impact of the dose to structures not routinely constrained in the radiation treatment (RT) of stage III NSCLC, such as the left anterior descending artery (LAD) and the estimated dose to the immune cells (EDIC).

Materials/Methods

The dataset of all consecutive patients affected by stage III NSCLC and treated with definitive RT between 2012 and 2022 at our Institution were retrieved in this IRB approved retrospective study. The LAD was manually contoured following published guidelines. EDIC was calculated according to the model by Jin. Overall Survival (OS) was the primary endpoint of the study. Kaplan-Meier and Cox logistic regression statistics were used. Receiver operating characteristic curves evaluated cutoff DVH values.

Results

108 patients were analyzed with a median follow-up of 52 months (95% CI 27.5–76.5). Median RT dose was 60 Gy (54–69.3 Gy). Median OS was 29 months (95% CI 10.2–47.8). On multivariate analysis, when evaluating the EDIC as a continuous variable, EDIC (HR 1.4, 95% CI 1.08–1.8, P = 0.009) and mean heart dose (MHD) < 10 Gy (HR 1.8, 95% CI 1.0-3.2, P = 0.03) were significantly associated with OS. When the EDIC was evaluated as a categorical variable, EDIC ≥ 5 Gy (HR 2.2, 95% CI 1.2–4.1, P = 0.008), MHD < 10 (HR 1.9, 95% CI 1.1–3.2, P = 0.01) and LAD V15 < 47.5% (HR 0.5, 95% CI 0.2–0.9, P = 0.02) were significantly associated with OS. PTV was the only parameter significantly associated with EDIC > 5 Gy (HR 16,6 95% CI 4.5–61.3, P < 0.001), with a cutoff value of 457.1 cc.

Conclusions

The relevant impact of the EDIC on the outcome of stage III NSCLC patients treated with RT is confirmed. The EDIC threshold of 5 Gy here reported merits prospective validation. Our results also reinforced the importance of LAD contouring and optimization in the clinical practice of RT for stage III NSCLC patients.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13014-026-02835-8.

Introduction

Lung cancer was the leading cancer worldwide in terms of incidence and mortality in 2022, with 2.5 million new cases representing 12.4% of all cancer cases and more than 1.8 million cancer-related deaths [1].

Non-small cell lung cancer (NSCLC), which accounts for more than 85% of lung cancer subtypes, is often diagnosed as locally advanced stage III or metastatic stage IV disease [2].

Stage III NSCLC comprises a heterogeneous group of diseases varying in terms of tumor size and nodal involvement, for which, in recent years, a wide variety of integrated treatment options have emerged [3–5].

The role of radiotherapy (RT) in the context of unresectable, locally advanced NSCLC was reinforced by the landmark phase III randomized Pacific trial, which compared conventional radio-chemotherapy (CRT) with the same treatment scheme followed by immunotherapy (IT) with the anti-programmed death ligand 1 antibody durvalumab; the results of the study among more than 700 participants revealed that the experimental treatment significantly prolonged overall survival (OS) and progression-free survival (PFS) compared with conventional CRT, and set CRT followed by IT as the new gold standard for stage III unresectable disease [4].

Conversely, the RTOG phase III trial 0617 failed to demonstrate a clinical benefit of radiation dose escalation (60 Gy vs. 74 Gy) in the setting of stage III unresectable NSCLC. The study enrolled 544 patients in a two-by-two factorial design, with radiation therapy dose as one treatment factor and cetuximab as the other. OS was significantly prolonged in the standard dose arm, with no benefit of the addition of cetuximab to concurrent CRT. These results established that 60 Gy in 30 fractions, which is unlikely adequate for eradicating macroscopic cancer, should remain the standard dose for the radical treatment of stage III NSCLC [6].

Excess normal tissue exposure, which ultimately limits the therapeutic index of RT in this setting, has been reported to explain these negative outcomes [7, 8]. Interestingly, two separate post hoc secondary analyses of the RTOG 0617 trial reported a significant association between new distinct dosimetric parameters and clinical outcomes. A recent study by Mckenzie et al. revealed a significant relationship between the dose to the left anterior descending (LAD) coronary artery, a structure that has rarely been considered and even contoured for RT planning in NSCLC in clinical practice, and all-cause mortality in the 0617 clinical database [9]. Moreover, the pioneering analysis by Jin et al. on the RTOG 0617 trial dataset contributed to improving the knowledge of the relationship between RT and the immune system [10].

