The association between toxicity and efficacy of immune checkpoint inhibitors in older adults with NSCLC

Abstract

Aim: This cohort study evaluated the association between immune checkpoint inhibitors (ICIs)-induced immune-related adverse events (irAEs) and mortality as well as ICI discontinuation among older adults with NSCLC.

Methods: 2007–2019 Surveillance, Epidemiology and End Results-Medicare linked database was used and survival analysis with time-varying exposure of irAEs was applied to estimate the associations.

Results & conclusion: A total of 8,175 individuals were included, with 46.8% of whom developed an irAE. Cox regression models showed the occurrence of any irAEs was associated with increased risk of mortality (HR: 1.73, 95% CI: 1.63–1.82) and treatment discontinuation (HR: 1.87, 95% CI: 1.78–1.97). Some variability was observed in the effect on the two outcomes depending on the type of irAE.

Plain Language Summary

A few studies have suggested that certain adverse events related to the immune system (called immune-related adverse events, or irAEs) following treatment of immune checkpoint inhibitors (ICIs) are linked to better clinical outcomes associated with ICI treatment. In contrast, this study of older adults with lung cancer found that patients suffering from irAEs were more likely to die and discontinue ICI treatment. Adverse events such as pneumonitis, arrhythmia, acute kidney injury, hepatitis and colitis were found to be associated with worse outcomes, while hypothyroidism and dermatologic irAEs were not. To prevent life-threatening outcomes for older adults with lung cancer, it is important to closely monitor for the development of irAEs following the initiation of ICI therapy.

Plain language summary

Article highlights.

• An increasing number of studies using real-world data show that the development of irAEs is associated with improved clinical outcomes; however, many of them failed to appropriately account for immortal time bias.

IrAEs & outcomes in older adults

• Using a large database of US Medicare beneficiaries, our study focused on older patients (aged ≥65 years) and assessed the association between irAE occurrence and health outcomes.

IrAEs associated with worse outcomes

• We found that older NSCLC patients with ICI-induced irAEs had higher risk of death as compared with who did not develop.

• Older patients who had irAEs were also more likely to discontinue their ICI treatment.

Findings varied among specific irAEs

• The occurrence of pneumonitis, arrhythmia, AKI, hepatitis and colitis can lead to worse health outcomes, while hypothyroidism and dermatologic irAEs were found to be associated with improved outcomes.

Conclusion

• Our study findings suggest that older adults should be closely monitored for the development of irAEs following initiation of ICI treatment to prevent life-threatening outcomes.

1. Introduction

Immunotherapy with immune checkpoint inhibitors (ICIs) has transformed the treatment paradigm for non-small-cell lung cancer (NSCLC), especially for patients without targetable oncogene mutations. ICIs are now recommended as the standard of care for NSCLC [1,2] and adoption has been rapidly increasing in routine clinical practice [3].

ICIs can cause immune-related adverse events (irAEs), which include a unique spectrum of toxicities, such as dermatological, hematologic, endocrine, pulmonary, renal, gastrointestinal, rheumatologic, cardiovascular and central nervous system (CNS) toxicities [4,5]. The management of irAEs usually requires more frequent monitoring and severe irAEs may lead to hospitalization, treatment discontinuation, morbidity and mortality [6,7].

The onset of irAEs is postulated to be linked with clinical benefit. An increasing number of studies using real-world data show that the development of irAEs is associated with improved clinical outcomes, including overall survival (OS), progression-free survival (PFS) and time to treatment failure (TTF) [8–15]. One explanation is that ICIs can unleash T-cells that target antigens with subsequent production of proinflammatory cytokines and tumor necrosis factor, which may result in excessive inflammation and autoimmune adverse events in addition to effects on shrinking the tumor [9]. While some evidence demonstrates that irAEs are associated with better survival outcomes, these results remain controversial [16–18]. Furthermore, studies assessing the association between irAEs and clinical outcomes have several limitations. Most of the existing studies have used single-center data [9,10,12–14,16–18], which are limited by small sample sizes as well as lack of generalizability. In addition, a number of studies have focused on a single ICI regimen or a specific type of irAE and evaluated short-term outcomes [8,11–14,16,17]. More importantly, a majority of existing studies failed to appropriately account for immortal time bias, which may lead to biased estimation of the association between irAEs and outcomes [19].

