Dose, Schedule, and Duration De-Escalation of Immune Checkpoint Inhibitors in Advanced Melanoma: Pharmacological Rationale and Clinical Evidence

1. Introduction

Long-term follow-up has established immune checkpoint inhibitors (ICIs) as a cornerstone of advanced melanoma therapy. In CheckMate 067, after a minimum of ten years of follow-up, median overall survival was 71.9 months with nivolumab plus ipilimumab, 36.9 months with nivolumab, and 19.9 months with ipilimumab [1]. Earlier pivotal studies established the survival benefit of ipilimumab in metastatic melanoma [2,3]. As durable disease control has become achievable for a substantial proportion of patients, treatment optimisation increasingly concerns not only efficacy, but also the dose, schedule, and duration required to maintain benefit while limiting toxicity, treatment burden, and cost. The pharmacological development of anti-PD-1 therapy differed from the traditional maximum-tolerated-dose paradigm. For nivolumab, no maximum tolerated dose was identified across the evaluated range, and pharmacodynamic and exposure-response analyses suggested substantial target engagement with relatively flat efficacy relationships across clinically active exposures [4]. A recent scoping review identified dose attenuation, schedule modification, treatment-duration reduction, and response-adapted strategies as major domains of melanoma treatment de-intensification [5]. Building on that broader framework, the present review focuses specifically on intentional ICI dose, schedule, and duration de-escalation in advanced melanoma and incorporates emerging 2025-2026 randomised, prospective, real-world, and guideline evidence. Pembrolizumab also demonstrated limited dose-response separation within clinically active ranges; in a randomised KEYNOTE-001 cohort, 2 mg/kg and 10 mg/kg every three weeks produced identical objective response rates of 26% [6]. These findings provide a rationale for studying lower treatment intensity, but do not by themselves establish that less drug or shorter treatment is clinically equivalent. De-escalation is particularly relevant because combination ICI therapy can cause substantial toxicity. In CheckMate 067, grade 3-4 treatment-related adverse events occurred in 59% of patients receiving nivolumab plus ipilimumab [7]. Older patients may discontinue treatment because of toxicity more frequently, and frailty may increase the clinical consequences of immune-related adverse events even when severe toxicity rates are not clearly higher [8,9]. This review therefore evaluates four intentional de-escalation strategies in unresectable stage III or stage IV melanoma: reduced-dose CTLA-4 inhibition within combination therapy, extended-interval anti-PD-1 dosing, anti-PD-1 dosing below approved exposure targets, and treatment-duration reduction or elective discontinuation.

Importantly, de-escalation comprises distinct interventions that should not be treated as interchangeable. Reducing the dose administered per cycle changes drug exposure, whereas extended-interval dosing may preserve overall exposure despite fewer visits. Shortening treatment duration reduces cumulative exposure over time, while response-adapted treatment changes intensity or duration in selected responders. Each approach has different pharmacological assumptions and requires a separate assessment of comparative efficacy, toxicity, and external validity.

2. Materials and Methods

This narrative review was prepared with reference to the Scale for the Assessment of Narrative Review Articles (SANRA) [10]. A structured literature search was conducted in PubMed/MEDLINE, Embase, and the Cochrane Central Register of Controlled Trials (CENTRAL) to identify clinical evidence relevant to dose, schedule, and treatment-duration de-escalation of immune checkpoint inhibitors in advanced melanoma. The search covered publications from January 2015 through 12 September 2026, with additional targeted searches and reference-based identification of earlier seminal studies.

The search strategy combined terms related to melanoma and immune checkpoint inhibition with concepts of de-escalation, dose reduction, alternative dosing schedules, very-low-dose therapy, treatment duration, and discontinuation. To improve the identification of studies not explicitly described as de-escalation or deintensification, regimen-specific terminology was also considered, including nivolumab, ipilimumab, pembrolizumab, N3I1, N1I3, nivolumab 3 mg/kg plus ipilimumab 1 mg/kg, nivolumab 1 mg/kg plus ipilimumab 3 mg/kg, 400 mg, 480 mg, every four weeks, every six weeks, and response-adapted treatment.

Additional targeted searches addressed pivotal clinical trials, including CheckMate 511, EORTC 1612–EBIN, KEYNOTE-029, KEYNOTE-555, ADAPT-IT, DANTE, and Safe Stop, as well as observational melanoma cohorts and pharmacokinetic/pharmacodynamic investigations. Clinical trial registries, relevant conference proceedings, and contemporary melanoma guidelines were considered when appropriate [11,12,13,14,15]. Reference lists of relevant clinical trials, systematic reviews, meta-analyses, and guidelines were screened to identify additional potentially relevant publications.

