Pneumonitis after chemoradiotherapy and/or immunotherapy for stage III and IV non-small cell lung cancer is characterized by specific radiologic and quantitative CT features

Abstract

Background

Patients with unresectable stage III non-small cell lung cancer (NSCLC) receive a combination of chemotherapy (CT), thoracic radiation therapy (RT) and immune checkpoint inhibition (ICI). Stage IV NSCLC with a high PD-L1 tumor proportion score receive first line treatment with ICI. These new therapies and combinations can give rise to pneumonitis due to RT but also due to ICI, both etiologies being notoriously difficult to disentangle. In this work follow-up CT scans with pneumonitis after these treatments were explored in a radiologic pattern and a quantitative CT feature analysis.

Materials and methods

This retrospective, single-center study included patients with pneumonitis of any grade who received chemoradiotherapy with or without durvalumab (RT-only group and RT + ICI group) for unresectable stage III NSCLC between 2011 and 2021, as well as patients with pneumonitis who received pembrolizumab as first-line therapy (ICI-only group) for stage IV NSCLC between 2017 and 2021. Clinical data and computed tomography (CT) scans were collected at baseline and follow-up. A radiologist compared CT characteristics among the three groups to identify patterns of radiotherapy-related and ICI-related pneumonitis. Furthermore, quantitative imaging features of lung damage derived from follow-up CT scans were compared between different lung regions within the same patient and across treatment groups.

Results

Seventy-one patients developed pneumonitis: 36/163 in the RT-only group (22.1%), 18/116 in the ICI-only group (15.5%), and 17/36 in the RT + ICI group (47.2%, p < 0.001). The ICI-only group showed involvement of more lobes vs. RT + ICI and RT-only (4.06 vs. 3.41 vs. 3.21 respectively; p = 0.03) and a trend toward more bilateral disease (83%, 65%, and 58%; p = 0.06). RT-only patients exhibited more bronchiectasis vs. RT + ICI and ICI-only (47%, 18%, and 6% respectively; p = 0.003), while consolidations were less frequent in the ICI-only group (50%, 75%, and 75%; p = 0.04). Low-grade pneumonitis (grade < 3) was more often associated with an organizing pneumonia pattern (p = 0.01), whereas higher-grade cases showed other radiologic patterns.

Risk factors for high-grade pneumonitis included pre-existing interstitial lung disease (ILD; 14% vs. 2%, p = 0.03) and lower baseline diffusion capacity (52.2 ± 15.3% vs. 71.0 ± 21.4%; p = 0.004). High-grade pneumonitis developed earlier (52.5 ± 49.5 vs. 133.0 ± 101.6 days; p < 0.001) and was more frequently bilateral (100% vs. 59%; p = 0.003).

Quantitative CT analysis revealed significant differences between lung regions receiving high (>10 Gy) and low (<10 Gy) radiation doses across all evaluated features. In high-dose regions, the RT + ICI group exhibited more severe lung damage than the RT-only group (e.g., entropy, p = 0.017), whereas no differences were observed in low-dose regions (p > 0.20). Imaging features from the ICI-only group most closely resembled those from high-dose regions of the RT-only group which may suggest a synergistic effect of radiotherapy and ICI in developing lung injury in the RT + ICI group.

Conclusion

A significantly higher incidence of pneumonitis was observed in patients treated with combined RT and ICI. High-grade pneumonitis was more often bilateral, presented with distinct radiologic patterns, developed earlier, and was associated with pre-existing ILD and reduced diffusion capacity. Quantitative CT analysis confirmed that the combination of moderate- to high-dose lung irradiation with ICI therapy leads to greater severity of lung damage.

Highlights

• •Patients in Stage III or IV NSCLC often receive thoracic radiotherapy and/or immunotherapy during the course of their treatment.
• •Pneumonitis is an important side effect for both
• •Clinically it is clinically difficult to differentiate the causative agent
• •In depth analysis of the CT patterns can help us to distinguish between radiotherapy and immunotherapy-induced pneumonitis.

1. Introduction

Lung cancer remains one of the most frequent malignancies and the leading cause of cancer-related death worldwide [1]. About 80 to 90% of lung cancers are non-small cell lung cancers (NSCLC) and stage III and IV NSCLC account for over half of all the new diagnoses [2].

