Impact of adjuvant therapy on melanoma-specific survival in elderly patients with high-risk resected melanoma: a multicenter international retrospective study

Abstract

Purpose

Adjuvant systemic therapy improves recurrence-free survival in resected high-risk melanoma, but elderly patients remain markedly under-represented in clinical trials; their competing risk of death from other causes needs to be considered. Consequently, the clinical benefit of adjuvant therapy in older populations remains uncertain. We conducted an international real-world study to evaluate survival outcomes of patients aged ≥75 years with resected stage III-IV melanoma managed with adjuvant systemic therapy or observation.

Methods

In this multicenter retrospective cohort study, patients aged ≥75 years with completely resected stage III-IV cutaneous melanoma were identified across 14 international melanoma centers. Patients received either standard adjuvant systemic therapy (anti-programmed cell death protein 1 immunotherapy or BRAF/MEK-targeted therapy) or observation. The primary endpoint were melanoma-specific survival (MSS) and overall survival (OS). Secondary endpoints included recurrence-free survival (RFS). Survival outcomes were estimated using Kaplan-Meier methods and evaluated with multivariable Cox regression models adjusted for age, stage, sex, and mutation status.

Results

A total of 580 patients were included (median age 79 years; range, 75–99), of whom 222 (38%) received adjuvant therapy and 358 (62%) underwent observation. Adjuvant therapy was associated with significantly improved MSS, with 5-year MSS of 74% compared with 60% in the observation group (adjusted HR 0.58; 95% CI 0.40 to 0.85; p=0.005), as well as OS (adjusted HR 0.66, 95% CI 0.51 to 0.87; p=0.002). Adjuvant therapy was also associated with improved RFS (HR 0.57, 95% CI 0.45 to 0.72; p<0.0001). Among patients managed with observation, disease recurrence occurred in 69%, yet only 47% of those patients received systemic therapy as first-line management of recurrence. Notably, among observation patients who remained recurrence-free or received systemic therapy after recurrence, MSS was comparable to that observed in adjuvant-treated patients (HR 0.83; 95% CI 0.55 to 1.25).

Conclusion

In this large international real-world cohort of elderly patients with resected melanoma, adjuvant therapy was associated with clinically meaningful improvements in MSS, OS and RFS. Although immunotherapy retained efficacy when administered at recurrence, many elderly patients did not receive systemic therapy as first-line management of recurrence. Adjuvant therapy may therefore represent an important strategy to ensure timely access to effective treatment in carefully selected older patients.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Adjuvant anti-programmed cell death protein 1 immunotherapy and BRAF/MEK-targeted therapy improve recurrence-free survival in resected high-risk melanoma, but no pivotal trial has demonstrated a significant overall survival benefit after nearly a decade of follow-up. Patients aged 75 years and older were markedly under-represented in these trials, yet they account for a disproportionate share of melanoma mortality, and in this age group comorbidity, competing non-melanoma mortality and treatment toxicity may erode any absolute benefit. Real-world evidence in the very elderly is sparse and conflicting.

WHAT THIS STUDY ADDS

• In 580 patients aged 75 years and older treated across 14 international centers, adjuvant therapy was independently associated with improved melanoma-specific survival (5-year 74% vs 60%; adjusted HR 0.58, 95% CI 0.40 to 0.85), as well as with overall and recurrence-free survival. Tolerability was less favorable than in trial populations: grade 2 or higher immune-related adverse events occurred in 42% of treated patients and 55% did not complete the planned 12 months, yet the survival benefit was preserved. Among patients managed by observation who either remained recurrence-free or received systemic therapy after recurrence, melanoma-specific survival was comparable to that of adjuvant-treated patients; however, 130 of the 247 patients who recurred (53%) did not receive systemic therapy as first-line management of recurrence.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Deferring immunotherapy until recurrence is a defensible strategy only for older patients who will still be treatable when they relapse; adjuvant therapy may preserve access to effective treatment for those who would otherwise become ineligible in the interval between surgery and recurrence. The preservation of benefit despite frequent early discontinuation supports the evaluation of shorter, response-adapted perioperative strategies in this age group. Chronological age should not by itself determine treatment: prospective adjuvant studies in older adults should embed geriatric assessment as a stratification factor and prespecified effect modifier.

Introduction

Adjuvant systemic therapy has become a standard of care for patients with high-risk resected melanoma. Immune checkpoint inhibitors (ICIs) such as nivolumab and pembrolizumab, as well as targeted therapy with dabrafenib plus trametinib in BRAF-mutated disease, have demonstrated significant improvements in recurrence-free survival (RFS) compared with placebo or previous standards of care.^1–3 These therapies are typically administered for 1 year and are widely adopted in routine clinical practice for patients with stage III-IV and selected stage II melanoma.^{4 5}

To date, despite long-term follow-up, a clear benefit to overall survival (OS) in the adjuvant setting has not been shown in any clinical trial or real-world population. Although long-term follow-up from pivotal trials confirms durable improvements in RFS and distant metastasis-free survival (DMFS), differences in OS remain modest or statistically non-significant, likely reflecting the availability of highly effective systemic therapies at recurrence.⁶ For example, updated analyses of CheckMate-238 showed sustained improvements in RFS with nivolumab compared with ipilimumab but no substantial difference in OS despite nearly a decade of follow-up.⁶