Little is known about the development of irAEs and subsequent outcomes among older NSCLC patients, despite NSCLC being predominantly a disease of older adults and this population being the main users of ICIs in the treatment of NSCLC [20,21]. Therefore, the purpose of this study is to evaluate the association between the occurrence of ICI-induced irAEs and clinical outcomes among older adults with NSCLC.

2. Methods

2.1. Study design & data source

A cohort study was conducted using the Surveillance, Epidemiology and End Results (SEER)-Medicare linked database to evaluate the association between ICI-induced irAEs and health outcomes of Medicare beneficiaries with NSCLC receiving treatment with ICIs. This study was approved by the Institutional Review Board at the University of Mississippi (protocol # 21-036).

The SEER-Medicare linked database comprises population-based data from cancer registries in the SEER program and matched Medicare administrative claims data, providing detailed information about demographics, clinical variables related to cancer, as well as healthcare services covered by Medicare at the patient level [22]. The current study used the 2020 SEER-Medicare linkage dataset, which bridges 1999–2017 SEER data to 1999–2019 Medicare claims data.

2.2. Patients

Patients were included if they had a diagnosis of NSCLC as a primary site of cancer between 2007 and 2017, were aged 65 or older at diagnosis and received at least one administration of nivolumab, pembrolizumab or atezolizumab between March 2015 and December 2018, with or without other anticancer agents. International Classification of Diseases for Oncology, third edition (ICD-O-3) site codes and ICD-O-3 histology codes were used to identify patients with NSCLC from the SEER-Medicare Cancer file. Patients were excluded if the diagnosis of NSCLC was made by death certificate or autopsy. The administration of ICIs was identified by the Healthcare Common Procedure Coding System (HCPCS) code (C9453 and J9299 for nivolumab, C9027 and J9271 for pembrolizumab and C9384 and J9022 for atezolizumab). The date of the first administration of ICI was set as the index date for each patient. Patients were required to have continuous enrollment in Medicare Parts A, B and D from the first diagnosis of NSCLC to the index ICI initiation or for at least 12 months prior to the index date, whichever period is longer. Patients were further excluded if they were found to be enrolled in a health maintenance organization (HMO) in the pre-ICI initiation period.

2.3. Outcomes of interest

The primary outcome was OS, defined as the period between the index date and the date of all-cause death. The secondary outcome was time to treatment discontinuation (TTD). TTD has been found to be associated with PFS in randomized controlled trial (RCT) data [23] and is commonly used as a surrogate end point in real-world settings [24]. In this study, TTD was defined as the period between the index date and the date of the last dose of ICI-containing regimen before patients permanently discontinued ICI treatment [25]. ICI discontinuation was identified through the medical claims if the patient started a subsequent line of therapy after the initial ICI-containing regimens, or if the patient died while on the ICI regimen, or if there was a gap in possession of the ICI regimen for more than 120 days. Patients were followed from the index date until the occurrence of an outcome of interest, with data censored in the case of disenrollment in Medicare Parts A, B, or D, initiating enrollment in an HMO, or at the end of the follow-up period (December 31st, 2019), whichever occurred earlier.

2.4. Immune-related adverse events

Patients were identified as having an incident irAE if they had a new medical claim with an irAE diagnosis code at any time between the index date and the 90th day after the last dose of ICI. Restricting irAE occurrence to a 90-day period after ICI administration helps attribute these events to ICI therapy and is based on the approach used in previous studies [26–28]. The first medical claim from any inpatient or outpatient setting with International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM)/International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) diagnosis code for an irAE (Supplementary Table S1) at the primary or secondary position was identified as first occurrence of an irAE. Only events that had no prior occurrence in the previous 12 months were considered incident irAEs. As mentioned in the statistical analysis section, occurrence of irAEs was modeled as a time-varying predictor (i.e., during model estimation, irAE exposure was allowed to vary over time as opposed to taking on a fixed value assigned based on whether an irAE occurred for an individual at any point during the entire study follow-up period) [29]. As described elsewhere, such an approach is aimed at preventing immortal time bias [19].