Patients with stage III unresectable NSCLC are treated with a platinum doublet chemotherapy in combination with radiotherapy. This has proven to be superior to radiotherapy in terms of overall survival [3]. Furthermore, concurrent chemoradiotherapy showed an additional benefit over a sequential approach, making this the standard of care [4], [5]. More recently, the PACIFIC trial proved a statistically significant improvement of median progression-free survival and overall survival in patients treated with 1 year of consolidation with the PD-L1 immune checkpoint inhibitor (ICI) durvalumab [6], [7]. Since October 2018 durvalumab is approved by EMA as monotherapy for the treatment of locally advanced, unresectable stage III NSCLC in adults whose tumor express PD-L1 on at least 1% of tumor cells and whose disease has not progressed following platinum based chemoradiotherapy.

For patients with ALK and EGFR negative stage IV NSCLC and a tumor PD-L1 expression of at least 50%, the KEYNOTE-024 phase III trial demonstrated a benefit of pembrolizumab, a PD-1 inhibitor, over 4 to 6 cycles of standard platinum-doublet chemotherapy in terms of overall survival and progression-free survival [8]. Thus, pembrolizumab is considered the standard of care for this patient population in the absence of contraindications for immunotherapy [2].

The everyday use of these PD-L1 and PD-1 inhibitors gives rise to an increased frequency of new immune-related adverse events. In particular pneumonitis, a potential life-threatening side effect, was observed in NSCLC patients receiving ICI [9]. The estimated prevalence of pneumonitis associated with PD-L1/PD-1 therapy used as a monotherapy is lower than 10% [10], but pneumonitis is the most common cause of treatment-related death [11]. ICI-related pneumonitis is a diagnosis of exclusion made by a combination of clinical and radiological elements. Typical CT patterns associated with ICI-related pneumonitis are non-specific interstitial pneumonia (NSIP), organizing pneumonia (OP), hypersensitivity pneumonitis (HP) or acute interstitial pneumonia (AIP) [12], with an OP pattern being the most common [13]. On the other hand, radiotherapy to the lung is a potential cause of pneumonitis as well with the relative lung volume irradiated to moderate to high doses as an important risk factor [14], [15]. Furthermore, in combination with ICI, radiotherapy may induce local inflammation with a possible synergistic effect in the development of pneumonitis [16], [17]. With this in mind and the increasing number of patients receiving thoracic radiotherapy and ICI in daily clinical practice, we may see a rise in the incidence of pneumonitis. This induces some difficult clinical challenges: it remains unclear which components of pneumonitis should be attributed to radiotherapy and which to ICI in patients receiving both. This is an important question as ICI-related pneumonitis requires a higher dose and longer duration of corticosteroids than radiotherapy-induced pneumonitis [14], [18], [19]. Moreover, recurrent ICI-related pneumonitis may lead do discontinuation of ICI, thus potentially resulting in loss of tumor control and worse prognosis [20]. Attempts have been made to distinguish different etiologies of pneumonitis using morphological features derived from CT scans and machine learning [21]. More bilateral disease and number of lobes involved were observed in ICI-related pneumonitis while pneumonitis presented with sharper borders in RT-related pneumonitis. Despite these promising results, the real life decisions are often much tougher. In this work we describe clinical and radiologic features of patients with pneumonitis and investigate quantitative CT features in an attempt to find patterns linked to the etiology of either ICI-related or RT-related pneumonitis.

2. Methods and materials

2.1. Patient datasets

In this retrospective study we included three patient cohorts from the university hospital of Leuven (UZL). To start, we identified 199 patients with unresectable stage III NSCLC treated at UZL between January 2011 and December 2021. Among these, a first cohort of 163 patients received chemoradiotherapy without consolidation ICI (RT-only group; 51–70 Gy, normofractionated, delivered using 3DCRT or IMRT), while a second cohort of 36 patients received durvalumab for up to one year following chemoradiotherapy (RT + ICI group; 55–66 Gy, normofractionated, with IMRT). In addition, we included a third cohort of 118 patients at UZL with stage IV NSCLC who received pembrolizumab as first-line monotherapy (ICI-only group).

