Prognostic significance of the radiologic features of pneumonitis induced by anti‐PD‐1 therapy

Abstract Background Interstitial lung disease (ILD) induced by anti‐programmed‐cell death‐1 (PD‐1) and anti‐PD‐ligand 1 (PD‐L1) is potentially life‐threatening and is a common reason of the discontinuation of therapy. In contrast, an enhancement in antitumor effects was reported in patients who developed immune‐related adverse events, including ILD. Although recent evidence suggests that radiologic patterns of ILD may reflect the severity of ILD and the antitumor immune responses to anti‐PD‐1/PD‐L1 therapies, the association between radiologic features and clinical outcomes remains unclear. Methods Patients with advanced non‐small‐cell lung cancer who were treated with 1st to 3rd line anti‐PD‐1 therapy from January 2016 through October 2017 were identified at multiple institutions belonging to the Niigata Lung Cancer Treatment Group. ILD was diagnosed by the treating physicians, and chest computed tomography scans were independently reviewed to assess the radiologic features of ILD. Results A total of 231 patients who received anti‐PD‐1 therapy were enrolled. Thirty‐one patients (14%) developed ILD. Sixteen patients were classified as having ground glass opacities (GGO), 16 were classified as having cryptogenic organizing pneumonia (COP), and one was classified as having pneumonitis not otherwise specified. Patients with GGO had significantly worse overall survival time compared to patients with COP (7.8 months (95% CI: 2.2‐NE) versus not reached (95% CI: 13.2‐NE); P = 0.0175). Multivariate analysis of all 231 patients also revealed that PS = 1 and ≥2 and GGO were significant predictors of a worse overall survival. Conclusions This study demonstrated that patients who developed GGO exhibited worse outcomes among non‐small‐cell lung cancer patients receiving anti‐PD‐1 therapies.


| INTRODUCTION
Immune checkpoint inhibitors (ICIs), including anti-programmed-cell death-1 (PD-1) and anti-PD-ligand 1 (PD-L1) antibodies, have demonstrated promising and durable benefits in non-small-cell lung cancer (NSCLC) patients. [1][2][3][4][5] Anti-PD-1 and anti-PD-L1 antibodies have become the new standard of care for NSCLC patients. Because interactions between PD-L1and PD-1 maintain immune tolerance to normal tissues peripherally, blockade of PD-L1 and PD-1 results in the activation of the immune system and often causes immune-related adverse events (irAEs). 6 Previous studies have shown that clinical efficacies of anti-PD-1/ PD-L1 therapies were augmented in patients with irAEs, suggesting that anti-PD-1/PD-L1 antibodies are capable of stimulating immune responses against both normal cells and tumor cells. 7,8 Previous phase III studies have reported that the incidence of interstitial lung disease (ILD) induced by anti-PD-1/PD-L1 therapies was 1%-6%. [1][2][3][4][5] ILD is potentially fatal and often results in the discontinuation of therapy. 3,9 Indeed, several studies have shown that ILD was the most common irAE leading to the discontinuation of anti-PD-1 therapies. 1,2 On the other hand, the augmentation of antitumor effects was reported in patients with ILD, similar to other irAEs. [10][11][12] Several radiologic patterns of ILD caused by ICIs have been reported. [13][14][15] Although these radiologic patterns may reflect the severity of ILD and the antitumor immune response enhanced by anti-PD-1/PD-L1 therapies, the correlation between radiologic patterns and clinical outcomes, especially the augmentation of antitumor effects, remains unclear. The aim of this study is to elucidate the correlation between the radiologic features of ILD induced by anti-PD-1 antibodies and clinical course of patients with advanced NSCLC.

| Study design and patients
The medical records of all consecutive patients with advanced NSCLC who were treated with anti-PD-1 as 1st to 3rd line therapy at multiple institutions belonging to the Niigata Lung Cancer Treatment Group from January 2016 through October 2017 were retrospectively reviewed. We started this study after approval from the institutional review board of each participating institution.