2.5. Covariates

Covariates included in this study were patient demographics, cancer-related characteristics, ICI-related characteristics and clinical history. Patient demographics included age during index, sex, race, ethnicity, marital status, dual enrollment in Medicaid and Medicare, geographic region of registry, urban residence and percentage of poverty in the census-tract of patient's residence. Cancer-related characteristics included cancer stage at NSCLC diagnosis, tumor histology, history of cancer-related surgery and radiotherapy and the presence of central nervous system (CNS) metastases (identified in the medical claims based on diagnosis codes) in the 6-month period prior to index date. ICI-related characteristics included the index ICI agent, line of therapy for index ICI [30], use of chemotherapy in combination with index ICI and time in months from NSCLC diagnosis to ICI initiation. Patient clinical history included a history of autoimmune disorders, National Cancer Institute (NCI) Comorbidities Index [31], use of corticosteroids and disability status [32].

2.6. Statistical analysis

Frequencies and percentages were reported for categorical variables and as medians and interquartile ranges (IQRs) for continuous variables. Mann-Whitney U tests were conducted for between-group comparisons for continuous variables. Chi-square tests were conducted for between-group comparisons for categorical variables.

Hazard ratios (HR) and corresponding 95% confidence intervals (95% CI) for study outcomes were estimated using a multivariable Cox proportional hazards model with the occurrence of irAE included as a time-varying predictor, after adjusting for covariates [29]. Associations between the occurrence of any irAE and the outcomes were first assessed. Subgroup analyses were conducted to evaluate whether the association of any irAE occurrence and outcomes differed across the following characteristics by fitting interaction terms in the survival models: age, sex, race, ICI type, cancer stage and the presence or absence of an autoimmune disorder. In order to provide a comprehensive overview of all patients with NSCLC who underwent ICI treatment, the primary analysis of this study includes all patients who were diagnosed with NSCLC from 2007. A subgroup analysis was also performed for patients who were more recently diagnosed with NSCLC (2015–2017) and received ICI treatment. In addition, separate models for the occurrence of selected types of irAEs (i.e., pneumonitis, arrhythmia, hypothyroidism, acute kidney injury [AKI], hepatitis, colitis and dermatologic irAEs) and outcomes were estimated.

All analyses were performed using SAS version 9.4 (SAS Institute).

3. Results

This study included 8,175 older patients with NSCLC who received at least one dose of an ICI. The majority (88.0%) of participants were White and almost half (49.0%) were males. Among the included patients, 46.8% developed an irAE after ICI treatment initiation (Table 1 ). Patients experiencing irAEs more commonly used ICIs as first-line cancer treatment at index (32.1 vs. 27.2%, p < 0.001), received pembrolizumab (40.6 vs. 33.3%, p < 0.001) and were taking chemotherapy concomitant to ICI treatment (11.8 vs. 9.4%, p < 0.001). Those who developed irAEs had a higher NCI comorbidity score (0.55 vs. 0.53, p = 0.03), a history of autoimmune disorders (21.2 vs. 19.1%, p = 0.02) and a lower likelihood of having CNS metastases at NSCLC diagnosis (15.0 vs. 18.6%, p < 0.001).

Table 1.

Patient characteristics by study cohort.