In their work, the authors developed a model to quantify the effective dose to immune cells in circulating blood (EDIC), which is considered a new organ at risk (OAR), and reported a significant association between EDIC and survival in the 0617 trial population. These findings could be of the utmost importance in light of the remarkable positive clinical outcomes associated with the use of IT in treating NSCLC.

However, the impact and management of these new dosimetric parameters in clinical practice should be determined before their implementation, considering that incorporating the LAD in the NSCLC RT planning process represents a time-consuming operation (even when autocontouring software capable of LAD contouring, as well as autoplanning systems, are available) and that implementing EDIC management in the daily clinical routine could be even more challenging.

In light of this, the aim of the present work was to analyze the impact of these new dosimetric parameters on the clinical outcome of stage III NSCLC patients treated with RT in clinical practice.

Materials and methods

Dataset

The Institutional Ethics Committee approved this retrospective study (June, 26th, 2024, protocol number 8959-RS76/IRE/24). The dataset of all consecutive patients affected by stage III unresectable NSCLC (staged according to the American Joint Committee on Cancer’s Staging Manual, seventh edition for patients treated before 2017 and eighth edition for patients treated thereafter), evaluated by the internal multidisciplinary tumor board and treated with RT with definitive intent at our institution between 2012 and 2022 was retrieved. Patient demographic, clinical, and treatment-related data were retrieved, along with toxicity and outcome data recorded during regular institutional follow-up visits.

Treatment and OAR volumes

The original treatment plan and OAR volumes were considered for the analysis, with the exception of the LAD, which was not included in the original dataset and was de novo manually contoured for all patients following published guidelines [11]. Common dose objectives that remained unchanged during the study period were a planning target volume (PTV) D95 ≥ 95%, a PTV D95 ≥ 95%, a mean whole-lung (minus GTV) dose < 20 Gy, a v20 < 35%, a mean heart dose (MHD) < 20 Gy, a mean esophageal dose < 34 Gy, and a maximum spinal cord dose < 45 Gy.

EDIC calculation

EDIC was calculated according to the model developed by Jin et al. [10] which considers the patient’s Integral dose, the mean heart dose (MHD) and the mean lung dose (MLD) according to the following equation:

A minimum of 25 treatment fractions were required for the EDIC calculation.

Endpoints and statistical analysis

OS was the primary endpoint of the study. Secondary endpoints included local progression-free survival (LPFS), defined as the time to progression at the original tumor site, and PFS, defined as the time to progression at any site. The time-to-event for each endpoint was calculated from the beginning of RT.

The data are summarized using basic descriptive statistics. The chi-square test or Fisher’s exact test was used to compare categorical variables when appropriate.

Hazard ratios (HRs) with 95% confidence intervals (95% CIs) were estimated for each variable of interest, including the EDIC, which was evaluated in separate models as a continuous and a categorical variable. A P value < 0.05 was considered significant. Only variables that were significant in the univariate analysis were included in the multivariate analysis. Multivariate logistic regression and proportional hazard models were developed using stepwise regression (forward selection, enter limit and remove limit, p = 0.10 and p = 0.15, respectively) to identify independent predictors of OS, LPFS and PFS. Variables with potential collinearity with the EDIC, such as the MLD and MHD, were evaluated using variance inflation factors (VIFs).

In the case of quantitative variables, to determine the best cutoff for dividing patients into two subgroups with different outcomes, the analysis was carried out using the maximally selected log-rank statistics method and receiving operator curve (ROC) analysis. Survival curves were generated by the Kaplan–Meier method. The log-rank test was used to evaluate the differences between curves. Statistical evaluation was performed using SPSS 29.0 (SPSS Inc., Chicago, IL, USA) for Windows, MedCalc V 20.027 (Ostend, Belgium) and R version 4.2.2 (Vienna, Austria).

Results

Patient characteristics

A total of 108 patients were identified. The patients’ clinical characteristics are summarized in Table 1. As expected, most patients (65%) were men; 55% of patients were designated as having stage IIIB disease. Tumor location was well distributed between the lobes, with the upper right lobe and lower left lobe representing the most frequent locations. Adenocarcinoma was the most diagnosed cancer subtype (52.7%), and the vast majority of patients (93%) were treated with integrated chemoradiation: concurrent and sequential schedules were well distributed with a slight preference (60%) for the latter. Consolidation IT with durvalumab was administered to 24 (22.2%) patients.