| Study assessment
For all patients, the following data were collected retrospectively: demographics, oncologic therapy including anti-PD-1 therapy, and irAEs. For patients with ILD, the clinical features of ILD and treatment for ILD were obtained. ILD was diagnosed by the treating physicians. The chest computed tomography (CT) scans of patients with ILD were obtained and independently reviewed by two pulmonologists (MH and SW) and one radiologist (HI) blinded to patient clinical data. ILD was classified into 5 subtypes: cryptogenic organizing pneumonia (COP), ground glass opacities (GGO), interstitial, hypersensitivity and pneumonitis not otherwise specified (NOS) according to previous reports. 14,[16][17][18] If there was disagreement, a consensus read was performed. ILD was graded according to the Common Terminology Criteria for Adverse Events version 4.0. Progression-free survival (PFS) was calculated from the start of anti-PD-1 therapy until progressive disease or death due to any cause. Overall survival (OS) was measured as the time between the start date of anti-PD-1 therapy and death due to any cause.

| Statistical analysis
Kaplan-Meier survival curves were constructed for PFS and OS, and differences between groups were identified using the log-rank test. Continuous variables are presented as the median (range) and were compared by two-sided t tests. Categorical variables were compared by Fisher's exact test or chi-square test. To evaluate prognostic factors for OS, we used multivariate Cox proportional hazards models. Multivariable analysis was performed, including age, sex, smoking status, histology, Eastern Cooperative Oncology Group performance status, irAE and radiologic features. To minimize lead-time bias associated with time-dependent factors, we performed landmark analysis including only patients Conclusions: This study demonstrated that patients who developed GGO exhibited worse outcomes among non-small-cell lung cancer patients receiving anti-PD-1 therapies.

K E Y W O R D S
immune checkpoint inhibitors, immune-related adverse event, interstitial lung disease, NSCLC, PD-1 who were alive or whose disease was under control at 43 days after anti-PD-1 therapy, which is the median time of onset of ILD, for OS (n = 214) and PFS (n = 172). Additionally, we performed landmark analysis at 6 weeks after anti-PD-1 therapy to evaluate the difference in OS between patients with and those without irAE using multivariate Cox proportional hazards models. All the reported p-values were 2-sided, and P < 0.05 was considered significant. Statistical analysis was performed using JMP 9.0.2 statistical software (SAS Institute, Cary, NC).

| Patient characteristics
Overall, 231 patients were enrolled in this study. Among these patients, 33 patients (14%) developed ILD (CONSORT diagram [ Figure 1]). Baseline characteristics at the initiation of anti-PD-1 therapy for patients with and those without ILD are presented in Table 1. The frequency of squamous carcinoma was significantly higher in patients with ILD than in patients without ILD. There were no significant differences in terms of age, sex, smoking status, Eastern Cooperative Oncology Group performance status, line of anti-PD-1 therapy and type of anti-PD-1 antibody.

| Baseline characteristics in patients with ILD
Of 33 patients with ILD, 16 were classified as having GGO, 16 were classified as having COP, and one was classified as having pneumonitis NOS (Table 1). Representative chest CT scans of patients with GGO, COP and pneumonitis NOS are shown in Figure S1. Table 2 presents the characteristics of patients with ILD. The number of patients who received nivolumab was significantly higher in patients with COP than in those with GGO (P = 0.0068). Patients with GGO received anti-PD-1 therapy earlier than those with COP (P = 0.0055). The expression levels of PD-L1 on tumor cells were significantly higher in patients with GGO than that in patients with COP (P = 0.0496). The median time of onset of ILD was significantly earlier among patients with GGO than among patients with COP (P = 0.041). The number of patients who received steroid therapy was significantly higher in the GGO group than in the COP group (P = 0.029). Most patients were successfully treated with corticosteroids, but one patient with GGO died from ILD. There were no significant differences in terms of age, sex, smoking status, PS, overall response rate, grade of ILD, other irAEs, or number of patients who received subsequent chemotherapy after anti-PD-1 therapy.

| Radiographic patterns of ILD and OS
The Kaplan-Meier survival curves for patients with COP, GGO and pneumonitis NOS are shown in Figure 3A. The OS was significantly worse among patients with GGO than among those with COP (7.8 months (95% CI: 2.2-NE) versus not reached (95% CI: 13.2-NE); P = 0.0175). There were no significant difference in PFS between patients with GGO and those with COP (7.6 months (95% CI: 0.9-NE) versus not reached (95% CI: 2.9-NE); P = 0.1602) ( Figure S2). Previous studies indicated that the use of steroids at baseline decreased the clinical benefit of anti-PD-1 therapy. 19 Thus, we evaluated whether steroid administration for the treatment of ILD decreased the antitumor effects of anti-PD-1 therapy. Steroid therapies for ILD did not affect the survival times (10.1 months (95% CI: 3.1-NE) versus not reached (95% CI: 5.3-NE); P = 0.3202) ( Figure 3B). Because a correlation between the development of irAEs and a clinical efficacy of anti-PD-1 therapy was reported, we assessed OS in patients with irAE other than ILD. 7,8 The OS for patients with irAE other than ILD was significantly better than that for patients with ILD (not reached (95% CI: NE-NE) versus 14.8 (95% CI: 7.6-NE); P = 0.0052) ( Figure 3C). As shown in Table 2, the median time of onset of ILD was significantly earlier in patients with GGO than in patients with COP; thus, we evaluated OS after the onset of ILD ( Figure 3D). The OS after the onset of ILD in patients with GGO tended to be shorter than that in patients with COP (7.3 months (95% CI: 1.7-NE) versus not reached (95% CI: 5.5-NE); P = 0.0503).

