Previous systematic reviews (Hao et al., 2017[6]; Karlsson and Saleh, 2017[14]; Yun et al., 2016[32]) assessed immune checkpoint inhibitors using interim results from early trials such as CheckMate 066 (Robert et al., 2015[21]) and CheckMate 067 (Wolchok et al., 2017[30]). However, updated long-term follow-up data, including outcomes up to 10 years, now provide more mature insights into safety and efficacy (Wolchok et al., 2024[31]). Earlier reviews did not include LAG-3 directed regimens in their comparative analyses, as their scope was limited to PD-1 and CTLA-4 inhibitors. Although PD-1 + LAG-3 therapy has been evaluated separately in advanced melanoma (Hasanzadeh et al., 2024[7]), this work does not place LAG-3 within a broader comparative context alongside other checkpoint classes. To address this gap, the present review provides a comprehensive synthesis of current immunotherapy strategies in advanced melanoma, including PD-1, CTLA-4 and LAG-3 regimens across monotherapy and combination approaches.

This systematic review aims to evaluate and compare the efficacy (OS, PFS, ORR) and safety (grade ≥ 3 AEs) of immune checkpoint inhibitors, including PD-1 and CTLA-4 monotherapies, PD-1 + CTLA-4 and PD-1 + LAG-3 combinations in advanced unresectable melanoma.

Between PD-1 and CTLA-4 inhibitor monotherapies, which regimen demonstrates better efficacy and safety compared with traditional systemic therapies (e.g., chemotherapy, vaccines) in advanced unresectable melanoma?

Population (P): Adults with advanced unresectable or metastatic melanoma

Studies were included if they enrolled adults with advanced unresectable melanoma and evaluated immune checkpoint inhibitors (ICIs) as monotherapy or in combination, compared to traditional therapies or other ICI regimens. Eligible studies had to report at least one key clinical outcome: overall survival (OS), progression-free survival (PFS), objective response rate (ORR), or adverse events (AEs). Only English-language randomized controlled trials (RCTs) were considered. When multiple publications existed for the same trial, the most recent analysis was prioritized. Exclusion criteria included non-randomized studies, observational designs, case reports, reviews, adjuvant-only trials, or those lacking relevant survival or safety outcomes.

Final systematic literature search was conducted on 18 March 2025 across PubMed and the Cochrane Central Register of Controlled Trials (CENTRAL) databases. Medical Subject Headings (MeSH) terms and related keywords were used, including “Melanoma,” “Programmed Cell Death 1 Receptor (PD-1),” “CTLA-4 Antigen,” “LAG-3 Protein,” as well as drug names such as “nivolumab”, “pembrolizumab”, “ipilimumab”, “tremelimumab”, and “relatlimab.” Boolean operators were applied to combine terms appropriately. The full search string was:

(“Melanoma”[Mesh] OR melanoma[tiab]) AND ((“Programmed Cell Death 1 Receptor”[Mesh] OR PD-1[tiab] OR PD1[tiab] OR nivolumab[tiab] OR pembrolizumab[tiab]) OR (“CTLA-4 Antigen”[Mesh] OR CTLA-4[tiab] OR CTLA4[tiab] OR ipilimumab[tiab] OR tremelimumab[tiab]) OR (“Lymphocyte Activation Gene-3”[Mesh] OR LAG-3[tiab] OR LAG3[tiab] OR relatlimab[tiab]))

Studies published from 2006 to 18 March 2025 (PubMed) and from database inception to 18 March 2025 (CENTRAL) were considered. The PubMed search, filtered for randomized controlled trials (RCTs), identified 213 records, while the Cochrane CENTRAL search identified 1,721 records after applying an English language filter.

Following title and abstract screening, 63 studies were assessed in full text. During full-text screening, 52 studies were excluded for the following reasons: duplicate or older versions (n = 20), non-rct studies (n = 15), studies evaluating resectable melanoma (n = 11), and other reasons (n = 6). The selection process is summarized in the PRISMA flow diagram (Figure 2(Fig. 2)).