Variable	Total sample (n = 8,175)		irAEs group (n = 3,826)		No irAEs group (n = 4,349)		p-value
	Median/n	IQR/%	Median/n	IQR/%	Median/n	IQR/%
Demographic characteristics
Median age at ICI initiation	75.0	71.0–80.0	75.0	71.0–80.0	75.0	71.0–79.0	0.01
Age categories							0.21
65–69	1,457	17.8%	646	16.9%	811	18.7%
70–74	2,414	29.5%	122	29.3%	1,292	29.7%
75–79	2,255	27.6%	1,068	27.9%	1,187	27.3%
80–84	1,312	16.1%	637	16.7%	675	15.5%
85 years +	737	9.0%	353	9.2%	384	8.8%
Sex							0.23
Male	4,008	49.0%	1,903	49.7%	2,105	48.4%
Female	4,167	51.0%	1,923	50.3%	2244	51.6%
Race							0.87
White	7,193	88.0%	3,373	88.2%	3,820	87.8%
Black	531	6.5%	242	6.3%	289	6.6%
Others/Unknown	441	5.5%	211	5.5%	240	5.6%
Ethnicity (Spanish-Hispanic-Latino)	364	4.5%	162	4.2%	202	4.6%	0.37
Marital status							0.12
Live independently	2,410	29.5%	1,117	29.2%	1,293	29.7%
Live with spouse or partner	3,543	43.3%	1,628	42.6%	1,915	44.0%
Unknown	2,222	27.2%	1,081	28.3%	1,141	26.2%
Dual enrolled in Medicaid and Medicare	1,449	17.7%	686	17.9%	763	17.5%	0.65
Census tract poverty							0.03
0% -5% poverty	1,996	24.4%	960	25.1%	1,036	23.8%
5% -10% poverty	2,059	25.2%	992	25.9%	1,067	24.5%
10% -20% poverty	2,106	25.8%	956	25.0%	1,150	26.4%
20% -100% poverty	1,186	14.5%	515	13.5%	671	15.4%
Unknown	828	10.1%	403	10.5%	425	9.8%
Metropolitan residence	7,077	86.6%	3,338	87.3%	3,739	86.0%	0.09
Geographic region							<0.001
West	2,678	32.8%	1,167	30.5%	1,511	34.7%
Northeast	3,326	40.7%	1,671	43.7%	1,655	38.1%
Midwest	652	8.0%	320	8.4%	332	7.6%
South	1,519	18.6%	668	17.4%	851	19.6%
Cancer-related characteristics
Stage							0.40
Non-metastatic	3,790	46.4%	1,789	46.7%	2,001	46.0%
Metastatic	4,214	51.6%	1,950	51.0%	2,264	52.1%
Unknown	171	2.1%	87	2.3%	84	1.9%
Year of NSCLC diagnosis							0.63
Before 2009	108	1.3%	48	1.3%	60	1.4%
2010	75	0.9%	32	0.9%	43	1.0%
2011	115	1.4%	58	1.5%	57	1.3%
2012	229	2.8%	111	2.9%	118	2.7%
2013	469	5.7%	226	5.9%	243	5.6%
2014	898	11.0%	407	10.6%	491	11.3%
2015	1,780	21.8%	804	21.0%	976	22.4%
2016	2,126	26.0%	987	25.8%	1,139	26.2%
2017	2,375	29.1%	1,153	30.1%	1,222	28.1%
Histology							0.33
Adenocarcinoma	4,747	58.1%	2,242	58.6%	2,505	57.6%
Squamous cell	2,286	28.0%	1,058	27.6%	1,228	28.2%
Large cell	85	1.0%	32	0.8%	53	1.2%
Other/not otherwise specified	1,057	12.9%	494	12.9%	563	13.0%
ICI-related characteristics
Months from diagnosis to ICI initiation							0.52
Less than 12 months	4,584	56.1%	2,135	55.8%	2,449	56.3%
12 to 24 months	1,902	23.3%	887	23.2%	1,015	23.3%
24 to 36 months	795	9.7%	365	9.5%	430	9.9%
Over 36 months	894	10.9%	439	11.5%	455	10.5%
Median	11.0	5.0–21.0	11.0	4.0–21.0	11.0	5.0–21.0	0.32
ICI regimens							<0.001
Nivolumab	4,861	59.5%	2,158	56.4%	2,703	62.2%
Pembrolizumab	3,003	36.7%	1,553	40.6%	1,450	33.3%
Atezolizumab	311	3.8%	115	3.0%	196	4.5%
Chemotherapy combined with ICI	860	10.5%	452	11.8%	408	9.4%	<0.001
Therapy line							<0.001
1L	2,411	29.5%	1,228	32.1%	1,183	27.2%
2L	4,026	49.2%	1,851	48.4%	2,175	50.0%
3L+	1,738	21.3%	747	19.5%	991	22.8%
Median ICI treatment duration (months)	3.2	1.2–8.4	5.0	1.6–12.8	2.3	0.9–5.6	<0.001
Cancer-related therapies history
Radiation therapy	4,841	59.2%	2,226	58.2%	2,615	60.1%	0.07
Cancer-directed surgery	1,760	21.5%	845	22.1%	915	21.0%	0.25
CNS metastasis	1,382	16.9%	574	15.0%	808	18.6%	<0.001
Clinical history
NCI comorbidity index	0.54	0.46	0.55	0.47	0.53	0.46	0.30
Disability status							0.22
Good (0)	4,287	52.4%	2,034	53.2%	2,253	51.8%
Poor (1)	3,888	47.6%	1,792	46.8%	2,096	48.2%
Autoimmune disorder	1,640	20.1%	811	21.2%	829	19.1%	0.02
Any use of corticosteroids	5,158	63.1%	2,382	62.3%	2,776	63.8%	0.14