Table 1.

Patients’ and treatment characteristics

Variable	N (range,%)
Median Age (range)	68(46/85)
Sex
male	71 (65.7)
female	37 (34.3)
PS ECOG
0	62 (57.4)
1	39 (36.1)
2	5 (4.6)
Smoking History
unknown	2 (1.8)
Never	3 (2.7)
Former	48 (44.4)
Current	55 (50.9)
Tumor Location
RUL	42 (38.8)
ML	3 (2.7)
RIL	13 (12)
LUL	35 (32.4)
LIL	15 (13.8)
AJCC Stage
IIIA	45 (41.6)
IIIB	60 (55.5)
IIIC	3 (2.7)
Histology
Adenocarcinoma	57 (52.7)
Squamous cell carcinoma	37 (34.2)
Other NSCLC	14 (12.9)
PD-L1 status
< 1%	17 (15.7)
>=1%<50%	33 (30.5)
>=50%	16 (14.8)
unknown	42 (38.8)
Chemotherapy Modality
no chemotherapy	3 (2.7)
Concomitant	42 (38.8)
Sequential	63 (58.3)
Chemotherapy Regimen
Carboplatin + Paclitaxel	37 (34.3)
Other	68 (62.9)

Abbreviations: RUL: right upper lobe, ML: medium lobe, RIL: right inferior lobe, LUL: left upper lobe, LIL: left inferior lobe

Radiation treatment characteristics

Radiation treatment details are described in Table 2. The median prescription dose was 60 Gy (range 54–69.3 Gy), which was delivered in most cases (92%) with intensity-modulated techniques (IMRT-VMAT). The median planning target volume (PTV) was approximately 400 cc.

Table 2.

Radiotherapy treatment details

Variable	N (range)
Median RT dose	60 Gy (55–69,3)
RT Technique
IMRT/VMAT 3DCRT	100 8
GTV	71,1 cc (8–321,9)
CTV	201,4 cc (19–762)
PTV	397,1 cc (111,5–1213,5)
Median MLD	13,59 Gy (1,57 − 20,50)
Median MHD	7,31 Gy (0–33,61)
Median MED	21,45 Gy (0,11–54,60)
Median Mean LAD dose	6,88 Gy (0,39–79,32)
Median LAD V15	9% (0-100)
Median EDIC	4,01 Gy (1.91–7.45)

Abbreviations: MLD: mean lung dose, MHD, mean heart dose, MED: mean esophagus dose, EDIC: estimated dose to immune cells

OAR doses

Table 2 reports the main dosimetric data for the OARs of interest in the present study. The median mean heart (MHD) and lung (MLD) doses were 7.31 Gy (range, 0–33.61) and 13.59 Gy (range, 1.57–20.5), respectively. The median LAD V15 was 9% (range, 0–100). The median EDIC was 4.01 Gy (range, 1.91–7.45).

Outcome analysis

The median follow-up was 52 months (95% CI 27.5–76.5). At the time of last follow-up, 67 patients had died. In the entire population, the median and 2-year OS were 29 months (95% CI 10.2–47.8) and 56.2%, respectively (supplementary material). The VIF analysis revealed low values for the MHD and MLD, with consequently low risk of multicollinearity; therefore, these variables were not omitted from the multivariate analysis. The results of the OS univariate and multivariate analyses are reported in Table 3. On univariate analysis, both the EDIC (both continuous and categorical) and LAD V15 were associated with better OS. ROC curve analysis revealed that the best cutoff for the LAD V15 was 47.5% in our series. In the model accounting for the EDIC as a continuous variable, the EDIC (HR 1.4, 95% CI 1.08–1.8; P = 0.009) and an MHD < 10 Gy (HR 1.8, 95% CI 1.0–3.2; P = 0.03) were independent predictors of OS. When the EDIC was evaluated as a categorical variable, an EDIC ≥ 5 Gy (HR 2.2, 95% CI 1.2–4.1; P = 0.008), an MHD < 10 (HR 1.9, 95% CI 1.1–3.2; P = 0.01) and an LAD V15 < 47.5% (HR 0.5, 95% CI 0.2–0.9; P = 0.02) were significantly associated with OS according to the multivariate analysis. The median OS was 51 months (95% CI 24.9–77.1) vs. 10 months (95% CI 2.2–17.8; P = 0.0001) for patients with an EDIC < 5 Gy vs. ≥ 5 Gy, respectively (Fig. 1).