Data extraction was conducted using a standardized form. Two summary tables were created: one described study and patient characteristics (study name, region, median age by arm, sample sizes, intervention/control treatments, ECOG performance status, and BRAF mutation status); the other summarized clinical outcomes. Outcomes were extracted separately for experimental and control arms, including median OS with 95 % CIs, HRs for OS and PFS, median PFS with 95 % CIs, ORR, and incidence of grade ≥ 3 AEs. Hazard ratios (HRs) for time-to-event outcomes quantify the relative rate at which events occur over time in the treatment group compared with the control group, whereas risk ratios (RRs) compare the overall probability of an event between groups. For trials with multiple publications, the most recent and complete datasets were prioritized to ensure mature data. Where key outcomes such as survival outcomes or AE profiles were incomplete, earlier interim reports were used to supplement missing information.

Risk of bias was assessed using the Cochrane Risk of Bias 2.0 (RoB 2) tool, which evaluates five domains: Randomization process, deviations from intended interventions, missing outcome data, outcome measurement, and selection of reported results. Each domain was rated as “low risk,” “some concerns,” or “high risk” following Cochrane Handbook guidelines. An overall judgment was assigned based on domain-level ratings. Assessments were summarized narratively and tabulated using a color-coded system: green for low risk, yellow for some concerns, and red for high risk as detailed in Table 1(Tab. 1) (References in Table 1: Di Giacomo et al., 2024[2]; Hamid et al., 2017[5]; Hodi et al., 2010[10]; Hodi et al., 2016[9]; Larkin et al., 2018[15]; Ribas et al., 2013[19]; Robert et al., 2011[22]; Robert et al., 2020[20]; Tawbi et al., 2024[25]; VanderWalde et al., 2023[27]; Wolchok et al., 2024[31]).

Data synthesis was conducted using Review Manager (RevMan) version 5.4.1. Hazard ratios (HRs) with 95 % confidence intervals (CIs) were pooled for time-to-event outcomes (OS and PFS), and risk ratios (RRs) for binary outcomes (ORR and grade ≥ 3 AEs). For all binary outcomes, raw event counts and total participants in each study arm were extracted and entered directly into RevMan, which calculated RRs using the Mantel-Haenszel random-effects model. A random-effects model was used to account for clinical and methodological heterogeneity. Statistical heterogeneity was assessed using the I² statistic, with values > 50 % indicating substantial heterogeneity. Where meta-analysis was not feasible due to limited or inconsistent data, results were summarized narratively.

A total of 11 randomized controlled trials were included in this systematic review and meta-analysis, as illustrated in the PRISMA flow diagram (Figure 2(Fig. 2)).

The included studies enrolled adults with advanced unresectable or metastatic melanoma and evaluated immune checkpoint inhibitors as monotherapy or in combination. Three studies assessed CTLA-4 inhibitors (ipilimumab or tremelimumab) versus traditional therapies such as chemotherapy or vaccines. Three evaluated PD-1 inhibitors (nivolumab or pembrolizumab) against chemotherapy. Five investigated combination therapies: four assessed nivolumab plus ipilimumab (dual PD-1 and CTLA-4 blockade) versus ipilimumab alone, and one evaluated nivolumab plus relatlimab (dual PD-1 and LAG-3 inhibition) versus nivolumab. Sample sizes ranged from 53 to 714, with median ages typically between 56 and 64. Most studies included patients with ECOG performance status 0-1, and BRAF-mutated melanoma was present in 0-42 % of participants. Study characteristics are summarized in Table 2(Tab. 2) (References in Table 2: Di Giacomo et al., 2024[2]; Hamid et al., 2017[5]; Hodi et al., 2010[10]; Hodi et al., 2016[9]; Larkin et al., 2018[15]; Ribas et al., 2013[19]; Robert et al., 2011[22]; Robert et al., 2020[20]; Tawbi et al., 2024[25]; VanderWalde et al., 2023[27]; Wolchok et al., 2024[31]).