1L: First line; 2L: Second line; 3L: Third line; ICI: Immune checkpoint inhibitor; IQR: Interquartile range.

Among the patients in the study, the percentage of patients with incident pneumonitis (16.5%) was the highest, followed by arrhythmia (11.2%), hypothyroidism (10.5%), AKI (5.8%), dermatologic irAEs (5.3%), colitis (2.8%) and hepatitis (2.1%). Most patients with AKI (63.9%) and pneumonitis (59.3%) had their first irAE event after the last dose of ICI (Supplementary Table S2). In most patients with hypothyroidism (88.4%) and dermatologic irAEs (77.5%), the first event occurred during ICI treatment.

According to the Cox regression model with incident irAEs as a time-varying predictor, the occurrence of irAE of any type was associated with an increased hazard of death in older adults with NSCLC, after accounting for all proposed covariates (HR: 1.73, 95% CI: 1.63 to 1.82). For the subgroup analysis limited to patients diagnosed with NSCLC in 2015–2017 (n = 6,281), consistent results were found as compared with the full cohort (HR: 1.77, 95% CI: 1.66 to 1.88). Subgroup analyses were further conducted stratified by patients' age (65–74 years and over 75 years), sex (male and female), race (White, Black and other), ICI agent, cancer stage (non-metastatic and metastatic) and the presence or absence of an autoimmune disorder. Results were mostly consistent across subgroups, such that irAE occurrence was generally associated with an increased risk of death in each subgroup (Supplementary Figure S1). For the subgroup analysis based on different ICI agents, the association of irAE occurrence with mortality did vary significantly (p = 0.001 for the interaction); all agents showed increased mortality with incident irAEs other than atezolizumab (HR: 1.19, 95% CI: 0.87 to 1.65) (Supplementary Figure S1). Specific irAEs associated with a greater mortality risk included pneumonitis (HR: 2.40, 95% CI: 2.24 to 2.57), arrhythmia (HR: 1.98, 95% CI: 1.82–2.14), AKI (HR: 2.36, 95% CI: 2.12 to 2.63) and hepatitis (HR: 1.55, 95% CI: 1.29–1.87) (Figure 1). In contrast, hypothyroidism (HR: 0.78, 95% CI: 0.70 to 0.86) and dermatologic irAEs (HR: 0.71, 95% CI: 0.62 to 0.82) were associated with a decreased risk of death. Colitis occurrence was not found be significantly associated with mortality.

Figure 1.

Multivariable Cox regression models predicting mortality.

Note: The following covariates have been controlled in each regression model: age, sex, race, ethnicity, marital status, geographic region, urban residence, dual Medicare and Medicaid enrollment, a history of cancer-related surgery, radiotherapy, time since NSCLC diagnosis to ICI initiation, cancer stage, tumor histology, ICI agents, line of therapy for ICIs, in combination of chemotherapy, presence of CNS metastases, a history of autoimmune disorders, disability status and NCI Comorbidities Index.

AKI: Acute kidney injury; CNS: Central nervous system; ICI: Immune checkpoint inhibitor; irAE: Immune-related adverse event; NCI: National Cancer Institute; NSCLC: Non-small-cell lung cancer.