Table 3.

Univariate and multivariate cox regression analysis for OS

Variable	Univariate		Multivariate (EDIC as continous)		Multivariate (EDIC as categorical)
	HR (95%CI)	P-value	HR (95%CI)	P-value	HR (95%CI)	P-value
Age	1.0 (0.9-1.03)	0.7	-	-	-	-
Sex
Female	Ref		-	-	-	-
Male	1.1 (0.6-2.1)	0.7	-	-	-	-
Smoking History
Never	Ref
Former	2.96 (0.4-21.8)	0.2	-	-	-	-
Current	2.9 (0.39-21.2)	0.2	-	-	-	-
PS
0-1	Ref
2	1.2 (0.7-19)	0.4	-	-	-	-
Histology
Squamous	Ref
Adenocarcinoma	1.1 (0.6-1.9)	0.6	-	-	-	-
Other	1.3 (0.6-2.6)	0.4	-	-	-	-
Tumor location
Upper lobes	Ref
other	1.0 (0.6-1.7)	0.8	-	-	-	-
Stage
IIIA	Ref
IIIB-IIIC	1.2(0.7-2.1)	0.3	-	-	-	-
Chemotherapy
Carboplatin-paclitaxel	Ref
Other	1.6 (0.9-2.6)	0.08	-	-	-	-
Chemotherapy scheme
Concomitant	Ref
Sequential	1.1 (0.6-1.9)	0.6	-	-	-	-
RT Technique
IMRT-VMAT	Ref
3DCRT	1.3 (0.6-2.9)	0.4	-	-	-	-
CTV	1 (0.9-1)	0.7	-	-	-	-
PTV	1 (1-1.002)	0.01	-	-	-	-
Lung V20	1.05 (1.01-1.09)	0.001	-	-	-	-
MLD	1.1 (1.04-1.2)	0.003	-	-	-	-
MHD	1.07 (1.03-1.1)	0.0001	-	-	-	-
MHD <10 Gy	2.6 (1.6-4.3)	0.0001	1.8-1.0-3.2	0.03	1.9 (1.1-3.2)	0.01
LAD V15	1.01 (1.0-1.02)	0.003	-	-	-	-
LAD V15 <10%	1.3 (0.8-2.2)	0.1	-	-	-	-
LAD V15 <47.5	1.01 (1.0-1.02)	0.003	-	-	0.5 (0.2-0.9)	0.02
EDIC (continuous)	1.6 (1.2-1.9)	0.0001	1.4 (1.08-1.8)	0.009	-	-
EDIC (categorical)
EDIC ≥ 3 Gy	2.1 (1.0-4.6)	0.03	-	-	-	-
EDIC ≥ 4 Gy	2.5 (1.5-4.1)	0.001	-	-	-	-
EDIC ≥ 5 Gy	3.4 (1.9-6)	0.001	-	-	2.2 (1.2-4.1)	0.008
EDIC ≥ 6 Gy	1.4 (0.4-4.7)	0.5	-	-	-	-
Esophageal Toxicity
< G2	Ref
≥ G2	1.05 (0.6-1.7)	0.8	-	-	-	-

Abbreviations: PS: performance status, RT: radiotherapy, IMRT: intensity modulated radiotherapya, VMAT: volumetric arc therapy, 3DCRT: conformal radiotherapy, MLD: meal lung dose, MHD, mean heart dose, LAD: left anterior descending coronary artery, EDIC: estimated dose to the immune cells

Fig. 1.

Kaplan-Meier Overall survival curves according to EDIC > vs. < = 5 Gy

Potential predictors of an EDIC > 5 Gy were investigated among clinical and treatment variables such as the tumor location, tumor stage, T and N classification, RT technique, PTV, and lung and heart volumes. In this context, the PTV was the only parameter significantly associated with an EDIC > 5 Gy (HR 16.6, 95% CI 4.5–61.3; P < 0.001). The cutoff PTV for patients with an EDIC > 5 was 457.1 cc (Fig. 2).

Fig. 2.