Risk of bias was assessed using the Cochrane Risk of Bias 2.0 (RoB 2) tool for RCTs, across five domains. Five studies (Robert et al., 2011[22]; Robert et al., 2020[20]; Hodi et al., 2016[9]; Wolchok et al., 2024[31]; Tawbi et al., 2024[25]) were rated low risk across all domains. Four (Ribas et al., 2013[19]; Larkin et al., 2018[15]; Hamid et al., 2017[5]; Di Giacomo et al., 2024[2]) were high risk, mainly due to deviations from intended interventions and lack of blinding or selective reporting. Two (Hodi et al., 2010[10]; VanderWalde et al., 2023[27]) had some concerns, primarily related to randomization or outcome assessment. Detailed assessments are presented in Table 1(Tab. 1).

Immune checkpoint inhibitors showed clear efficacy advantages over traditional treatments in advanced melanoma. Among monotherapy regimens, PD-1 inhibitors consistently demonstrated improved progression-free survival and objective response rates across trials. While the pooled overall survival result did not reach statistical significance, narrative synthesis of the included studies showed generally favorable outcomes with PD-1 treatment compared to control. This discrepancy may be explained by considerable heterogeneity among studies. In Hamid et al. (2017[5]) post-progression crossover to PD-1 therapy was allowed in the control arm, likely reducing the measurable survival difference between groups. Still, the consistency of direction across individual studies supports a meaningful clinical effect of PD-1 monotherapy. In contrast, CTLA-4 inhibitors showed more mixed results. While pooled analyses indicated significant improvement in PFS and OS, ORR did not reach statistical significance, and the results across trials were variable. Some studies showed only modest gains with CTLA-4 monotherapy, suggesting a less predictable therapeutic impact. These inconsistencies may reflect both clinical and biological factors, including trial population differences and the broader mechanism of immune activation associated with CTLA-4 inhibition. PD-1 and CTLA-4 inhibitors differ in their immune targets and functional effects, which may account for the differences observed in clinical performance. PD-1 blockade reactivates previously primed, antigen-experienced T cells within the tumor microenvironment, restoring localized cytotoxic immune activity against tumor cells. This targeted mechanism is more likely to produce sustained and tumor-specific responses. In contrast, CTLA-4 inhibition acts earlier in the immune response by promoting the expansion of naïve T cells in lymphoid tissues. While this broadens immune activation, it may result in less focused anti-tumor activity and contributes to the variability in treatment outcomes observed with CTLA-4 monotherapy (Buchbinder and Desai, 2016[1]).

Blocking two immune checkpoints simultaneously has emerged as a powerful strategy to improve treatment outcomes in advanced melanoma. Combination regimens consistently outperformed monotherapy across key efficacy endpoints, providing stronger and more durable clinical benefit. Dual inhibition of PD-1 and CTLA-4 produced particularly robust improvements, reflecting the synergy achieved by targeting different stages of the T-cell activation cycle. While PD-1 plus LAG-3 also demonstrated improved outcomes compared to PD-1 monotherapy, the survival advantage was statistically significant but more modest in scale. Nonetheless, both approaches underscore the therapeutic advantage of disrupting multiple immunosuppressive pathways rather than relying on a single mechanism.

The enhanced efficacy of PD-1 and CTLA-4 combination therapy reflects their complementary roles in immune regulation, targeting distinct stages of T-cell activation. This dual approach enables broader and more durable anti-tumor responses. In contrast, PD-1 and LAG-3 are often co-expressed on exhausted CD8+ T cells within the tumor microenvironment, where they suppress immune function through parallel pathways. LAG-3 primarily reduces cytotoxicity and cytokine production, while PD-1 limits T-cell proliferation. Blocking both pathways reactivates effector function by enhancing TCR signaling, increasing IFN-γ production, and restoring cytolytic activity. These mechanistic distinctions help explain why PD-1 + CTLA-4 blockade produces the most robust efficacy, while PD-1 + LAG-3 offers a more moderate, yet meaningful, clinical benefit (Rotte, 2019[23]; Qiu et al., 2024[17]).

