Safety and Biomarker Analysis of NRG-LU004: Phase I Trial of Accelerated or Conventionally Fractionated Radiotherapy Combined With Durvalumab in PD-L1 High Locally Advanced Non-Small Cell Lung Cancer

Abstract

Background.

Patients with advanced non-small cell lung cancer (NSCLC) and high PD-L1 (>50%) expression show improved outcomes with immune checkpoint inhibitors compared to chemotherapy. We hypothesized that concurrent durvalumab and radiotherapy (RT) without concurrent chemotherapy would be safe and feasible in patients with locally advanced NSCLC (LA-NSCLC).

Methods.

Stage II-III LA-NSCLC patients with PD-L1 >50% received durvalumab (1500 mg Q4 weeks x 13 cycles) with either accelerated fractionated RT (ACRT, 60 Gy/15 fractions) or standard fractionated RT (SDRT, 60 Gy/30 fractions), as initial and expansion cohorts. Primary objective was safety assessed through dose-limiting toxicities (DLTs). Feasibility was defined as ≥80% of patients receiving ≥80% of planned durvalumab in the first eight weeks. Peripheral blood specimens were collected at baseline and throughout treatment for exploratory immune profiling analyses to evaluate biomarkers related to treatment and/or adverse events.

Results.

Among 24 evaluable patients (n=12 each), one DLT occurred (SDRT arm). Both initial cohorts advanced to expansion. Most patients received RT per protocol (83%). At time of analysis, 24% had completed all durvalumab cycles. In ACRT, adverse events included four grade 3, one grade 4 (lymphopenia), and one grade 5 (lung infection, unrelated). In SDRT, there were eight grade 3 and one grade 5 (respiratory failure, unrelated) events reported. Durvalumab feasibility was 85% in ACRT and 75% in SDRT. RT was completed per protocol in all 12 ACRT, and 10 of 12 for SDRT. Soluble proteomic analysis revealed significant immune changes particularly in the SDRT arm and potential biomarkers of immune-related adverse events.

Conclusions.

Chemotherapy-free thoracic RT with either standard fractionated or accelerated courses with concurrent durvalumab is safe in PD-L1 high LA-NSCLC.

Introduction

The current standard of care for unresectable locally advanced non-small cell lung cancer (LA-NSCLC) is concurrent chemoradiation followed by one year of consolidation treatment with anti-PD-L1 agent durvalumab^{1, 2}. This has substantially improved the overall survival (OS) above chemoradiation using conventional radiation dosing and fractionation over 6 weeks. Concurrent chemotherapy and radiotherapy substantially increase esophagus and cardiovascular toxicities, with only modest improvement in OS above sequential chemotherapy and radiation³. Based upon recent data demonstrating that immunotherapy by itself is highly active in patients with advanced NSCLC with PD-L1> 50%⁴, and preclinical data that immunotherapy can enhance the activity of radiation^{5, 6}, we hypothesized that it is plausible to replace chemotherapy with immune checkpoint inhibitors (ICIs) to combine with radiation therapy (RT). The approach has the potential to enhance clinical outcomes without the toxicity of chemoradiotherapy.

Under the same premise that ICIs could be more efficacious than chemotherapy under certain circumstances, studies have been designed to substitute chemotherapy with ICIs in locally advanced NSCLC who are otherwise eligible for chemoradiation. The SPRINT trial⁷ (NCT03523702) is a single institutional study that enrolled patients with locally advanced NSCLC, with differential treatment based on PD-L1 expression. In Tumor Proportion Score (TPS) ≥ 50% tumors, patients were treated with induction pembrolizumab followed by response-determined radiotherapy alone followed by consolidation pembrolizumab. Tumors with TPS <50% were treated with standard chemoradiation. The treatment was well tolerated with excellent 1-year PFS and OS. The DOLPHIN trial (jRCT2080224763) is a Japanese study that enrolled patients with TPS ≥1% and were treated with durvalumab and concurrent standard fractionated (60 Gy in 30 fractions) radiotherapy without chemotherapy⁸. The regimen was well tolerated with excellent 1-year disease free survival (DFS) and OS, comparable to the SPRINT trial. These studies demonstrate the feasibility and preliminary efficacy of substituting concurrent chemotherapy for immunotherapy in combination with radiotherapy alone.

While 60 Gy in 30 fractions is the conventional dosing scheme established since 1980 for locally advanced NSCLC⁹, accelerated course radiotherapy is commonly employed in situations when concurrent chemotherapy is not used, particularly in patients with performance statuses that would not be amenable for chemotherapy^10–12. The biologically effective dose of 60 Gy in 15 fractions is higher than conventional dosing, with the potential that the higher dose per fraction could increase the proinflammatory and antigenic potential to improve better tumor control¹³. Even if it is equivalent, the shorter course of therapy enhances patient convenience and could potentially reduce severity of lymphopenia which is related to the duration of treatment¹⁴. Reduction of lymphopenia may enhance the efficacy of ICIs in this setting¹⁵.