At the same time, melanoma incidence is increasingly shifting toward older populations. Epidemiologic studies from Europe and North America have reported a steady rise in the proportion of patients diagnosed after the age of 75, a group that now represents a substantial fraction of melanoma cases and accounts for a disproportionate share of melanoma-related mortality.^7–9 Despite this demographic shift, elderly patients remain under-represented in randomized clinical trials. In CheckMate 238, the median age was 56 years (range, 19–83); in COMBI-AD, only about 3% of patients were older than 75 years as expected given the low frequency of BRAF ^V600E in older patients¹⁰; and in KEYNOTE-054, patients older than 65 years represented only approximately one quarter of the trial population.^{6 11 12}

The clinical decision to offer adjuvant therapy in elderly patients therefore raises several unresolved questions. On the one hand, aging is associated with cumulative ultraviolet exposure and a higher tumor mutational burden, factors that may enhance the immunogenicity of melanoma and potentially increase responsiveness to ICIs.¹³ On the other hand, older patients frequently present with comorbidities, reduced physiologic reserve, and competing risks of death, which may attenuate the absolute survival benefit derived from 1 year of systemic therapy and increase the clinical relevance of treatment-related toxicity, cross-drug interactions, and quality-of-life impairment.

Furthermore, with the availability of highly effective systemic therapies for advanced/recurrent disease,^{14 15} an alternative strategy, such as active surveillance followed by treatment at disease recurrence, may represent a reasonable option for selected elderly patients. Whether early treatment in the adjuvant setting translates into a meaningful survival advantage in this population remains uncertain.

To address this knowledge gap, we conducted a large international retrospective cohort study evaluating outcomes of patients aged ≥75 years with completely resected stage III-IV melanoma who were managed with either adjuvant systemic therapy or observation. The primary aim of this study was to assess melanoma-specific survival (MSS) and OS in this population and to estimate the clinical benefit of adjuvant therapy, in the context of competing risks and real-world treatment patterns.

Methods

Study design and patient population

This was a retrospective, multicenter, international, observational cohort study conducted at Melanoma Institute Australia (MIA) and collaborating 14 international centers across Australia, USA and Europe. We identified patients aged ≥75 years with completely resected stage III or IV cutaneous melanoma (American Joint Committee on Cancer [AJCC] eighth edition) who received either standard adjuvant systemic therapy or observation. All deceased patients were eligible for inclusion regardless of follow-up duration; for patients still alive, a minimum of 3 years of follow-up was required. Patients with non-cutaneous melanoma (uveal or mucosal), incomplete treatment or survival data, or loss to follow-up before 3 years (if alive) were excluded.

Data collection

Clinical data were extracted from the electronic medical records starting from January 2010, of the institutional databases at participating centers. All patients either provided written informed consent for the use of their clinical data, or the requirement for consent was waived by the relevant ethical review board. This study was approved by the Melanoma Institute Australia Research Committee (MIA2024/506), conducted under the Sydney Local Health District HREC–approved Molecular Protocol X17−0312/2019/ETH07604 and MIA Research Database protocol X15−0311/2019/ETH06854. Patient information was de-identified prior to analysis. Collected variables included patient demographics, primary melanoma characteristics (histological subtype, Breslow thickness, ulceration, mitotic rate), Eastern Cooperative Oncology Group (ECOG) performance status (PS), follow-up status (alive or deceased, with cause of death), adjuvant treatment details (agent, duration, treatment-related grade ≥2 adverse events, and endocrine adverse events), recurrence events (date, pattern, and site), and subsequent treatments, responses in the advanced setting, and cause of death (melanoma vs non-melanoma related). No formal geriatric evaluation was available for any patient: neither a comprehensive geriatric assessment (CGA) nor an abbreviated screening or comorbidity instrument (G8, Charlson Comorbidity Index, Cumulative Illness Rating Scale-Geriatric [CIRS-G]) was routinely performed or recorded in the source medical records of the participating centers over the study period, and these scores could not be reconstructed retrospectively from the available documentation.

Patients were classified into two groups: those who received standard adjuvant therapy (anti-programmed cell death protein 1 (PD-1) immunotherapy (nivolumab or pembrolizumab) or targeted therapy with dabrafenib plus trametinib, for BRAF^V600-mutated tumors) and those who underwent active surveillance (observation).

Endpoints

The primary endpoints were MSS and OS. MSS was defined as time from the date of stage III/IV surgery to death specifically attributable to melanoma. OS was defined as time from the date of surgery to death from any cause. RFS was defined as time from the date of surgery to disease recurrence (as determined by the treating physician) or death from any cause. Patients who did not experience the event of interest were censored at their last available follow-up date.

Statistical analysis

Patient characteristics were summarized using medians and percentages, stratified by receipt of adjuvant therapy (yes vs no). Between-group differences were assessed using χ² tests for categorical variables and Wilcoxon rank-sum tests for continuous variables, as appropriate. RFS, OS, and MSS were summarized using the Kaplan-Meier method. Landmark survival rates at 1, 3, and 5 years post-surgery, along with their 95% CIs, are reported. Differences in survival between groups were assessed using the log-rank test. The independent association of adjuvant therapy with each outcome (RFS, OS, and MSS) was evaluated by univariate and multivariate Cox proportional hazards regression using the Wald test, adjusting for clinically relevant baseline confounders including age, sex, AJCC stage and mutation status. To address potential collinearity between composite AJCC stage and its individual components, two prespecified multivariable models were compared for each endpoint: Model A included composite AJCC stage (IIIA/IIIB/IIIC/IIID/IV no evidence of disease), whereas Model B decomposed stage into its constituent components: T category (used as Breslow-thickness tier), ulceration status, and AJCC N classification (N class N1/N2/N3 and N subcategory a/b/c). Both models were also fitted with mutation status as a sensitivity analysis (complete-case). Discrimination between Models A and B was compared using Harrell’s concordance (C-) index. A two-sided p value of <0.05 was considered statistically significant. All analyses were performed using SAS V.9.4 (SAS Institute, Cary, North Carolina, USA) and R V.4.5.2 (R Core Team, Vienna, Austria).

