Key Points
Question
What risk factors and inflammatory biomarkers are associated with the development of chronic immune-related adverse events (irAEs) from adjuvant anti–programmed cell death protein 1 (anti–PD-1) treatment?
Findings
In this cohort study of 303 patients treated with adjuvant anti–PD-1 treatment for high-risk melanoma, corticosteroid treatment and later onset of irAEs were associated with risk of developing chronic irAEs. Multiple cytokines had higher expression among patients with chronic irAEs.
Meaning
These findings suggest that corticosteroid treatment and later onset of irAEs may identify patients at higher risk of developing chronic irAEs, and elevated circulating cytokines associated with chronic irAEs suggest possible candidate inflammatory pathways.
Abstract
Importance
Adjuvant anti–programmed cell death protein 1 (anti–PD-1) treatment improves outcomes in high-risk resected melanoma but causes chronic immune-related adverse events (irAEs) in more than 40% of patients. Risk factors associated with chronic vs acute irAEs remain lacking.
Objective
To identify factors and possible candidate inflammatory pathways associated with the development of chronic nonendocrine irAEs.
Design, Setting, and Participants
This retrospective cohort study was conducted from January 1, 2015, to December 31, 2024, at 7 institutions in the US and Australia. Patients had stage III to IV melanoma treated with adjuvant anti–PD-1 therapy and more than 12 months of follow-up after discontinuation of anti–PD-1 treatment.
Exposures
Potential risk factors (age, sex, time to irAE onset, corticosteroid exposure, time to corticosteroid initiation, peak corticosteroid dose, and anti–PD-1 treatment status) for chronic irAEs.
Main Outcomes and Measures
Incidence, type, and risk factors for chronic (persisting ≥3 months after anti–PD-1 therapy cessation) irAEs and cytokine levels at 12 months after immune checkpoint inhibitor initiation were assessed.
Results
Among 303 patients (median age at anti–PD-1 therapy initiation, 64 years [IQR, 55-72 years]; 184 men [61%]), 220 (73%) developed acute irAEs and 146 (48%) developed chronic irAEs. The most common chronic irAEs were hypothyroidism (45 of 146 [31%]), arthritis (25 of 146 [17%]), and dermatitis (17 of 146 [12%]). Chronic irAEs had later onset (median, 153.0 days [IQR, 67.0-294.0 days] vs 99.0 days [IQR, 32.2-228.8 days]; P = .006) and more frequent corticosteroid exposure (35% [34 of 96] vs 24% [48 of 204]; P = .03) compared with nonchronic irAEs. For exploratory risk factor analysis of 300 nonendocrine irAE events, corticosteroid treatment and onset time had a significant interaction; corticosteroid-treated patients with earlier-onset irAEs had higher odds of developing chronic irAEs (odds ratio, 3.00; 95% CI, 1.45-6.22), adjusted for age, sex, and anti–PD-1 therapy continuation. Eight different cytokines had higher circulating plasma levels among patients with chronic irAEs after false discovery rate (FDR) adjustment (including VEGFA and CCL19), and 19 cytokines had significantly higher circulating plasma levels prior to FDR adjustment. No cytokine levels were higher among controls.
Conclusions and Relevance
In this cohort study of patients with melanoma treated with adjuvant anti–PD-1 therapy, chronic irAEs were common, persistent, and associated with elevated circulating cytokines, suggesting possible candidate inflammatory pathways for future validation. Later irAE onset and corticosteroid-treated irAEs were associated with chronicity of irAEs. Given the long-term survival of these patients, monitoring and managing chronic irAEs is crucial.
This cohort study assesses risk factors and inflammatory biomarkers associated with the development of chronic immune-related adverse events from adjuvant anti–programmed cell death protein 1 treatment (anti–PD-1).
Introduction
Melanoma is an aggressive skin cancer that had poor prognosis until recently.1 In the past 10 to 15 years, immune checkpoint inhibitors (ICIs) have improved survival outcomes among patients with metastatic melanoma.2 ICIs are monoclonal antibodies that target programmed cell death protein 1 (PD-1) and cytotoxic T-lymphocyte–associated protein-4 (CTLA-4).3
Anti–PD-1 ICIs have shown durable benefit among patients with advanced melanoma. For example, the CheckMate-067 phase 3 trial studied 945 patients with metastatic melanoma and demonstrated a 10-year overall survival of nearly 50% for both nivolumab monotherapy and nivolumab combined with ipilimumab.4 This success led to adjuvant and neoadjuvant studies for resectable stage 3 melanoma,5 where anti–PD-1 monotherapy is now a standard of care.
ICIs can trigger T-cell activation and lead to immune-related adverse events (irAEs), which are autoinflammatory toxic effects that can impact many organ systems. Severe irAEs occur in 15% to 20% of patients treated with anti–PD-1 monotherapy and up to 50% of patients treated with combination anti–PD-1 and anti–CTLA-4 therapy.6 Previous studies have shown that irAEs become chronic (defined as persisting ≥3 months after treatment cessation) in approximately 40% of patients treated with adjuvant anti–PD-1 therapy for melanoma; most events persisted after more than 1.5 years of follow-up.7,8
Because anti–PD-1 ICIs are increasingly used across many tumor types and can lead to long-term survivorship, it is critical to further assess their chronic effects. Currently, we lack reliable ways to identify patients at risk for developing chronic irAEs. To address this knowledge gap, we conducted a large, multicenter cohort study that built on a recent investigation of patients with and without irAEs from adjuvant anti–PD-1 therapy.7 Here, we (1) assess the incidence, characteristics, and long-term outcomes (including resolution vs persistence) of chronic irAEs from adjuvant anti–PD-1 therapy with additional follow-up; (2) evaluate potential risk factors to assess factors associated with the development of chronic irAEs; and (3) evaluate whether specific cytokines are upregulated among patients with chronic irAEs.