We conducted a phase I study to evaluate the safety and feasibility of these two RT dosing and fractionation schemes in combination with anti-PD-L1 monoclonal antibody durvalumab in tumors with TPS ≥ 50%. We report here the primary safety and feasibility data, as well as biomarker analyses using longitudinally collected blood samples tested for targeted proteomics and cellular composition to correlate with treatment effect and adverse events.

Material and Methods

Study Participants

Eligible patients had pathologically confirmed, unresectable or medically inoperable stage II–III NSCLC (AJCC 8th edition) or locally recurrent N1–N3 disease post-surgery without prior RT. Inclusion required adequate hepatic/renal function, FEV1 ≥ 0.8 L or ≥ 35% predicted, and DLCO ≥ 40%, with or without bronchodilators. Tumors had to be PD-L1 TPS ≥ 50%, assessed via Dako 22C3 or Ventana SP263. All participants were ≥ 18 years old with ECOG 0–1 within 30 days of registration.

Study Design and Procedures

Patients received RT plus durvalumab in four cohorts (n = 6 each): Cohorts 1 and 3: Accelerated RT (ACRT; 4 Gy × 15, total 60 Gy in 3 weeks); Cohorts 2 and 4: Standard RT (SDRT; 2 Gy × 30, total 60 Gy in 6 weeks). Cohorts 1 and 2 served as initial non-randomized safety arms, followed by randomization in Cohorts 3 and 4 if both regimens met safety thresholds. Expansion proceeded only if ≤1 dose-limiting toxicities (DLT) occurred in six patients. DLT was defined as having any of the following safety events within the toxicity evaluation period (12 weeks for ACRT and 8 weeks for SDRT) that are attributed to protocol treatment: 1) grade 4–5 non-hematologic serious adverse events (SAEs); 2) any adverse events (AEs) possibly, probably or definitely related to protocol treatment that led to prolonged dose delays (defined as skipping at least 2 doses of either durvalumab, during the same monitoring period for each arm); 3) permanent discontinuation of durvalumab due to toxicity related to protocol treatment within the first 30 days of starting durvalumab; or 4) SAEs related to RT (including grade 3–4 congestive heart failure, pneumonitis, arrhythmias, and myocardial infarction, grade 4 esophagitis, and any AEs that lead to hospitalization) that causes either an interruption or early termination of RT.

Durvalumab

Durvalumab (1500 mg IV over 60 min) started Day −14 (±48h) pre-RT and continued Q4W for 13 cycles or until progression, toxicity, or death. Infusion vitals were monitored every 30 min. RT interruptions did not alter durvalumab schedule. Delayed doses resumed when criteria were met; missed doses were not made up.

Radiotherapy

RT began within 30 days of registration and 14 days (±48h) after durvalumab initiation. Plans were reviewed pre-treatment. Motion management via 4DCT or equivalent was required. RT was delivered using photon-based 3D-conformal radiation therapy (3D-CRT) or Intensity-Modulated Radiation Therapy (IMRT), with beam energies of 6–10 MV. Daily image guidance and volumetric planning were used. The gross tumor volume (GTV) was adjusted based on the motion management technique used. The clinical tumor volume (CTV) was a 0.5–1 cm expansion from the GTV, and planning tumor volume added 0.5 cm to the CTV. Central review ensured compliance.

Blood Collection

Blood was collected at baseline (Day −14), Day 15, Day 30, and 3-months post-treatment (8.5 mL in ACD-A or sodium heparin tubes). Samples were shipped same-day to the NRG Columbus biobank for processing and storage of plasma and peripheral blood mononuclear cells (PBMCs) for exploratory studies.

Sample Size and Power for Safety Analysis

Each arm included up to 12 evaluable patients. If ≤1/6 patient had a DLT, the arm expanded; safety was confirmed if <4/12 patients experienced a SAE. This design offered ≥77% power to detect a 40% toxicity rate and ≥70% power to confirm safety if toxicity was ≤18%.

Feasibility

A regimen was feasible if ≥10/12 patients received ≥80% of planned durvalumab (≥3000 mg, or ≥2 doses) within the first 8 weeks following the initial dose.

Trial Objectives

The primary objective of this study was to assess safety of adding durvalumab to ACRT or SDRT. Secondary objectives assessed feasibility and toxicities, and progression-free survival (which will be reported separately). Exploratory objectives evaluated immune responses via longitudinal blood sampling and tumor assessments per Immune-Related Response Criteria (irRC).