We included patients in the current analysis after diagnosis of pneumonitis, based on clinical and radiologic findings. We graded degree of severity pneumonitis according to the Common Terminology Criteria for Adverse Events (CTCAE), version 5.0. We extracted patient characteristics from their electronic medical file, including age, sex, smoking history, comorbidities, date of diagnosis, tumor stage and histology, baseline pulmonary function, and treatments received. Time to pneumonitis was calculated as the interval from the last day of radiotherapy in the RT-only and RT + ICI groups, or from the first day of pembrolizumab treatment in the ICI-only group.

2.2. Radiologic patterns and quantitative CT features

For all pneumonitis cases, chest CT scans performed at the time of diagnosis were analyzed by an experienced radiologist blinded to patient characteristics, subgroup allocation, and clinical pneumonitis grade. Each scan was scored for the number of lobes involved, the distribution pattern, and specific radiological features, including ground-glass opacities, reticulations, bronchiectasis, consolidations, and nodules. A general pattern was assigned to each case: usual interstitial pneumonia (UIP), nonspecific interstitial pneumonia (NSIP), hypersensitivity pneumonitis (HP), acute interstitial pneumonia (AIP), organizing pneumonia (OP), or radiofibrosis. When no distinct pattern could be identified, the label undefined pneumonitis was set.

Consequently, in a quantitative analysis, three first-order radiomic features characterizing the CT density distribution within the lungs were extracted from follow-up scans in a subset of patients from each treatment group [22]. Patients who had undergone previous radiotherapy treatment in or close to thoracic region were excluded as well as two cases with adaptive replanning hindering an accurate lung dose map reconstruction. After these exclusions 12, 15 and 28 pneumonitis patients with exportable CT scans remained for quantitative CT analysis in the ICI only, RT + ICI, and RT only groups respectively. The extracted radiomic features included entropy, representing the randomness of gray-level intensity within the lungs; median lung voxel density; and the 95th percentile of voxel density (Q95) in Hounsfield Units (HU). For patients treated with radiotherapy (RT-only and RT + ICI groups), all features were calculated separately for high-dose (>10 Gy) and low-dose (<10 Gy) lung regions, based on our previous lung dose – RT damage response work [23]. An overlay of the dose distribution, which was calculated on the planning CT scan, was performed after rigid registration (MIM software, Cleveland, OH) of the follow-up CT scan. For the ICI only group, the features were calculated for the left and right lung in order to have two separate lung volumes analyzed in this group as well.

2.3. Statistical analysis

In the case of continuous variables, patient characteristics were analyzed using student's t-test after an F-test for equality of variances. Nominal and ordinal variables were analyzed using the Pearson's chi-square test. Differences in the CT density feature values between lung regions of the same patient were assessed by a Wilcoxon paired signed rank test. Differences in feature values between treatment groups were assessed by a Mann-Whitney U test.

3. Results

3.1. Pneumonitis incidence and risk factors

A total of 71 pneumonitis cases were analyzed (Table 1): 36 in the RT-only group and 17 in the RT + ICI group (stage III NSCLC), and 18 in the ICI-only group (stage IV NSCLC). Five patients in the ICI-only group had prior thoracic radiotherapy: three received palliative doses (5 × 4 Gy) and two had full-dose thoracic RT (33 × 2 Gy and 23 × 2.75 Gy) years before metastatic relapse.

Table 1.

