Incidence and risk factors of immune‐related adverse events induced by immune checkpoint inhibitors among older adults with non‐small cell lung cancer

Abstract Background Immune checkpoint inhibitor (ICI) treatment has been linked to a variety of immune‐related adverse events (irAEs), which can affect any organ system. The incidence and risk factors of irAEs have not been adequately evaluated among older adults with NSCLC. Methods A cohort study was conducted using 1999–2019 SEER‐Medicare data among beneficiaries aged ≥65 years with a diagnosis of NSCLC who received nivolumab, pembrolizumab, or atezolizumab. Incident irAEs were identified post‐ICI initiation. Demographic, cancer‐related characteristics, and clinical history risk factors of irAEs were evaluated with competing events considered. Results A total of 8175 older NSCLC patients were included (with 46.8% experiencing irAEs). Pneumonitis (16.5%), hypothyroidism (10.5%), arrhythmia (11.18%), and acute kidney injury (AKI) (5.8%) were the most common irAEs. The median time to first irAE was 82 days (IQR: 29–182 days). The earliest onset of irAE occurrence was for hematologic irAEs, while the latest were gastrointestinal, dermatologic, and musculoskeletal irAEs. Fine–Gray regression modeling revealed significantly greater hazards of irAE occurrence in patients who received pembrolizumab at index, did not have CNS metastases, had a history of autoimmune disorder, and had chemotherapy in combination with ICI. Race, socioeconomic status, previous radiation therapy, and comorbidity burden were found to be associated with the occurrence of certain type of irAEs. Conclusion A significant proportion of older patients with NSCLC develop an irAE after receiving ICI treatment. Factors related to cancer and treatment as well as demographics contribute to the increased risk of irAEs. Close monitoring and prediction of irAE among older patients receiving ICI is warranted.


| INTRODUCTION
Immune checkpoint inhibitors (ICIs) have rapidly become the new standard of care for both first-and second-line treatment of advanced non-small-cell lung cancer (NSCLC). 1,25][16][17] Most irAEs are mild to moderate in severity 18 ; however, fatal irAEs are reported in 0.3%-1.3% of treated patients, with a majority being myocarditis, followed by pneumonitis, hepatitis, myositis, nephritis, neurologic, and hematologic adverse effects. 19espite a growing body of research focused on ICIinduced irAEs, the toxicity profile in older patients with NSCLC remains unclear.Older adults are usually underrepresented in RCTs due to age restriction, the presence of comorbidities, poor performance status, or lack of sufficient social support. 20Only 31.7%-41.0% of RCT participants are older than 65 years, while more than 60% of patients with NSCLC who received ICIs in realworld practice are older than 65. 21Moreover, existing observational studies targeting older adults with NSCLC are limited by small sample sizes, 6,7,15 short follow-up time, 6,7 or single ICI regimens. 15,224][25] The management of irAEs for older patients is further challenged by potential comorbidity, frailty, dysfunction of organs, and polypharmacy. 20,25It is necessary to understand the nature, characteristics, and impact of irAEs for older NSCLC patients undergoing ICI therapy in routine practice at a population level.This population-based study aimed to characterize the incidence and clinical features of ICI-induced irAEs in a large cohort of older adults with NSCLC.Demographic, cancer-related characteristics, and clinical history risk factors of irAE incidence are also evaluated.

| Study design and data source
Using the Surveillance, Epidemiology, and End Results (SEER)-Medicare linked database, a cohort study was conducted to characterize features of irAEs as well as assess their risk factors among Medicare beneficiaries with NSCLC treated with ICIs.The SEER database collects newly diagnosed cancer patients' data from 21 cancer registries across various geographical areas of the United States covering approximately 34.6% of the US population.The SEER-Medicare linked database comprises population-based data of cancer patients in the SEER program and their matched Medicare administrative claims data, which allows a longitudinal assessment of diagnoses, treatments, and service use before and after the cancer diagnosis.This study used the 2020 linkage dataset, which includes 1999-2017 SEER data linked with 1999-2019 Medicare claims data.This study was approved by the Institutional Review Board at the University of Mississippi (protocol #21-036).