Biomarker Assays

Quantification of soluble protein analytes from peripheral blood plasma samples was performed using the Olink Proximity Extension Immunoassay (PEA) platform, specifically employing the Immuno-Oncology (IO) panel. Data obtained from the Olink platform were subsequently normalized into NPX (Normalized Protein Expression) values on a log2 scale. For more information on this panel, please see the manufacturer’s webpage (https://olink.com/products/olink-target-96, accessed date on 08 January 2025). Mass Cytometry by Time-of-Flight (CyTOF) was conducted on PBMCs which were barcoded, stained with metal-conjugated antibodies, normalized, and analyzed for immune subsets via Astrolabe Diagnostics and manual gating. Major immune-cell subsets were identified using a hierarchical clustering approach¹⁶ (Astrolabe Diagnostics, Inc) and further validated through manual gating. The resulting datasets provided cell counts, cell frequencies, and marker expression quantiles.

Analyses were performed in R¹⁷. For quality control, Olink analytes with zero variance or >50% below the limit of detection were excluded; CyTOF cell populations unassigned by Astrolabe were removed. Principal Component and Principal Variance Component Analyses (via dream package¹⁸) identified technical variation.

Linear mixed-effects models (LMEMs) were fit using dream function from the variancePartition package¹⁸ to assess changes in Olink protein expression across timepoints, with fixed effects for arm, time, and their interaction, and random intercepts per patient. Since the Bartlett test failed to reject the null hypothesis that variance was homogeneous across arms, models were fitted assuming homogenous variance. Normality was ascertained via Q-Q plots. Hypotheses were tested using the Empirical Bayes method¹⁹. Resultant p-values for the comparisons of interest were adjusted for multiple testing of Olink analytes via Benjamini-Hochberg²⁰, controlling the false discovery rate (FDR). Results were visualized using heatmaps with hierarchical clustering based on Euclidean distance and Ward D2 agglomeration method. No inter-arm comparisons were performed.

Time to first ≥Grade 3 or serious adverse event was estimated via Kaplan-Meier and stratified by baseline biomarker (high vs. low, median split), with comparisons by log-rank test. Cox regression assessed hazard ratios, with proportional hazards tested via Schoenfeld residuals. Data from both arms were pooled.

Prognostic value of Day 15 biomarkers was assessed using Cox models starting from Day 15, including only event-free participants. Biomarkers were dichotomized (high/low) based on median values, with values above the median considered high and those at or below the median low.

Results

Patient cohorts.

Between January 2019 and June 2021, 25 eligible (24 evaluable) patients were enrolled in the study (Figure 1). Initially, seven patients were accrued to Cohort 1 (durva+ACRT) and six patients to Cohort 2 (durva+SDRT). One patient in Cohort 1 died during the safety monitoring period without experiencing DLTs and was therefore not considered evaluable and was substituted. The study was paused for patient enrollment on April 13, 2020, to allow safety analysis of Cohort 1 and 2. The one recorded grade 5 SAE in cohort 2 based on the initial protocol version occurred while the protocol was undergoing an amendment to require DLTs be related to treatment. Thus, this DLT no longer met the DLT criteria and was removed after the amendment was approved. Since both cohorts experienced 0–1 DLTs, both were deemed safe to proceed to the randomized expansion phase on July 16, 2020 to Cohorts 3 and 4. Thirteen patients were randomized, seven to Cohort 3 and six to Cohort 4. One patient who enrolled to Cohort 3 withdrew consent prior to treatment so was not counted towards the target sample size. Patient characteristics by treatment group are provided in Table 1.

Figure 1 |. Flow chart of the assigned cohorts, numbers who completed protocol treatments and safety assessment of trial participants.

Table 1.

Patient and Tumor Characteristics for All Eligible Patients Grouped by Radiation Therapy Cohort