Methods
Patient Selection
This cohort study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. Institutional review board or ethics committee approval was obtained from Vanderbilt University Medical Center, Melanoma Institute of Australia, New York University Langone Comprehensive Cancer Center, Georgetown Lombardi Comprehensive Cancer Center, and Massachusetts General Hospital. Across all institutions, informed consent was waived due to patient data being deidentified.
Retrospective deidentified data were collected from 7 participating academic medical centers in the US and Australia. Patients who had resected stage III to IV melanoma treated with adjuvant anti–PD-1 therapy from January 1, 2015, to December 31, 2024, were included. Those without evaluable follow-up for at least 12 months after discontinuation of anti–PD-1 therapy were excluded.
Data Collection
Patient demographic characteristics (melanoma subtype, comorbidities, genetic mutation status, and stage), treatment details (dose, frequency, duration, and reason for discontinuation), and disease recurrence details were collected from the electronic medical record from each institution. Type, grade, duration, treatment, and resolution details of irAEs were also recorded. Acute irAEs were defined as those arising during anti–PD-1 therapy, while chronic irAEs were defined as acute events that persisted at least 3 months after cessation of anti–PD-1 therapy. Banked plasma from patients with (n = 17) or without (n = 10) chronic nonendocrine irAEs at a routine visit nearest to 12 months after anti–PD-1 therapy initiation was analyzed using a Target 96 Inflammation Panel (V.3027; Olink) to quantify markers of inflammation.
Statistical Analysis
Descriptive statistics were summarized as median (IQR) or mean (SD) for continuous variables and as frequency (%) for categorical variables. Summary statistics were further stratified by status of chronic irAEs (chronic irAEs vs chronic nonendocrine irAEs). We used the Wilson score method for calculating 95% CIs for proportions, whereas we calculated 95% CIs for continuous variables using the normal approximation. Missing data were minimal; among the 301 toxicity events, 1 event was excluded from the analysis for missing key clinical variables. A complete-case analysis was performed; therefore, this event was excluded, resulting in a final analytic sample of 300 toxicity events. No additional missingness was present in the variables included in the analyses.
In the univariable analysis, the Kruskal-Wallis test was applied to compare continuous variables, including age, time to irAE onset from initiation of anti–PD-1 therapy, corticosteroid exposure, time to corticosteroid initiation from irAE onset, and peak corticosteroid dose between groups. Effect estimates are reported as Hodges-Lehmann location differences (chronic minus nonchronic) with 95% CIs. For categorical variables, including sex (female vs male), corticosteroid exposure (yes vs no), and anti–PD-1 treatment status (defined as anti–PD-1 therapy discontinuation within 30 days of irAE onset; continued therapy vs early discontinuation), 2-sample proportion tests were used and are summarized by differences in proportions with Wilson 95% CIs.
Redundancy analysis with a coefficient of determination threshold of R2 = 0.8 was performed to evaluate multicollinearity among potential risk factors, and no variables were found to be highly explained by others. Multivariable logistic regression was used to assess the association between potential risk factors and status of chronic irAEs (chronic vs nonchronic), adjusted for age and sex. Odds ratios (ORs) with IQRS or ORs with 95% CIs were reported (eMethods in Supplement 1). All statistical tests were 2-sided, and statistical significance was defined as P < .05.
In the proteomic analysis, data normality was assessed using the Shapiro-Wilk test, which indicated that the protein expression values were nonnormally distributed. Consequently, group comparisons for each biomarker were conducted using the Mann-Whitney test. P values were adjusted using the Benjamini-Hochberg procedure to control the false discovery rate (FDR). Statistical significance was defined as an adjusted P < .05 (eMethods in Supplement 1).
Results
We included 303 patients (median age at anti–PD-1 therapy initiation, 64 years [IQR, 55-72 years]; 184 men [61%] and 119 women [39%]) (Table 1). A total of 220 patients (73%) developed acute irAEs and 146 (48%) developed chronic irAEs; 61 patients (20%) had chronic endocrine irAEs, 95 (31%) had chronic nonendocrine irAEs, and 10 (3%) had both. Median follow-up was 67.0 months (IQR, 57.5-74.4 months) (median, 57.1 months [IQR, 47.6-66.4 months] after treatment discontinuation). Among the 146 patients with chronic irAEs, the most common were hypothyroidism (45 [31%]), arthritis (25 [17%]), dermatitis (17 [12%]), xerostomia (11 [8%]), and adrenal insufficiency (9 [6%]) (Table 2). At last follow-up, 102 patients (34%) had ongoing chronic irAEs; 49 (16%) with endocrine irAEs, 47 (16%) with nonendocrine irAEs, and 8 (2%) with both (Table 3). Of the 102 patients with ongoing chronic irAEs at last follow-up, 17 (17%) had arthritis. Management at last follow-up included corticosteroids (n = 4), nonsteroidal anti-inflammatory drugs (n = 4), hydroxychloroquine (n = 3), methotrexate (n = 2), and adalimumab (n = 1). Twenty-five additional patients (8%) experienced chronic toxic effects outside of classical irAEs, most often fatigue (20 [7%]), orthostasis (2 [1%]), and headache (2 [1%]) (Table 2). Only 4 of these 25 chronic toxic effects (16%) outside of classical irAEs had documented resolution at last follow-up.