Evidence selection and synthesis. Evidence was grouped into four approaches: (1) reduced-dose ipilimumab in combination therapy; (2) extended-interval anti-PD-1 administration; (3) very-low-dose anti-PD-1 therapy; and (4) elective or response-adapted discontinuation. Randomised and prospective trials in advanced melanoma received greatest weight for causal interpretation; observational cohorts, pharmacokinetic/pharmacodynamic studies, and systematic reviews were evaluated in the context of their study designs. Adjuvant melanoma and mixed-malignancy evidence was treated as indirect rather than definitive for advanced melanoma. Overlapping primary studies across meta-analyses were not regarded as independent replication. Findings were synthesised qualitatively because populations, interventions, end points, and selection criteria were heterogeneous. No new quantitative pooling, duplicate independent screening, or formal risk-of-bias assessment was undertaken.

3. Dose and Schedule De-Escalation of Immune Checkpoint Inhibitors

Approved and experimental de-escalation strategies should be distinguished by what they are designed to change. Pembrolizumab target-engagement modelling from KEYNOTE-001 suggested that peripheral-blood PD-1 engagement approached saturation at doses of at least 1 mg/kg every three weeks, with 2 mg/kg predicted to achieve approximately 95% target engagement at steady state [16]. Population pharmacokinetic and exposure-response analyses further found limited clinically relevant variability and a relatively flat relationship between pembrolizumab exposure and tumour-size response within the 2-10 mg/kg range [17,18]. These data support investigation of lower intensity, but they do not define a minimum effective dose.

Accordingly, the distinction between an exposure-matched schedule and a genuinely lower drug dose is central to clinical interpretation. Pharmacokinetic plausibility cannot establish therapeutic non-inferiority, particularly for CTLA-4 inhibition, for which exposure, antitumour activity, and immune-mediated toxicity may be more closely associated [19].

The same pharmacological framework underlies approved extended-interval schedules. Nivolumab 480 mg every four weeks provides a similar time-averaged steady-state exposure to 3 mg/kg every two weeks, despite a lower trough and higher peak concentration [20]. Pembrolizumab 400 mg every six weeks was selected through exposure matching and later supported by melanoma-specific data from KEYNOTE-555 [21]; the US Food and Drug Administration authorised this schedule in 2020 based on pharmacokinetic modelling, exposure-response bridging, and emerging clinical data [22]. These schedules therefore maintain exposure within clinically established ranges rather than intentionally reducing total anti-PD-1 exposure. In contrast, ipilimumab has a clearer dose-response and dose-toxicity relationship, making dose reduction a potential efficacy-toxicity trade-off [19].

3.1. Reduced-Dose Ipilimumab in Combination Therapy

CheckMate 511 provides the strongest randomised evidence for reducing the ipilimumab component of combination therapy. The trial randomised 360 previously untreated patients with unresectable stage III or IV melanoma to nivolumab 3 mg/kg plus ipilimumab 1 mg/kg (NIVO3+IPI1) or nivolumab 1 mg/kg plus ipilimumab 3 mg/kg (NIVO1+IPI3), every three weeks for four doses followed by nivolumab maintenance [23]. Grade 3-5 treatment-related adverse events were significantly less frequent with NIVO3+IPI1 (33.9% vs 48.3%; P=0.006). Objective response rates were 45.6% and 50.6%, respectively, and median progression-free survival was 9.9 versus 8.9 months. However, efficacy endpoints were secondary and descriptive; the study was not designed to establish non-inferiority. Mechanistically, CTLA-4 blockade interferes with CTLA-4 interactions with CD80/CD86 and can inhibit ligand transendocytosis, but such pathway-level activity does not define the exposure required for optimal antitumour efficacy [24]. Three-year follow-up remained consistent with the primary analysis, with overall survival of 59% versus 61%, but did not resolve the absence of a formal efficacy comparison [25].

Cross-tumour meta-analyses strengthen the toxicity signal, although their applicability to advanced melanoma is indirect. Somekawa et al. pooled 48 populations from 40 articles, encompassing 4,677 patients with various malignancies. Grade ≥3 adverse events were reported more frequently with NIVO1+IPI3 than NIVO3+IPI1 (64.3% vs 35.7%), although between-study differences limit a direct regimen-level causal comparison [26]. Xu et al. likewise found fewer grade 3-4 treatment-related adverse events with NIVO3+IPI1 than NIVO1+IPI3 (31.3% vs 55.9%; OR 0.52, 95% CI 0.32-0.87) and did not identify a statistically significant difference in objective response rate [27]. Chen et al. found a lower incidence of grade ≥3 treatment-related adverse events with NIVO3+IPI1, while some response measures favoured NIVO1+IPI3 [28]. These analyses include overlapping primary studies and heterogeneous tumour types. The absence of a statistically significant difference in response is not evidence of efficacy non-inferiority.