Results

Patient population

A total of 580 patients aged ≥75 years with completely resected stage III-IV melanoma were identified across 14 international participating centers (including Australia, USA and Europe). Of these, 222 (38%) received adjuvant systemic therapy and 358 (62%) underwent observation.

Baseline characteristics are summarized in table 1. The median age at surgery was 78 years (range 75–93) in the adjuvant group and 80 years (range 75–99) in the observation group (p<0.001). Patients who received adjuvant therapy were more likely to be aged ≤85 years (93% vs 80%; p<0.001), reflecting the selection of fitter patients for treatment in clinical practice. Sex distribution was balanced between groups (68% male in both; p=0.75). The predominant stage was IIIC, representing 59% of the overall cohort. Stage distribution differed slightly between groups (p=0.024). Median follow-up was 41.1 months (range 3.4–133.8) in the adjuvant group and 38.8 months (range 1.6–151.4) in the observation group (p=0.92). ECOG PS was available for 322 patients (55%), and 43% had ECOG PS ≥1 (and was similar between groups). Its availability, however, differed markedly between groups: ECOG PS was documented for 205 of 222 adjuvant-treated patients (92%) but for only 117 of 358 observation patients (33%) (table 1). Within the observation cohort the gradient followed access to medical oncology rather than disease status: PS was recorded for only 29 of the 130 patients whose recurrence was not managed with first-line systemic therapy (22%), reflecting the fact that many patients managed by observation were never referred to a medical oncologist and therefore never had a performance status formally recorded. Mutational status was available for 408 patients (70%), with BRAF mutations identified in 36% and 29% of adjuvant and observation patients, respectively (p=0.24).

Table 1. Baseline characteristics of the included population.

Characteristics	All (N=580)	Observation (N=358)	Treatment (N=222)	P value
Age (years) at stage III surgery
Median (range)	79.0 (75.0–99.0)	80.5 (75.0–99.0)	78.0 (75.0–93.0)
Sex
Female	183/578 (31.7%)	111/356 (31.2%)	72/222 (32.4%)	0.753
Male	395/578 (68.3%)	245/356 (68.8%)	150/222 (67.6%)
Stage
IIIA	23/580 (4.0%)	20/358 (5.6%)	3/222 (1.4%)	0.024
IIIB	161/580 (27.8%)	90/358 (25.1%)	71/222 (32.0%)
IIIC	343/580 (59.1%)	220/358 (61.5%)	123/222 (55.4%)
IIID	33/580 (5.7%)	18/358 (5.0%)	15/222 (6.8%)
IV NED	20/580 (3.4%)	10/358 (2.8%)	10/222 (4.5%)
Breslow thickness (AJCC T category)
T0	52/580 (9.0%)	31/358 (8.7%)	21/222 (9.5%)	0.807
T1	32/580 (5.5%)	19/358 (5.3%)	13/222 (5.9%)
T2	76/580 (13.1%)	46/358 (12.8%)	30/222 (13.5%)
T3	177/580 (30.5%)	113/358 (31.6%)	64/222 (28.8%)
T4	205/580 (35.3%)	130/358 (36.3%)	75/222 (33.8%)
Tis	4/580 (0.7%)	2/358 (0.6%)	2/222 (0.9%)
Tx	34/580 (5.9%)	17/358 (4.7%)	17/222 (7.7%)
Ulceration
Absent	198/580 (34.1%)	122/358 (34.1%)	76/222 (34.2%)	0.578
Present	286/580 (49.3%)	181/358 (50.6%)	105/222 (47.3%)
Unknown	96/580 (16.6%)	55/358 (15.4%)	41/222 (18.5%)
AJCC N category
N0	8/580 (1.4%)	3/358 (0.8%)	5/222 (2.3%)	0.110
N1a	128/580 (22.1%)	79/358 (22.1%)	49/222 (22.1%)
N1b	100/580 (17.2%)	73/358 (20.4%)	27/222 (12.2%)
N1c	85/580 (14.7%)	44/358 (12.3%)	41/222 (18.5%)
N2a	30/580 (5.2%)	17/358 (4.7%)	13/222 (5.9%)
N2b	68/580 (11.7%)	43/358 (12.0%)	25/222 (11.3%)
N2c	44/580 (7.6%)	26/358 (7.3%)	18/222 (8.1%)
N3a	7/580 (1.2%)	3/358 (0.8%)	4/222 (1.8%)
N3b	46/580 (7.9%)	32/358 (8.9%)	14/222 (6.3%)
N3c	31/580 (5.3%)	21/358 (5.9%)	10/222 (4.5%)
Unknown	33/580 (5.7%)	17/358 (4.7%)	16/222 (7.2%)
Follow-up time (months)
Median (range)	39.5 (1.6–151.4)	38.8 (1.6–151.4)	41.1 (3.4–133.8)
ECOG
0	184/322 (57.1%)	62/117 (53.0%)	122/205 (59.5%)	0.499*
1	117/322 (36.3%)	45/117 (38.5%)	72/205 (35.1%)
2	16/322 (5.0%)	7/117 (6.0%)	9/205 (4.4%)
3	5/322 (1.6%)	3/117 (2.6%)	2/205 (1.0%)
Not available	258/580 (44.5%)	241/358 (67.3%)	17/222 (7.7%)	<0.0001†
Mutation
BRAF	131/408 (32.1%)	64/220 (29.1%)	67/188 (35.6%)	0.245
KIT	9/408 (2.2%)	5/220 (2.3%)	4/188 (2.1%)
NRAS	108/408 (26.5%)	66/220 (30.0%)	42/188 (22.3%)
Wild type	158/408 (38.7%)	83/220 (37.7%)	75/188 (39.9%)
Year of diagnosis				<0.001
2010–2013	112/580 (19.3%)	111/358 (31.0%)	1/222 (0.5%)
2014–2016	96/580 (16.6%)	87/358 (24.3%)	9/222 (4.1%)
2017–2019	169/580 (29.1%)	75/358 (20.9%)	94/222 (42.3%)
≥2020	203/580 (35.0%)	85/358 (23.7%)	118/222 (53.2%)