Patient or Tumor Characteristic	Durvalumab + ACRT (n=13)	Durvalumab + Standard RT (n=12)	Total (n=25)
Age (years)
50 – 59	1 (7.7%)	1 (8.3%)	2 (8.0%)
60 – 69	9 (69.2%)	4 (33.3%)	13 (52.0%)
≥ 70	3 (23.1%)	7 (58.3%)	10 (40.0%)
Median	69	72	69
Min - Max	53 – 78	54 – 79	53 – 79
		66 – 76	63 – 73
Gender
Female	6 (46.2%)	8 (66.7%)	14 (56.0%)
Male	7 (53.8%)	4 (33.3%)	11 (44.0%)
Race	Q1 - Q3	63 – 69
American Indian or Alaska Native	0 (0.0%)	1 (8.3%)	1 (4.0%)
Black or African American	3 (23.1%)	3 (25.0%)	6 (24.0%)
White	10 (76.9%)	8 (66.7%)	18 (72.0%)
Ethnicity
Not Hispanic or Latino	12 (92.3%)	11 (91.7%)	23 (92.0%)
Unknown	1 (7.7%)	1 (8.3%)	2 (8.0%)
Zubrod Performance Status
0	8 (61.5%)	10 (83.3%)	18 (72.0%)
1	5 (38.5%)	2 (16.7%)	7 (28.0%)
Primary tumor location
Lower Lobe (left)	1 (7.7%)	1 (8.3%)	2 (8.0%)
Lower Lobe (right)	4 (30.8%)	2 (16.7%)	6 (24.0%)
Middle Lobe (right)	1 (7.7%)	0 (0.0%)	1 (4.0%)
Other	1 (7.7%)	1 (8.3%)	2 (8.0%)
Upper Lobe (left)	3 (23.1%)	2 (16.7%)	5 (20.0%)
Upper Lobe (right)	3 (23.1%)	6 (50.0%)	9 (36.0%)
PDL1 Value
50	2 (15.4%)	2 (16.7%)	4 (16.0%)
55	0 (0.0%)	1 (8.3%)	1 (4.0%)
60	0 (0.0%)	1 (8.3%)	1 (4.0%)
70	2 (15.4%)	1 (8.3%)	3 (12.0%)
75	1 (7.7%)	0 (0.0%)	1 (4.0%)
76	1 (7.7%)	0 (0.0%)	1 (4.0%)
80	2 (15.4%)	2 (16.7%)	4 (16.0%)
85	1 (7.7%)	0 (0.0%)	1 (4.0%)
90	0 (0.0%)	3 (25.0%)	3 (12.0%)
99	0 (0.0%)	1 (8.3%)	1 (4.0%)
100	4 (30.8%)	1 (8.3%)	5 (20.0%)
AJCC Stage
IIIA	6 (46.2%)	3 (25.0%)	9 (36.0%)
IIIB	4 (30.8%)	7 (58.3%)	11 (44.0%)
IIIC	2 (15.4%)	2 (16.7%)	4 (16.0%)
N1-Undetectable primary	1 (7.7%)	0 (0.0%)	1 (4.0%)
T-Stage
T1a	1 (7.7%)	1 (8.3%)	2 (8.0%)
T1b	0 (0.0%)	2 (16.7%)	2 (8.0%)
T1c	3 (23.1%)	2 (16.7%)	5 (20.0%)
T2a	2 (15.4%)	1 (8.3%)	3 (12.0%)
T2b	1 (7.7%)	1 (8.3%)	2 (8.0%)
T3	1 (7.7%)	2 (16.7%)	3 (12.0%)
T4	4 (30.8%)	3 (25.0%)	7 (28.0%)
TX	1 (7.7%)	0 (0.0%)	1 (4.0%)
N-Stage
N0	1 (7.7%)	1 (8.3%)	2 (8.0%)
N1	2 (15.4%)	1 (8.3%)	3 (12.0%)
N2	6 (46.2%)	4 (33.3%)	10 (40.0%)
N3	4 (30.8%)	6 (50.0%)	10 (40.0%)

Adverse events and clinical trial safety outcomes.

Overall adverse events (AEs) by treatment group regardless of attribution (Supplementary Table 1) and those attributed to treatment are summarized in Table 2. The primary endpoint of safety based on DLTs was met for both radiation arms. In the durva+ACRT cohort no DLTs out of 12 were reported. Four patients (30.8%) experienced grade 3 AEs, and one patient (7.7%) had grade 4 lymphopenia. A single grade 5 AE occurred in a patient from Cohort 1, who developed a lung infection shortly after receiving the initial durvalumab dose and before the initiation of RT. This patient was hospitalized and subsequently died from the infection, which was assessed as unrelated to treatment. Since the patient didn’t start RT, the patient was deemed unevaluable and was replaced in the study.

Table 2.

Distribution of Highest-Grade Adverse Event by System Organ Class Grouped by Radiation Cohort For All Reported Adverse Events that are Possibly, Probably, or Definitely Related to Treatment