Table 1. Patient Demographic Characteristics.
| Characteristic | No. (%) [95% CI]a | |
|---|---|---|
| No chronic toxic effects (n = 157) | Chronic toxic effects (n = 146) | |
| Age, y | ||
| Mean (SD) | 61.8 (13.5) | 61.3 (13.4) |
| Median (IQR) | 64.0 (59.7-63.9) | 64.0 (59.1-63.5) |
| Duration of follow-up, y | ||
| Mean (SD) | 5.4 (1.3) | 5.1 (1.7) |
| Median (IQR), y | 5.4 (5.2-5.6) | 5.6 (4.9-5.4) |
| Sex | ||
| Male | 99 (63) [55-70] | 85 (58) [50-66] |
| Female | 58 (37) [30-45] | 61 (42) [34-50] |
| Primary location | ||
| Cutaneous or unknown | 148 (94) [90-97] | 134 (92) [86-95] |
| Acral | 7 (4) [2-9] | 10 (7) [4-12] |
| Mucosal | 2 (1) [0-5] | 2 (1) [0-5] |
| Stage | ||
| I or II | 4 (2) [1-6] | 5 (3) [2-8] |
| IIIA or B | 69 (44) [36-52] | 66 (45) [37-53] |
| IIIC or D | 70 (45) [37-52] | 60 (41) [33-49] |
| IV (resected) | 14 (9) [5-14] | 15 (10) [6-16] |
| Gene mutation | ||
| BRAF V600 mutated | 33 (21) [15-28] | 30 (20) [15-28] |
| NRAS mutated | 21 (13) [9-20] | 33 (23) [17-30] |
| BRAF/NRAS WT/unknown | 103 (66) [70-83] | 80 (55) [47-63] |
| Preexisting autoimmune disorder | 17 (11) [17-30] | 17 (12) [7-18] |
| Treatment | ||
| Nivolumab | 121 (77) [48-63] | 116 (79) [72-85] |
| Pembrolizumab | 36 (23) [37-52] | 30 (21) [15-28] |
| Center | ||
| US | 87 (55) [41-56] | 94 (64) [56-72] |
| Australian | 70 (45) [42-57] | 52 (36) [28-44] |
| Recurrent disease | 76 (48) [50-66] | 52 (36) [28-44] |
| irAEs | ||
| Acute | 78 (50) [42-57] | 142 (97) [93-99]b |
| Chronicc | ||
| All | NA | 146 (48) [43-54] |
| Endocrine irAEs | NA | 61 (20) [16-25] |
| Nonendocrine irAEs | NA | 95 (31) [26-37] |
| Endocrine and nonendocrine | NA | 10 (3) [2-6] |
| Ongoing irAE at last follow-upd | NA | 102 (34) [29-39] |
| Chronic toxic effects outside of classical irAEs | NA | 25 (8) [6-12] |
Abbreviations: irAE, immune-related adverse event; NA, not applicable; WT, wild type.
The 95% CIs for proportions were calculated using the Wilson score method, whereas for continuous variables, the 95% CIs were calculated by normal approximation.
A total of 4 patients had irAEs characterized as delayed irAEs.
Percentage of full cohort (N = 303).
Median, 67.0 months (IQR, 56.7-73.6 months).
Table 2. Chronic irAEs and Chronic Toxic Effects Outside of Classical irAEs.
| irAE or toxic effect | No. of patients (% of patients with chronic irAEs) [95% CI]a,b |
|---|---|
| Hypothyroidism | 45 (31) [24-38] |
| Arthritis or arthralgias | 25 (17) [12-24] |
| Dermatitis or pruritus | 17 (12) [7-18] |
| Xerostomia | 11 (8) [4-13] |
| Adrenal insufficiency | 9 (6) [3-11] |
| Hypophysitis | 7 (5) [2-10] |
| Pneumonitis | 7 (5) [2-10] |
| Neuropathy | 7 (5) [2-10] |
| Vitiligo | 6 (4) [2-9] |
| Other neurotoxic effects | 6 (4) [2-9] |
| Colitis or diarrhea | 4 (3) [1-7] |
| Hepatitis | 4 (3) [1-7] |
| Nephritis | 4 (3) [1-7] |
| Ocular toxic effects | 4 (3) [1-7] |
| Sicca syndrome | 3 (2) [1-6] |
| Otherc | 3 (2) [1-6] |
| Bullous pemphigoid | 2 (1) [0-5] |
| Cough | 2 (1) [0-5] |
| Gastritis or enteritis | 2 (1) [0-5] |
| Mucositis | 2 (1) [0-5] |
| Nausea | 2 (1) [0-5] |
| Diabetes | 1 (1) [0-4] |
| Dysgeusia | 1 (1) [0-4] |
| Hypogonadism | 1 (1) [0-4] |
| Pancreatitis | 1 (1) [0-4] |
| Psoriasis | 1 (1) [0-4] |
| Raynaud syndrome | 1 (1) [0-4] |
| Retinal vasculitis | 1 (1) [0-4] |
| Sjögren syndrome | 1 (1) [0-4] |
| Uveitis | 1 (1) [0-4] |
| Chronic toxic effects outside of classical irAEd | |
| Fatigue | 20 (7) [4-10] |
| Orthostasis | 2 (1) [0-2] |
| Headache | 2 (1) [0-2] |
| Gastrointestinal symptoms | 1 (0.33) [0-2] |
Abbreviation: irAE, immune-related adverse event.
Denominator is the total number of patients with chronic irAEs in the study (n = 146).
The 95% CIs for proportions were calculated using the Wilson score method without continuity correction.
Any other irAE than the ones listed in the table.
Denominator is the total number of patients in the study (N = 303).
Table 3. Chronic irAEs Persisting With Long-Term Follow-Up.
| Chronic irAE | No. of patients (% of all patients with chronic irAEs) [95% CI]a,b | No. of patients with ongoing chronic irAE at last follow-up (% of patients with the same type of chronic irAE) [95% CI]b,c |
|---|---|---|
| Hypothyroidism | 45 (31) [24-39] | 42 (93) [82-98] |
| Arthritis or arthralgias | 25 (17) [12-24] | 17 (68) [48-83] |
| Dermatitis or pruritus | 17 (12) [7-18] | 13 (76) [53-90] |
| Adrenal insufficiency and hypophysitis | 16 (11) [7-17] | 16 (100) [81-100] |
| Neuropathy and other neurotoxic effects | 13 (9) [5-15] | 7 (54) [29-77] |
| Xerostomia | 11 (8) [4-13] | 8 (73) [43-90] |
| Pneumonitis | 7 (5) [2-10] | 5 (71) [36-92] |
| Vitiligo | 6 (4) [2-9] | 5 (83) [44-97] |
| Colitis or diarrhea | 4 (3) [1-7] | 3 (75) [30-95] |
| Hepatitis | 4 (3) [1-7] | 2 (50) [15-85] |
| Nephritis | 4 (3) [1-7] | 2 (50) [15-85] |
| Ocular toxic effects | 4 (3) [1-7] | 3 (75) [30-95] |
| Other | 25 (17) [12-24] | 14 (56) [37-73] |
Abbreviation: irAE, immune-related adverse event.