P values were calculated using the two-sample t-test for continuous variables and the χ² test (or Fisher’s exact test where cell counts <5) for categorical variables.

ECOG PS was available for 322 of 580 patients (55.5%).

Mutational status for 408 of 580 (70.3%); including KRAS-mutant patients (n=2).

^*Comparison of ECOG PS distribution restricted to the 322 patients with a documented score.

^†Comparison of availability (documented vs not available).

AJCC, American Joint Committee on Cancer; ECOG, Eastern Cooperative Oncology Group; NED, no evidence of disease; PS, performance status.

Adjuvant treatment, tolerability, and post-recurrence management

Among the 222 adjuvant-treated patients, the most common agents were the anti-PD-1 nivolumab and pembrolizumab (89%), with dabrafenib plus trametinib in 7% of patients (table 2). Median treatment duration was 9.5 months (range 0–35). Treatment was discontinued early in 122 patients (55%), including 50 (23%) due to disease recurrence, 52 (23%) due to treatment-related toxicity, 2 (1%) due to non-treatment-related death, and 18 (8%) for other elective reasons, including deterioration in general condition unrelated to treatment, development of intercurrent illnesses, or patient choice.

Table 2. Adjuvant treatment characteristics and irAEs ≥grade 2.

Characteristic (N=222)	N (%)
Type of adjuvant therapy
Nivolumab	135 (61)
Pembrolizumab	62 (28)
Dabrafenib+trametinib	15 (7)
Other*	10 (4)
Treatment duration	N=219
Median, months (range)	9.5 (0.0–35.0)
irAEs ≥grade 2, any	94 (42)
Endocrine
Thyroid dysfunction	23 (10)
Adrenal insufficiency	15 (7)
Gastrointestinal
Colitis	14 (6)
Hepatitis	4 (2)
Pulmonary
Pneumonitis	10 (4)
Dermatologic
Skin rash/psoriasis	11 (5)
Bullous dermatitis	4 (2)
Cardiac
Myocarditis†	5 (2)
Musculoskeletal
Myositis	4 (2)
Other (each <2%)	‡

All percentages are calculated on the total treated population (N=222). irAE counts reflect cumulative events across all toxicity episodes per patient.

^*Other includes: clinical trial agent (BMSCA2098FC), dabrafenib monotherapy, vemurafenib, ipilimumab+nivolumab, pembrolizumab+mRNA-4157.

^†One patient had fatal (grade 5) myocarditis.

^‡Includes fatigue, fever, hepatitis, myasthenia gravis, nephritis, peripheral neuropathy, thrombocytopenia, thrombosis, encephalitis, esophagitis, arthritis, gastritis, sarcoidosis.

irAE, immune-related adverse event.

Immune-related adverse events (irAEs) of grade ≥2 occurred in 94 patients (42%). The most frequent irAEs were thyroid dysfunction (10%), adrenal insufficiency (7%), colitis (6%), and skin rash or psoriasis (5%). Pneumonitis occurred in 10 patients (4%), myocarditis in 5 (2%), including one case of fatal myocarditis. No new or unexpected safety signals were identified in this elderly cohort.

Among the 358 observation patients, disease recurrence occurred in 247 (69%), of whom 117 (47%) received systemic therapy as first-line management of recurrence. The most common systemic agents were anti-PD-1 monotherapy (63%), BRAF/MEK inhibitors (18%), and ipilimumab-containing regimens (12%) (table 3). The remaining 130 patients did not receive first-line systemic therapy for their recurrence; of these, 26 received systemic therapy later, after initial locoregional treatment, and 104 never received it. Initial post-recurrence management in this subgroup included locoregional treatment in 91 patients (70%) and best supportive care in 26 (20%); it was not recorded in 13 (10%). The median age of patients who did not receive first-line systemic therapy despite recurrence was 81 years. Among these patients, recurrence had a distant component in 40 (31%). Notably, only 64 of the 128 recurrences with a recorded date (50%) occurred in the post-2015 era, and 17 patients had BRAF-mutant disease (26% of those with known mutational status). The specific reasons for not initiating systemic therapy were not available due to the retrospective nature of the study. Importantly, the proportion of patients receiving systemic therapy at any time after recurrence increased markedly between the pre-immunotherapy era (2010–2013, 14.8%) and the subsequent periods (2014–2016, 68.7%; 2017–2019, 74.5%; ≥2020, 68.5%; Cochran-Armitage trend test, p<0.0001), reaching a plateau from 2014 onwards without further significant gains over time; nonetheless, roughly one-third of patients relapsing in the modern era never received systemic treatment (online supplemental table S1).

Table 3. Recurrence and management in the observation cohort.