Occurrence of any irAE was also associated with an increased risk of treatment discontinuation (HR: 1.87, 95% CI: 1.78 to 1.97), after accounting for all proposed covariates. Consistent result was found among patients diagnosed with NSCLC in 2015–2017 (HR: 1.92, 95% CI: 1.81 to 2.03). Similar findings were noted among other subgroups; however, the effect of irAE occurrence on TTD was statistically significant, but smaller for atezolizumab (HR: 1.36, 95% CI: 1.03 to 1.79) relative to nivolumab (HR: 1.79, 95% CI: 1.68 to 1.91) and pembrolizumab (HR: 2.08, 95% CI: 1.91 to 2.27) (Supplementary Figure S2). ICI discontinuation was found to be less likely with hypothyroidism (HR: 0.87, 95% CI: 0.80 to 0.94) and more likely with pneumonitis (HR: 2.49, 95% CI: 2.34 to 2.66), arrhythmia (HR: 1.90, 95% CI: 1.77 to 2.05), AKI (HR: 2.82, 95% CI: 2.55 to 3.11), colitis (HR: 1.56, 95% CI: 1.35 to 1.81) and hepatitis (HR: 1.82, 95% CI: 1.55 to 2.14) (Figure 2). There was no significant association between dermatologic irAEs and ICI discontinuation.

Figure 2.

Multivariable Cox regression models predicting treatment discontinuation.

Note: The following covariates have been controlled in each regression model: age, sex, race, ethnicity, marital status, geographic region, urban residence, dual Medicare and Medicaid enrollment, a history of cancer-related surgery, radiotherapy, time since NSCLC diagnosis to ICI initiation, cancer stage, tumor histology, ICI agents, line of therapy for ICIs, in combination of chemotherapy, presence of CNS metastases, a history of autoimmune disorders, disability status and NCI Comorbidities Index.

AKI: Acute kidney injury; CNS: Central nervous system; ICI: Immune checkpoint inhibitor; irAE: Immune-related adverse event; NCI: National Cancer Institute; NSCLC: Non-small-cell lung cancer.

4. Discussion

This study assessed the association between the occurrence of incident irAE and cancer-related survival outcomes, including time to all-cause mortality and time to ICI treatment discontinuation, in patients aged 65 years and older with NSCLC. Generally, the development of irAEs was associated with an increased risk of mortality and treatment discontinuation. Patients who developed pneumonitis, arrhythmia, AKI and hepatitis were observed to be more likely to die or discontinue their ICI treatment. In contrast, occurrence of hypothyroidism and dermatologic irAEs were found to be associated with improved outcomes. The occurrence of colitis was associated with a higher risk of treatment discontinuation, but was not significantly associated with a higher risk of death. Given the presence of biases or limitations with previous studies in this area, these findings are critical for extending knowledge about use of ICIs in patients with NSCLC.

A limited number of studies have evaluated the association between irAE occurrence and clinical outcomes in older patients with NSCLC. Most studies concluded that irAEs were associated with extended OS and PFS in NSCLC patients, regardless of age or settings [8–15]. There were, however, two studies that failed to find a relationship between irAEs and survival. In a retrospective study of 175 advanced NSCLC patients who were nivolumab treated outside the clinical trials, Kothari et al. found no significant differences in OS and PFS [33]. In a single-center study of NSCLC patients who received single-agent immunotherapy, Owen et al. found no significant relationship between irAEs and OS [17]. In older adults, though, irAEs were more likely to precipitate treatment discontinuation and death. According to several studies, older patients are more vulnerable to ICI discontinuation and hospitalization during ICI treatment [34–36]. Even though it has been reported that the objective response rate (ORR) and disease control rate among patients with irAEs were higher than those without [8,10,11,14], our study indicates that the occurrence of irAEs can potentially offset the effect of ICIs on OS for older patients. Notably, subgroup analyses showed that the association of irAEs with mortality varied significantly according to the ICI agent used, with the association being non-significant for atezolizumab. However, this finding may be the result of fewer patients in the dataset using atezolizumab. Additionally, it is noteworthy that association between irAE occurrence and the outcomes were not appreciably different between those with and without autoimmune disorders. Future research should explore these findings in more detail.