Receiver operating characteristic curve for EDIC > 5 Gy as a function of PTV

The median OS was 48 months (95% CI 27.4–68.6) vs. 8 months (95% CI 1.1–9; P = 0.0001) for patients with an LAD V15 < 47.5% vs. ≥ 47.5%, respectively (Fig. 3).

Fig. 3.

Kaplan-Meier Overall survival curves according to LAD V15 > vs. < = 47.5%

Among the secondary endpoints, the median and 2-year PFS were 13 months (95% CI 6.2–19.8) and 37.1%, respectively (supplementary material). The results of the PFS univariate and multivariate analyses are reported in Table 4. The EDIC (continuous and categorical) and LAD V15 were associated with better PFS according to the univariate analysis, whereas the only variables associated with PFS according to the multivariate analysis were the MHD (HR 1.06, 95% CI 1.02–1.1; P = 0.003) and PTV (HR 1.02, 95% CI 1.01–1.03; P = 0.006). The median PFS was 20 months (95% CI 13.8–26.2) vs. 7 months (95% CI 4.5–9.5; P = 0.001) with an MHD < 10 Gy vs. ≥ 10 Gy (supplementary material).

Table 4.

Univariate and multivariate cox regression analysis for PFS

Variable	Univariate		Multivariate (EDIC as continous)		Multivariate (EDIC as categorical)
	HR (95%CI)	P-value	HR (95%CI)	P-value	HR (95%CI)	P-value
Age	0.9 (0.9-1.03)	0.9	-	-	-	-
Sex
Female	Ref		-	-	-	-
Male	1.0 (0.6-1.8)	0.7	-	-	-	-
Smoking History
Never	Ref
Former	1.0 (0.2-4.5)	0.9	-	-	-	-
Current	1.0 (0.2-4.3)	0.9	-	-	-	-
PS
0-1	Ref
2	1.05 (0.6-1.7)	0.8	-	-	-	-
Histology
Squamous	Ref
Adenocarcinoma	1.1 (0.7-1.9)	0.6	-	-	-	-
Other	0.6 (0.3-1.5)	0.3	-	-	-	-
Tumor location
Upper lobes	Ref
other	1.0 (0.6-1.6)	0.9	-	-	-	-
Stage
IIIA	Ref
IIIB-IIIC	1.3(0.8-2.2)	0.2	-	-	-	-
Chemotherapy
Carboplatin-	Ref
paclitaxel Other	1.6 (0.9-2.7)	0.053	-	-	-	-
Chemotherapy scheme
Concomitant	Ref
Sequential	1.1 (0.7-1.9)	0.5	-	-	-	-
RT Technique
IMRT-VMAT	Ref
3DCRT	1.0 (0.4-2.5)	0.9	-	-	-	-
CTV	1 (1.0-1.003)	0.04	-	-	-	-
PTV	1 (1.001-1.002)	0.001	1.02 (1.01-1.003)	0.006	-	-
Lung V20	1.02 (0.9-1.06)	0.2	-	-	-	-
MLD	1.1 (1.02-1.2)	0.01	-	-	-	-
MHD	1.07 (1.03-1.1)	0.0001	-	-	-	-
MHD <10 Gy	2.1 (1.3-3.5)	0.002	1.06 (1.0.2-1.1)	0.003
LAD V15	1.01 (1.0-1.02)	0.01	-	-	-	-
LAD V15 <10%	1.1 (0.7-1.9)	0.4	-	-	-	-
LAD V15 <47.5	2.4 (1.4-4.4)	0.002	-	-	-	-
EDIC (continuous)	1.5 (1.1-1.9)	0.001	-	-	-	-
EDIC (categorical)
EDIC ≥ 3 Gy	2.4 (1.1-5.1)	0.01	-	-	-	-
EDIC ≥ 4 Gy	1.7 (1.0-2.7)	0.02	-	-	-	-
EDIC ≥ 5 Gy	2.8 (1.6-5.1)	0.001	-	-	-	-
EDIC ≥ 6 Gy	4.4 (1.5-12.3)	0.005	-	-	-	-
Esophageal Toxicity
< G2	Ref
≥ G2	1.1 (0.7-1.8)	0.5	-	-	-	-

Abbreviations: PS: performance status, RT: radiotherapy, IMRT: intensity modulated radiotherapya, VMAT: volumetric arc therapy, 3DCRT: conformal radiotherapy, MLD: meal lung dose, MHD, mean heart dose, LAD: left anterior descending coronary artery, EDIC: estimated dose to the immune cells