The international, randomised phase II EORTC 1612-EBIN trial adds prospective information on NIVO3+IPI1 in an explicitly melanoma-specific treatment sequence [29]. In 271 previously untreated patients with unresectable stage III or IV BRAF V600E/K-mutant melanoma, 12 weeks of encorafenib-binimetinib induction followed by NIVO3+IPI1 was compared with immediate NIVO3+IPI1. Median progression-free survival was 9 months in both groups (HR 0.87; 90% CI 0.67-1.12; P=0.36); grade 3-5 treatment-related adverse events occurred in 42% versus 32% of treated patients. The experimental variable was preceding targeted therapy rather than ipilimumab dose. EBIN therefore informs feasibility, outcomes, and sequencing with the reduced-ipilimumab regimen but does not establish comparability with NIVO1+IPI3.

Other studies support the broader concept of lowering CTLA-4 exposure while maintaining anti-PD-1 therapy. KEYNOTE-029 Part 1B reported an objective response rate of 62.1% with pembrolizumab 2 mg/kg plus ipilimumab 1 mg/kg, with grade 3-4 treatment-related adverse events in 47.1% [30]. In cohort 1C, pembrolizumab 200 mg every three weeks was combined with ipilimumab 50 mg every six weeks or 100 mg every 12 weeks; objective response rates were 55% and 61%, while grade 3-5 treatment-related adverse events occurred in 24% and 39%, respectively [31]. ADAPT-IT evaluated response-adapted shortening of induction: after two NIVO1+IPI3 doses, 68% of patients with a favourable early radiographic effect omitted further ipilimumab, although grade 3-5 toxicity remained 57% [32]. Final follow-up showed durable activity, but the small single-arm design and historical comparisons limit routine application [33]. Overall, reduced-dose ipilimumab has convincing evidence for lower toxicity, whereas preservation of efficacy has not been formally demonstrated.

Real-world evidence supports the toxicity signal but is less definitive for comparative efficacy. In a Swedish registry study of 399 patients, grade 3-5 immune-related adverse events occurred in 30.6% with NIVO3+IPI1 and 51.1% with NIVO1+IPI3; adjusted analyses also favoured the lower-ipilimumab regimen for progression-free and overall survival [34]. Because treatment allocation was non-randomised and the regimens were used during different periods with baseline imbalances, residual confounding limits causal interpretation. A smaller German cohort likewise reported grade 3-4 treatment-related adverse events in 33% of 79 patients treated with NIVO3+IPI1, but lacked a concurrent comparator [35].

A Polish five-centre retrospective series provides context regarding the challenges of combination immunotherapy outside registration trials [36]. Among 50 patients with metastatic melanoma treated with nivolumab plus ipilimumab, adverse events were frequent and treatment cessation occurred for progression, toxicity, and death. However, the series did not directly compare NIVO3+IPI1 against NIVO1+IPI3 and should not be used to attribute its clinical outcomes to dose reduction. Together with comparative registries, it reinforces the need to distinguish real-world feasibility from causal dose-effect evidence.

A melanoma-specific systematic review by Aleid et al. also compared nivolumab-ipilimumab combination treatment with monotherapy [37]. This addresses a different comparator than the NIVO3+IPI1 versus NIVO1+IPI3 question and was therefore considered contextual rather than direct evidence of dose reduction. Its safety conclusions should be interpreted cautiously against the underlying randomised studies, given differences in regimens and endpoint definitions.

3.2. Extended-Interval Anti-PD-1 Dosing

Extended-interval anti-PD-1 dosing is the most clinically established form of de-escalation because it reduces infusion frequency while maintaining modelled exposure within validated therapeutic ranges. In KEYNOTE-555 cohort B, 101 previously untreated patients with advanced melanoma received pembrolizumab 400 mg every six weeks; the objective response rate was 50.5%, median progression-free survival 13.8 months, and grade 3-4 treatment-related adverse events occurred in 12.9% [21]. As a single-arm study, these findings support consistency with established pembrolizumab experience rather than randomised efficacy equivalence.

Real-world data are broadly reassuring but methodologically limited. In a UK multicentre retrospective study of 277 patients with advanced or resected melanoma, grade ≥3 toxicity was similar with pembrolizumab every six weeks and every three weeks (14.9% vs 15.5%); wide confidence intervals and baseline imbalances precluded a clear comparative efficacy conclusion [38]. A 2026 retrospective cohort comparing standard- and extended-interval nivolumab in 236 patients found grade ≥3 immune-related adverse events in 8.9% and 7.8%, respectively (P=0.763), but the treatment settings and baseline characteristics differed and follow-up was short [39]. Extended-interval schedules may reduce visits, travel, chair time, and resource use, but less frequent infusions should not automatically imply less frequent clinical or laboratory surveillance; monitoring should remain symptom- and treatment-phase based, consistent with immune-toxicity guidance [40].