The findings of this study reinforce the role of specific irAEs in predicting survival in older patients with NSCLC. As in other studies focusing on patients with NSCLC, dermatologic irAEs and hypothyroidism were related to improved survival outcomes in older patients. Cortellini et al. conducted a landmark study of Italian patients with advanced NSCLC treated with PD-1 inhibitors and found that endocrine and skin adverse events were significantly correlated with improved survival [10]. A study of advanced NSCLC patients receiving nivolumab in a second-line setting also found that endocrine irAEs and dermatologic irAEs were associated with improvements in both OS and PFS [12]. A recent review echoes the finding that the development of thyroid irAE was associated with a favorable outcome in NSCLC patients, even after accounting for immortal time bias, suggesting that thyroid irAEs might serve as a surrogate marker for ICI response [37]. Interestingly, dermatologic irAEs in our study were significantly associated with prolonged overall survival, but not treatment discontinuation. It would be beneficial to conduct further research on the relationship between dermatologic irAEs and ICI response in older adults.

ICI treatment duration and survival time were found to be shorter for older NSCLC patients who developed pneumonitis in the present study. The incidence of pneumonitis during ICI treatment ranges from 7–22% in NSCLC patients [16,18] and ICI-induced pneumonitis occurs more frequently in NSCLC than in other cancers [38,39]. Despite the fact that pneumonitis has been found to be associated with a better ORR in advanced NSCLC patients [39], most existing studies have found that ICI-associated pneumonitis is associated with adverse survival outcomes [15,18,39]. Using data from ICI-treated NSCLC patients, Suresh et al. developed a multistate model to find that incidence of pneumonitis independently increased the risk of mortality [40]. According to a retrospective analysis of the World Health Organization pharmacovigilance database, a total of 35% of PD-1/PD-L1 inhibitors-related deaths were caused by pneumonitis [41]. Moreover, in a retrospective analysis of a global metastatic NSCLC cohort, pneumonitis was found to be the most common irAE associated with permanent ICI discontinuation [13]. In addition, pneumonitis was reported to be the most common irAE requiring hospitalization [35]. Our findings validate the toxicity profile of pneumonitis on ICI treatment maintenance and mortality in older patients.

This study discovered a relatively higher occurrence of incident arrhythmia and AKI among older adults with NSCLC, as compared with what has been reported in other studies, but poor outcomes were associated with them. As a result of comorbidities and polypharmacy, older patients may be more likely to develop cardiac and renal toxicity, but information regarding their impact on patients' outcomes is limited. Arrhythmia onset during ICI treatment is less often reported in both RCTs and observational studies. It can, however, be linked to myocarditis, atrial fibrillation and other ventricular events [42], all of which have been associated with a high mortality rate [43]. It is also known that lung cancer increases the likelihood of cardiac adverse events [44]. Cardiac irAEs have also previously been shown to increase the risk of ICI discontinuation [45]. Immunotherapy, in addition, was reported as an independent risk factor for decreased OS in older patients with NSCLC who died from cardiovascular disease [46]. Our study is in line with the existing evidence indicating the possible harmful impact of arrhythmia incidence in older patients. Similarly, worse outcomes were also observed in patients with AKI. The incidence of AKI among cancer patients treated with ICI was previously reported to be 3% [47] and 16.5–17% [48–50] in RCTs and real-world data, respectively. In studies of both a French and a Canadian cohort, there was no association between AKI and mortality even when the time-varying nature of exposure was taken into account [48,50]. Notably, the current study demonstrates that new-onset AKI has serious consequences and the strength of its effect (HR: 2.36) was relatively high. Given the high occurrence of arrhythmia and AKI and weaker health status in older patients of our study sample, it is imperative to closely monitor their occurrence and the relevant health outcomes during the immunotherapy.

This study also evaluated survival outcomes associated with colitis and hepatitis. The incidence of colitis and hepatitis with anti-PD-1/PD-L1 monotherapy is generally less than 5% [51] and lower in older patients [52]. Current literature suggests no significant relationship between hepatitis and PFS, OS, or ORR [11,15,39]. However, hepatitis appears to be linked to shorter survival time and a greater likelihood of ICI cessation in this study, despite a low incidence observed in older adults. In regard to colitis, some studies found it was associated with a significant increase in PFS and OS [10,14], while other studies suggested no significant association [11,15] or even worse OS [53]. In our study, no significant association was found between colitis and OS in older patients, but colitis was found to result in an increased risk of discontinuation of ICI. As compared with patients treated continuously, previous research has shown that patients who interrupted ICI due to an irAE had a lower median OS [53]. However, this finding should be interpreted with caution, given the fact that incidence of colitis among older patients is observed relatively later post-treatment initiation compared with other irAEs. A further evaluation examining gaps in ICI treatment with longer follow-up data may be warranted to evaluate the association of colitis and mortality.