Baseline patient characteristics (mean and SD or number and proportions).^⁎

	RT only (St III)	RT + ICI (St III)	ICI only (St IV)	p-value
N° pneumonitis	36	17	18
N° screened	163	36	116
Pneumonitis rate (any grade)	22.1%	47.2%	15.5%	p < 0.001
Age (year)	63.2 +/− 9.1	64.7 +/− 7.9	68.1 +/− 8.4	p = 0.07⁎
Male, n (%)	23 (64)	10 (59)	13 (72)	p = 0.63
KI at diagnosis, n (%)
100%	5 (14)	2 (12)	1 (6)
90%	26 (72)	11 (65)	5 (28)
80%	5 (14)	3 (18)	7 (39)
70% or lower	0 (0)	1 (6)	5 (28)	p = 0.003
Smoking history, n (%)
Current	14 (39)	6 (35)	6 (33)
Former	20 (55)	11 (65)	12 (67)
Never	2 (6)	0 (0)	0 (0)	p = 0.67
Packyear	33 (0–70)	34 (15–94)	34 (1–60)
Comorbidity, n (%)
COPD	18 (50)	9 (53)	7 (39)	p = 0.66
Cardiovascular	11 (31)	7 (41)	6 (33)	p = 0.75
Diabetes Mellitus	4 (11)	2 (12)	2 (11)	p = 0.99
CKD	3 (8)	3 (18)	6 (33)	p = 0.07
ILD	1 (3)	1 (6)	1 (6)	p = 0.83
Auto-immune	3 (8)	0 (0)	0 (0)	p = 0.22
Other malignancy	7 (19)	5 (29)	4 (22)	p = 0.72
Lung function
FVC (L)	3.69 +/− 0.87	3.21 +/− 0.91	3.43 +/− 1.02	p = 0.69⁎
FEV1 (L)	2.40 +/− 0.64	2.14 +/− 0.74	2.29 +/− 0.83	p = 0.88⁎
DLCO (%)	69.1 +/− 20.4	70.9 +/− 18.5	62.4 +/− 26.2⁎⁎	p = 0.22⁎
Tumor relapse at inclusion, n (%)	0 (0)	2 (12)	4 (22)	p = 0.02
Histology, n (%)
Adenocarcinoma	19 (53)	7 (41)	13 (72)
Squamous cell	13 (36)	9 (53)	3 (17)
Other	4 (11)	1 (6)	2 (11)	p = 0.26
EGFR+, n (%)	4 (11)	1 (6)	0 (0)	p = 0.32

^⁎p-value comparing RT +/− ICI vs. ICI only.

^⁎⁎DLCO not measured in 1 patient.

Incidence of pneumonitis was highest in the RT + ICI group (47.2%) compared with RT-only (22.1%) and ICI-only (15.5%; p < 0.001). Age, gender and smoking history were similar across groups, although stage IV patients had a lower Karnofsky index (p = 0.003). COPD was the most common comorbidity (50% RT-only, 53% RT + ICI, 39% ICI-only; p = 0.66), with no major differences in other comorbidities, baseline pulmonary function, or tumor histology. A total of 6 patients had disease relapse at inclusion (2 RT + ICI, 4 ICI-only).

No difference was seen in time to onset of pneumonitis and overall severity between all groups (111.5 ± 71.7 days RT-only, 109.6 ± 94.1 days RT + ICI, 139.9 ± 144.2 days ICI-only; p = 0.43; Table 2). Thirteen patients developed severe (grade ≥ 3) pneumonitis, with one death in the ICI-only group. In the RT + ICI group, durvalumab was put on hold in 10 patients due to pneumonitis; seven were rechallenged, and three relapsed after which durvalumab was stopped definitively.

Table 2.

Pneumonitis data in all patients (median and range or number and proportions).^⁎

Pneumonitis	St III – RT only	St III – RT + ICI	St IV- ICI only	p-value
N° pneumonitis	36	17	18
Thoracic RT
Patients, n (%)	36 (100)	17 (100)	5 (28)
Dose (Gy)	65.1 (55–70)	63.1 (55–66)	37.9 (20–66)
Onset pneumonitis (days)	111.5 +/− 71.7	109.6 +/− 94.1	139.9 ±144.2	p = 0.43*
Pneumonitis grade, n (%)
1	10 (28)	5 (29)	3 (17)
2	19 (53)	8 (47)	12 (67)
3	7 (19)	4 (24)	2 (11)
4	0 (0)	0 (0)	0 (0)
5	0 (0)	0 (0)	1 (6)	p = 0.74
Treatment, n (%)
Corticosteroids, n (%)	26 (72)	11 (71)	15 (83)	p = 0.45
High dose (1 > mg/kg)	5 (14)	3 (18)	8 (44)	p = 0.03

^⁎p-value comparing RT +/− ICI vs. ICI only.