System Organ Class/Term	Durvalumab + ACRT (n=13)					Durvalumab + Standard RT (n=12)
	n and (%) of Patients by Grade					n and (%) of Patients by Grade
	1	2	3	4	5	1	2	3	4	5
OVERALL HIGHEST GRADE	2	5	3	1	0	1	4	5	0	0
	(15.4)	(38.5)	(23.1)	(7.7)	(0.0)	(8.3)	(33.3)	(41.7)	(0.0)	(0.0)
GASTROINTESTINAL DISORDERS	5	2	1	0	0	4	2	0	0	0
	(38.5)	(15.4)	(7.7)	(0.0)	(0.0)	(33.3)	(16.7)	(0.0)	(0.0)	(0.0)
Esophagitis	1	2	1	0	0	0	0	0	0	0
	(7.7)	(15.4)	(7.7)	(0.0)	(0.0)	(0.0)	(0.0)	(0.0)	(0.0)	(0.0)
INFECTIONS AND INFESTATIONS	0	1	0	0	0	0	1	1	0	0
	(0.0)	(7.7)	(0.0)	(0.0)	(0.0)	(0.0)	(8.3)	(8.3)	(0.0)	(0.0)
Lung infection	0	0	0	0	0	0	0	1	0	0
	(0.0)	(0.0)	(0.0)	(0.0)	(0.0)	(0.0)	(0.0)	(8.3)	(0.0)	(0.0)
INVESTIGATIONS	1	1	2	1	0	0	4	4	0	0
	(7.7)	(7.7)	(15.4)	(7.7)	(0.0)	(0.0)	(33.3)	(33.3)	(0.0)	(0.0)
Alanine aminotransferase increased	0	0	1	0	0	0	3	0	0	0
	(0.0)	(0.0)	(7.7)	(0.0)	(0.0)	(0.0)	(25.0)	(0.0)	(0.0)	(0.0)
Aspartate aminotransferase increased	1	0	1	0	0	0	3	0	0	0
	(7.7)	(0.0)	(7.7)	(0.0)	(0.0)	(0.0)	(25.0)	(0.0)	(0.0)	(0.0)
Lymphocyte count decreased	1	1	1	1	0	0	1	4	0	0
	(7.7)	(7.7)	(7.7)	(7.7)	(0.0)	(0.0)	(8.3)	(33.3)	(0.0)	(0.0)
RESPIRATORY, THORACIC AND MEDIASTINAL DISORDERS	4	3	2	0	0	2	2	0	0	0
	(30.8)	(23.1)	(15.4)	(0.0)	(0.0)	(16.7)	(16.7)	(0.0)	(0.0)	(0.0)
Dyspnea	3	1	1	0	0	1	1	0	0	0
	(23.1)	(7.7)	(7.7)	(0.0)	(0.0)	(8.3)	(8.3)	(0.0)	(0.0)	(0.0)
Pneumonitis	2	2	1	0	0	1	2	0	0	0
	(15.4)	(15.4)	(7.7)	(0.0)	(0.0)	(8.3)	(16.7)	(0.0)	(0.0)	(0.0)

Adverse events were graded with CTCAE version 5.

In the durva+SDRT group, a total of 1 of 12 developed DLT. The one DLT was in a patient from cohort 2 who developed grade 3 bronchopulmonary hemorrhage and grade 2 dyspnea within 30 days of starting durvalumab that led to permanent discontinuation of durvalumab. Overall, eight patients (66.7%) in the SDRT arm experienced grade 3 AEs, but no grade 4 AEs were reported. A grade 5 AE occurred in Cohort 4, where a patient developed respiratory failure due to a large obstructing mass with post-obstructive pneumonia. This event was deemed unrelated to the treatment.

At the time of analysis, 76% of patients had discontinued durvalumab, whereas six patients (24%) completed all 13 cycles. Nearly half the patients (48%) had durvalumab modified; 11 modifications were due to AEs, and eight were for other reasons. As a secondary objective, feasibility was assessed based on patients receiving at least 80% of the planned durvalumab dose (≥3000 mg) within the first eight weeks of treatment. In the ACRT arm, 11 of 13 patients (85%) met this criterion. One patient died unexpectedly after the first infusion without receiving RT, and another developed grade 2 shingles, resulting in a skipped second dose before resuming treatment in cycle 3 and completing up to cycle 13. In the SDRT arm, nine of 12 patients (75%) received at least two doses of durvalumab. Three patients received only the first 1500 mg dose and missed the second dose: one due to bronchopulmonary hemorrhage and subsequent withdrawal from the study; a second due to reactivation of previously undiagnosed viral hepatitis and had skipped second dose but resumed durvalumab at the 3^rd dose 4 weeks later; the third patient missed the second dose due to grade 4 respiratory failure and atrial fibrillation and didn’t resume durvalumab, but all AEs were deemed unrelated to treatment that combines with radiation and not scored as DLTs.

Regarding RT compliance, one patient from each treatment arm received the first dose of durvalumab but did not receive RT due to death from pneumonia or disease-related declining performance status, both of which were not related to treatment. All but 1 patient received 60 Gy as a total dose. All patients who were to receive ACRT received 15 fractions whereas 10 of the 11 patients who were to receive SDRT received 30 fractions (Supplementary Table 2). Thus, of the 23 eligible patients who received RT, almost all patients (95.6%) received RT per protocol or with an acceptable variation (Supplementary Table 3). In terms of contour review, overall score, overall tumor volume, and overall organs at risk were all per protocol or with an acceptable variation (Supplementary Table 4). For dose volume analysis, >90% were per protocol or had acceptable variation for overall tumor volume and overall organs at risk (Supplementary Table 5).

Upregulation of protein markers associated with immune cytotoxicity, cell activation, and trafficking upon treatment.

All 92 unique proteins in the 96 Immuno-Oncology Olink panel were detected in more than 92.5% of blood samples. Of these 92 proteins, 8 were removed due to zero variance (FGF2, IL-1a, IL-13, IL-2, IL-4, IL-5, NOS3, and PTN). Following treatment initiation, 32 analytes (38%) were found to be differentially expressed (FDR < 0.05). Longitudinal clustering revealed distinct patterns among cytokines, chemokines, receptors, and other circulating proteins, showing either consistent expression changes across both arms or divergent dynamics specific to SDRT or ACRT (Figure 2A).