In a separate Japanese single-centre retrospective analysis, Ito et al. evaluated 71 melanoma patients receiving anti-PD-1 monotherapy with conventional intervals (n=50), extended intervals (n=14), or a switch from conventional to extended intervals (n=7) [41]. Immune-related adverse events occurred in 32.0%, 35.7%, and 71.4% of patients, respectively, with no statistically significant difference in event probability by Kaplan-Meier analysis. All events in the switch group occurred before the switch, underscoring the potential for timing and selection effects. The limited sample size and mixed treatment settings preclude reliable efficacy equivalence claims.

3.3. Very-Low-Dose Anti-PD-1 Therapy

Very-low-dose anti-PD-1 therapy is conceptually different from approved extended-interval dosing because cumulative drug exposure is intentionally reduced, sometimes by more than an order of magnitude. Relatively flat exposure-response relationships provide a pharmacological rationale for studying this approach, including nivolumab exposure-response analyses in advanced melanoma [42], but do not identify a minimum effective exposure for individual patients.

Melanoma-specific evidence remains sparse. In a non-randomised phase II adjuvant cohort after complete resection of melanoma metastases, 21 patients received nivolumab 10 mg for approximately one year, with a 12-month relapse-free survival rate of 85.7% [43]. The regimen delivered a markedly lower cumulative dose than contemporary schedules, but the cohort was small, lacked a standard-dose comparator, and involved patients without measurable unresectable disease. Cross-tumour evidence is similarly preliminary. A 2024 pharmacokinetic study comparing nivolumab 40 mg with 3 mg/kg in 25 patients with heterogeneous advanced malignancies showed much lower absolute exposure with the low-dose regimen, but its size and heterogeneity precluded reliable efficacy inference [44]. A 2025 systematic review identified only four comparative studies involving 435 patients with advanced solid tumours; all were retrospective and considered low-quality evidence with high risk of bias [45]. Very-low-dose therapy may have important access and health-economic implications, but it should remain investigational until prospective dose-ranging and comparative studies establish preserved clinical efficacy.

Evidence supporting very-low-dose anti-PD-1 therapy remains preliminary and substantially weaker than that for dose-schedule modification. Small studies of reduced milligram doses, including those conducted in the adjuvant setting or other tumour types, cannot define the minimum effective exposure for patients with measurable advanced melanoma.

4. Treatment-Duration De-Escalation and Elective Discontinuation

The rationale for shortening ICI therapy is the durability of immune-mediated tumour control after treatment has stopped. In KEYNOTE-001, 105 patients with metastatic melanoma achieved a complete response, and 67 discontinued pembrolizumab for observation; estimated 24-month disease-free survival from complete response was 89.9% [46]. These data showed that durable remission can persist after cessation in selected responders but did not define the minimum treatment duration required.

Observational evidence has since expanded. A 2025 systematic review and meta-analysis reported progression after ICI discontinuation for reasons other than progression in 20.9% of 2,542 patients across 34 studies, with disease control in 57.3% of retreated patients [47]. Another 2025 meta-analysis reported pooled progression-free survival of 86% at one year and 71% at three years after cessation; outcomes were most favourable after complete response and after elective rather than toxicity-related discontinuation [48]. These estimates are clinically informative but are derived largely from non-randomised or post hoc cohorts subject to heterogeneity, selection, and guarantee-time bias.

Amiot et al. addressed time-dependent biases by emulating sequential target trials within the prospective MelBase cohort [49]. Discontinuation at six months was associated with poorer overall survival compared with treatment continuation for at least three additional months, whereas analyses at 12 and 18 months showed no clear survival advantage for either strategy. At 24 months, estimates favoured discontinuation, although residual confounding and model dependence limit causal interpretation. These findings suggest that the optimal treatment duration may vary according to the timing of discontinuation but do not establish a universally safe stopping threshold.

DANTE was the first phase III randomised test of a fixed stopping strategy. Patients who remained progression-free after approximately one year of first-line anti-PD-1-based therapy were randomised to stop treatment, with retreatment permitted at progression, or continue for at least two years [50]. The trial planned to enrol 1,208 patients but closed early after only 166 were randomised. At a median 29.1 months of follow-up, one-year progression-free survival after randomisation was 80.2% in the discontinuation arm and 87.6% in the continuation arm (HR 1.76; 90% CI 1.03-3.03). Because accrual was far below target, the trial was underpowered for a definitive non-inferiority conclusion; it therefore does not establish that universal stopping at one year preserves efficacy.

Safe Stop evaluated a different, response-guided approach. This prospective Dutch multicentre study enrolled 200 patients with unresectable stage III or metastatic melanoma who had a confirmed complete or partial response to first-line anti-PD-1 monotherapy and discontinued treatment after a median of 24 weeks [51]. At 24 months from treatment initiation, the ongoing response rate was 81.8%, overall survival 95.9%, and melanoma-specific survival 97.4%. However, the primary comparison used a historical benchmark rather than a concurrent continuation arm, and patients were selected based on demonstrated treatment sensitivity. Safe Stop therefore provides prospective support for response-guided early discontinuation but not proof of equivalence to continued treatment.