Denominator is the total number of patients with chronic irAEs in the study (n = 146).
The 95% CIs for proportions were calculated using the Wilson score method without continuity correction.
Denominator is the total number of patients who had the chronic irAE of interest (including those that persisted with long-term follow-up and those that did not) in the given table row.
We then assessed potential risk factors for chronic irAEs compared with acute, resolving irAEs. For this analysis, we excluded endocrine irAEs because nearly all became chronic, resulting in 300 total nonendocrine irAEs with evaluable data. In the univariate analysis, chronic irAEs had later onset (median, 153.0 days [IQR, 67.0-294.0 days] vs 99.0 days [IQR, 32.2-228.8 days]; P = .006) and increased percentage treated with corticosteroids (35% [34 of 96] vs 24% [48 of 204]; P = .03) compared with nonchronic irAEs (Table 4). Time to corticosteroid initiation from irAE onset was longer but not statistically significant in chronic irAE events compared with nonchronic irAE events (median, 7.0 days [IQR, 0.0-29.0 days] vs 3.0 days [IQR, 0.0-16.5 days]; P = .34). Peak corticosteroid dose was lower but not statistically significant in chronic irAEs vs nonchronic irAEs (median, 40.0 mg [IQR, 16.2-60.0 mg] vs 55.0 mg [IQR, 25.0-100.0 mg]; P = .08). Rates of early anti–PD-1 therapy discontinuation were similar for chronic vs nonchronic irAEs (8% [8 of 96 vs 8% [16 of 204]; P = .88). Median age at diagnosis (63.0 years [IQR, 53.5-69.0 years] vs 65.0 years [IQR, 58.0-74.0]; P = .13) and the percentage of men (57% [55 of 96] vs 58% [118 of 204]; P = .93) were also similar for chronic vs nonchronic irAEs.
Table 4. Risk Factors for Nonchronic vs Chronic Nonendocrine irAEs .
| Exposure | irAEs | Difference (95% CI)a | P value | |
|---|---|---|---|---|
| Nonchronic (n = 204) | Chronic (n = 96) | |||
| Age, median (IQR), y | 65.0 (58.0 to 74.0) | 63.0 (53.5 to 69.0) | −1.8 (−4.9 to 1.4) | .13 |
| Sex (% male) | 58 (118 of 204) | 57 (55 of 96) | −0.6 (−12.5 to 11.4) | .93 |
| Time to irAE onset from anti–PD-1 therapy initiation, median (IQR), d | 99.0 (32.0 to 228.8) | 153.0 (67.0 to 294.0) | 36.0 (9.0 to 67.0) | .006 |
| Steroid exposure, % (No./total No.) | 24 (48 of 204) | 35 (34 of 96) | −11.9 (−23.1 to −0.7) | .03 |
| Time to corticosteroid initiation from irAE onset, median (IQR), y | 3.0 (0.0 to 16.5) | 7.0 (0.0 to 29.0) | 0.0 (−1.0 to 7.0) | .34 |
| Peak corticosteroid dose, median (IQR), mg | 55.0 (25.0 to 100.0) | 40.0 (16.2 to 60.0) | −10.0 (−35.0 to 0.0) | .08 |
| Anti–PD-1 treatment status, % (No./total No.)b | 8 (16 of 204) | 8 (8 of 96) | 0.5 (−6.2 to 7.1) | .88 |
Abbreviations: anti–PD-1, anti–programmed cell death protein 1; irAE, immune-related adverse event.
Continuous variables were compared using the Kruskal-Wallis rank-sum test and are summarized by Hodges-Lehmann location differences (chronic minus nonchronic) with 95% CIs. Categorical variables were compared using a 2-sample proportion test, with Wilson 95% CIs without continuity correction for differences in proportions.
Percentage of irAEs that prompted early discontinuation.
In the multivariable analysis using logistic regression, a likelihood ratio test comparing a model containing 3 two-way interactions—(1) log(time to irAE onset + 1) × corticosteroid exposure, (2) log(time to irAE onset + 1) × anti–PD-1 treatment status, and (3) corticosteroid exposure × anti–PD-1 treatment status—with a model additionally including the corresponding 3-way interaction found no evidence of a 3-way interaction among time to irAE onset, corticosteroid exposure, and anti–PD-1 treatment status. The interaction between corticosteroid exposure and anti–PD-1 treatment status was subsequently removed because its inclusion did not improve model fit. Therefore, the final model included 2 two-way interactions: (1) between time to irAE onset and corticosteroid exposure and (2) between time to irAE onset and anti–PD-1 therapy treatment status, adjusting for age and sex (eTable 1 in Supplement 1).