Figure 2 |. Proteomic marker changes post-treatment in NSCLC trial patients measured by Olink.

(a) Heatmap depicting the mean fold change in normalized protein expression (NPX) relative to baseline at Days 15, 30, and 90 across study arms. Color shading represents the fold changes per analyte (in log2-scale). Statistical significance is indicated for adjusted or nominal p-values. The blue bracket highlights markers increased following durvalumab in combination with SDRT. The pink bracket denotes markers upregulated early during ACRT treatment (Day 15) and later in treatment (Day 90) with SDRT treatment. The green bracket encompasses markers upregulated in both arms and associated with nivolumab treatment. Box plots in b-d are representative analytes, with the marginal mean at each time point (red) superimposed with the individual data points (black). P-value annotations indicate statistical significance for comparisons between time points and baseline values. (b) Selected markers exhibiting significant increases relative to baseline following SDRT, including markers of cytotoxicity and Th1 signaling. (c) Markers upregulated early (Day 15) in the ACRT arm and later (Day 90) in the SDRT arm. (d) Circulating markers upregulated in both arms, including direct targets of nivolumab and ICI-associated markers such as CXCL9 and CXCL10.

Overall, there was an increase in Th1 associated markers relative to baseline in both the SDRT arm and ACRT arm (bottom cluster highlighted in green in Figure 2A). Several markers associated with Th1, immune cell activation, and trafficking were upregulated following treatment relative to baseline in both arms at Day 30 or Day 90 post-treatment, including IFNg, IL15, CD70, CXCL10, and CXCL11 (p or adj. p < 0.05) (Figure 2D). The strongest and earliest increase post-baseline was observed for soluble PD-L1, the target of durvalumab, which could be attributed in part to stabilization of the protein by the drug.

In the durva+SDRT arm, many additional proteins related to antigen presentation and activation (LAMP3, TNFRSF4/OX40, CD83, IL12Rb1, PD-L2, ICOSLG), inflammation (IL8, IL18, CXCL9, CCL23, MCP-2, IL10), and exhaustion (PDCD1/PD-1, LAG3) were significantly increased post-treatment at Day 30 and/or Day 90 (p or adj. p < 0.05) (Figure 2A, B, D, green and blue highlights). However, in the durva+ACRT arm, these proteins remained unchanged or showed only marginal increases without reaching significance (Figure 2A, B, D).

In the durva+SDRT arm, markers of T and NK cell cytotoxicity (GZMA, GZMH, CRTAM, KLRD1, CD244), were significantly upregulated from baseline at Day 90 (adjusted p-value <0.05) (Figure 2D). In contrast, in the durva+ACRT arm, these cytotoxicity markers (including additionally CD8 and KIR3DL1) showed a transient but significant decrease from baseline at Day 15, with a return to baseline level by Day 90 (p or adj. p < 0.05) (Figure 2A, B, C, pink highlight).

A small number of proteins showed gradual decrease in both arms, including MMP12, CXCL13, and MUC16 (the latter more pronounced with durva+ACRT treatment).

Contrary to our original hypothesis, these results highlight an unexpectedly limited, if not reduced, profile of immune activation and cytotoxicity with durva+ACRT, while durva+SDRT led to widespread increases in inflammatory and T-cell related markers.

Immune changes in plasma associated with the incidence of adverse events.

Next, we evaluated the relationship between circulating immune markers and the occurrence of serious and/or significant (grade 3 or higher) treatment-related adverse events within the first six months of therapy. In the durva+SDRT arm, several markers were downregulated by Day 15 in patients who experienced adverse events (Figure 3A, blue and Figure 3B). These protein reductions at Day 15 were not due to treatment alone (almost no significant change occurred in these markers at Day 15 compared to baseline with durva+SDRT), suggesting they were specifically associated with AEs in this arm. These proteins represented a diverse array of pathways (cytotoxicity, antigen co-stimulation, neutrophil markers such CXCL5 or CXCL1, tumor markers such as ADA or CAIX) that gradually normalized at Day 30 and Day 90. Three markers had a trend of lower expression at baseline in patients eventually experiencing AEs in the durva+SDRT arm: CD40L, CD5, and TRAIL. Among these, reduced CD5 expression was associated at baseline and at Day 15 with an increased hazard of experiencing AEs regardless of arm (Figure 3C). The predictive value of soluble CD5 as a biomarker of AE may be reflecting its possible role as a negative regulator to T and B cell immunity. The only other significant predictive marker of toxicity was high baseline soluble LAG3 expression (Figure 3C).

Figure 3 |. Proteomic marker changes associated with adverse events during the first six months of treatment.

