Molecular and Clinical Features of Hospital Admissions in Patients with Thoracic Malignancies on Immune Checkpoint Inhibitors

Simple Summary Lung cancer immunotherapy has many complications and hospitalizations that often occur in non-small cell lung cancer (NSCLC) while on immunotherapy due to adverse events or other factors. The molecular and clinical profiles of these patients are often not well-defined, and the aim of our retrospective study is to better understand these clinical and molecular features. We evaluated a cohort of 90 stage IV thoracic malignancy patients who had hospital admissions after treatment with immune checkpoint inhibitors. We identified a relationship between immune-related adverse events (irAEs) and molecular markers that showed unique survival outcomes, as well as a significant overall survival improvement in patients who required discontinuation or interruption of immunotherapy due to irAEs. Abstract Lung cancer patients undergoing systemic treatment with immune checkpoint inhibitors (ICIs) can lead to severe immune-related adverse events (irAEs) that may warrant immediate hospitalization. Patients with thoracic malignancies hospitalized at City of Hope while undergoing treatment with ICIs were identified. Pathology and available next-generation sequencing (NGS) data, including the programmed death-ligand 1 (PD-L1) status and clinical information, including hospitalizations, invasive procedures, and the occurrence of irAEs, were collected. Unpaired T-tests, Chi-square/Fisher’s exact test, and logistic regression were used to analyze our cohort. The overall survival (OS) was calculated and compared using univariate and multivariate COX models. Ninety patients with stage IV lung cancer were admitted after ICI treatment. Of those patients, 28 (31.1%) had documented irAEs. Genomic analyses showed an enrichment of LRP1B mutations (n = 5/6 vs. n = 7/26, 83.3% vs. 26.9%; odds ratio (OR) (95% confidence interval (CI): 13.5 (1.7–166.1); p < 0.05) and MLL3 mutations (n = 4/6, 66.7% vs. n = 5/26, 19.2%; OR (95% CI): 8.4 (1.3–49.3), p < 0.05) in patients with irAE occurrences. Patients with somatic genomic alterations (GAs) in MET (median OS of 2.7 vs. 7.2 months; HR (95% CI): 3.1 (0.57–17.1); p < 0.05) or FANCA (median OS of 3.0 vs. 12.4 months; HR (95% CI): 3.1 (0.70–13.8); p < 0.05) demonstrated a significantly shorter OS. Patients with irAEs showed a trend toward improved OS (median OS 16.4 vs. 6.8 months, p = 0.19) compared to hospitalized patients without documented irAEs. Lung cancer patients who required treatment discontinuance or interruption due to irAEs (n = 19) had significantly longer OS (median OS 18.5 vs. 6.2 months; HR (95% CI): 0.47 (0.28–0.79); p < 0.05). Our results showed a significant survival benefit in lung cancer patients hospitalized due to irAEs that necessitated a treatment interruption. Patients with positive somatic GAs in MET and FANCA were associated with significantly worse OS compared to patients with negative GAs.

