Document resource
WHAT IS ALREADY KNOWN ON THIS TOPIC Dual immunotherapy (eg, nivolumab plus relatlimab (NIVO+RELA)) and BRAF/MEK inhibitor combinations are both approved first-line treatments for BRAF-mutant advanced melanoma, but there are no head-to-head randomised trials comparing these two treatment strategies.WHAT THIS STUDY ADDS Using matching-adjusted indirect comparisons, this study found that NIVO+RELA was associated with improved long-term overall survival and progression-free survival from 12 months onward compared with multiple BRAF/MEK inhibitor regimens.NIVO+RELA also demonstrated a more favourable or comparable safety profile compared with the BRAF/MEK inhibitors.HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY These findings support the use of NIVO+RELA as a viable first-line treatment option offering long-term survival benefit in BRAF-mutant advanced melanoma, which can help guide treatment selection.Introduction Melanoma is one of the deadliest skin cancers, 1 affecting approximately 100 640 people with 8290 attributable deaths in the USA in 2024.2 In addition to environmental factors (ie, ultraviolet light exposure), oncogenic gene mutations influence the aetiology of melanoma, with 41%–55% of metastatic melanomas harbouring BRAF mutations.3 4 In previous treatment eras (2006 to 2016), BRAF-mutant melanoma was characterised as aggressive, particularly in advanced stages, and was associated with a 23% higher mortality risk compared with melanomas not harbouring the mutation.5Treatments for BRAF-mutant melanoma have included monotherapy and combination therapy directed at programmed cell death protein 1 receptor (PD-1) and its ligand (PD-L1), cytotoxic T-lymphocyte-associated protein 4 (CTLA-4), lymphocyte-activation gene 3 (LAG-3) or inhibition of mitogen-activated protein kinase (MEK) enzymes or the active component of the mutant BRAF V600 protein.6 Three targeted therapies that are BRAF/MEK inhibitor combinations have been approved by the US Food and Drug Administration (FDA) for patients with BRAF-mutant advanced melanoma: dabrafenib+trametinib (DAB+TRAM) in 2014,7–9 vemurafenib+cobimetinib (VEM+COBI) in 201510–12 and encorafenib+binimetinib (ENCO+BINI) in 2018.13–15 In addition, atezolizumab (ATEZO), a PD-L1 inhibitor, was approved in combination with VEM+COBI for patients with BRAF V600 mutation-positive unresectable or metastatic melanoma in 2020.16 17Nivolumab (NIVO), an anti-PD-1 monoclonal antibody that relieves T-cell dysfunction and restores cytotoxic function,18 has been a standard of care in the treatment of unresectable or metastatic melanoma (alone or in combination with ipilimumab (IPI)). In 2022, the FDA approved a fixed-dose combination of NIVO plus relatlimab (NIVO+RELA)—a human LAG-3 blocking antibody that restores the effector function of dysfunctional T-cells19—for metastatic or inoperable melanoma regardless of BRAF mutation status.20 21 The approval of NIVO+RELA was based on the results of the randomised, double-blinded RELATIVITY-047 trial, which demonstrated that patients with previously untreated metastatic or unresectable stage III or IV melanoma achieved significantly longer progression-free survival (PFS) when treated with NIVO+RELA compared with NIVO alone (10.1 vs 4.6 months).18 19 Among the BRAF-mutant population, NIVO+RELA was