| DISCUSSION
The current study demonstrated the relationship between the radiologic patterns of ILD caused by anti-PD-1 therapy and OS in NSCLC patients. Patients with GGO had a significantly

T A B L E 1 Patient Characteristics at
PD-1 therapy worse survival time than those with COP ( Figure 3A). Multivariate analysis also demonstrated that the development of GGO was significantly correlated with poor prognosis (Table 3). In contrast, COP was not correlated with OS (Table 3). Previous studies did not clearly show the association between the development of ILD and patient outcomes after anti-PD-1 therapy. Fujimoto et al reported that NSCLC patients with ILD had longer PFS with nivolumab therapy. 10 On the other hand, Shresh et al showed that the development of ILD decreased survival in NSCLC. 20 Our study demonstrated that patients with ILD exhibited prolonged PFS; however, patients with ILD had similar OS compared to patients without ILD (Figure 2A and B). Anti-PD-1 therapies cause several types of radiologic patterns of ILD, and our results indicate that patients with ILD could be divided into good and poor prognosis groups according to radiologic features. As there is no laboratory test to determine the severity of ILD, we must consider intensive therapy for ILD based on clinical findings, including cough, fever, dyspnea and hypoxemia. Our findings on the correlation between radiologic patterns and clinical outcomes will help physicians make decisions regarding the administration of steroids and immunosuppressants for ILD. There are several possible explanations for the worse prognosis of patients with GGO. The time of onset of GGO was significantly earlier than that of COP, and the median treatment cycles of anti-PD-1 therapy tended to be shorter in patients with GGO than in patients with COP (Table 2). Moreover, 12 out of 16 patients with GGO discontinued anti-PD-1 therapy after no more than 3 cycles. The relationship between the discontinuation of PD-L1/PD-1 therapy and poor outcome has been indicated previously. 21,22 Early discontinuation of anti-PD-1 therapy could be correlated with worse prognosis in patients with GGO. Another explanation may be the exposure to systemic steroids. Patients with GGO were treated with steroids significantly more frequently than patients with COP (Table 2). Although our data did not demonstrate that steroid therapy decrease OS among patients with ILD, previous studies have demonstrated that patients treated with systemic steroid during the first cycle of nivolumab had a shorter survival time ( Figure 3B). 19 Previous studies have also demonstrated that patients with irAEs have a more favorable prognosis than patients without irAEs. 7,8 Similar to these studies, the current study showed that patients with irAEs other than ILD had significantly better survival times compared to patients with ILD ( Figure 3C).
The limitations of the current study include a relatively small number of patients with ILD and its retrospective nature. Although all medical records of consecutive NSCLC patients receiving anti-PD-1 therapy were assessed, radiology review was performed with only clinically determined ILD cases, not all cases treated with anti-PD-1 therapy. In our study, no patients were diagnosed with ILD based on pathological evidence.
To the best of our knowledge, this is the first report to demonstrate the correlation between radiologic features of ILD and OS in NSCLC patients. Our findings suggest that patients with GGO require intensive therapy more often than patients with COP. A prospective study is required to establish the treatment strategy for ILD according to radiologic patterns. The fact that some patients with GGO responded to anti-PD-1 therapy indicates that damage to normal lung tissues might be different even in patients with similar radiologic features of ILD. Evaluation of lung tissues from patients with ILD and assessment of the correlation between radiologic patterns and histology are warranted. Abbreviations: CI, confidence interval; COP, cryptogenic organizing pneumonia; GGO, ground glass opacities; ILD, interstitial lung disease; irAE, immune-related adverse events; PS, performance status.