However, hospital admission during treatment is common in cancer patients undergoing systemic treatment. Distinct toxicity profiles and immune-related adverse events (irAEs) due to ICIs have been widely reported, including skin reactions, thyroid disorders, pneumonitis, colitis, hepatitis, hypophysitis, and myocarditis [15,16]. Other severe adverse events not related to ICIs can emerge as well during treatment and lead to hospitalization. In a meta-analysis of 35 clinical trials involving ICIs, irAEs of grade 3 and above were reported in 14% of patients treated with monotherapy ICIs, 34% with anti-CTLA-4 antibodies, 46% with combination ICI-chemotherapy, and 55% with ICIs combinations [17].
Several prominent oncologic societies, such as the National Comprehensive Cancer Network (NCCN), American Society of Clinical Oncology (ASCO), and Society for Immunotherapy of Cancer (SITC), have published guidelines on the management of irAEs in the standard clinical setting. However, irAE management in patients who require hospitalization and are steroids-refractory remains problematic [19][20][21]. The characterization of clinical features regarding irAEs and non-irAEs in hospitalized patients may facilitate the understanding and management of toxicities in this setting.
Previous studies have described associations between several tumor genomic features and the tumor response to ICIs. Notably, a poor tumor response was reported in patients on ICIs with molecular alterations in EGFR or MET [22][23][24]. In addition to a poor response, the development of severe irAEs (especially within 3 months) has been described in a retrospective analysis of EGFR mutated NSCLC patients (15%; 6/41) treated with ICIs followed by osimertinib, although the underlying mechanisms are still poorly understood [25]. However, the development of irAEs has not been established as a predictive marker in measuring responsiveness to ICIs.
The clinical characterization of irAEs and full assessment of genomic data is necessary to optimize the patient selection criteria for ICI treatment, understand the underlying mechanisms of irAE development, and develop novel strategies to avoid irAEs while maintaining the anti-tumor efficacy [26]. In our retrospective analysis, we collected clinical and molecular information on 90 patients diagnosed with thoracic malignancies who received ICI treatment and were subsequently hospitalized in order to characterize irAE and non-irAE development, evaluate the management of irAEs, and analyze the survival outcomes.

Patients
Patients with metastatic thoracic malignancies who were hospitalized after receiving ICI treatment (pembrolizumab, nivolumab, atezolizumab, and ipilimumab/nivolumab) in different treatment settings, including standard of care, compassionate use, and clinical trials at City of Hope were reviewed. Ninety patients with histologies, including SCLC, NSCLC, and other thoracic malignancies were identified. Demographic, clinical, and pathological information was collected with approval by the City of Hope institutional review board (IRB #18529). The overall survival (OS) was measured from the start of the ICI treatment to the date of death and calculated, if available, at the study time point. The data cutoff date was 8 November 2018.

Clinical and Molecular Information Collection
Tumor genomic alterations (GAs) were extracted from the available clinical data on next-generation sequencing (NGS) via several platforms, including FoundationOne (Foundation Medicine, Cambridge, MA, USA), Caris (Caris Life Sciences, Phoenix, AZ, USA), Paradigm (Paradigm Diagnostics, Phoenix, AZ, USA), Guardant360 (Guardant, Redwood City, CA, USA), NeoGenomics (NeoGenomics Laboratories, Fort Myers, FL, USA), and City of Hope gene sequencing panels. The PD-L1 (22C3) expression by immunohistochemistry was reported as the tumor proportion score (TPS), which is defined as the percentage of viable tumor cells showing partial or complete membrane staining of ≥1% relative to all viable tumor cells present in the sample.
Negative PD-L1 expression was defined as <1% of viable tumor cells showing membranous staining. The tumor mutational burden (TMB) was reported and categorized as low (≤5 Muts/Mb), intermediate (6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19) Muts/Mb), or high (≥20 Muts/Mb) by Foundation Medicine. Somatic GAs were sorted by the detected positive rate of GAs among all tested patients (the number of tested patients for each gene varied due to different gene panels in the testing platforms). IrAEs were defined as treatment-related toxicities documented by the admitting physician or primary oncologist and independently confirmed by another physician who reviewed the patient medical charts, including the laboratory, imaging, and pathological evidence.
The severity of irAEs was documented from grade 1 to 5 as per the National Institute of Health Common Terminology Criteria for Adverse Events (CTCAE), version 4.03. Clinical information, such as lines of therapy; the length of stay (LOS) in hospital; the status of metastatic disease in the brain; the therapy regimen of ICIs; the management of irAEs, including invasive, diagnostic, and therapeutic procedures; and any interruption or discontinuation of ICIs due to irAEs was also collected.