also associated with longer PFS (HR: 0.72, 95% CI 0.54 to 0.96) compared with NIVO alone.22The emergence of novel treatments for BRAF-mutant advanced melanoma motivates the need to assess the comparative clinical efficacy of approved options. Importantly, the DREAMseq trial in advanced BRAF-mutated melanoma found that patients randomised to receive NIVO+IPI had significantly better 2-year overall survival (OS) compared with those receiving BRAF/MEK inhibitor DAB+TRAM (71.8% vs 51.5%, p=0.01).23 However, a head-to-head trial has not yet been conducted comparing NIVO+RELA with BRAF/MEK inhibitors in this setting; therefore, it is unknown whether a similar survival benefit exists. In the absence of head-to-head randomised trials, matching-adjusted indirect comparisons (MAICs) leveraging individual patient-level data from the trial of the treatment of interest and aggregate-level data from the comparators’ pivotal trials can be used to assess relative treatment efficacy and adjust for differences in patient populations. Accordingly, this study used MAICs to compare the efficacy and safety of NIVO+RELA vs DAB+TRAM, ENCO+BINI, VEM+COBI and ATEZO+VEM+COBI for first-line treatment of BRAF-mutant advanced melanoma.Methods Data sources Individual patient-level data for NIVO+RELA were sourced from the phase II/III RELATIVITY-047 trial (ClinicalTrials.gov ID: NCT03470922; data cut-off of October 2023). The subset of patients with BRAF mutation who received NIVO+RELA (median follow-up: 40 months; n=136) was included in this analysis. As this was a post hoc analysis of previously collected data, patients and the public were not involved in the design or conduct of the research.Summary data for the comparators were extracted from their pivotal trial publications, as follows. For DAB+TRAM, data were extracted from the pooled COMBI-d (NCT01584648) and COMBI-v trial (NCT01597908) publication (median follow-up: 22 months).24 Data for ENCO+BINI were extracted from the COLUMBUS trial (NCT01909453) publications (safety and overall response rate (ORR) median follow-up, 48.8 months; OS minimum follow-up, >93 months; and PFS median follow-up, 16.7 months).25–28 Data for VEM+COBI were extracted from the coBRIM trial (NCT01689519) publications (median follow-up, 21.2 months).29 30 Finally, data for ATEZO+VEM+COBI were extracted from the IMspire150 trial (NCT02908672) publications (median follow-up, 29.1 months).31 32 The data cut-offs from each trial were identified as having the follow-up period that most closely aligned with that of the available RELATIVITY-047 data. The details of the comparator trials’ exclusion/inclusion criteria and populations have been previously published; a brief summary of the comparator trials’ study designs is provided in the online supplemental material.SP110.1136/bmjonc-2025-000912.supp1Supplementary dataSummary aggregate data and digitised versions of the Kaplan-Meier (KM) curves for the comparators were extracted from the publications listed above, and pseudo-patient-level data were generated following the method outlined by Guyot et al.33 The appropriateness of the pseudo-patient-level data was verified by (1) conducting a visual inspection of the KM reconstructed curves against the published curves and (2) comparing the reconstructed numbers at risk and median survival (and its