A significant interaction was observed between corticosteroid exposure and time to irAE onset (β [SE] = –0.5374 [0.2263]; P = .02; eTable 1 in Supplement 1), indicating that the association between corticosteroid exposure and chronic irAEs varied according to the timing of irAE onset (eFigure 1 in Supplement 1). At earlier irAE onset (first quartile of 42 days), the adjusted odds of developing chronic irAEs were higher in corticosteroid-requiring events than in those not treated with corticosteroids (adjusted OR, 3.00; 95% CI, 1.45-6.22). However, this association decreased with increasing time to irAE onset; the adjusted OR was 1.77 (95% CI, 1.02-3.08) at a median onset time of 113.5 days (IQR, 42-259 days) and 1.14 (95% CI 0.60-2.16) at later onset (third quartile, 259 days). This finding suggests that the association between corticosteroid exposure and chronic irAEs was strongest with earlier irAE onset and weakened with later onset. In contrast, there was no evidence that the association between anti–PD-1 treatment status and chronic irAEs varied according to time to irAE onset (β [SE] = –0.3397 [0.4801]; P = .48; eTable 1 in Supplement 1). The adjusted ORs comparing continued vs discontinued anti–PD-1 therapy were 2.05 (95% CI, 0.45-9.32) at a first quartile of 42 days, 1.47 (95% CI, 0.56-3.90) at a median of 113.5 days (IQR, 42-259 days), and 1.11 (95% CI, 0.38-3.22) at a third quartile of 259 days (eFigure 2 in Supplement 1). In the final logistic regression model, the association of time to irAE onset was evaluated separately according to corticosteroid exposure and anti–PD-1 treatment status. Among irAEs not treated with corticosteroids with continued anti–PD-1 therapy (n = 206), increasing time to irAE onset from a first quartile of 42 days to a third quartile of 259 days was associated with significantly increased odds of chronic irAEs (OR, 2.36; 95% CI, 1.39-3.98; P = .001) (eTable 2 in Supplement 1). Although larger point estimates were observed for events with corticosteroid exposure and discontinuation of anti–PD-1 therapy (OR, 4.34; 95% CI, 0.77-24.54), the CI was wide, and the association was not statistically significant (P = .10). No significant association between time to irAE onset and chronic irAEs was observed with corticosteroid treatment, regardless of anti–PD-1 treatment status. An example of a 65-year-old man (median of age) shows increasing probability of chronic irAEs without corticosteroid exposure and with anti–PD-1 discontinuation over time (eFigure 3 in Supplement 1).
To assess the molecular features associated with chronic irAEs, we identified 17 patients with ongoing symptomatic, chronic nonendocrine irAEs with plasma collection 12 months after starting treatment and compared their cytokine levels with 10 patients lacking irAEs at 12 months (eTable 3 in Supplement 1 and eTable 4 in Supplement 2). Eight of the 92 studied cytokines had significantly higher expression in patients with chronic irAEs compared with patients without chronic irAEs after FDR adjustment, while 19 cytokines had significantly higher expression before FDR adjustment) (Figure; eTable 5 in Supplement 2 reports full results). Cytokines with significant differences after FDR adjustment included FGF-23, MMP-1, CSF-1, CCL19, Flt3L, VEGFA, CCL25, and MMP-10; several cytokines with validated targets also had nominally significant differences (eg, tumor necrosis factor [TNF], interleukin [IL]-8, IL-17, programmed cell death 1 ligand 1 [PD-L1]). None of the 92 cytokines had lower expression in patients with chronic irAEs. Significance was not associated with excluding 1 sample with quality control flags, 3 patients with immunosuppression, or 2 patients with fatigue only (eTable 6 in Supplement 1).
Figure. Volcano Plot and Box Plot of Differential Plasma Protein Abundance in Patients With Vs Without Chronic Immune-Related Adverse Events (irAEs).
Plasma from patients with (n = 17) and patients without (n = 10) an active chronic nonendocrine irAE was profiled using the Target 96 Inflammation panel (v.3027; Olink), and the 2 groups were compared analyte-by-analyte using the Mann-Whitney test with Benjamini-Hochberg correction across the 73 analytes meeting the detection threshold (≥50% of samples above the plate-specific limit of detection). NPX indicates normalized protein expression. A, False discovery rate (FDR)-significant points are significant at an adjusted P < .05; and nominally significant points are significant only at an unadjusted P < .05 but nonsignificant at an adjusted P ≥ .05. The 8 FDR-significant analytes are labeled, along with tumor necrosis factor (TNF), a nominally significant and potentially therapeutically tractable target. Most differentially abundant proteins were elevated in chronic irAE. B, NPX distributions for the 8 FDR-significant analytes, ordered by adjusted P value, shown by group. To display all analytes on a common scale, each analyte’s NPX values were centered on the analyte means (solid lines); values therefore represent deviation from the analyte mean (log2 scale) rather than absolute NPX. Boxes indicate the median and IQR, whiskers extend to 1.5× the IQR, and points represent individual patients, dodged by group. All 8 analytes were elevated in chronic irAE, although group distributions overlapped, consistent with these analytes representing candidate signals rather than individually discriminating biomarkers.
Discussion
Incidence and Characteristics of Chronic irAEs
In this cohort study of 303 patients with melanoma treated with adjuvant anti–PD-1 therapy, chronic irAEs were common and persistent. With additional follow-up (median, 67 months), 34% of patients had ongoing irAEs. Endocrine irAEs (hypothyroidism, adrenal insufficiency, and hypophysitis) were the most common and nearly always persisted to the last follow-up. In contrast, other more symptomatic irAEs, including ocular, rheumatologic, oral, and cutaneous toxic effects, often resolved but individually persisted at low rates. Generally, these events were low grade but could negatively impact quality of life.
With additional follow-up, this study contributes to the finding that chronic irAEs are frequent and often persistent.7,9,10 A previous study has hypothesized whether chronic irAEs reflect a gradual autoimmune process or a “burnout” phenotype, in which the cells of interest have been largely destroyed, causing manifestations to persist.11 Specifically in the cases of endocrine irAEs, their near-universal persistence indicates that permanent tissue damage may occur. For these patients, rechallenging with ICIs is usually low risk, although patients with prior hypophysitis may still experience flares (with headaches and signs of pituitary inflammation).12 In contrast, the fact that some chronic irAEs resolve, while others flare upon rechallenging with ICIs, suggests that ongoing inflammation persists in certain patients.7 Arthritis can be a morbid and challenging chronic irAE, frequently requiring long-term immune modulations; 17 patients had arthritis at last follow-up and were frequently still receiving treatment with corticosteroids and biologics.