Statistical Analysis
First, the association of clinical and molecular features with the OS was analyzed using the univariate Cox proportional hazards model. Based on the results of the univariate analysis, clinically and biologically relevant features with statistical significance (cutoff p-value of 0.05 with the number of patients, n ≥ 5) were selected for the multivariate Cox proportional hazards model. TMB was categorized by Foundation Medicine molecular testing reports. PD-L1 expression was categorized as negative (<1%), and positive (grouped as 1-49% and ≥50%).
We used the Kaplan-Meier method and log-rank test to estimate the OS, and we compared the survival curves, respectively. The chi-square test, Fisher's exact test, and logistic regression were used for comparison between patient groups (i.e., patients who had irAEs vs. patients who did not have irAEs). Statistical analyses and data visualization were performed using GraphPad Prism 8 (GraphPad Software, Version 8, Graphpad Holdings, LLC, San Diego, CA, USA) and R (version 3.6.2, R Foundation for Statistical Computing, Vienna, Austria) [27]. All tests were two-sided, and p < 0.05 was considered statistically significant.
However, no statistically significant difference was found in the MLL3 or LRP1B mutation status corresponding with irAE occurrence in our multivariate logistic regression analysis ( Table 5). The most frequent GAs and patient demographic information are visualized in the oncoplot in Figure 1