corresponding 95% CI) to the values reported in the publication.Sample selection The target population included adults (aged ≥18 years) with previously untreated BRAF-mutant advanced melanoma. Some key differences were identified in the eligibility criteria between RELATIVITY-047 and each of the comparator trials. Therefore, additional eligibility criteria were applied to patients in RELATIVITY-047 to better emulate the populations of each of the comparator trials. Specifically, RELATIVITY-047 enrolled patients regardless of BRAF-mutation status; thus, the BRAF-mutant subset of the NIVO+RELA arm in RELATIVITY-047 was selected for all comparisons. Further, because no prior immuno-oncology (IO) or BRAF/MEK inhibitor treatments were allowed in IMspire150, the NIVO+RELA arm was further restricted to patients who had not previously received these treatments in the adjuvant setting.Outcomes Efficacy outcomes assessed in the MAIC included OS, PFS per investigator and confirmed ORR per investigator. All outcomes were defined consistently across trials. PFS (determined per investigator) was defined as the time from randomisation to the date of progressive disease (per RECIST 1.1) or death due to any cause, whichever came first. OS was defined as the time from randomisation to the date of death due to any cause. Confirmed ORR (determined per investigator) was defined as the proportion of patients with confirmed complete response/partial response according to RECIST 1.1. Safety outcomes included the rates of any adverse event (AE), grade 3/4 AEs, AEs leading to discontinuation and specific AEs (rash, diarrhoea, fatigue, arthralgia and pruritus). Safety outcomes were summarised as the proportion of patients with the event.Statistical methods Descriptive statistics were used to summarise the baseline characteristics for the NIVO+RELA BRAF-mutant subset of RELATIVITY-047 and the populations from COMBI-d/v, COLUMBUS, coBRIM and IMspire150.Separate unanchored MAICs were conducted to compare outcomes between NIVO+RELA vs each of the four comparators: DAB+TRAM, ENCO+BINI, VEM+COBI and ATEZO+VEM+COBI. For each pairwise comparison, individual patients in the BRAF-mutant subset (or the BRAF-mutant, IO, and BRAF/MEK naïve subset for the comparison versus ATEZO+VEM+COBI) of the NIVO+RELA arm from RELATIVITY-047 were assigned weights such that the weighted baseline characteristics of patients in the NIVO+RELA arm exactly matched those reported for patients in each comparator arm. Each patient’s weight was based on their estimated odds (propensity) of enrolment in the comparator trial versus the RELATIVITY-047 trial. The baseline characteristics that were matched on for each comparison are listed in online supplemental table S1. These were determined based on data availability and clinical input and importance (ie, clinically relevant for disease prognosis). Effective sample sizes (ESS) after matching were calculated for the NIVO+RELA arm for each comparison. The 12- and 24-month thresholds were chosen prior to the MAIC analyses based on clinical relevance and for consistency with prior publications.34Before and after matching, OS and PFS per investigator were summarised using KM curves, and HRs with corresponding 95% CIs were calculated using Cox proportional hazards models. Due to violation of the proportional