(a) Bar plot summarizing differences in analyte expression at each time point between patients who experienced an adverse event and those who did not. Markers with higher expression in patients with an adverse event are shown in red, while those with lower expression are shown in blue. Differences are presented as NPX values measured by Olink, per study arm. Statistical significance is indicated for adjusted or nominal p-values, as in Figure 2, comparing patients with and without adverse events at each time point within each study arm. (b) Longitudinal box plots of selected markers showing NPX values at each time point for patients who experienced an adverse event (green) and those who did not (grey). Marginal means for each group are represented in the corresponding group color, with statistical significance for each comparison displayed above the respective time points. (c-d) Kaplan–Meier plots for the entire study cohort (both arms combined), stratified by baseline marker expression either at day 0 (C) or day 15 (D). Patients with marker expression above the median at the respective timepoint are shown in blue and those below the median are shown in yellow. P-values and false discovery rates (FDR) from the log-rank test indicate statistical significance between low and high expression groups.

Protein dynamics associated with AEs seemed to be mostly driven by SDRT (Figure 3A, red). Instead, TRAIL, CAIX, CD70, MMP12, PD-L1, and IL-15 behaved in the opposite direction, where higher levels were associated with toxicity, suggesting that SDRT and ACRT treatment toxicities maybe linked to different mechanisms.

Reduced circulating lymphocytes post-treatment.

Following treatment, both trial arms showed an overall reduction in circulating lymphocytes among PBMCs, as measured by CyTOF, compared to myeloid cell populations, as expected with RT-induced lymphopenia (Figure 4A, with negative percent change from baseline in nearly all lymphoid subsets). When considering individual immune subsets with significant changes (unadjusted or adjusted p < 0.05), naïve and memory B cells exhibited the sharpest decrease early upon treatment that was sustained until Day 90 especially in the durva+SDRT arm and less so in the durva+ACRT arm (Figure 4A–C). Other cell types decreasing in both arms with treatment included CD4⁺ naïve T cells, NKT cells, and gamma-delta T cells (Figure 4A, 4D). As anticipated, there was more limited reduction in circulating immune subsets in ACRT as compared to SDRT. A reduction in regulatory T cells (Tregs) and various memory T cell subsets only achieved significance at Day 30 in the durva+SDRT arm. The most significant post-treatment increases were seen in granulocytes at Day 30 in the durva+SDRT arm, and in plasmacytoid dendritic cells (pDC) and HLA-DR-expressing lineage-negative myeloid cells (Figure 4A).

Figure 4 |. Longitudinal changes in immune cell subsets post-treatment in NSCLC trial patients measured by CyTOF.

(a) Heatmap illustrating changes in immune cell populations relative to baseline, stratified by parent cell type. Color shading indicates the relative increase or decrease in cell frequency at each specified time point, in percentages. (b–d) Longitudinal box plots showing the frequencies of selected immune cell subsets at each indicated time point across study arms. Marginal means are shown in red, and statistical significance for comparisons between each time point (Day 15, Day 30, Day 90) and baseline is indicated above the box and whisker bars. Individual data points are represented in black.

When exploring the association between circulating cell subsets and AEs, no cell type showed significant associations except for an increase in conventional DCs at Day 30 in the durva+SDRT arm (p < 0.05) (Supplementary Figure 1A). Elevated pDC were the only cell type also significantly associated at baseline or Day 15 with worse toxicities when considering time-to event analyses of AEs across both arms, while elevated transitional B cells and NK CD56⁺CD16⁻ cells only showed trends in that direction (Supplementary Figure 1B). A lower level of effector memory CD8⁺ T cells at Day 15 was also trending towards a worse hazard of experiencing AEs, though not reaching significance.

Discussion

The current standard of care for the curative management of unresectable locally advanced NSCLC is a conventional dose of 60 Gy in 30 fractions with concurrent chemotherapy followed by 1 year of durvalumab¹. Due to the significantly increased normal-tissue toxicities of concurrent chemotherapy with RT vs. RT alone, the potential of using RT alone without compromising disease outcomes in select settings would be desirable, if the benefits of chemotherapy could be preserved by substituting with immunotherapy. We designed this trial before the completion and reporting of the PACIFIC-2²¹, Checkmate 73L²², and EA5181²³, clinical trials which all asked the similar question if concurrent ICI with CRT will benefit compared to CRT alone (PACIFIC-2) or CRT followed by ICI consolidation (Checkmate 73L and EA5181). While the results were all negative with respect to the benefit of concurrent ICI with CRT, it is still unclear if the lack of benefit extends to radiotherapy alone without concurrent chemotherapy, to the setting of hypofractionated radiotherapy, or in tumors with high PD-L1 levels ≥50%. For this trial, we have demonstrated that in PD-L1 high tumors, substituting low dose chemotherapy with durvalumab to combine with RT was safe and well tolerated, using either the standard or accelerated fractionated radiation regimens. Many of the documented SAEs were not considered attributable to the protocol regimen given the acute nature of the SAEs relative to the infusion date of the durvalumab, especially coming from infections or hemorrhage after one dose of durvalumab before start of RT. However, we also cannot fully exclude the possibility that these are not immune-related SAEs due to durvalumab. While technically SDRT did not meet the threshold of feasibility due to our predefined protocol definition, we do not believe the conventional dose regimen was any less feasible relative to ACRT due to non-treatment related cause of delay or withdrawal of therapy before receiving the second dose of durvalumab. In one of the 3 patients in the SDRT arm, durvalumab resumed after skipping the second dose, and continued on for the full year afterwards.