| Study population
This study included Medicare beneficiaries who entered the SEER registry from 1999 through 2017, were diagnosed with NSCLC as primary site of cancer, received ICI treatment, and were at least 65 years of age at their first diagnosis with NSCLC based on International Classification of Diseases for Oncology, third edition (ICD-O-3) codes.Patients were excluded if the diagnosis of NSCLC was made by death certificate or autopsy.Patients were identified as having received ICI therapy if they received at least one cycle of either pembrolizumab, nivolumab, or atezolizumab from March, 2015 (the time that the first ICI agent was approved by the FDA for treating NSCLC) to December, 2018, either as monotherapy or combined with chemotherapy or other anticancer agents.Healthcare Common Procedure Coding System (HCPCS) codes (C9453 and J9299 for nivolumab, C9027 and J9271 for pembrolizumab, and C9384 and J9022 for atezolizumab) were used to identify the administration of ICIs.The date of medical claim indicating the first dose of ICI administration defined the index date for each patient.To ensure that patients were incident users of ICIs for treating NSCLC, we excluded those who had any use of ICIs prior to March, 2015.Continuous enrollment in Medicare Part A, Part B, and Part D was required from the month of diagnosis or for at least 12 months before the first dose of ICI, whichever was longer, through the index date.Patients were excluded if they were found to be enrolled in a health maintenance organization (HMO) in the pre-ICI initiation period.The model treated disenrollment from Medicare Part A, B, and D coverage, enrollment in HMO, switching to other ICI agents, or reaching the end of the study period without an event (December, 2019) as censored.

| Outcome of interest
The outcome variable for this study was the incident occurrence of irAEs post the initiation of ICI treatment.Immune-related adverse events are defined as unique adverse events that have an autoimmune basis requiring more frequent monitoring and potential intervention with immune suppression therapy. 26In this study, an irAE was defined as present if a patient had a new medical claim diagnosis (no same event occurred within 12 months before the onset of irAE) with a corresponding irAE diagnosis code 16 S1) were used to identify any and each specific type of irAE by scanning the primary or secondary diagnoses on inpatient claims (Medicare Part A) and physician visit claims (Medicare Part B).The primary endpoint of this study was the occurrence of incident irAEs (any or specific type), and the secondary endpoint was the timing of irAE incidence after ICI initiation.

| Risk factor variables
Risk factors for irAE incidence evaluated in this study included patient demographics, cancer-related characteristics, and clinical history.Patients' demographic information included age, sex, race, ethnicity, marital status, geographic region, urban residence, census tract poverty status, and dual Medicare and Medicaid eligibility.Cancer-related risk factors included time from first NSCLC diagnosis to ICI initiation, cancer stage at the time of diagnosis, tumor histology at diagnosis, and the presence of central nervous system (CNS) metastasis (identified in the medical claims based on diagnosis codes) in a 6-month period before ICI initiation.Cancer treatment history was also identified, including NSCLC-related surgery and radiation therapy from NSCLC diagnosis to ICI initiation.Adjuvant chemotherapy with index ICI treatment was identified if received within the 28 days (±14 days of index date) of the first administered ICI.The treatment line of ICI regimens was determined based on a published algorithm by Hess et al. 27 In addition, other important clinical history variables, such as any history of autoimmune disorders, the NCI Comorbidity Index, 28 and disability status, 29,30 were captured from the medical claims within 12 months before the initiation of ICIs.
Lastly, recent use of corticosteroids was measured in the 3-month period prior to the index date.

| Statistical analysis
Descriptive statistics were used to depict baseline patient characteristics and irAE characteristics.t-Tests or Mann-Whitney U tests were used to assess differences in continuous baseline variables between patients with irAEs and patients without irAEs, as appropriate.Chi-squared tests were used to assess differences among categorical variables between the two groups, as appropriate.Given death or initiation of a subsequent line of treatment may occur before patients develop an irAE, the occurrence of these events would preclude the observation of the irAEs.The competing risk of these events might bias the estimation of predictive effect of covariates in a conventional survival analysis.Therefore, the Fine-Gray regression model 31 was used to assess the association between risk factors and the time to develop the first occurrence of irAEs (any type) adjusting for competing risk of all-cause death and advancing to subsequent line of therapy in the cohort of patients with NSCLC who received ICI treatment.The subsequent line of therapy post-ICI-containing regimens was identified if patients started single-agent chemotherapy (e.g., pemetrexed, docetaxel, or gemcitabine), other ICI agents, or docetaxel plus ramucirumab.To further analyze the risk factors associated with first occurrence of a specific type of irAEs, separate multivariable models were estimated for select types of irAEs, including pneumonitis, hypothyroidism, arrhythmia, and acute kidney injury (AKI).All analyses were performed using SAS version 9.4 (SAS Institute).