hazards assumption based on the scaled Schoenfeld residuals, separate interval weighted Cox models with a boundary point at 12 months for DAB+TRAM, ENCO+BINI and VEM+COBI, and 24 months refers to ATEZO+VEM+COBI, were used to compare PFS per investigator and OS between treatments.For ORR per investigator, ORs with 95% CIs were calculated, and for safety outcomes, rate differences (RDs) with 95% CIs were estimated, both using Wald tests before and after matching. In IMspire150, safety outcomes were collected among the all-treated population (n=231), which differed from the overall population (n=256). The absence of reported baseline characteristics for the all-treated population precluded matching to this subpopulation; therefore, weights from matching to the overall population were used in the safety outcome comparisons versus ATEZO+VEM+COBI.Differences between groups were assessed using 95% CIs; however, strict significance levels were not applied, and p values were not calculated, as this analysis was exploratory in nature. All analyses were conducted using R V.3.6.3. Additional details on the MAIC methodology are provided in the online supplemental methods.Results Sample size The initial sample size of the NIVO+RELA arm in RELATIVITY-047 was 355. A total of 136 patients with a BRAF mutation from the NIVO+RELA arm of RELATIVITY-047 were included in the MAICs versus DAB+TRAM, ENCO+BINI and VEM+COBI. After excluding patients with prior IO or BRAF/MEK inhibitor use, 132 patients from the NIVO+RELA arm were included in the MAIC versus ATEZO+VEM+COBI. The DAB+TRAM, ENCO+BINI, VEM+COBI and ATEZO+VEM+COBI groups from the comparator trials included 563, 192, 247 and 256 patients, respectively.Baseline characteristics Prior to matching, several baseline characteristics differed between patients in the NIVO+RELA and comparator cohorts ( table 1, online supplemental table S2). Relative to the comparator cohorts, the NIVO+RELA cohort had lower proportions with prior immunotherapy (11.8% vs 20.8% (DAB+TRAM) and 29.7% (ENCO+BINI); 9.1% vs 16.0% (ATEZO+VEM+COBI)) or stage M1c or M1d disease (44.1% vs 64.5% (DAB+TRAM), 64.1% (ENCO+BINI) and 59.1% (VEM+COBI); 45.5% vs 56.3% (ATEZO+VEM+COBI)). The NIVO+RELA cohort was also younger, with 40.4% (DAB+TRAM), 46.3% (ENCO+BINI) and 42.6% (VEM+COBI) of patients above the comparator cohorts’ median age; the exception was that 59.1% of patients were above the median age of the ATEZO+VEM+COBI cohort. Further, the NIVO+RELA cohort had a lower proportion with lactate dehydrogenase above the upper limit of normal than the VEM+COBI cohort (33.8% vs 45.3%). There were missing data for Eastern Cooperative Oncology Group (ECOG) performance status for DAB+TRAM and VEM+COBI; for purposes of the analyses, the missing category and ECOG performance status of 1 were pooled, resulting in a conservative approach for the pair-wise comparisons vs NIVO+RELA.Table 1Patient baseline characteristics after matching for NIVO+RELA versus comparators*Baseline characteristicsNIVO+RELA vs DAB+TRAMNIVO+RELA vs ENCO+BININIVO+RELA vs VEM+COBINIVO+RELA vs ATEZO+VEM+COBIAfter matchingAfter matchingAfter matchingAfter matchingDAB+TRAMn=563NIVO+RELAESS=100ENCO+BINIn=192NIVO+RELAESS=78VEM+COBIn=247NIVO+RELAESS=109ATEZO+VEM+COBIn=256NIVO+RELAESS=106Age (above comparator median)†282 (50.0%)50.0%96 (50.0%)50.0%124 (50.0%)50.0%128 (50.0%)50.00%Sex Female244 (43.3%)43.3%77 (40.1%)40.1%101 (40.9%)40.9%106 (41.4%)41.40% Male319 (56.7%)56.7%115 (59.9%)59.9%146 (59.1%)59.1%150 (58.6%)58.60%ECOG 0403 (71.6%)71.6%136 (70.8%)70.8%184 (74.5%)74.5%194 (75.8%)75.80% 1‡155 (27.5%)28.4%56 (29.2%)29.2%59 (23.9%)25.5%62 (24.2%)24.20% Missing5 (0.9%)0.0%0 (0.0%)0.0%4 (1.6%)0.0%0 (0.0%)0.0%LDH > ULN194 (34.5%)34.5%55 (28.6%)28.6%112 (45.3%)45.3%84 (32.8%)32.80% ≤ ULN365 (64.8%)65.5%137 (71.4%)71.4%130 (52.7%)54.7%172 (67.2%)67.20%Metastatic status M0 or M1a§94 (16.7%)16.7%35 (18.2%)18.2%61 (24.7%)24.7%55 (21.5%)21.50% M1b105 (18.7%)18.7%34 (17.7%)17.7%40 (16.2%)16.2%57 (22.3%)22.30% M1c or M1d363 (64.5%)64.7%123 (64.1%)64.1%146 (59.1%)59.1%144 (56.3%)56.30%Number of disease sites ≥3275 (48.8%)48.8%‒‒‒‒‒‒ <3287 (51.0%)51.2%‒‒‒‒‒‒Prior immunotherapy, yes117 (20.8%)20.8%57 (29.7%)29.7%‒‒41 (16.0%)16.00%History of brain metastases Yes‒‒‒‒1 (0.4%)0.4%5 (2.0%)2.00% No‒‒‒‒246 (99.6%)99.6%251 (98.0%)98.00%Sum of longest diameter <44 mm‒‒‒‒‒‒109 (42.6%)42.60% ≥44 mm‒‒‒‒‒‒147 (57.4%)57.40% Unknown‒‒‒‒‒‒0 (0.0%)0.00%*Additional results (before matching for NIVO+RELA) can be found in the expanded online supplemental table S2.†The median ages for COMBI-d/v, COLUMBUS, coBRIM and IMspire150 were 55, 57, 56 and 54 years, respectively.‡One patient in the VEM+COBI arm of coBRIM had an ECOG score of 2 after randomisation but before the first dose was received and was grouped into the category of ECOG 1 for this analysis. All patients in the NIVO+RELA arm of RELATIVITY-047 had an ECOG score of 0 or 1.§A total of 3.2% of patients in the VEM+COBI arm of coBRIM had stage IIIc melanoma and were grouped in the metastatic status M0 or M1a category.ATEZO, atezolizumab; BINI, binimetinib; COBI, cobimetinib; DAB, dabrafenib; ECOG, Eastern Cooperative Oncology Group; ENCO, encorafenib; ESS, effective sample size; LDH, lactate dehydrogenase; NIVO, nivolumab; RELA, relatlimab; TRAM, trametinib; ULN, upper limit of normal; VEM, vemurafenib.After matching, all matched-on baseline characteristics were balanced between NIVO+RELA and each comparator cohort. For example, the NIVO+RELA cohort had 50.0% of patients above the comparator median age, an equal sex distribution (56.7% (DAM+TRAM) to 59.6% (ENCO+BINI) males) and similar proportions with M1c or M1d status (56.3% (ATEZO+VEM+COBI) to 64.7% (DAM+TRAM)) in all comparisons. The ESSs were 100, 78, 109 and 106, respectively, for the DAB+TRAM, ENCO+BINI, VEM+COBI and ATEZO+VEM+COBI comparisons.Efficacy outcomes OS Overall, NIVO+RELA was associated with greater OS benefits compared with DAB+TRAM both before (overall HR (95% CI): 0.58 (0.44 to 0.77)) and after (0.61 (0.43 to 0.85)) matching ( figure 1A). This trend for OS was consistent both during and after the first 12 months of treatment.Figure 1OS comparisons of NIVO+RELA versus (A) DAB+TRAM, (B) ENCO+BINI, (C) VEM+COBI and (D) ATEZO+VEM+COBI, before and after matching.a ATEZO, atezolizumab; BINI, binimetinib; COBI, cobimetinib; ENCO, encorafenib; ESS, effective sample size; DAB, dabrafenib; NIVO, nivolumab; OS, overall survival; RELA, relatlimab; TRAM, trametinib; U, unweighted; VEM, vemurafenib, W, weighted. Note: aOnly the CI for the comparator and weighted NIVO+RELA curves are included in the figures.In the first 12 months of treatment, the