Corticosteroids were used in most patients, with higher use of high-dose therapy (≥1 mg/kg) in the ICI-only group (44%) versus RT + ICI (18%) and RT-only (14%; p = 0.03).

Subgroup analysis showed that high-grade pneumonitis was associated with pre-existing interstitial lung disease (15% vs. 2% in low-grade; p = 0.03), lower baseline diffusion capacity (52.2 ± 15.3% vs. 71.0 ± 21.4%; p = 0.004), earlier onset (52.5 ± 49.5 days vs. 133.0 ± 101.6 days; p < 0.001), involvement of more lobes (4.62 vs. 3.22; p = 0.002), and more frequent bilateral disease (100% vs. 59%; p = 0.003). High-grade cases also required corticosteroids more often overall (100% vs. 69%; p = 0.02) and high-dose corticosteroids (62% vs. 14%; p < 0.001).

3.2. Radiologic pattern analysis

Analysis of follow-up CT scans revealed differences in radiologic patterns across treatment groups. The ICI-only group showed a non-significant trend toward more bilateral involvement (p = 0.06) and involvement of a greater number of lobes compared with the RT ± ICI groups (4.06 vs. 3.29 lobes; p = 0.03). Overall distribution patterns were similar among groups (Table 3), with organizing pneumonia (OP), usual interstitial pneumonia (UIP), or undefined pneumonitis being most common. Radiofibrosis was observed only in the RT-only group).

Table 3.

Chest CT characteristics in all patients (number and proportions between brackets).^⁎

Chest CT	St III – RT only	St III – RT + ICI	St IV- ICI only	p-value
Bilateral, n (%)	21 (58)	11 (65)	15 (83)	p = 0.06*
N° lobes, n (%)
1	2 (6)	1 (6)	1 (6)
2	9 (25)	2 (12)	1 (6)
3	10 (28)	6 (35)	4 (22)
4	9 (25)	5 (29)	2 (11)
5	6 (17)	3 (18)	10 (56)	p = 0.03*
Average	3.22	3.41	4.06
Distribution, n (%)
(Multi)focal	10 (28)	7 (41)	3 (17)
Diffuse	26 (72)	10 (59)	15 (83)	p = 0.27
Elements, n (%)
Ground glass	29 (81)	16 (94)	16 (89)	p = 0.38
Reticulation	12 (33)	11 (65)	9 (50)	p = 0.09
BRECT	17 (47)	3 (18)	1 (6)	p = 0.003
Consolidation	27 (75)	13 (76)	9 (50)	p = 0.04*
Noduli	4 (11)	2 (12)	5 (28)	p = 0.10*
Pattern, n (%)
UIP	2 (6)	2 (12)	4 (22)
NSIP	0 (0)	2 (12)	2 (11)
HP	0 (0)	0 (0)	1 (6)
AIP	5 (14)	1 (6)	1 (6)
OP	11 (31)	6 (35)	6 (33)
Radiofibrosis	5 (14)	0 (0)	0 (0)
Undefined pneumonitis	13 (36)	6 (35)	4 (22)	p = 0.15

^⁎p-value comparing RT +/− ICI vs. ICI only.

Looking at radiologic elements of the CT scan, ground-glass opacities were predominantly present across all groups. Reticulations were more present in the RT + ICI group, though not statistically significant (33% RT-only, 65% RT + ICI, 50% ICI-only; p = 0.09). Nodules were more common in the ICI-only group without statistical significance. Bronchiectasis and consolidations were more frequent in patients receiving RT (RT-only and RT + ICI) compared with the ICI-only group (bronchiectasis: 47% and 18% vs. 6%; p = 0.003; consolidations: 75% and 76% vs. 50%; p = 0.04).

When comparing low- versus high-grade pneumonitisTable 4, differences in individual radiologic elements were not significant, though reticulations were more frequent in the high-grade subgroup in a non-statistically significant way (69% vs. 40%; p = 0.05). Overall radiologic patterns differed by severity, with low-grade cases more often showing OP, and high-grade pneumonitis exhibiting more UIP, NSIP, and AIP patterns (p = 0.01).

Table 4.