DANTE and Safe Stop investigated different discontinuation questions. DANTE examined stopping after a predefined treatment period among patients without progression, whereas Safe Stop selected patients who had already attained a confirmed complete or partial response. The populations, treatment-decision rules, and comparator structures differed; consequently, their reported outcomes should not be juxtaposed as though they were a randomised comparison of fixed-duration versus response-guided discontinuation. Selection of responders, guarantee-time effects, and variable retreatment access remain important obstacles to causal inference.

Taken together, current evidence does not support a single stopping threshold for all patients. Depth and durability of response, residual disease burden, prior disease kinetics, toxicity, patient preference, surveillance feasibility, and access to retreatment should inform individual decisions. Prospective trials are still needed to determine whether response-, imaging-, or biomarker-guided strategies can identify patients in whom further drug exposure adds little clinical benefit.

5. Clinical Implications, Limitations, and Future Directions

De-escalation should be interpreted according to the strength and directness of evidence. Approved extended-interval pembrolizumab and nivolumab are established exposure-matched options; reduced-dose ipilimumab lowers severe toxicity but lacks randomised evidence of efficacy non-inferiority; very-low-dose anti-PD-1 therapy remains experimental; and early discontinuation is supported in selected responders without a universal stopping standard (Table 2).

Clinical context remains important. Older age alone does not consistently predict lower ICI efficacy or substantially higher rates of severe immune-related toxicity, but frailty, multimorbidity, polypharmacy, and reduced functional reserve may magnify the consequences of treatment complications [52]. Geriatric assessment may support individualisation but is not validated to select a specific de-escalation strategy [53]. Sex, body composition, and pre-existing autoimmune disease may modify toxicity patterns, yet current evidence does not support empiric dose or duration reduction on these factors alone [54,55,56]. Resource constraints justify investigation of low-dose regimens but do not establish therapeutic equivalence.

Guidelines broadly mirror this evidence hierarchy (Table 1): ESMO and SEOM-GEM allow selective consideration of NIVO3+IPI1 when toxicity is a major concern, whereas ASCO and SITC retain standard combination dosing as the reference approach, particularly for melanoma brain metastases [11,12,13,14]. NCCN provides the broader treatment framework; de-escalation recommendations should not be inferred beyond published guidance [15].

Recurring limitations include retrospective or single-arm designs, underpowered comparisons, indirect evidence, and under-representation of frail patients and patients with central nervous system disease. Future research should prioritise adequately powered comparison of NIVO3+IPI1 versus NIVO1+IPI3, prospective dose-ranging and pharmacokinetic/pharmacodynamic studies of very-low-dose anti-PD-1 therapy, and response- or biomarker-guided stopping strategies with predefined surveillance and rechallenge pathways.

The evidence hierarchy is particularly important when interpreting dose comparisons (Table 3). CheckMate 511 provides direct randomised evidence for a toxicity advantage, whereas broad oncology meta-analyses support that signal only indirectly. EBIN provides prospective data using NIVO3+IPI1, but its targeted-therapy sequencing design does not test dose non-inferiority. Observational series and extended-interval cohorts add information on feasibility and safety but remain vulnerable to clinical selection and time-related confounding.

Future prospective studies should use prespecified clinically acceptable non-inferiority margins for efficacy, adequately powered safety comparisons, and consistent adverse-event definitions. They should include response depth, central nervous system involvement, frailty, patient-reported outcomes, surveillance burden, and retreatment availability. Pharmacokinetic and circulating tumour DNA approaches may inform patient selection, but neither should be presented as an established basis for empiric treatment reduction without prospective validation.

Table 1. Selected guideline positions relevant to ICI de-escalation in advanced melanoma.

Guideline	Position relevant to de-escalation	Implication for de-escalation
ESMO 2025 [11]	NIVO3+IPI1 may be discussed in selected frail patients because of lower toxicity; general use is not recommended because CheckMate 511 was designed to assess safety rather than efficacy non-inferiority.	Selective consideration in frail patients; efficacy equivalence has not been established.
SEOM-GEM 2024 [12]	Low-dose ipilimumab (1 mg/kg) combined with standard-dose nivolumab may be considered when toxicity is a particular concern.	Toxicity-driven treatment option in selected patients, with a lower strength of recommendation than standard first-line regimens.
ASCO 2023 update [13]	CheckMate 511 is recognised as demonstrating reduced toxicity with NIVO3+IPI1; the recommended combination dosing schedule remains NIVO1+IPI3.	Reduced-dose ipilimumab is not established as an efficacy-equivalent replacement for standard combination dosing.
SITC v3.0, 2023 [14]	Alternate ipilimumab-nivolumab dosing is discussed; for melanoma brain metastases, ipilimumab 3 mg/kg plus nivolumab 1 mg/kg is specifically recommended when combination therapy is used.	Evidence supporting standard-dose ipilimumab is strongest in patients with melanoma brain metastases.
NCCN v2.2026 [15]	The publicly available JNCCN publication provides selected guideline content and does not reproduce the complete systemic-treatment algorithm.	No specific de-escalation recommendation should be inferred beyond the content explicitly available in the published version.