Characteristics	All (N=580)	Observation (N=358)	Treatment (N=222)	P value
First recurrence type
Distant	121/360 (33.6%)	75/247 (30.4%)	46/113 (40.7%)	0.094
Regional	205/360 (56.9%)	145/247 (58.7%)	60/113 (53.1%)
Both	34/360 (9.4%)	27/247 (10.9%)	7/113 (6.2%)
Initial management of first recurrence (n=247)*
Systemic therapy alone		87 (35%)
Surgery alone		72 (29%)
Surgery+systemic		20 (8%)
BSC		26 (11%)
Radiotherapy alone		9 (4%)
Surgery+RT		8 (3%)
RT+systemic		7 (3%)
Surgery+RT+systemic		2 (1%)
Local therapy (TVEC/diphencyprone)		2 (1%)
Systemic+TVEC		1 (<1%)
Not recorded		13 (5%)
Systemic therapy for recurrence (n=247)
As first-line management		117 (47%)
Only later, after locoregional treatment		26 (11%)
Never		104 (42%)
First-line systemic regimen (n=117)†
Anti-PD-1 monotherapy		74 (63%)
BRAF/MEK inhibitors		21 (18%)
Ipilimumab+anti-PD-1		7 (6%)
Ipilimumab monotherapy		7 (6%)
Nivolumab+relatlimab		5 (4%)
Anti-PD-1+experimental agent		3 (3%)

^*Treatment modality, or combination of modalities, used first to manage the first recurrence. Categories are mutually exclusive; percentages are calculated among the 247 observation patients with recurrence. Post-recurrence management was not collected for the adjuvant cohort.

^†Percentages are calculated among the 117 patients who received systemic therapy as first-line management of recurrence

BRAF/MEK, serine/threonine-protein kinase B-Raf/mitogen-activated protein kinase; BSC, best supportive care; PD-1, programmed cell death protein 1; RT, radiotherapy; TVEC, talimogene laherparepvec.

Melanoma-specific survival and overall survival

Death without documented melanoma recurrence was observed in both groups. In the observation cohort, 66 patients (18%) died without a reported distant recurrence, including 43 (12%) who had no recorded melanoma recurrence of any type. Similarly, among the 84 deaths observed in the adjuvant-treated group, 26 (12%) occurred without prior distant recurrence and 18 (8%) without any documented recurrence. In the observation group, 50 deaths were attributed to non-melanoma-related causes (14%); while in the adjuvant-treated group, 24 deaths were attributed to non-melanoma-related causes (11%).

Adjuvant therapy was associated with a significant improvement in MSS in the overall study population (figure 1A). At 1, 3, and 5 years, MSS probabilities were 96.7% (95% CI 94.6% to 99.2%), 81.6 (95% CI 76.3 to 87.1), and 74.3 (95% CI 67.8 to 81.4) in the adjuvant group, compared with 88.5 (95% CI 85.2 to 91.9), 70.2 (95% CI 65.4 to 75.3), and 60.0 (95% CI 54.4 to 66.1) in the observation group, representing a 14.3 percentage point absolute difference in 5-year MSS. On multivariable Cox regression adjusted for age, sex, AJCC stage, and mutation status, adjuvant therapy was independently associated with improved MSS (HR 0.58, 95% CI 0.40 to 0.85; p=0.005). Wild-type mutation status was also independently associated with better MSS (HR 0.64, 95% CI 0.44 to 0.94; p=0.022). A complementary multivariable specification (Model B), in which composite AJCC stage was decomposed into T category, ulceration status and AJCC N classification (including the a/b/c subcategories), confirmed the independent prognostic effect of adjuvant therapy (HR 0.62, 95% CI 0.44 to 0.87; p=0.006) and identified ulceration (HR 1.68, 95% CI 1.17 to 2.41) and N3 disease (HR 3.63, 95% CI 2.44 to 5.39) as the strongest independent prognostic factors, while T tier and the N a/b/c subclassification contributed little additional prognostic information once N class was accounted for. Model B showed numerically higher discrimination than Model A (Harrell’s C-index 0.70 vs 0.64 for MSS; 0.67 vs 0.64 for OS). Detailed univariable and multivariable estimates for both models and across all endpoints are reported in online supplemental table S2.

Figure 1. MSS by treatment group in patients aged ≥75 years with resected stage III melanoma. (A) Kaplan-Meier curves for MSS in the overall study population. (B) Kaplan-Meier curves for MSS comparing adjuvant-treated patients with observation patients who remained recurrence-free or received systemic therapy after recurrence. Numbers at risk and landmark survival probabilities (95% CIs) at 1, 3, and 5 years are shown below each panel. MSS, melanoma-specific survival.

To evaluate whether adjuvant therapy conferred an MSS benefit beyond that achievable with systemic therapy at recurrence, a prespecified subgroup analysis was performed comparing all adjuvant-treated patients with a selected subgroup of observation patients. This subgroup included patients who remained recurrence-free during follow-up as well as those who experienced disease recurrence and received systemic therapy at any time thereafter, either as first-line management or later. Patients who never received systemic therapy after recurrence were not included in this comparison (figure 1B).

Overall, 247 of 358 observation patients (69%) experienced a recurrence during follow-up, of whom 143 (58%) received systemic therapy at any time. In this composite comparator cohort, MSS at 1, 3, and 5 years was 93.4% (95% CI 90.4% to 96.6%), 77.7% (95% CI 72.4% to 83.3%), and 70.1% (95% CI 63.9% to 76.8%), respectively, comparable to the adjuvant group. On multivariable analysis, adjuvant therapy was not significantly associated with MSS in this comparison (HR 0.83, 95% CI 0.55 to 1.25; table 4). Nodal category remained the dominant prognostic factor in this analysis. Ulcerated primary tumors were associated with worse MSS compared with non-ulcerated tumors (HR 1.97, 95% CI 1.26 to 3.08), and increasing nodal burden (N2–N3 vs N1) was associated with progressively inferior outcomes (N3: HR 2.85, 95% CI 1.72 to 4.73) (table 4).