Subgroup analysis by pneumonitis grade (low grade (<3) versus high grade(≥3)).^⁎

Grade analysis	Grade 1–2	Grade 3–5	p-value
Patients, n	58	13
Age (year)	64.1 +/− 8.9	68.0 +/− 7.7	p = 0.15
KI at diagnosis, n (%)
100%	8 (14)	0 (0)
90%	34 (59)	8 (62)
80%	11 (19)	4 (31)
70% or lower	5 (9)	1 (8)	p = 0.47
Smoking history, n (%)
Ever	56 (97)	13 (100)
Never	2 (3)	0 (0)	p = 0.50
Comorbidity, n (%)
COPD	27 (47)	7 (54)	p = 0.63
Cardiovascular	17 (29)	7 (54)	p = 0.09
Diabetes Mellitus	6 (10)	2 (15)	p = 0.60
CKD	9 (16)	3 (23)	p = 0.51
ILD	1 (2)	2 (15)	p = 0.03
Auto-immune	3 (5)	0 (0)	p = 0.40
Other malignancy	13 (22)	3 (23)	p = 0.96
Lung function
FVC (L)	3.48 +/− 0.87	3.64 +/− 1.18	p = 0.58
FEV1 (L)	2.32 +/− 0.72	2.28 +/− 0.71	p = 0.86
DLCO (%)	71.0 +/− 21.4	52.2 +/− 15.3*	p = 0.004
Thoracic RT, n (%)	48 (82)	10 (77)	p = 0.62
Immunotherapy, n (%)	29 (50)	6 (46)	p = 0.80
Time to pneumonitis (days)	133.0 +/− 101.6	52.5 +/− 49.5	p < 0.001
Treatment, n (%)
Corticosteroids	40 (69)	13 (100)	p = 0.02
High dose corticosteroids	8 (14)	8 (62)	p < 0.001
Bilateral, n (%)	34 (59)	13 (100)	p = 0.004
Number of lobes, n (%)
1	4 (7)	0 (0)
2	12 (21)	0 (0)
3	20 (34)	0 (0)
4	11 (19)	5 (38)
5	11 (19)	8 (62)	p = 0.002
Average	3.22	4.62
Distribution, n (%)
(Multi)focal	19 (33)	1 (8)
Diffuse	39 (67)	12 (92)	p = 0.07
Elements, n (%)
Ground glass	49 (84)	12 (92)	p = 0.46
Reticulation	23 (40)	9 (69)	p = 0.05
BRECT	17 (29)	4 (31)	p = 0.92
Consolidation	41 (71)	8 (62)	p = 0.52
Noduli	10 (17)	1 (8)	p = 0.39
Pattern, n (%)
UIP	5 (9)	3 (23)
NSIP	2 (3)	2 (15)
HP	1 (2)	0 (0)
AIP	3 (5)	4 (31)
OP	22 (38)	1 (8)
Radiofibrosis	5 (9)	0 (0)
Undefined pneumonitis	20 (34)	3 (23)	p = 0.01

^⁎DLCO not measured in 1 patient.

3.3. Quantitative CT feature analysis

For quantitative features derived from the voxelized density distributions on CT scans (Table 5), marked differences between high- and low-dose lung regions were observed in the RT-only group (Fig. 1C), with all three features reaching statistical significance (p < 0.007). Similar patterns were identified in the RT + ICI group (Fig. 1A). However, the magnitude of these differences was greater in the RT + ICI group than in the RT-only group, as reflected by lower p-values and higher feature values (e.g., entropy within the 10 Gy isodose: 6.33 in the RT-only group vs. 6.49 in the RT + ICI group, p = 0.017). These findings suggest more pronounced lung damage in high-dose regions following combined RT and ICI compared with RT alone.

Table 5.

Quantitative CT density features.

CT density features
RT only	>10 Gy	<10 Gy	Difference	p-value
Entropy	6.33	6.08	0.25	<0.001
Median density (HU)	−793.6	−830.4	36.8	0.020
Q-95% density (HU)	−245.0	−409.9	164.9	0.002
RT + ICI	>10 Gy	<10 Gy	Difference	p-value
Entropy	6.49	6.11	0.38	<0.001
Median density (HU)	−790.9	−841.8	50.9	0.006
Q-95% density (HU)	−165.8	−466.4	300.6	<0.001
ICI only	Left lung	Right lung	Difference	p-value
Entropy	6.41	6.41	0.0	0.64
Median density (HU)	−776.7	−789.6	12.9	0.93
Q-95% density (HU)	−260.3	−277.8	17.6	0.81

Fig. 1.