The median and 2-year LPFS were 40 months (95% CI 12–68) and 55.4%, respectively (supplementary material). Univariate and multivariate analyses of LPFS are reported in Table 5. On univariate analysis, the EDIC (continuous and categorical), the LAD V15 and an MHD < 10 Gy were associated with better LPFS. Multivariate analysis revealed that an MHD < 10 Gy was associated with better LPFS (HR 2, 95% CI 1.1–3.7; P = 0.01). An EDIC > 3 Gy was marginally significant in the multivariate analysis (supplementary material). The median LPFS was 56 months (95% CI 5.5–106.4) vs. 10 months (95% CI 2.2–21.8; P = 0.001) with an MHD < 10 Gy vs. ≥ 10 Gy (supplementary material). All outcome predictors that were identified remained unchanged when adjusted for the use of consolidation IT.

Table 5.

Univariate and multivariate cox regression analysis for LPFS

Variable	Univariate		Multivariate (EDIC as continous)		Multivariate (EDIC as categorical)
	HR (95%CI)	P-value	HR (95%CI)	P-value	HR (95%CI)	P-value
Age	0.9 (0.9-1.01)	0.2	-	-	-	-
Sex
Female	Ref		-	-	-	-
Male	1.1 (0.6-2.1)	0.6	-	-	-	-
Smoking History
Never	Ref
Former	2.1 (0.2-15)	0.4	-	-	-	-
Current	2.3 (0.3-17.2)	0.4	-	-	-	-
PS
0-1	Ref
2	1.02 (0.5-1.8)	0.9	-	-	-	-
Histology
Squamous	Ref
Adenocarcinoma	1.6 (0.8-2.9)	0.1	-	-	-	-
Other	1.2 (0.4-2.8)	0.7	-	-	-	-
Tumor location
Upper lobes	Ref
other	1.0 (0.5-1.8)	0.8	-	-	-	-
Stage
IIIA	Ref
IIIB-IIIC	1.03 (0.5-1.8)	0.9	-	-	-	-
Chemotherapy
Carboplatin- paclitaxel	Ref
Other	1.5 (0.8-2.7)	0.1	-	-	-	-
Chemotherapy scheme
Concomitant	Ref
Sequential	1.07 (0.5-1.9)	0.8	-	-	-	-
RT Technique
IMRT-VMAT	Ref
3DCRT	1.5 (0.6-3.9)	0.3	-	-	-	-
CTV	1 (0.9-1)	0.6	-	-	-	-
PTV	1 (1-1.002)	0.1	-	-	-	-
Lung V20	1.01 (0.9-1.05)	0.5	-	-	-	-
MLD	1.0 (0.9-1.1)	0.2	-	-	-	-
MHD	1.07 (1.02-1.1)	0.001	-	-	-	-
MHD <10 Gy	2.4 (1.3-4.3)	0.002	2.5 (1.4-4.5)	0.001	2 (1.1-3.7)	0.01
LAD V15	1.01 (1.0-1.02)	0.03	-	-	-	-
LAD V15 <10%	1.2 (0.6-2.1)	0.5	-	-	-	-
LAD V15 <47.5	2.1 (1.1-4.3)	0.02
EDIC (continuous)	1.3 (1.0-1.7)	0.01			-	-
EDIC (categorical)
EDIC ≥ 3 Gy	3.6 (1.3-10.2)	0.01	-	-	0.3 (0.1-	0.059
EDIC ≥ 4 Gy	1.1 (0.6-2.0)	0.6	-	-	1.04)	-
EDIC ≥ 5 Gy	2.3 (1.1-4.6)	0.01	-	-	-	-
EDIC ≥ 6 Gy	2.2 (0.6-7.2)	0.1	-	-	- -	-
Esophageal Toxicity
< G2	Ref
≥ G2	1.5 (0.8-2.8)	0.1	-	-	-	-

Abbreviations: PS: performance status, RT: radiotherapy, IMRT: intensity modulated radiotherapya, VMAT: volumetric arc therapy, 3DCRT: conformal radiotherapy, MLD: meal lung dose, MHD, mean heart dose, LAD: left anterior descending coronary artery, EDIC: estimated dose to the immune cells

Discussion

Our study contributes to improving the understanding the effect of dose on structures still not routinely considered in RT for stage III unresectable NSCLC, such as the immune system and the LAD. Our results on an unselected, consecutive dataset of patients treated at a single institution, strengthen the clinical significance of assessing and limiting the radiation dose to such structures.