Table 4. Cox proportional hazards analysis of melanoma-specific survival among observation patients who remained recurrence-free or received systemic therapy after recurrence, compared with all adjuvant-treated patients.

Variable	Univariable		Multivariable*
	HR	P value	HR	P value
Adjuvant
No	1	0.4598	1	0.3675
Yes	0.87 (0.60 to 1.26)		0.83 (0.55 to 1.25)
Age (years) at stage III surgery	1.02 (0.98 to 1.06)	0.3990
Age at stage III surgery
≤85 years	1	0.5965	1	0.7964
>85 years	1.16 (0.67 to 2.03)		0.91 (0.43 to 1.92)
Sex
Female	1	0.6264	1	0.9222
Male	1.10 (0.74 to 1.65)		1.02 (0.66 to 1.60)
T-stage
Non-ulcerated primary	1	0.0030	1	0.0029
Ulcerated primary	1.89 (1.24 to 2.88)		1.97 (1.26 to 3.08)
N-Stage
N1	1	<0.0001	1	0.0002
N2	1.74 (1.11 to 2.71)		1.81 (1.11 to 2.95)
N3	3.45 (2.21 to 5.38)		2.85 (1.72 to 4.73)
Mutation status
BRAF/other	1	0.0073	1	0.0051
Wild type	0.56 (0.37 to 0.85)		0.51 (0.31 to 0.82)

^*Multivariable model adjusted for age at stage III surgery, sex, ulceration, AJCC N category and mutation status. Mutation status was available in a subset of patients; those with missing data were excluded from the multivariable model.

AJCC, American Joint Committee on Cancer; NED, no evidence of disease.

Adjuvant therapy was associated with improved OS (figure 2B). 5-year OS was 60% (95% CI 53.3% to 67.9%) in the adjuvant group versus 45% (95% CI 39.8% to 50.9%) in the observation group. On multivariable analysis, adjuvant therapy was independently associated with reduced mortality from any cause (HR 0.66, 95% CI 0.51 to 0.87; p=0.002).

Figure 2. RFS and OS by treatment group in patients aged ≥75 years with resected stage III melanoma. (A) Kaplan-Meier curves for RFS. (B) Kaplan-Meier curves for OS. Numbers at risk and landmark survival probabilities (95% CIs) at 1, 3, and 5 years are shown below each panel. OS, overall survival; RFS, recurrence-free survival.

Recurrence-free survival

Adjuvant therapy was associated with a significant and clinically meaningful improvement in RFS (figure 2A). 5-year RFS was 44% (95% CI 36.6% to 51.7%) in the adjuvant group compared with 28% (95% CI 23.0% to 33.1%) in the observation group. On multivariable analysis, adjuvant therapy was independently associated with reduced risk of recurrence or death (HR 0.57, 95% CI 0.45 to 0.72; p<0.0001).

Sensitivity analysis

To account for the possibility that a proportion of observation patients had undergone surgery in an era preceding the availability of ICIs and therefore would not have been eligible for salvage systemic therapy at recurrence, we performed a sensitivity analysis excluding patients who underwent stage III surgery before January 2015. In this restricted cohort, representing patients treated entirely within the modern immunotherapy era, the survival estimates, and treatment effect remained fully consistent with the primary analysis (online supplemental figure S1).

Discussion

To our knowledge, this is the largest real-world analysis of patients aged ≥75 years with resected stage III-IV melanoma treated in the modern immunotherapy era. Adjuvant systemic therapy was associated with a significant improvement in MSS (the primary endpoint), with a 14 percentage-point absolute benefit at 5 years (HR 0.58), as well as in RFS and OS. With a median follow-up of 40 months, our study provides the longest follow-up yet reported in this population.

The pivotal trials that established adjuvant systemic therapy as standard of care enrolled substantially younger populations than those encountered in routine practice.^{6 11 12} Median ages were 54–58 years, with patients aged ≥75 years representing only a small minority (≈3% in COMBI-AD; in KEYNOTE-054, only ≈25% were older than 65 years).^{6 11 12} Despite durable RFS and DMFS gains, none has demonstrated a significant OS advantage after nearly a decade of follow-up.^{6 11 12} Against this backdrop, the OS signal in our cohort is noteworthy, although the attenuation from MSS (HR 0.58) to OS (HR 0.66) is consistent with competing non-melanoma mortality in this age group. Given the non-randomized design, treatment selection bias likely influenced outcomes: adjuvant-treated patients and those receiving systemic therapy at recurrence almost certainly represented a clinically fitter subgroup with more favorable disease kinetics, whereas a proportion of observation patients experienced early deterioration or death following recurrence that precluded systemic treatment. Comorbidity and frailty data were not uniformly available across participating centers, and ECOG PS was documented in only 55% of patients; PS and competing comorbidity are likely to have weighed heavily in both physician and patient decision-making regarding systemic therapy, and residual confounding by these unmeasured factors cannot be excluded. The observed HRs are therefore best regarded as an upper bound of the true treatment effect, and our data are consistent with, but cannot establish, a causal benefit of earlier intervention in this population.