In terms of safety, the pooled analysis for grade ≥ 3 adverse events was not statistically significant in either the PD-1 or CTLA-4 subgroup. Both subgroups showed high heterogeneity, particularly CTLA-4. This likely reflects differences in trial design, especially background therapies. For instance, ipilimumab was combined with dacarbazine in Robert et al. (2011[22]) and with the gp100 vaccine in Hodi et al. (2010[10]), both of which are independently associated with toxicity and may have inflated adverse event rates in CTLA-4 arms. Narrative synthesis indicated that CTLA-4 monotherapy was more frequently associated with severe immune-related adverse events, whereas PD-1 inhibitors tended to be better tolerated across trials. As previously discussed, CTLA-4 blockade promotes broad immune activation by enhancing naïve T-cell priming in lymphoid tissues. This widespread stimulation increases the likelihood of off-target inflammation, particularly in barrier organs like the gastrointestinal tract, where immune-related colitis is common. In contrast, PD-1 inhibition acts locally within the tumor microenvironment, reactivating exhausted T cells without broadly stimulating the immune system. This tumor-specific reactivation contributes to the lower incidence of systemic immune-related adverse events observed with PD-1 (Wang et al., 2023[29]). Given these mechanistic differences, combining PD-1 and CTLA-4 blockade amplifies immune activation both during the early T-cell priming in lymphoid tissues (CTLA-4) and at the tumor site where PD-1 regulates exhausted T cells., which likely accounts for the significantly increased toxicity observed in pooled analyses. No heterogeneity was detected, suggesting this effect was consistent across studies. In contrast, PD-1 plus LAG-3 demonstrated a more favorable safety profile. As outlined previously, LAG-3 and PD-1 blockade acts more selectively on exhausted T cells in the tumor microenvironment, which may limit systemic immune activation. These findings support a clear toxicity gradient, with PD-1 in monotherapy being the most tolerable, followed by PD-1 + LAG-3 in the combination therapy (Wang et al., 2023[29]; Maruhashi et al., 2020[16]).

The findings of this systematic review and meta-analysis demonstrate that immunotherapy is more effective than traditional treatments in advanced melanoma. Single ICIs improved OS, PFS, and ORR, with PD-1 inhibitors consistently outperforming CTLA-4 inhibitors in both efficacy and safety. CTLA-4 inhibitors were associated with modest clinical benefits and a higher rate of severe adverse events, while PD-1 inhibitors showed greater improvements across all endpoints with a comparatively better safety profile. Combination therapy with PD-1 and CTLA-4 inhibitors provided the most substantial benefit in terms of survival and response, but this came with significantly increased toxicity. The PD-1 + LAG-3 combination, which represents a more recent approach targeting a novel immune checkpoint, demonstrated favorable efficacy while being better tolerated than PD-1 + CTLA-4. This suggests that PD-1 + LAG-3 may offer a more clinically viable option, especially for patients who may not tolerate the high toxicity associated with dual PD-1 and CTLA-4 blockade. These findings are consistent with previous studies reporting similar trends for PD-1 and CTLA-4 inhibitors (Hao et al., 2017[6]; Karlsson and Saleh, 2017[14]; Yun et al., 2016[32]).

The key strength of this systematic review with meta-analysis is the inclusion of only randomized controlled trials with generally low risk of bias. It captured the most widely used immune checkpoint inhibitors in melanoma, CTLA-4 and PD-1 as well as the newer PD-1 + LAG-3 combination, offering a broad and clinically relevant evaluation of current treatment strategies. The analysis included long-term follow-up where available, including studies with survival data up to 10 years, such as Wolchok et al. (2024[31]) enabling a more complete understanding of treatment durability. Efficacy and safety were comprehensively assessed through key outcomes including OS, PFS, ORR, and grade ≥ 3 AEs. Subgroup analyses enabled direct comparison across ICI classes.