Beyond these classical irAEs, our study also identified nonclassical chronic toxic effects in 25 patients (8%), with the most common being fatigue, orthostasis, and headache. This finding aligns with evidence that has identified nonclassical, off-target, long-term effects associated with ICIs. For example, in a large retrospective multicenter study of 517 patients receiving anti–PD-1 treatment, fatigue was one of the most common ICI-induced toxic effects, with early fatigue seeming to be a negative prognostic indicator.13 Although less common, dysautonomia has also been reported in case reports as a result of ICI therapy.14 We and others are currently conducting patient-reported outcome studies to quantify nonclassical irAEs and assess for a syndrome analogous to long COVID after anti–PD-1 therapy (conceptualized as “long-PD-1”).
Potential Risk Factors for Chronic irAEs
In our study, we identified several potential risk factors for chronic irAEs. On univariate analysis, corticosteroid treatment was associated with chronicity, which likely corresponds with more severe irAEs. Later onset was associated with chronicity in univariable analysis. Multivariable analyses suggested an interaction between time and corticosteroid treatment: early irAEs were more likely to become chronic if they were treated with corticosteroids compared with those that were not treated with corticosteroids. It is possible this represents the limitations of modest sample size but could also suggest a speculative model where: (1) early-onset irAEs that are severe enough to require corticosteroids may be high risk for chronicity, while (2) milder, later-onset toxic effects may not be treated with corticosteroids as frequently, thus never resolving and becoming chronic. Similar findings were seen with anti–PD-1 treatment status: early discontinuation was associated with less risk of chronic toxic effects whereas the association was less pronounced later in the treatment course.
Developing comprehensive clinical or molecular factors associated with irAEs has been a challenge; only younger age and preexisting autoimmune disease have shown modest associations with increased risk of severe acute irAEs.15 Current guidelines for the management of irAEs have thus far focused more on early identification of irAEs and prompt initiation of immunosuppressive treatment while also providing recommendations for escalation of treatment in cases of severe and/or corticosteroid-refractory irAEs.15 Further research should refine our potential risk factors in larger, independent cohorts and integration into multivariable models—possibly incorporating genomics, cytokine signatures, and immune cell profiling—to move toward clinically actionable predictive tools.
Cytokine Signatures: Mechanistic and Biomarker Implications
We observed elevated levels of specific cytokines in patients experiencing chronic irAEs, which argues that these patients likely have ongoing inflammation associated with their symptoms. Other studies that have largely focused on pretreatment samples or samples of patients early in treatment broadly mirror our findings; for example, elevated vascular endothelial growth factor A has been observed in cases of ICI-induced inflammatory arthritis,16 and elevated baseline and on-treatment levels of IL-17 have been associated with severe colitis in patients treated with CTLA-4 inhibitors.17 High baseline soluble PD-L1 and larger treatment-associated fluctuations have been linked with severe irAEs.18 Similarly, IL-2 and IL-13 were highly correlated with severe irAEs via incorporation into the cytokine toxicity score,19,20 and early increases in IL-6, IL-10, and CXCL9-11 during treatment were associated with irAEs.17,19,21 In affected tissue, IL-1β upregulation in gastrointestinal tissues during ICI-induced colitis and in arthritic lesions suggests its role in chronic irAEs.22 Our findings add to this body of literature by suggesting (for the first time, to our knowledge) that persistent cytokine elevation may be a hallmark of chronic irAEs. Larger series are needed to validate this finding and assess whether specific irAEs correlate with distinct cytokine profiles.
Cytokine Signatures: Therapeutic Implications
Our identification of elevated cytokines in chronic irAEs could suggest precision-targeted therapies that could reduce the burden of toxic effects while preserving the effectiveness of ICIs. Systemic corticosteroids are the current mainstay for managing nearly all nonendocrine irAEs. Corticosteroids provide broad immunosuppression but cause well-documented long-term adverse effects that affect nearly every organ system.23
In contrast, targeted therapies against specific cytokines could offer a more focused mechanism of action. Preclinical murine models demonstrate that blocking IL-25 or IL-17RA alongside anti–PD-1 and anti–CTLA-4 therapy prevents irAEs such as pneumonitis and hepatitis while preserving antitumor effectiveness.24 A case series in 2 patients presented evidence showing that IL-17A blockade was associated with improved irAEs.25 A case of pembrolizumab-induced psoriasiform dermatitis also resolved with systemic IL-17A blockade (secukinumab) without impacting antitumor activity.26 The activity of TNF inhibitors such as infliximab as second-line biologics provides additional evidence for targeted cytokine modulation.27 TNF-α inhibitors have also demonstrated safety and efficacy without an obvious association with the antitumor responses.28,29 These signatures would need to be validated, and elevated expression does not necessarily indicate causation.
Limitations
This study has limitations. Relying on clinical data and notes in the electronic medical record to review toxic effects may not match the precision of formal clinical trial adverse event reporting, which could result in either overreporting or underreporting of chronic irAEs. However, a recent clinical trial with formal annotation of toxic effects found even higher rates of chronic irAEs at 5 years (69%) with the use of a combination of ipilimumab and nivolumab therapy, thus generally supporting our analysis.30 , Because this study is observational, we cannot definitively establish causal relationships between risk factors and the development of chronic vs acute irAEs. Last, patients with shorter survival (eg, early death from cancer progression) were not captured. However, this potential limitation likely had minor effect given the favorable outcomes of adjuvant anti–PD-1 therapy in resected melanoma.4,31
Conclusions
In this cohort study of patients with melanoma treated with adjuvant anti–PD-1 therapy, we found that chronic irAEs were common and persistent, with most events failing to resolve even with prolonged follow-up. As adjuvant anti–PD-1 therapy continues to extend survival for patients with high-risk resected melanoma, attention must increasingly turn to understanding irAEs. Our findings highlighted that living longer increasingly equates to living with—and needing management for—chronic ICI-induced toxic effects. Our study also showed that corticosteroid treatment and later irAE onset may identify patients at higher risk of developing chronic irAEs. Elevated circulating cytokines associated with chronic irAEs may also suggest possible candidate inflammatory pathways. Given the long-term survival of patients who receive adjuvant anti–PD-1 therapy, further research should aim to identify patients at higher risk of irAEs and investigate novel treatment approaches.