The relationship between the dose to the immune system and the treatment outcome in stage III unresectable NSCLC patients represents an emerging research area since the initial observation, in a secondary analysis of RTOG 0617, that radiation-induced immunosuppression, rather than pulmonary and cardiac toxicity, could explain the significantly lower OS registered in the 74 Gy high dose arm of the trial [10].

To our knowledge, few studies have evaluated the impact of the EDIC in NSCLC patients treated with radiotherapy [10, 12–14].

All the studies reported a significant association between a greater EDIC and OS and LPFS (when evaluated), whereas some differences were registered in the association between the EDIC and PFS. PFS was significantly associated with the EDIC in the studies by Mccall [12] and Ladbury [13], whereas it was not in the original study by Jin [10]. In the study by Friedes the EDIC and PFS were significantly correlated only in the statistical model considering the EDIC as a continuous variable (and not when evaluating the EDIC as a categorical variable).

The results of our study confirm the strong association between the EDIC and OS, which was the primary endpoint of our work; moreover, the EDIC cutoff value of 5 Gy that we identified in our study is in the range of that reported in the original work by Jin (6 Gy) and in the other more recent studies by Mccall (6 Gy) and Friedes (4.7 Gy). In our work a nonsignificant trend toward better LPFS with a lower EDIC and no clear improvement in PFS was observed. The differences in these outcomes between studies could have different explanations, such as their being less precisely assessable parameters compared with OS, especially in retrospective studies, differences in statistical modeling (i.e., evaluating the EDIC as a continuous or categorical variable, or both), slight differences in EDIC model parameters between the Ladbury and other studies (including the present work), and the heterogeneity of the patient population due to the long observation period reported. Notably, PFS, LPFS and the EDIC were significantly correlated in the study by Mccall, who specifically evaluated the impact of the EDIC in a population of NSCLC patients treated with consolidation durvalumab according to the Pacific scheme. An excessive dose to circulating immune cells that decreased the benefit of durvalumab could explain these findings and corroborate the belief that a strong interaction exists between radiotherapy and the immune system [15].

Currently, with the continuous emergence of new studies defining the role of neoadjuvant perioperative immunotherapies for locally advanced resectable or even borderline resectable nonmutated NSCLC [3, 16, 17], the role of RT in this setting is unclear, and the risks need to be considerably reduced in the future. Admittedly, although the use of consolidation durvalumab significantly improved the outcomes of patients with NSCLC treated with integrated chemo-RT [8, 18], further improvement of the outcomes of RT for unresectable patients eligible for IT, not to mention the outcomes of Pdl-1-negative patients or patients with oncogenic-addicted NSCLC, is needed.

In this context, the increasing knowledge of the impact of the EDIC on patient outcomes deserves attention and merits validation in prospective studies, i.e., studies evaluating whether EDIC reduction approaches could revitalize the RT dose-escalation strategy for NSCLC, which has been unsuccessful [6, 19]. Multicenter registries reporting associations of the EDIC with outcomes would also be desirable [20].

In clinical practice, the first steps should include (1) implementation of the EDIC in the routine plan evaluation process for NSCLC patients; (2) assessment of optimization strategies aiming to reduce the EDIC to a defined dose/objective threshold (i.e., 5 Gy as a result of the present work); and (3) evaluation of the advantages and disadvantages of these strategies in terms of plan quality and doses to OARs.

Modern RT could play a key role in this setting. Indeed, we recently reported that compared with conventional X-ray VMAT, the use of active scanning proton therapy significantly halved the EDIC in stage III NSCLC patients [21]. The recent work by Friedes confirmed these results, showing a significant association between the use of PT and a lower EDIC [14]. Our finding that larger PTVs were associated with an EDIC > 5 Gy could aid in the selection of patients eligible for advanced (and costly) treatments such as PT.

Moreover, we hypothesize that the use of modern optimization techniques in conventional X-ray VMAT, such as PTV-less, robust optimization techniques [22] based on true clinical target volumes, could reduce unnecessary doses to OARs with a presumed EDIC reduction aimed at improving oncological patient outcomes; the implementation of this technique for lung cancer represents an area of active research at our institution.