| DISCUSSION
In our study, nearly half of the older patients with NSCLC experienced at least one irAE after ICI initiation.
Pneumonitis, arrhythmia, hypothyroidism, and AKI were among the most common irAEs.The median interval from ICI initiation to irAE onset was found to be around 3 months, but one-fourth of patients developed their first irAE after 6 months.After accounting for competing risks of death and treatment discontinuation, greater hazards of irAE incidence were particularly associated with patients who received certain type of ICI at index, did not have CNS metastases, had a history of autoimmune disorder, and had chemotherapy in combination with ICI for older adults with NSCLC.Unique risk factors were also found specifically be associated with the occurrence of certain types of irAEs.To the best of our knowledge, this is the first study to investigate the incidence and risk factors of irAE incidence among older Medicare beneficiaries who have been diagnosed with NSCLC and initiated treatment with ICIs.
Few studies have investigated the incidence of irAEs after ICI initiation among older patients with NSCLC in a real-world setting.A study by Muchnik et al. conducted on 75 older patients with NSCLC from one academic hospital found that 37% experienced irAEs of any grade. 7neumonitis and thyroiditis were the most common irAEs, followed by colitis and dermatitis. 7Our findings are consistent with this spectrum, although we observe a higher occurrence of incident irAEs in the current study.This may be due to the discrepancy in study design and data source.Although the limitations of the database used for the current study prevent robust direct comparisons between older and younger adults, it is worth noting that the incidence of any irAEs in our study was similar to that reported in a large retrospective study of younger NSCLC patients. 163][34] Older NSCLC patients in our study, however, were observed to have a higher incidence of pneumonitis, hypothyroidism, AKI, and arrhythmia compared with the study evaluating younger patients. 16his suggests unique characteristics of the incidence irAEs among older patients that may be attributed to the impairment of their immune systems or dysfunction in particular organs.It is important to monitor such events and provide early intervention during ICI treatment.In addition, our study observed a low incidence (~1%) of some irAEs, such as adrenal insufficiency, type 1 diabetes, hypophysitis, cardiomyopathy, and myocarditis, comparable to other observational studies in patients of mixed ages and tumor types, suggesting that these irAEs are similarly rare among older patients with NSCLC.These events may, however, be serious and fatal; further research is necessary to assess their severity and impact on older patients. 16,35,36parse data are available to describe the timing of irAEs in older adults with NSCLC.Based on our study, the median time to occurrence of any type of irAEs was T A B L E 2 Fine-Gray model predicting the incidence of any irAEs.82 days after ICI initiation at the median, and pneumonitis-the most common irAE-occurred 106 days after ICI initiation.This is consistent with Daniello et al.'s findings in German NSCLC patients at stage IV of any age, in which the median irAE time-to-onset from ICI start was 3.1 months, with 105 days for lung irAEs. 17Nevertheless, another real-world study, which evaluated patients with various cancers, including NSCLC, and reported a shorter time to onset (median: 9.8 weeks). 37This indicates that older patients with NSCLC may experience a relatively late onset of irAEs compared to patients with other cancers.We found that irAE incidence increased cumulatively up to 12 months and occurred even after treatment discontinuation, which is in line with the results of other retrospective studies. 16It has been reported that patients in advanced stage of NSCLC had a short time to onset of heart and liver irAEs and a longer onset of endocrine and kidney irAEs. 17Contrary to these findings, our study observed that the cardiovascular and liver irAEs developed later, but endocrine and renal toxicities were present earlier in our study cohort.This may be due to our study was conducted in a patient population with specific baseline characteristic and from a difference healthcare system.Given the fact that around half of our study sample were diagnosed with NSCLC at the nonmetastatic stage, these findings may indicate that older patients and/or patients at an earlier stage of the disease may experience a different pattern of irAEs onset.9][40][41][42] Meanwhile, we offer evidence in areas where there is controversy in the literature, including whether autoimmune disorders (AID), radiation therapy, and chemotherapy all contribute to irAEs development.A history of AID was previously reported to be associated with an increased flare-up of AID or an increased risk for irAEs. 43Yet, in a different study of NSCLC patients, there was no significant association between baseline AID and either the development, grade, or time to irAE. 44An analysis of retrospective data in Mexico showed that patients with radiation therapy treated with second-line immunotherapy had a significantly higher incidence of pneumonitis.6][47] In Daniello et al.'s study, no significant difference was found between chemoimmunotherapy and mono-ICI therapy. 17In contrast to this, a meta-analysis of RCTs in NSCLC demonstrated that the rate of most irAEs could be reduced by combining ICI with chemotherapy. 48This may reflect that the impact of autoimmune diseases, prior cancer treatment, and immunochemotherapy can be detrimental to older patients who are eligible for immunotherapy.