pre- and post-matching OS HRs for NIVO+RELA vs ENCO+BINI were 0.75 (95% CI 0.46 to 1.22) and 0.90 (0.49 to 1.63), respectively (figure 1B). NIVO+RELA demonstrated greater OS benefit after 12 months from treatment initiation compared with ENCO+BINI both before (HR (95% CI: 0.55 (0.37 to 0.82)) and after (0.51 (0.32 to 0.83)) matching.For the first 12 months following treatment initiation, the pre- and post-matching OS HRs for NIVO+RELA versus VEM+COBI were 0.69 (95% CI 0.43 to 1.11) and 0.85 (0.51 to 1.42), respectively (figure 1C). After 12 months of treatment, NIVO+RELA was associated with longer OS compared with VEM+COBI both before (HR (95% CI): 0.44 (0.30 to 0.66)) and after (0.41 (0.26 to 0.62)) matching.NIVO+RELA demonstrated an OS benefit compared with ATEZO+VEM+COBI both before (overall HR (95% CI): 0.70 (0.51 to 0.97)) and after (0.68 (0.48 to 0.98)) matching (figure 1D).PFS NIVO+RELA demonstrated comparable PFS to the other therapies during the early treatment period (first 12 or 24 months) but was associated with better long-term PFS in most comparisons.During the first 12 months of treatment, NIVO+RELA and DAB+TRAM had similar PFS before (HR (95% CI): 1.16 (0.88 to 1.53)) and after (1.36 (0.96 to 1.94)) matching (figure 2A). However, from 12 months onward, NIVO+RELA had better PFS compared with DAB+TRAM both before (HR (95% CI): 0.54 (0.34 to 0.86)) and after (0.56 (0.33 to 0.93) matching.Figure 2PFS per investigator comparisons of NIVO+RELA versus (A) DAB+TRAM, (B) ENCO+BINI, (C) VEM+COBI and (D) ATEZO+VEM+COBI, before and after matching.a ATEZO, atezolizumab; BINI, binimetinib; COBI, cobimetinib; ENCO, encorafenib; ESS, effective sample size; DAB, dabrafenib; NIVO, nivolumab; PFS, progression-free survival; RELA, relatlimab; TRAM, trametinib; U, unweighted; VEM, vemurafenib; W, weighted. Note: aOnly the CI for the comparator and weighted NIVO+RELA curves are included in the figures.During the first 12 months of treatment, NIVO+RELA had worse PFS compared with ENCO+BINI with pre- and post-matched HRs of 1.44 (95% CI 1.03 to 2.01) and 1.72 (1.13 to 2.61), respectively. After 12 months, NIVO+RELA was associated with better PFS compared with ENCO+BINI, with prematching HR of 0.49 (0.24 to 1.00) and post-matching HR of 0.63 (0.28 to 1.45) (figure 2B).The pre- and post-matching PFS HRs for NIVO+RELA versus VEM+COBI in the first 12 months of treatment were 1.22 (0.90 to 1.66) and 1.27 (0.89 to 1.81), respectively (figure 2C). After 12 months, NIVO+RELA was associated with better PFS compared with VEM+COBI both before (0.45 (0.28 to 0.74)) and after (0.41 (0.24 to 0.70)) matching.PFS during the first 24 months was similar for NIVO+RELA and ATEZO+VEM+COBI (prematching HR (95% CI), 1.17 (0.88 to 1.54); post-matching, 1.21 (0.87 to 1.68)) (figure 2D). However, NIVO+RELA was associated with better PFS after 24 months in the post-matching analysis (0.45 (0.19 to 1.07)).ORR All comparators were associated with higher ORR per investigator than NIVO+RELA before and after matching. The ORs and corresponding 95% CIs before and after matching were 0.43 (0.33 to 0.57) and 0.37 (0.27 to 0.50), respectively, for DAB+TRAM, 0.30 (0.21 to 0.42) and 0.25 (0.17 to 0.38) for ENCO+BINI, 0.40 (0.29 to 0.55) and 0.37 (0.27 to 0.52) for VEM+COBI and 0.46 (0.34 to 0.63) and 0.45 (0.32 to 0.63) for