A population-based Swedish registry study found no clear MSS or OS improvement at the population level despite widespread PD-1 use, with age-stratified analyses suggesting attenuation of effect in older patients and no clear benefit in the oldest.¹⁶ In contrast, a nationwide Danish registry of patients receiving adjuvant anti-PD-1 therapy reported improved survival without significant age-related differences.¹⁷ Together, these registries illustrated the heterogeneity of real-world outcomes in the adjuvant setting, particularly in older patients. Our findings in a highly selected ≥75 year cohort suggest that differences in patient selection and treatment eligibility may partially account for these divergent population-level observations.

Our findings are consistent with emerging real-world evidence on ICIs, while extending available data to a substantially older population. Observational studies have generally reported outcomes comparable to pivotal trials, with manageable toxicity.⁷ However, much of this literature derived from the metastatic setting, limiting its applicability to the adjuvant context. In the German ADOREG cohort, ICI efficacy in patients aged ≥75 years was not significantly impaired compared with younger patients, and anti-PD-1 monotherapy had a more favorable toxicity profile than combination ICIs.¹⁸ The Dutch Melanoma Treatment Registry similarly reported that, among elderly patients with metastatic disease, MSS was broadly comparable to that of younger patients despite lower treatment rates.¹⁹ These findings support preserved ICI activity in older adults but do not directly address the minimal residual disease setting characteristic of adjuvant therapy.

Dedicated data on adjuvant ICIs in patients aged ≥75 years remain sparse. A dedicated cohort of 456 patients receiving adjuvant pembrolizumab or nivolumab reported shorter DFS in those aged ≥75 years despite broadly comparable severe-toxicity rates.²⁰ That study also reported a median treatment duration of only 7 months in the elderly subgroup and frequent premature discontinuation for comorbidity or performance-status decline, mirroring our own findings.²⁰ A complementary international cohort of 928 patients aged ≥80 years receiving single-agent ICIs across tumor types reported comparable grade ≥3 irAE rates across age strata, but ICI discontinuation for toxicity was more than twice as frequent in those ≥90 versus <90 years (30.9% vs 15.1%; p=0.008),²¹ reinforcing that tolerability and treatment persistence, rather than intrinsic efficacy, may drive outcome differences in older patients. Our cohort contributes to addressing this evidence gap.

Our study focused exclusively on patients aged ≥75 years, with a median age of ≈80 years at surgery, extending the evidence base to the oldest and most clinically relevant segment encountered in routine practice. The low use of adjuvant dabrafenib plus trametinib in our cohort is consistent with the age-related decline in BRAF ^V600 mutation prevalence and the proportional shift toward V600K in chronically sun-damaged patients, which may derive comparable or greater benefit from immunotherapy than from targeted therapy.

A sensitivity analysis restricted to patients undergoing stage III surgery from 2015 onward (ie, those who would have had access to effective systemic therapy at recurrence) yielded survival estimates and treatment effects fully consistent with the primary analysis, indicating that the observed benefit was not driven by historical treatment limitations in the observation cohort but reflected a true outcome difference in contemporary practice.

A central clinical question is whether to withhold adjuvant therapy in elderly patients and reserve ICIs for recurrence, avoiding upfront toxicity. Our data showed that among observation patients who either remained recurrence-free or received systemic therapy after recurrence, MSS was comparable to that of adjuvant-treated patients (HR 0.83, 95% CI 0.55–1.25), suggesting that effective systemic therapy at recurrence can largely offset the lack of upfront treatment in this selected subgroup. This comparator combined, by design, two biologically distinct populations: patients who never recurred and patients who recurred and were fit enough to receive systemic therapy, because restricting it to the latter alone would have conditioned the comparison on a post-baseline event and introduced guarantee-time bias. The estimate should accordingly be read as describing the outcome of a deferred-treatment strategy as a whole, and not as a measure of the efficacy of salvage therapy in recurrent disease. However, this scenario applied to only 143 of the 358 observation patients (40%).

Tolerability was broadly acceptable: grade ≥2 irAEs occurred in 42% of treated patients, most commonly thyroid dysfunction (10%), adrenal insufficiency (7%), and colitis (6%), with one treatment-related death from myocarditis (0.5%). Notably, 55% did not complete the planned 12 months of therapy, substantially higher than the 10–20% rate in pivotal trials,^{6 11 12} reflecting the greater comorbidity burden of elderly patients. Despite this, meaningful MSS and RFS benefits were preserved, consistent with Danish registry data showing comparable outcomes between patients discontinuing early for toxicity and those completing the full course.¹⁷ The randomized phase 3 Grand-SLAM trial (NCT06488482) is now evaluating whether a 6-month course of adjuvant PD-1-based therapy is non-inferior to the standard 12-month duration.