According to our study data, older adults should be closely monitored for the development of irAEs following ICI treatment initiation in order to avoid life-threatening outcomes. As part of treatment selection, the possibility of the occurrence of irAEs needs to be carefully assessed and a comprehensive discussion of the benefits and risks of treatment with the patient and their caregivers should take place. Based on our research in the same study cohort, several significant risk factors for the development of irAEs have been identified [54]. Together with the results of this previous study, we hope that findings from the current study can provide evidence to support clinical decisions for older cancer patients.

Some limitations to this study should also be mentioned. First, the SEER-Medicare database does not track genetic mutation status, smoking status, or body mass index (BMI). Related to this, the SEER-Medicare database does not contain information on the severity of irAEs, such as the Common Terminology Criteria for Adverse Events (CTCAE) grading for adverse events. Second, this study did not include patients enrolled in Medicare Advantage plans who may have variable treatment patterns – therefore generalizations to other populations must be made with caution. Third, study findings should be considered in light of the inherent limitations of retrospective claims analyses. In claims data, there is no information on ICI response, disease progression, or failure of the treatment. However, we used time to treatment discontinuation as a proxy end point to address this limitation. Fourth, since there are no specific diagnosis codes for irAEs, the sensitivity and specificity of our methods for identifying irAEs are unknown. However, given the careful refinement of the approach used for irAE identification, as well as the consistency of results across various types of irAEs, our findings are likely to be robust. This study also did not include irAEs such as arthralgia-type reactions, as there was no specific diagnosis code available to determine whether the joint pain following ICI initiation was an immune-mediated reaction. Fifth, medical claims submitted by health providers reflect the time when patients sought care for their conditions as opposed to the actual time or date that the irAE occurred. However, given the nature of many of the irAEs examined in this study, it is unlikely that this affected study results. Sixth, misclassification bias may have occurred because of the use of ICD-9-CM/ICD-10-CM diagnosis codes and CPT/HCPSC procedure codes. Finally, the National Cancer Institute cautions against using claims data to identify site-specific metastases after cancer diagnosis due to the potential for underreporting. However, this study only captured and controlled for CNS metastases post NSCLC diagnosis, of which the identification in Medicare claims data has been shown to have a high sensitivity and specificity [55].

5. Conclusion

Older patients with NSCLC who received ICI treatment and developed irAEs were at higher risk of death and/or treatment discontinuation. Associations were particularly apparent in the cases of pneumonitis, arrhythmia, AKI, hepatitis and colitis, whereas for hypothyroidism and dermatologic irAEs, improvements in health outcomes were noted. IrAEs should be closely monitored in older patients after ICI initiation. Further research is necessary in order to determine if irAEs are a sign of immunotherapy response.

Supplemental material

Supplemental data for this article can be accessed at https://doi.org/10.1080/1750743X.2024.2394382

Author contributions

Y Rong: conceptualization; methodology; software; formal analysis; investigation; roles/writing – original draft. S Ramachandran: conceptualization; methodology; supervision; writing – review & editing. Y Yang: writing – review & editing. K Bhattacharya: writing – review & editing. YZ: writing – review & editing. S Earl: writing – review & editing. JP Bentley: conceptualization; methodology; supervision; writing – review & editing.

Financial disclosure

The authors have no financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Competing interests disclosure

The authors have no competing interests or relevant affiliations with any organization or entity with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, stock ownership or options and expert testimony.

Writing disclosure

No writing assistance was utilized in the production of this manuscript.

Data availability statement

The data for this study were obtained through a data use agreement with the National Cancer Institute and cannot be shared publicly. Data can be accessed, subject to approval and data use agreement, from the Healthcare Delivery Research Program at the National Cancer Institute (http://appliedresearch.cancer.gov/seermedicare/obtain/requests.html).