Abbreviations: ASCO, American Society of Clinical Oncology; ESMO, European Society for Medical Oncology; ICI, immune checkpoint inhibitor; IPI, ipilimumab; JNCCN, Journal of the National Comprehensive Cancer Network; NCCN, National Comprehensive Cancer Network; NIVO, nivolumab; NIVO1+IPI3, nivolumab 1 mg/kg plus ipilimumab 3 mg/kg; NIVO3+IPI1, nivolumab 3 mg/kg plus ipilimumab 1 mg/kg; SEOM-GEM, Spanish Society of Medical Oncology-Spanish Multidisciplinary Melanoma Group; SITC, Society for Immunotherapy of Cancer.

Table 1. Selected guideline positions relevant to ICI de-escalation in advanced melanoma.

Guideline	Position relevant to de-escalation	Implication for de-escalation
ESMO 2025 [11]	NIVO3+IPI1 may be discussed in selected frail patients because of lower toxicity; general use is not recommended because CheckMate 511 was designed to assess safety rather than efficacy non-inferiority.	Selective consideration in frail patients; efficacy equivalence has not been established.
SEOM-GEM 2024 [12]	Low-dose ipilimumab (1 mg/kg) combined with standard-dose nivolumab may be considered when toxicity is a particular concern.	Toxicity-driven treatment option in selected patients, with a lower strength of recommendation than standard first-line regimens.
ASCO 2023 update [13]	CheckMate 511 is recognised as demonstrating reduced toxicity with NIVO3+IPI1; the recommended combination dosing schedule remains NIVO1+IPI3.	Reduced-dose ipilimumab is not established as an efficacy-equivalent replacement for standard combination dosing.
SITC v3.0, 2023 [14]	Alternate ipilimumab-nivolumab dosing is discussed; for melanoma brain metastases, ipilimumab 3 mg/kg plus nivolumab 1 mg/kg is specifically recommended when combination therapy is used.	Evidence supporting standard-dose ipilimumab is strongest in patients with melanoma brain metastases.
NCCN v2.2026 [15]	The publicly available JNCCN publication provides selected guideline content and does not reproduce the complete systemic-treatment algorithm.	No specific de-escalation recommendation should be inferred beyond the content explicitly available in the published version.

Abbreviations: ASCO, American Society of Clinical Oncology; ESMO, European Society for Medical Oncology; ICI, immune checkpoint inhibitor; IPI, ipilimumab; JNCCN, Journal of the National Comprehensive Cancer Network; NCCN, National Comprehensive Cancer Network; NIVO, nivolumab; NIVO1+IPI3, nivolumab 1 mg/kg plus ipilimumab 3 mg/kg; NIVO3+IPI1, nivolumab 3 mg/kg plus ipilimumab 1 mg/kg; SEOM-GEM, Spanish Society of Medical Oncology-Spanish Multidisciplinary Melanoma Group; SITC, Society for Immunotherapy of Cancer.

Table 2. Key evidence for ICI de-escalation strategies in melanoma.