Another aim of our work was to evaluate the effect of the dose to the LAD in our patient population. Among cardiac substructures, we focused on the LAD given the results of the post hoc analysis of the RTOG 0617 trial, which confirmed an LAD V15 < 10% as an independent predictor of OS [7]. The LAD dose-volume threshold of V15 < 10% was indeed first identified by the retrospective analysis by Atkins et al., who established a significant association between the LAD dose and major adverse cardiac events and all-cause mortality in a cohort of 701 patients with locally advanced NSCLC [7].

We were unable to exactly reproduce this finding in our dataset of patients. An LAD V15 < 10% was not associated with better OS in our analysis. However, in the model accounting for the EDIC as a categorical variable, the LAD V15 was significantly associated with better OS; in our dataset, the best cutoff value was a V15 < 47.5%. Patients with a V15 < 47.5% showed considerable improvement in median survival according to the results of the Kaplan‒Meier analysis. The observed variation in the LAD V15 cutoff value could depend on the uncertainties in LAD contouring (especially when performed on planning CT without contrast), which could remain even after the implementation of the cardiac atlas contouring guidelines. In fact, as reported in the original paper by Feng [11], the mean overlap rate with the LAD gold standard contour, although it improved with the use of the cardiac atlas, remained in the range of 60%, which is greater than the 90% achieved by a more standard and easy-to-contour structure such as the whole heart. Autocontouring packages capable of LAD contouring following consensus guidelines could help improve the reproducibility and reliability of delineating this structure among studies and ease the routine implementation of LAD contouring and optimization in the clinical practice of RT for stage III NSCLC.

Among the cardiac dosimetric parameters, in our study, the MHD had a significant effect on outcomes in stage III NSCLC patients. The observation of relevant differences in survival curves of patients who received an MHD < 10 Gy vs. ≥ 10 Gy prompted us to adopt an MHD < 10 Gy as a new soft (to be achieved when feasible without target undercoverage) constraint in RT planning for stage III NSCLC in our clinical practice, as already reported in other studies [23].

Our study involves several limitations related to its retrospective design. The lack of a per-protocol, structured follow-up precluded the retrieval of important toxicity data, such as the dynamic changes in the white cell count (including lymphopenia) and perceived cardiac events occurring during the study period. In this context, more in-depth and separate analyses of recurrence pattern and of the correlation between the EDIC and/or MHD and survival in patients with or without recurrence were beyond the scope of the study and could not be performed.

Moreover, data concerning the detailed causes of death were not available for all patients. However, at the time of last follow-up, 14 patients were dead without certain disease recurrence: EDIC > 5 and MHD remained statistically significant looking at cancer specific survival data (data not shown). In four of these patients, death could be attributed to the worsening of the respiratory function possibly related to RT: interestingly, the EDIC was > 5 in all of these patients.

Other limitations of the study included potential selection bias, variations in staging and treatment during the long observation period and other unknown confounding variables.

The EDIC itself is also affected by general limitations inherent to models.

In conclusion, our study confirms for the first time in Europe the relevant role of the EDIC in the outcomes of stage III NSCLC patients treated with RT. The implementation of routine EDIC assessments in clinical practice in this setting is advised. The EDIC threshold of 5 Gy we report based on our experience merits prospective validation. Our results also reinforce the importance of implementing LAD contouring and optimization in the clinical practice of RT for stage III NSCLC patients.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

F.D., G.M, G.S: conceptualization, data curation, writing, review & editing. I.S. writing, data curation, review and editing; L.G., A.N: Data curation, review & editing P.O, A.S., M.D, data curation, review and editing, G.S conceptualization, writing, review & editing.

Funding

This work was financially supported through funding from the institutional “Ricerca Corrente” granted by the Italian Ministry of Health. The authors have no additional conflicts of interest to disclose. Research data are stored in an institutional repository and will be shared upon request to the corresponding author.

Data availability

Research data are stored in an institutional repository and will be shared upon request to the corresponding author.

Declarations

Ethical approval and consent to participate

The Institutional Ethics Committee approved this retrospective study (June, 26th, 2024, protocol number 8959-RS76/IRE/24). The study was conducted in accordance with the Declaration of Helsinki. Formal informed consent was waived according to the nature of the study.

Consent for publication

Not required.

Competing interests

The authors declare no competing interests.