Neoadjuvant immunotherapy strategies may prove particularly suitable for elderly patients with resectable stage III melanoma. The phase III NADINA trial showed that two cycles of neoadjuvant ipilimumab plus nivolumab with response-adapted adjuvant therapy achieved 12-month event-free survival of 83.7% versus 57.2% for adjuvant nivolumab alone (HR 0.32; p<0.001), sparing major pathological responders from further adjuvant treatment.²² The phase II SWOG S1801 trial similarly showed that three cycles of neoadjuvant pembrolizumab plus adjuvant pembrolizumab improved 2-year event-free survival versus adjuvant therapy alone (72% vs 49%).²³ By concentrating immunological priming preoperatively and enabling de-escalation for pathological responders, these strategies could reduce cumulative treatment exposure and toxicity in elderly patients. Our observation that many patients did not complete a full year of adjuvant therapy yet still derived MSS and RFS benefit provides real-world support for shorter, response-guided perioperative approaches in this population. A related and clinically pressing question is whether a full 12 months of adjuvant therapy is preferable to surveillance followed by response-adapted neoadjuvant treatment at the time of (locoregional) recurrence. Deferral has genuine attractions in this age group. Adjuvant therapy increased 5-year RFS by 16 percentage points in our cohort (44% vs 28%), which corresponds to approximately six elderly patients treated for one recurrence or death avoided at 5 years; the remainder are exposed to up to 12 months of immunotherapy, with grade ≥2 irAEs in 42% of treated patients, discontinuation for toxicity in 23%, and one treatment-related death, without an individual benefit that can be identified, whether because they would never have recurred or because they recurred despite treatment. A two-cycle to three-cycle neoadjuvant course, by contrast, is delivered only to patients with proven macroscopic disease and permits treatment to be stopped altogether in major pathological responders.²² The counterargument is equally concrete in our data: deferral is only advantageous if the patient is still treatable at relapse, and 130 of the 247 observation patients who recurred (53%) did not receive first-line systemic therapy, at a median age of 81 years. The interval between resection and recurrence is precisely when older patients lose functional reserve, and a neoadjuvant approach is in any case only applicable to the subset with resectable macroscopic recurrence. On balance, deferral appears defensible for patients whose functional trajectory is stable and who can be reliably surveilled and rapidly re-accessed, whereas patients with borderline fitness, limited access to specialist follow-up, or high nodal burden are those most likely to lose their therapeutic opportunity while under observation. A prospective comparison of surveillance with neoadjuvant salvage versus upfront adjuvant therapy, with geriatric assessment embedded at baseline, would be of considerable value in this population.

This study has several limitations. The retrospective, non-randomized design carries selection bias: adjuvant-treated patients were on average 3 years younger than observation patients (median 78 vs 81 years), reflecting clinician-driven selection of fitter individuals. Detailed data on comorbidities, frailty, and the specific reasons for not initiating adjuvant therapy were not systematically available, particularly for patients treated up to 15 years ago, when retrieval of historical records was not consistently feasible. Although Cox models were adjusted for age, sex, AJCC stage, and mutation status, unmeasured confounders such as functional status and comorbidity burden may have contributed to outcome differences.

A fundamental limitation of any age-defined cohort is that chronological age is a poor surrogate for biological age: what weighs most in the treatment decision is the patient’s age-related clinical condition, and guidelines accordingly recommend that such decisions be informed by a CGA, or at minimum by a validated screening tool such as the G8.^{24 25} Our study could not incorporate these measures. ECOG PS, the only structured measure of fitness available, is coarse and, critically, not missing at random: it was recorded for 92% of adjuvant-treated but only 33% of observation patients, and for barely one in five of those whose recurrence was not managed with first-line systemic therapy. Adjusting for it would have restricted the analysis to a subgroup in which the two arms are represented in inverted proportions, excluding the frailest patients; we judged this would introduce more bias than it removed. Indirect signals nonetheless indicated a substantial unmeasured frailty burden: non-melanoma death accounted for 14% and 11% of the observation and adjuvant groups, adjuvant-treated patients were younger and fitter at selection (93% vs 80% aged ≤85 years), and 55% did not complete 12 months of therapy, 8% stopping for deterioration unrelated to treatment. Residual confounding by biological age therefore remains the principal uncertainty of this analysis, and prospective adjuvant trials in older adults should embed CGA or G8 screening as a stratification factor and prespecified effect modifier.

Post-recurrence management and subsequent systemic therapy were captured for the observation cohort only; equivalent data were not collected for adjuvant-treated patients who relapsed, so the two groups cannot be compared with respect to the treatment received after recurrence.

This selection pattern is, however, an inherent feature of real-world practice, being elderly patients already pre-selected for adjuvant therapy by fitness, so our data reflect the population in whom treatment is actually considered. The cohort also has a temporal asymmetry: adjuvant-treated patients were predominantly diagnosed from 2017 onward (reflecting regulatory approval), whereas observation patients were more evenly distributed, with a substantial proportion diagnosed before 2017 when adjuvant therapy was not standard of care. This may introduce immortal-time and period bias, as earlier observation patients had no opportunity to receive adjuvant treatment and may have been managed under different surveillance and salvage paradigms. Sentinel-lymph-node tumor burden was not systematically captured across the participating centers and therefore could not be incorporated into the multivariable models, although a complementary specification (Model B) decomposing AJCC stage into its constituent components (T category, ulceration, and N classification including the a/b/c subcategories) yielded results fully consistent with the primary Model A and is reported in detail in the online supplemental material.

In conclusion, our study provides the largest and longest real-world analysis to date of adjuvant immunotherapy in patients aged ≥75 years with resected stage III melanoma. Adjuvant treatment was associated with clinically meaningful improvements in MSS, OS, and RFS, although with a less favorable tolerability profile than in younger trial-selected populations. Among observation patients who remained recurrence-free or received systemic therapy at recurrence, MSS was comparable to that of adjuvant-treated patients, suggesting that ICIs remain effective when administered at recurrence.

Together, these findings suggest that delaying immunotherapy until recurrence may be a reasonable strategy for some elderly patients. However, a substantial proportion of observation patients did not receive first-line systemic treatment after relapse, a group whose baseline characteristics and reasons for non-treatment could not be fully reconstructed. These patients may have been frailer or more comorbid from the outset, such that withholding therapy at recurrence was as appropriate as withholding it in the adjuvant setting; alternatively, deterioration between surgery and recurrence may have foreclosed subsequent treatment. To the extent that interval deterioration accounts for some of these patients, earlier adjuvant therapy could preserve access to effective treatment in carefully selected older patients who would otherwise become ineligible by the time of recurrence.

Data availability statement

Data are available upon reasonable request.