eMethods.
eTable 1. Coefficient Table With Standard Error (S.E.), Wald Z Test and P-Value in the Final Model Including Main Effect and Interaction Terms
eTable 2. Odds Ratio of Developing Chronic irAEs per Interquartile Range (IQR) Increase in Log(Time to Onset + 1) Given the Combination of Steroid Exposure and Anti-PD-1 Treatment Status
eTable 3. Proteomic Cohort Characteristics
eTable 6. Sensitivity Analyses: Effect Size and FDR-Adjusted P
eFigure 1. Estimated Odds Ratios (ORs) and 95% Confidence Intervals for Chronic Immune-Related Adverse Events (irAEs) in Steroid Users Versus Non-Steroid Users as a Function of Time to irAE Onset
eFigure 2. Estimated Odds Ratios (ORs) and 95% Confidence Intervals for Chronic Immune-Related Adverse Events (irAEs) in Patients Who Continued Anti-PD-(L)1 Therapy Compared With Those Who Discontinued Anti-PD-(L)1 Therapy Early, as a Function of Time to irAE Onset
eFigure 3. Partial Effect Plots of the Predicted Probability of Chronic irAE From the Final Logistic Regression Model, Incorporating Two Interaction Terms: (1) Time to irAEs Onset × Steroid Exposure (Top Panel) and (2) Time to irAE Onset × Anti–PD-1 Discontinuation (Bottom Panel)
eTable 4.
eTable 5.
Data Sharing Statement
References
- 1.Ferlay J, Colombet M, Soerjomataram I, et al. Estimating the global cancer incidence and mortality in 2018: GLOBOCAN sources and methods. Int J Cancer. 2019;144(8):1941-1953. doi: 10.1002/ijc.31937 [DOI] [PubMed] [Google Scholar]
- 2.Ugurel S, Röhmel J, Ascierto PA, et al. Survival of patients with advanced metastatic melanoma: the impact of novel therapies—update 2017. Eur J Cancer. 2017;83:247-257. doi: 10.1016/j.ejca.2017.06.028 [DOI] [PubMed] [Google Scholar]
- 3.Vilgelm AE, Johnson DB, Richmond A. Combinatorial approach to cancer immunotherapy: strength in numbers. J Leukoc Biol. 2016;100(2):275-290. doi: 10.1189/jlb.5RI0116-013RR [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Wolchok JD, Chiarion-Sileni V, Rutkowski P, et al. ; CheckMate 067 Investigators . Final, 10-year outcomes with nivolumab plus ipilimumab in advanced melanoma. N Engl J Med. 2025;392(1):11-22. doi: 10.1056/NEJMoa2407417 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Grossmann KF, Othus M, Patel SP, et al. Adjuvant pembrolizumab versus IFNα2b or ipilimumab in resected high-risk melanoma. Cancer Discov. 2022;12(3):644-653. doi: 10.1158/2159-8290.CD-21-1141 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Dougan M, Luoma AM, Dougan SK, Wucherpfennig KW. Understanding and treating the inflammatory adverse events of cancer immunotherapy. Cell. 2021;184(6):1575-1588. doi: 10.1016/j.cell.2021.02.011 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Goodman RS, Lawless A, Woodford R, et al. Extended follow-up of chronic immune-related adverse events following adjuvant anti–PD-1 therapy for high-risk resected melanoma. JAMA Netw Open. 2023;6(8):e2327145. doi: 10.1001/jamanetworkopen.2023.27145 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Goodman RS, Di Guardo L, Maurichi A, et al. Long-term outcomes and persistent toxicities following BRAF/MEK inhibitor therapy for advanced melanoma. Eur J Cancer. 2023;194:113354. doi: 10.1016/j.ejca.2023.113354 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Patrinely JR Jr, Johnson R, Lawless AR, et al. Chronic immune-related adverse events following adjuvant anti–PD-1 therapy for high-risk resected melanoma. JAMA Oncol. 2021;7(5):744-748. doi: 10.1001/jamaoncol.2021.0051 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Fletcher KA, Goodman RS, Lawless A, et al. Characterizing chronic cutaneous immune-related adverse events following immune checkpoint inhibitors. JAMA Dermatol. 2025;161(5):555-557. doi: 10.1001/jamadermatol.2025.0025 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Johnson DB, Nebhan CA, Moslehi JJ, Balko JM. Immune-checkpoint inhibitors: long-term implications of toxicity. Nat Rev Clin Oncol. 2022;19(4):254-267. doi: 10.1038/s41571-022-00600-w [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Park BC, Jung S, Wright JJ, Johnson DB. Recurrence of hypophysitis after immune checkpoint inhibitor rechallenge. Oncologist. 2022;27(12):e967-e969. doi: 10.1093/oncolo/oyac220 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Cortellini A, Vitale MG, De Galitiis F, et al. Early fatigue in cancer patients receiving PD-1/PD-L1 checkpoint inhibitors: an insight from clinical practice. J Transl Med. 2019;17(1):376. doi: 10.1186/s12967-019-02132-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Tezuka T, Okuzumi S, Nakashima C, et al. Dysautonomia associated with immune checkpoint inhibitors. J Neurol. 2023;270(7):3413-3423. doi: 10.1007/s00415-023-11667-5 [DOI] [PubMed] [Google Scholar]
- 15.Schneider BJ, Naidoo J, Santomasso BD, et al. Management of immune-related adverse events in patients treated with immune checkpoint inhibitor therapy: ASCO guideline update. J Clin Oncol. 2021;39(36):4073-4126. doi: 10.1200/JCO.21.01440 [DOI] [PubMed] [Google Scholar]