(A) RT + ICI group example case with lung damage presentation in the region that received a 10 Gy dose (blue contour = 10 Gy isodose rigid overlay). (B) ICI only group example case with lung damage uniformly spread throughout both lungs. (B) RT only group example case with lung damage presentation in the region that received a 10 Gy dose (blue contour = 10 Gy isodose rigid overlay).

In contrast, no significant differences were observed between low-dose lung regions in the RT-only and RT + ICI groups (p > 0.20 for all features), indicating that pneumonitis-related lung damage predominantly localizes to high-dose regions regardless of treatment modality.

The ICI-only group (Fig. 1B) showed no significant differences between the left and right lungs for any feature (all p > 0.6). Feature values in this group were comparable to those observed in high-dose lung regions of both the RT + ICI group (all p > 0.05) and the RT-only group (all p > 0.2). In contrast, feature values in the ICI-only group differed significantly from those in low-dose lung regions of the RT-only group (p < 0.02 for all three features) and the RT + ICI group (p < 0.02 for all three features).

4. Discussion

This study proposed an in-depth analysis of pneumonitis in the context of NSCLC across different etiologies: radiotherapy, immune checkpoint inhibition or a combination of both. Current guidelines for treatment of ICI-related or RT-related pneumonitis differ: ICI-related pneumonitis generally requires a higher dose of corticosteroids and longer duration of treatment compared to RT-related pneumonitis [24]. Moreover, rechallenging an ICI-related pneumonitis case with the same checkpoint inhibitor carries a considerable risk, whereas this might not be the case if the pneumonitis is mainly radiotherapy-induced [25].

In this study we aimed to distinguish radiologic components attributable to radiotherapy from those associated with checkpoint inhibitors. Notably, patients who received both RT and ICI were more prone to develop any kind of pneumonitis compared to the RT-only and ICI-only group. The prevalence of pneumonitis in the RT-only group was relatively high (22.1%), which might be explained by the inclusion of patients treated as early as 2011 when the older 3DCRT technique was still standard of care.

Analyzing pneumonitis by severity grade, patients with high-grade events had a lower diffusion capacity at baseline. Despite the low prevalence, the high-grade subgroup also showed a significantly higher incidence of pre-existing interstitial lung disease. This is consistent with prior evidence of pre-existing interstitial lung disease being a risk factor for developing pneumonitis in all solid tumors [26]. Notably, high-grade pneumonitis occurred much faster than the low-grade cases. This observation aligns with previous reports in patients receiving checkpoint inhibitors, which have shown that steroid-refractory pneumonitis tends to develop early in the treatment course [27], [28].

ICI-related pneumonitis tended to present more bilaterally, involving multiple lobes, and was characterized with more noduli and reticulations on CT scan. In contrast bronchiectasis and consolidation seemed to be more present with RT-related pneumonitis. Remarkably, radiofibrosis was present in the RT-only group but absent in the RT + ICI group. Usual interstitial pneumonia (UIP) and non-specific interstitial pneumonia (NSIP) patterns were primarily seen in the ICI only and RT + ICI group. As expected, organizing pneumonia (OP) was the predominant pattern in all groups [29]. Furthermore, we observed distinct patterns between low- and high-grade pneumonitis as described in literature [12], [29]: An OP pattern was predominantly seen in low grade pneumonitis, whereas acute interstitial pneumonia (AIP) patterns, and to a lesser extent UIP and NSIP were more often seen in the high grade group. While definitive conclusions cannot be drawn solely from CT patterns, the findings suggest that pneumonitis arising from RT and ICI may represent an intermediate radiologic phenotype of pneumonitis depending on the grade of severity.