Strategy	Study	Design	Primary signal	Evidence interpretation
Flipped NIVO3+IPI1	CheckMate 511 [23]	Randomised phase IIIb/IV; n=360	Grade 3-5 TRAEs 33.9% vs 48.3%; efficacy endpoints descriptive	Randomised evidence for toxicity reduction; efficacy non-inferiority not established
Flipped NIVO3+IPI1	Björkström et al. 2026 [34]	Swedish registry; n=399	Grade 3-5 irAEs 30.6% vs 51.1%; adjusted PFS and OS favoured NIVO3+IPI1	Comparative real-world evidence; residual confounding limits causal efficacy inference
Extended pembrolizumab	KEYNOTE-555 [21]	Single-arm phase I; n=101	ORR 50.5%; median PFS 13.8 mo; grade 3-4 TRAEs 12.9%	Prospective melanoma-specific evidence supporting an exposure-matched approved schedule; no randomised comparator
Extended nivolumab	Malmberg et al. 2026 [39]	Retrospective cohort; n=236	Grade ≥3 irAEs 8.9% standard vs 7.8% extended; P=0.763	Supportive short-term safety evidence; retrospective design and cohort imbalances limit equivalence inference
Very-low-dose nivolumab	Schwarze et al. [43]	Small phase II adjuvant cohort; n=21 monotherapy	Nivolumab 10 mg; 12-mo RFS 85.7%	Preliminary clinical activity; indirect for unresectable or metastatic melanoma
Stop at 1 year	DANTE [50]	Phase III RCT; n=166; prematurely closed	1-yr PFS 80.2% stop vs 87.6% continue; HR 1.76	Randomised evidence, but underpowered for a definitive non-inferiority conclusion
Response-guided stop	Safe Stop [51]	Prospective single-arm study; n=200	Median treatment 24 wk; 24-mo ongoing response 81.8%	Prospective support for response-guided cessation; historical rather than concurrent comparator
Flipped NIVO3+IPI1	Nübel et al. 2023 [35]	Real-world cohort; n=79	Grade 3-4 TRAEs 33%; 2-yr OS 67% in first-line subgroup	Supportive real-world evidence; small cohort without a concurrent comparator
Extended pembrolizumab	Mehta et al. 2023 [38]	UK multicentre retrospective study; n=277	Grade ≥3 toxicity 14.9% Q6W vs 15.5% Q3W; no clear comparative efficacy conclusion because of wide confidence intervals and cohort imbalance	Supportive real-world feasibility and safety data; non-randomised mixed advanced/adjuvant population
NIVO3+IPI1 in BRAF-mutant melanoma	EORTC 1612-EBIN [29]	Randomised phase II; n=271	Median PFS 9 mo in both sequencing arms	Both arms used NIVO3+IPI1; not a dose comparison
Cross-tumour NIVO3+IPI1 vs NIVO1+IPI3	Somekawa 2022; Chen 2020; Xu 2019 [26,27,28]	Systematic reviews and meta-analyses	Consistent lower high-grade toxicity signal with NIVO3+IPI1	Tumour heterogeneity; overlapping primary studies
Extended anti-PD-1 intervals	Ito et al. 2023 [41]	Retrospective single centre; n=71	irAEs 32.0% conventional vs 35.7% extended	Small groups; no comparative efficacy equivalence
Combination ICI in clinical practice	Pacholczak-Madej et al. [36]	Multicentre real-world cohort; n=50	Substantial toxicity and treatment discontinuation	No alternative-dose comparator

Abbreviations: HR, hazard ratio; IPI, ipilimumab; irAE, immune-related adverse event; NIVO, nivolumab; NIVO3+IPI1, nivolumab 3 mg/kg plus ipilimumab 1 mg/kg; ORR, objective response rate; OS, overall survival; PFS, progression-free survival; Q3W, every 3 weeks; Q6W, every 6 weeks; RCT, randomised controlled trial; RFS, relapse-free survival; TRAE, treatment-related adverse event; yr, year; mo, month; wk, week.

Table 2. Key evidence for ICI de-escalation strategies in melanoma.

Strategy	Study	Design	Primary signal	Evidence interpretation
Flipped NIVO3+IPI1	CheckMate 511 [23]	Randomised phase IIIb/IV; n=360	Grade 3-5 TRAEs 33.9% vs 48.3%; efficacy endpoints descriptive	Randomised evidence for toxicity reduction; efficacy non-inferiority not established
Flipped NIVO3+IPI1	Björkström et al. 2026 [34]	Swedish registry; n=399	Grade 3-5 irAEs 30.6% vs 51.1%; adjusted PFS and OS favoured NIVO3+IPI1	Comparative real-world evidence; residual confounding limits causal efficacy inference
Extended pembrolizumab	KEYNOTE-555 [21]	Single-arm phase I; n=101	ORR 50.5%; median PFS 13.8 mo; grade 3-4 TRAEs 12.9%	Prospective melanoma-specific evidence supporting an exposure-matched approved schedule; no randomised comparator
Extended nivolumab	Malmberg et al. 2026 [39]	Retrospective cohort; n=236	Grade ≥3 irAEs 8.9% standard vs 7.8% extended; P=0.763	Supportive short-term safety evidence; retrospective design and cohort imbalances limit equivalence inference
Very-low-dose nivolumab	Schwarze et al. [43]	Small phase II adjuvant cohort; n=21 monotherapy	Nivolumab 10 mg; 12-mo RFS 85.7%	Preliminary clinical activity; indirect for unresectable or metastatic melanoma
Stop at 1 year	DANTE [50]	Phase III RCT; n=166; prematurely closed	1-yr PFS 80.2% stop vs 87.6% continue; HR 1.76	Randomised evidence, but underpowered for a definitive non-inferiority conclusion
Response-guided stop	Safe Stop [51]	Prospective single-arm study; n=200	Median treatment 24 wk; 24-mo ongoing response 81.8%	Prospective support for response-guided cessation; historical rather than concurrent comparator
Flipped NIVO3+IPI1	Nübel et al. 2023 [35]	Real-world cohort; n=79	Grade 3-4 TRAEs 33%; 2-yr OS 67% in first-line subgroup	Supportive real-world evidence; small cohort without a concurrent comparator
Extended pembrolizumab	Mehta et al. 2023 [38]	UK multicentre retrospective study; n=277	Grade ≥3 toxicity 14.9% Q6W vs 15.5% Q3W; no clear comparative efficacy conclusion because of wide confidence intervals and cohort imbalance	Supportive real-world feasibility and safety data; non-randomised mixed advanced/adjuvant population
NIVO3+IPI1 in BRAF-mutant melanoma	EORTC 1612-EBIN [29]	Randomised phase II; n=271	Median PFS 9 mo in both sequencing arms	Both arms used NIVO3+IPI1; not a dose comparison
Cross-tumour NIVO3+IPI1 vs NIVO1+IPI3	Somekawa 2022; Chen 2020; Xu 2019 [26,27,28]	Systematic reviews and meta-analyses	Consistent lower high-grade toxicity signal with NIVO3+IPI1	Tumour heterogeneity; overlapping primary studies
Extended anti-PD-1 intervals	Ito et al. 2023 [41]	Retrospective single centre; n=71	irAEs 32.0% conventional vs 35.7% extended	Small groups; no comparative efficacy equivalence
Combination ICI in clinical practice	Pacholczak-Madej et al. [36]	Multicentre real-world cohort; n=50	Substantial toxicity and treatment discontinuation	No alternative-dose comparator