- 16.Gray-Gaillard EF, Shah AA, Bingham Iii CO, et al. Higher levels of VEGF-A and TNFα in patients with immune checkpoint inhibitor–induced inflammatory arthritis. Arthritis Res Ther. 2025;27(1):74. doi: 10.1186/s13075-025-03546-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Les I, Martínez M, Pérez-Francisco I, et al. Predictive biomarkers for checkpoint inhibitor immune-related adverse events. Cancers (Basel). 2023;15(5):1629. doi: 10.3390/cancers15051629 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Oya K, Nakamura Y, Shen LT, Ishizuki S, Matsusaka S, Fujisawa Y. Soluble PD-L1 predicts tumor response and immune-related adverse events in patients with advanced melanoma treated with anti–PD-1 antibodies. J Dermatol. 2024;51(6):807-815. doi: 10.1111/1346-8138.17183 [DOI] [PubMed] [Google Scholar]
- 19.Henderson Berg MH, Del Rincón SV, Miller WH. Potential therapies for immune-related adverse events associated with immune checkpoint inhibition: from monoclonal antibodies to kinase inhibition. J Immunother Cancer. 2022;10(1):e003551. doi: 10.1136/jitc-2021-003551 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Brown AC, Quiroz J, Parikh DA, et al. Cytokines in PD-1 immune checkpoint inhibitor adverse events and implications for the treatment of uveitis. BMC Ophthalmol. 2024;24(1):312. doi: 10.1186/s12886-024-03575-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Liu X, Shi Y, Zhang D, et al. Risk factors for immune-related adverse events: what have we learned and what lies ahead? Biomark Res. 2021;9(1):79. doi: 10.1186/s40364-021-00314-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Reschke R, Sullivan RJ, Lipson EJ, Enk AH, Gajewski TF, Hassel JC. Targeting molecular pathways to control immune checkpoint inhibitor toxicities. Trends Immunol. 2025;46(1):61-73. doi: 10.1016/j.it.2024.11.014 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Oray M, Abu Samra K, Ebrahimiadib N, Meese H, Foster CS. Long-term side effects of glucocorticoids. Expert Opin Drug Saf. 2016;15(4):457-465. doi: 10.1517/14740338.2016.1140743 [DOI] [PubMed] [Google Scholar]
- 24.Hu X, Bukhari SM, Tymm C, et al. Inhibition of IL-25/IL-17RA improves immune-related adverse events of checkpoint inhibitors and reveals antitumor activity. J Immunother Cancer. 2024;12(3):e008482. doi: 10.1136/jitc-2023-008482 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Sierra-Davidson K, Boland GM. Targeting IL-17A to combat immune-related adverse events. Nat Cancer. 2024;5(9):1289-1291. doi: 10.1038/s43018-024-00804-2 [DOI] [PubMed] [Google Scholar]
- 26.Johnson D, Patel AB, Uemura MI, et al. IL17A blockade successfully treated psoriasiform dermatologic toxicity from immunotherapy. Cancer Immunol Res. 2019;7(6):860-865. doi: 10.1158/2326-6066.CIR-18-0682 [DOI] [PubMed] [Google Scholar]
- 27.Yin Q, Wu L, Han L, et al. Immune-related adverse events of immune checkpoint inhibitors: a review. Front Immunol. 2023;14:1167975. doi: 10.3389/fimmu.2023.1167975 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Perez-Ruiz E, Minute L, Otano I, et al. Prophylactic TNF blockade uncouples efficacy and toxicity in dual CTLA-4 and PD-1 immunotherapy. Nature. 2019;569(7756):428-432. doi: 10.1038/s41586-019-1162-y [DOI] [PubMed] [Google Scholar]
- 29.Johnson DH, Zobniw CM, Trinh VA, et al. Infliximab associated with faster symptom resolution compared with corticosteroids alone for the management of immune-related enterocolitis. J Immunother Cancer. 2018;6(1):103. doi: 10.1186/s40425-018-0412-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Hoeijmakers LL, Dimitriadis P, Wijnen SCMA, et al. Neoadjuvant ipilimumab plus nivolumab in melanoma: 5-year survival and biomarker analysis from the phase 2 PRADO-trial. Nat Med. 2026;32(3):952-963. doi: 10.1038/s41591-025-04158-9 [DOI] [PubMed] [Google Scholar]
- 31.Rosenzweig N, Dvir-Szternfeld R, Tsitsou-Kampeli A, et al. PD-1/PD-L1 checkpoint blockade harnesses monocyte-derived macrophages to combat cognitive impairment in a tauopathy mouse model. Nat Commun. 2019;10(1):465. doi: 10.1038/s41467-019-08352-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eMethods.
eTable 1. Coefficient Table With Standard Error (S.E.), Wald Z Test and P-Value in the Final Model Including Main Effect and Interaction Terms
eTable 2. Odds Ratio of Developing Chronic irAEs per Interquartile Range (IQR) Increase in Log(Time to Onset + 1) Given the Combination of Steroid Exposure and Anti-PD-1 Treatment Status
eTable 3. Proteomic Cohort Characteristics
eTable 6. Sensitivity Analyses: Effect Size and FDR-Adjusted P
eFigure 1. Estimated Odds Ratios (ORs) and 95% Confidence Intervals for Chronic Immune-Related Adverse Events (irAEs) in Steroid Users Versus Non-Steroid Users as a Function of Time to irAE Onset
eFigure 2. Estimated Odds Ratios (ORs) and 95% Confidence Intervals for Chronic Immune-Related Adverse Events (irAEs) in Patients Who Continued Anti-PD-(L)1 Therapy Compared With Those Who Discontinued Anti-PD-(L)1 Therapy Early, as a Function of Time to irAE Onset
eFigure 3. Partial Effect Plots of the Predicted Probability of Chronic irAE From the Final Logistic Regression Model, Incorporating Two Interaction Terms: (1) Time to irAEs Onset × Steroid Exposure (Top Panel) and (2) Time to irAE Onset × Anti–PD-1 Discontinuation (Bottom Panel)
eTable 4.
eTable 5.
Data Sharing Statement