One of the interesting aspects of this study is the collaboration with CIMAC on biomarker analysis of the two-radiation fractionation cohorts. Given the size of the cohorts, the intention is not to formally compare but to evaluate the interplay between the immune system and radiation modality individually. In our studies of immune changes in peripheral blood, we found that whereas both arms showed increases in Th1-associated markers (IFN-γ, IL-15, CXCL11), the durva+SDRT arm exhibited broader changes, with sustained elevations in inflammatory and cytotoxic proteins through Day 90, for example in CRTAM and IL-18. These markers have been shown to promote cytotoxic responses and induce CD8-T and NK cell activity especially when paired with IL-12, leading to enhanced IFN-γ production and more robust anti-tumor activity²⁴. Moreover, these soluble changes paralleled a rebound in effector memory and TEMRA T cells, suggesting either enhanced anti-tumor immunity or immune overactivation, supported by increased PD-1, LAG3, and IL-10.

In contrast, in the durva+ACRT arm, we saw early reductions in cytotoxic markers and minimal inflammatory changes. IL-8, for instance, was decreased post-ACRT but elevated with SDRT, aligning with the hypothesis that ACRT elicits less inflammatory damage. The limited activation signature in ACRT may indicate weaker immune priming, possibly due to local lymphocyte depletion at the tumor site, despite relatively stable circulating T cell numbers.

Radiation also affected B cells, with both regimens reducing their numbers—more markedly with SDRT—potentially compromising humoral immunity. While both arms showed decreased lymphocytes, the effect was more pronounced in SDRT. As fraction number is strongly associated with lymphopenia risk²⁵, this observation was expected to be more significant across the spectrum of peripheral immune cells in the SDRT cohort. However, formal testing between arms was not conducted, per protocol.

We also identified early biomarkers linked to AEs, notably reduced soluble proteins at Day 15 in the durva+SDRT arm (e.g., EGF, CAIX, ADA, CASP-8, granzymes), implicating endothelial damage or impaired immune mobilization. Soluble PD-L1, consistently elevated with treatment, was even higher in AE cases in the durva+ACRT arm, suggesting potential as a pharmacodynamic marker. The most consistent biomarker significantly associated with AE occurrence regardless of arms, when measured either at baseline or early in treatment (Day 15), was a reduction in soluble CD5. While membrane-bound CD5 is known for its role as a negative regulator of immune responses²⁶, the functions and implications of soluble CD5, particularly in the context of antitumor immunity and treatment-related AEs, are less well-defined. Surface expression of CD5 is known to negatively regulate BCR signaling by increasing the threshold of activation, thereby contributing to maintaining B-cell tolerance²⁷. Reduced expression of CD5 in CD8 T cells enhances their responsiveness to antigens, while up-regulation of CD5 has been shown to prevent cell death in autoreactive CD4 T-cells²⁸. Furthermore, CD5 expression is associated with the production of IL-10, an anti-inflammatory cytokine²⁹. In our study, elevated CD5 levels were linked to a reduced risk of adverse events and showed a positive correlation with IL-10 levels.

While our study is limited by the small number of patients to make any definitive conclusions about the efficacy of the approach as compared to the current standard of care, we have met the primary objective of demonstrating the safety of combining concurrent durvalumab with either SDRT or ACRT, while feasibility needs further assessment given the small numbers. Our in-depth biomarker analysis offers hypothesis-generating data that provides perspectives on how dose and fractionation of RT could impact inflammatory response in combination with immune checkpoint inhibitors.

Supplementary Material

Figure 1 Supplemental

Supplementary Figure 1. Changes in immune cell subsets associated with adverse events during the first six months of treatment.

Supplementary Table 1-5

Supplementary Table 1. Distribution of NRG-LU004 Patients by Highest Grade Adverse Event by System Organ Class Grouped by Radiation Cohort for All Reported Adverse Events without Regard to Attribution.

Supplementary Table 2. Radiation Therapy Delivery Summary.

Supplementary Table 3. Radiation Delivery Details and Compliance.

Supplementary Table 4. Radiation Therapy Contour Review.

Supplementary Table 5. Radiation Therapy Dose Volume Analysis.

Abbreviations

RT

Radiotherapy

ACRT

Accelerated Radiotherapy

SDRT

Standard Radiotherapy

LA-NSCLC

Locally Advanced Non-Small Cell Lung Cancer

CRT

Chemoradiation

PD-L1

Programmed Death Ligand-1