The quantitative results of the CT density feature analysis further support these observations. In both the RT-only and the RT + ICI group, we observed differences in feature values – hypothesized to serve as a surrogate of lung damage severity - between the lung regions receiving higher versus lower radiation doses. Within the moderate-to-high dose regions (i.e., the >10 Gy isodose), entropy values were higher in the RT + ICI group compared with the RT-only group. This provides a quantifiable indication of more severe lung damage at equivalent radiation doses when radiotherapy is combined with a checkpoint inhibitor. Moreover, for all evaluated radiomic features, the contrast between high- and low-dose regions was greater in the RT + ICI group than in the RT-only group. Furthermore, the CT density features of the highly irradiated zones of the RT-only group more closely resembled those features of the ICI-only group than those of the RT + ICI group. This pattern might suggest a synergistic effect between radiation and immune activation in developing lung injury.

Notably, no significant difference in CT density features between the low-dose regions of the RT + ICI and RT only groups were detected. This suggest that checkpoint inhibitors did not contribute to additional parenchymal lung damage outside the moderate-to-high radiated areas. Taken this into account, these results give weight to the hypothesis that RT + ICI related pneumonitis may represent a localized form of ICI-related pneumonitis, amplified by radiotherapy. Furthermore, no significant differences were observed in CT density features between the left and right lung in the ICI-only group, indicating no preference for lung damage to either side.

This study has several limitations. The relatively low number of pneumonitis cases in all three subgroups limited the statistical power and did not allow for multivariable analysis of the risk factors for pneumonitis (grade) or its radiologic elements. For the quantitative feature analysis, a 10 Gy cut-off was chosen based on prior studies to distinguish lung regions at risk of RT-induced injury. While differences between dose regions and treatment groups were significant, an alternative cut-off might prove more optimal in this respect. Moreover, as the follow-up CT scans were acquired during routine thoracic follow-up, some variability in acquisition settings (e.g. voltage ranging from 100 kV to 140 kV, slice thickness ranging from 3 mm to 5 mm and reconstruction kernels of different manufacturers) was present in the dataset. To rule out any potential impact this may have had on the extracted first-order features of the lungs, a confirmation study with a dedicated scan protocol seems warranted.

In conclusion, this retrospective study of stage III-IV NSCLC patients treated with radiotherapy, immune checkpoint inhibitors, or a combination of both demonstrated that pneumonitis occurs more frequent in the RT + ICI group. ICI-related pneumonitis was more bilateral, involving multiple lobes, and more often required high-dose steroids. Bronchiectasis or fibrosis were more seen in RT-induced pneumonitis. Patients with a pre-existing interstitial lung disease or reduced baseline diffusion capacity were at higher risk of developing severe pneumonitis, which tended to present earlier than the low-grade variant. Radiologically, an organizing pneumonia (OP) pattern was most often associated with low- grade pneumonitis, whereas acute interstitial pneumonia (AIP), and to a lesser extent usual interstitial pneumonia (UIP) and nonspecific interstitial pneumonia (NSIP) predominated the high-grade cases. Finally, the quantitative imaging findings suggest increased lung injury within the 10 Gy isodose region when radiotherapy is combined with immune-checkpoint inhibitors, suggesting RT + ICI pneumonitis remains a localized phenomenon.

Authors contribution

Prof dr. Els Wauters and dr. Xander Verbeke conceived the study. Dr. Xander Verbeke participated in data collection. Prof. dr Walter De Wever performed the radiologic assessments. Dr. Gilles Defraene and prof. dr. Maarten Lambrecht performed the quantitative CT feature extraction. Dr. Xander Verbeke and dr. Gilles Defraene performed the data analyses. Dr. Xander Verbeke and dr. Gilles Defraene wrote the manuscript. All authors read and approved the final manuscript.

CRediT authorship contribution statement

Xander Verbeke: Writing – review & editing, Writing – original draft, Validation, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Gilles Defraene: Writing – review & editing, Writing – original draft, Visualization, Software, Investigation, Formal analysis, Data curation. Walter De Wever: Formal analysis. Maarten Lambrecht: Writing – review & editing, Visualization, Supervision. Els Wauters: Writing – review & editing, Supervision, Methodology, Formal analysis, Conceptualization.

Funding

Dr. Gilles Defraene is postdoctoral fellow of the Research Foundation Flanders (FWO, project 1292021 N).

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.