While the findings are clinically meaningful and largely consistent with previous evidence, several limitations should be noted. Some outcomes, including OS for PD-1 monotherapy, did not reach statistical significance despite favorable trends. In Hamid et al. (2017[5]) two pembrolizumab doses (2 mg/kg and 10 mg/kg) were pooled, which may have introduced variability. Post-progression crossover occurred in multiple trials, including Hamid et al. (2017[5]) and Ribas et al. (2013[19]) potentially diluting treatment effects and underestimating survival benefit. Some monotherapy trials included ICIs with chemotherapy, such as ipilimumab with dacarbazine (Robert et al., 2011[22]) and nivolumab versus dacarbazine (Robert et al., 2020[20]), possibly confounding interpretation of ICI-only effects. Only 11 trials were included, limiting statistical power for subgroup analyses. Sensitivity analyses were not performed due to the small number of studies, and funnel plots were not generated, so publication bias could not be formally assessed. Heterogeneity was substantial in several analyses, particularly among single-agent comparisons, due to differences in trial design, treatment protocols, and patient populations. In the analysis of grade ≥ 3 AEs, the study by Di Giacomo et al. (2024[2]) was excluded as all other included studies were of high quality and showed consistent findings (I² = 0 %), while Di Giacomo et al. (2024[2]) used an uncommon comparator (ipilimumab plus fotemustine) and reported a distinct toxicity profile, which would have introduced unnecessary heterogeneity. Lastly, the meta-analysis relied on published aggregate data rather than patient-level data, limiting the ability to conduct adjusted or stratified analyses.

The results of this systematic review and meta-analysis support the use of PD-1 inhibitors as the preferred monotherapy option in advanced melanoma, given their consistent efficacy and favorable safety profile compared to CTLA-4 inhibitors. Dual checkpoint blockade with PD-1 and CTLA-4 provides the most substantial survival benefit but is associated with high toxicity, which limits its use in many patients. The PD-1 + LAG-3 combination showed promising results with meaningful survival benefit and improved tolerability and may be a more suitable alternative for patients who are unable to tolerate the toxicity of PD-1 + CTLA-4 therapy.

Future research should focus on optimizing treatment sequencing, identifying reliable predictive biomarkers, and developing strategies to overcome resistance. Not all patients benefit from immune checkpoint inhibition, as these therapies only block certain immune pathways that may not be active in all tumors. A large proportion of patients continue to show no meaningful clinical response, even with combination therapy, which exposes them to unnecessary toxicity. This highlights the need to better understand why some patients do not respond and to expand treatment options beyond the currently available targets. Improving biomarker-based selection could help personalize therapy, reduce exposure to adverse effects in non-responders, and ultimately lead to better outcomes across a wider group of patients. Further clinical trials are also needed to evaluate LAG-3 based combinations across different melanoma subtypes, confirm their long-term benefit, and define their role within first-line treatment strategies.

A preprint version of this manuscript was previously published on the Qeios platform under the title “Efficacy and Safety of CTLA-4, PD-1 and LAG-3 Immune Checkpoint Inhibitors as Monotherapy and Combination Therapy in Advanced Melanoma: A Systematic Review and Meta-Analysis.” doi: https://doi.org/10.32388/4AAZYE.

The content has not been peer-reviewed elsewhere, and this submission represents the final, revised version intended for journal publication.

This manuscript is based on the author's final-year dissertation completed as part of the Bachelor of Science (BSc) in Biomedical Science at the University of Buckingham, United Kingdom. The dissertation was submitted for academic evaluation only and has not been published elsewhere.

The author declares that they have no conflict of interest and no financial relationship with any organization that sponsored the research.

This work did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

This article does not contain any studies with human participants or animals performed by the author. All analyses were based on previously published data, and no new ethical approval was required. The study design, literature screening, data extraction, analysis, interpretation of results, and all scientific decisions were carried out entirely by the author. The author takes full responsibility for the content and conclusions of this work.

Artificial intelligence tools were used solely to assist with language editing and improvement of clarity.