Discussion
The use of ICIs in lung cancer treatment has drastically improved the outcomes of advanced NSCLC patients with an average five-year OS of 15.6% with nivolumab and 23.2% with pembrolizumab as a first-line therapy [28,29]. However, patients who undergo ICI treatment can experience hospital admissions due to severe irAEs and/or other comorbidities. As researchers continue to investigate ICI treatment in earlier-stage disease, it is necessary to explore strategies in minimizing toxicities and avoiding severe irAEs that could be long-lasting or fatal. In this study, we analyzed 90 patients with thoracic cancers who were hospitalized during ICI treatment. Of those, 28 patients (31.1%) experienced irAEs with the most common irAE being pneumonitis (n = 10/90, 11.1%). This is consistent with other reports demonstrating that 12% of emergency room visits and inpatient care were associated with irAE development in metastatic solid tumor patients undergoing ICI treatment [30]. This result is also consistent with a previous study that reported immune-related interstitial pneumonia as the most common irAE in 13.2% (n = 5/38) of lung cancer patients treated with nivolumab [31].
We also reported that patients with documented irAEs underwent more invasive diagnostic procedures but with no observed difference in the hospital LOS. The severity of irAEs may have caused further intensive interventions due to the risk of long-lasting effects. Sattar et al. described a correlative study between irAEs and efficacy in an older patient population treated with ICIs, and patients age ≥75 years did not present with excess toxicities [32], consistent with our findings of no associations between irAE development and age.
However, we did not observe an OS benefit between our irAE and non-irAE populations. Previous studies have demonstrated superior progression-free survival (PFS) and OS in patients with irAEs, while our study only demonstrated a trend toward significance for OS in our irAE group [31,[33][34][35]. In a large observational study, Grangeon et al. measured the survival outcomes in 270 patients with metastatic NSCLC treated with at least one dose of anti-PD-L1 or anti-PD-1 antibodies. The study stratified cohorts between patients who did and did not experience irAEs.
Interestingly, Haratani et al. showed that patients who required systemic corticosteroids for irAE management had superior survival outcomes, while Shafqat et al. [35]. demonstrated that irAEs were associated with improved PFS regardless of systemic corticosteroids use [35,37]. The 19 patients who had discontinuation or interruption of ICIs due to irAEs had significantly longer OS, which implied the positive correlations of irAEs with survival outcomes. However, in clinical practice, people might be more comfortable to stop treatment when their disease is better-controlled; therefore, this might be a highly selective patient subpopulation.
Our results demonstrated an OS benefit for patients who underwent ICI treatment as first-line compared to second-line or greater (p < 0.01). A study by Durbin et al. confirmed our results by showing a shorter OS in metastatic solid tumor patients who underwent ICI treatment as second-line or greater [38]. However, another study also analyzed the safety and efficacy of ICIs as second-line treatment in a real-world setting. Chen et al. described the association between the occurrence of irAEs and higher PFS in a patient population who received ICIs in the second-line setting and concluded that the presence of irAEs may act as a predictive marker for antitumor efficacy [39].
Next, our study revealed that patients who experienced an interruption of ICI treatment due to irAEs had significantly longer OS than those who continued treatment (p < 0.05), suggesting a positive correlation between irAE occurrence and survival outcomes. Conversely, Ksienski et al. showed that treatment interruptions in NSCLC patients undergoing treatment with pembrolizumab or nivolumab due to documented irAEs (n = 116/271, 42.8%) were associated with a worse OS [40].
A correlative study by Mouri et al. retrospectively analyzed 49 NSCLC patients treated with nivolumab that had treatment interruption due to a serious irAE. With patients stratified into a retreatment or discontinuation cohort, patients rechallenged with nivolumab displayed an ORR of 15%, without a significant increase in irAEs; however, the median OS and PFS did not differ significantly among the patient cohorts [41].
The difference among survival outcomes with varying ICIs used for treatment may also play a role in discontinuation if the patient experiences detrimental irAEs. Lastly, Jia et al. describe varying biomarkers that can predict irAEs based on specific and nonspecific symptoms. Due to irAE effects in every organ, ongoing investigation in regards to the application scope, benefit from treatment interruption, and selection of the treatment population for ICIs based on biomarkers is required [42].
Our analysis reported enrichment of somatic LRP1B and MLL3 mutations in patients with irAEs. Yet, it was not statistically significant in the multivariate analysis, likely due to the limited sample size and confounding factors of smoking and gender. LRP1B gene encodes for an LDL receptor and acts as a putative tumor suppressor in lung cancer whose function is only partially defined [43,44]. Chen et al. reported greater survival and higher TMB in melanoma and NSCLC patients with LRP1B mutations undergoing ICI treatment [45]. MLL3 gene encodes for histone 3 lysine 4 methyltransferases and acts as a tumor suppressor.
Mutated MLL3 (or KMT2C) proteins have been implicated in multiple cancers, including urothelial carcinoma, human lymphoid, and myeloid leukemia [46][47][48]. Further, we found that NSCLC patients with FANCA mutations had significantly worse OS compared to those without FANCA mutations. The FANCA gene is important for the repair of double-stranded DNA breaks and is involved in the cellular process known as the Fanconi anemia pathway. Outcomes have been assessed in patients treated with ICIs, and the results showed significantly higher objective response rate, longer median PFS, and longer median OS with patients on PD-(L)1 therapy [49].
The survival outcome that we observed was also previously confirmed in a larger cohort by our group, yet further investigation is required [50]. The MET gene encodes a transmembrane receptor tyrosine kinase, and its ligand hepatocyte growth factor (HGF) is involved in the MET/HGF signaling pathway. Patients in our cohort with MET GAs were associated with poor OS, and the statistical significance was retained in the multivariate analysis. This is consistent with the previous findings of MET mutated lung cancer and worsened outcomes with immunotherapy treatment [24,51]. We did not observe a correlation between TMB with irAEs or OS, which may be explained by the lack of TMB information in our cohort as only 18 patients had TMB information available. It is unclear how somatic mutations in tumors contribute to the development of irAEs, and more research is warranted to examine the role of genomic mutations in lung cancer immunotherapy.

Conclusions
Our retrospective analysis investigated the clinical and molecular features of lung cancer patients undergoing ICI treatment who were hospitalized. We observed that patients with irAE occurrences who required treatment interruption had a significantly longer OS. Further, patients with somatic GAs in FANCA and MET had a worse OS, which is consistent with previously reported studies. A limitation of this study is that the patient cohort was of limited sample size and from a single institution. Further investigation is required to analyze a larger and diverse population set. Strikingly, our findings indicated that the majority of patients who were hospitalized on ICI treatment did not have an irAE. Therefore, future clinical studies should focus on identifying and cataloging the variables that may be associated with hospitalization due to ICI treatment.