Abbreviations: HR, hazard ratio; IPI, ipilimumab; irAE, immune-related adverse event; NIVO, nivolumab; NIVO3+IPI1, nivolumab 3 mg/kg plus ipilimumab 1 mg/kg; ORR, objective response rate; OS, overall survival; PFS, progression-free survival; Q3W, every 3 weeks; Q6W, every 6 weeks; RCT, randomised controlled trial; RFS, relapse-free survival; TRAE, treatment-related adverse event; yr, year; mo, month; wk, week.

Table 3. Comparison of nivolumab–ipilimumab induction regimens in advanced melanoma.

Characteristic	NIVO3+IPI1	NIVO1+IPI3
Nivolumab induction	3 mg/kg Q3W × 4	1 mg/kg Q3W × 4
Ipilimumab induction	1 mg/kg Q3W × 4	3 mg/kg Q3W × 4
Grade 3–5 TRAEs (CheckMate 511)	33.9%	48.3%
ORR (CheckMate 511)	45.6%	50.6%
Median PFS (CheckMate 511)	9.9 months	8.9 months
Efficacy non-inferiority	Not established	Reference comparator
Evidence interpretation	Direct randomised toxicity benefit; no formal efficacy non-inferiority	Established comparator with mature efficacy evidence

Abbreviations: NIVO, nivolumab; IPI, ipilimumab; Q3W, every 3 weeks; TRAE, treatment-related adverse event; ORR, objective response rate; PFS, progression-free survival. Efficacy outcomes in CheckMate 511 were descriptive; the trial was not designed to demonstrate efficacy non-inferiority [23].

Table 3. Comparison of nivolumab–ipilimumab induction regimens in advanced melanoma.

Characteristic	NIVO3+IPI1	NIVO1+IPI3
Nivolumab induction	3 mg/kg Q3W × 4	1 mg/kg Q3W × 4
Ipilimumab induction	1 mg/kg Q3W × 4	3 mg/kg Q3W × 4
Grade 3–5 TRAEs (CheckMate 511)	33.9%	48.3%
ORR (CheckMate 511)	45.6%	50.6%
Median PFS (CheckMate 511)	9.9 months	8.9 months
Efficacy non-inferiority	Not established	Reference comparator
Evidence interpretation	Direct randomised toxicity benefit; no formal efficacy non-inferiority	Established comparator with mature efficacy evidence

Abbreviations: NIVO, nivolumab; IPI, ipilimumab; Q3W, every 3 weeks; TRAE, treatment-related adverse event; ORR, objective response rate; PFS, progression-free survival. Efficacy outcomes in CheckMate 511 were descriptive; the trial was not designed to demonstrate efficacy non-inferiority [23].

6. Conclusions

ICI de-escalation in advanced melanoma comprises strategies with markedly different levels of validation. Extended-interval anti-PD-1 schedules are clinically established exposure-matched alternatives; reducing the ipilimumab component of combination therapy lowers severe toxicity without proven efficacy equivalence; and very-low-dose anti-PD-1 therapy remains investigational. Treatment-duration reduction is promising, particularly when response-guided, but no universal stopping threshold has been established. Future optimisation should therefore focus on identifying which patients can safely receive less treatment, using prospective evidence that integrates efficacy, toxicity, pharmacology, biomarkers, patient-reported outcomes, and health-economic considerations.