Risk factor
In addition, we identified a few distinctive features, which have been rarely evaluated in the literature, of predicting the occurrence of specific type of irAEs, of which the incidence was found to be higher in older NSCLC patients.First, in our cohort, the presence of CNS metastases is linked to a lower risk of irAEs, especially pneumonitis, hypothyroidism, and AKI, which are the most common irAEs in our cohort.Randomized controlled trials typically exclude patients with active CNS metastases.Research suggests that ICIs can benefit NSCLC patients who have brain metastases with acceptable toxicity. 49tudies exploring mechanisms are needed for further prospective validation.Second, our study suggests that pembrolizumab may also increase the risk of irAEs, including pneumonitis, hypothyroidism, and AKI, as compared to other agents.The choice of a treatment plan for older patients should be made with caution.Furthermore, unlike other irAEs, hypothyroidism occurs more frequently among patients with a low NCI index score and at early cancer stages.Since the incidence of hypothyroidism has been found to be associated with survival benefit, 41,50 our study provided evidence in the hypothesis that the occurrence of hypothyroidism may reflect a response to ICI.Our study observed that AKI was much more prevalent than what was reported in RCTs; however, the incidence and risk factor of it was less investigated in the current literature. 13,51,52We also discovered that irAE rates vary based on patients' socioeconomic status.Black patients, dual-enrolled Medicaid individuals, and patients residing in nonmetropolitan areas had a higher risk of developing irAEs, especially AKI.There may be a connection between low socioeconomic status of these patient populations and their vulnerability in terms of their health and less accessibility to oncologists and medicine subspecialists.
The results of this study must be viewed in the context of certain limitations.First, although patients were well-characterized in demographic characteristics and NSCLC histology variables, SEER-Medicare data do not provide information regarding PD-L1 expression level, smoking status, body mass index (BMI), or the status of other potential biomarkers.Therefore, the association between these variables and the incidence of irAEs cannot be evaluated in the current study.According to recent studies, patients' human leukocyte antigen (HLA) expression may be predictive of their risk and severity of irAEs.4][55] There is also a lack of data about irAE severity (e.g., Common Terminology Criteria for Adverse Events grading) in the SEER-Medicare linked database.It is also necessary to conduct future studies to assess the incidence and risk factors associated with severe irAEs.Furthermore, some variables in the SEER Cancer file could not be adequately analyzed due to missing information, especially from three registries.Second, the analysis was limited to NSCLC patients enrolled in fee-forservice Medicare.Consequently, findings from this study are not generalizable to older adults with NSCLC enrolled in HMOs (also known as Medicare Advantage).Third, relying on ICD-9-CM/ICD-10-CM diagnosis codes and CPT/ HCPSC procedure codes may lead to misclassification bias, underestimating irAEs, ICI use, and risk factor classification (e.g., the identification of CNS metastasis based on ICD diagnosis codes might be incomplete or inaccurate).

| CONCLUSION
This real-world study highlighted the significant high incidence and unique spectrum of irAEs in older patients with NSCLC.Race, socioeconomic status, cancer status, treatment variables, and comorbidity burden were specifically associated with increased risk of a certain type of irAEs.Accordingly, it is imperative that older adults with NSCLC be closely monitored and the occurrence of irAEs should be anticipated.
Patients with stage I-IV non-small cell lung cancer as primary site of cancer in 1999-2017 N = 693,257 Patients with index ICI claims between March 1 st , 2015 to December 31 st , 2018 N = 14,681 º Patients without continuous enrollment in Part A and Part B, N = 980 º Patients with HMO enrollment, N= 1,209 º Patients without continuous enrollment in Part D, N = 3,961 Patient without a precise record of date of death information, N = 356 Patients aged 65 or older at diagnosis with NSCLC N = 579,794 Patients aged less than 65 at diagnosis with NSCLC Final cohort for analysis N = 8,175 Patients with continuous enrollment from diagnosis to ICI initiation N = 8,531 º Patients without claim of pembrolizumab, nivolumab, or atezolizumab from Mar 2015-Dec 2018, N = 565,013 º Patients whose NSCLC diagnosis was prior to 2006, N = 63 º Patents whose index date was prior to NSCLC diagnosis, N = 42

2
Significant risk factors for predicting the incidence of specific irAEs: (A) pneumonitis, (B) hypothyroidism, (C) arrhythmia, and (D) AKI Patient characteristics and irAE incidence.aftertreatmentinitiation.A comparison between the demographic and clinical characteristics of both the irAE and non-irAE cohorts is provided in Table1.Higher incidence of irAEs was observed among patients on first-line ICI treatment at index (50.9%vs. second-line: 46.0% and third-line or later: 43.0%, p < 0.001), receiving pembrolizumab (51.7% vs. nivolumab: 44.4% and atezolizumab: 37.0%, p < 0.001), and receiving ICI treatment combined with chemotherapy (52.6% vs. 46.1%,p < 0.001).The incidence of irAEs was value for univariable analyses for assessing the association between the given characteristic and whether or not an irAE was experienced.b For categorical characteristics, for the columns labeled "IrAEs group" and "No irAEs group," the percents are row percentages and thus represent what percentage of each category experienced or did not experience an irAE, respectively.