ATEZO+VEM+COBI.Toxicity After matching, NIVO+RELA was associated with lower absolute risk of grade 3/4 AEs than DAB+TRAM (RD (95% CI): −20.1% (−30.6 to −9.5)), ENCO+BINI (−29.2% (−42.2 to −16.3)) and VEM+COBI (−36.9% (−47.5 to −26.3)) ( table 2). Among individual AEs, NIVO+RELA was associated with a lower risk of rash (RD (95% CI): −8.8% (−16.8 to −0.8)) than DAB+TRAM, of diarrhoea than both ENCO+BINI (−11.4% (−22.8 to −0.1)) and VEM+COBI (−33.6% (−43.9 to −23.3]) and of arthralgia than VEM+COBI (−15.5% (−25.5 to −5.5]) after matching.Table 2Adverse event risk differences of NIVO+RELA versus DAB+TRAM, ENCO+BINI, and VEM+COBI, before and after matchingNIVO+RELA vs DAB+TRAMNIVO+RELA vs ENCO+BININIVO+RELA vs VEM+COBIRD, % (95% CI)RD, % (95% CI)RD, % (95% CI)Before matchingAfter matchingBefore matchingAfter matchingBefore matchingAfter matchingAny AE0.5 (−1.8 to 2.8)−0.1 (−3.3 to 3.0)0.1 (−2.6 to 2.8)−0.8 (−4.5 to 2.9)−0.7 (−3.0 to 1.6)0.1 (−1.6 to 1.8)Any grade 3/4 AE−21.0 (−30.1 to −11.9)−20.1 (−30.6 to −9.5)−30.0 (−40.5 to −19.5)−29.2 (−42.2 to –16.3)−37.6 (−47.3 to −27.8)−36.9 (−47.5 to −26.3)AE leading to discontinuation11.7 (3.4 to 20.0)10.0 (0.7 to 19.2)19.0 (10.2 to 27.8)16.4 (5.7 to 27.0)2.4 (−7.0 to 11.8)1.7 (−8.4 to 11.7)Specific AEs (any grade) Rash−5.5 (−13.4 to 2.5)−8.8 (−16.8 to −0.8)6.6 (−2.1 to 15.4)2.1 (−6.8 to 11.0)−19.1 (−28.5 to −9.8)−18.6 (−28.8 to −8.5) Diarrhoea−7.7 (−16.2 to 0.9)−6.5 (−16.1 to 3.0)−10.6 (−20.8 to −0.4)−11.4 (−22.8 to –0.1)−33.3 (−43.0 to −23.7)−33.6 (−43.9 to –23.3) Fatigue−3.1 (−11.9 to 5.7)−5.0 (−14.7 to 4.8)2.7 (−7.5 to 12.8)0.0 (−11.9 to 11.8)−5.1 (−15.1 to 4.8)−6.9 (−17.4 to 3.6) Arthralgia−6.0 (−14.0 to 2.0)−7.7 (−16.0, 0.6)−5.9 (−15.4 to 3.7)−7.2 (−17.4 to 3.0)−15.9 (−25.2 to −6.6)−15.5 (−25.5 to −5.5) Pruritus––8.1 (−0.2 to 16.3)9.9 (−0.8 to 20.5)0.0 (−8.4 to 8.5)−1.4 (−10.1 to 7.3)AE, adverse event; BINI, binimetinib; COBI, cobimetinib; DAB, dabrafenib; ENCO, encorafenib; NIVO, nivolumab; RD, risk difference; RELA, relatlimab; TRAM, trametinib; VEM, vemurafenib.NIVO+RELA also had numerically lower risk of diarrhoea (−6.5% (−16.1 to 3.0)), fatigue (−5.0% (−14.7 to 4.8)) and arthralgia (−7.7% (−16.0 to 0.6)) than DAB+TRAM (table 2). NIVO+RELA had numerically lower risk of arthralgia (−7.2% (−17.4 to3.0)), and numerically higher risk of rash (2.1% (−6.8 to 11.0)) and pruritus (9.9% (−0.8 to 20.5)), compared with ENCO+BINI. NIVO+RELA had numerically lower risk of rash (−18.6% (−28.8 to −8.5)), fatigue (−6.9% (−17.4 to 3.6)) and pruritus (−1.4% (−10.1 to 7.3)), as well as numerically higher risk of AEs leading to discontinuation (1.7% (−8.4 to 11.7)), compared with VEM+COBI. These findings should be interpreted with caution, as the CIs include the null hypothesis, indicating that the true effect may be limited or nonexistent.The rates of any-grade AEs were comparable for NIVO+RELA and ATEZO+VEM+COBI after matching (98.8% (95% CI 94.5 to 99.8) versus 100.0% (100.0 to 100.0)) (table 3). A lower proportion of the NIVO+RELA cohort compared with the ATEZO+VEM+COBI cohort reported grade 3/4 AEs after matching (40.0% (31.1 to 49.7) versus 74.9% (68.8 to 80.3)). The specific AEs with large differences in incidence between NIVO+RELA versus ATEZO+VEM+COBI after matching included diarrhoea (28.9% (20.9 to 38.3) versus 50.2% (43.6 to 56.8)), arthralgia (22.3% (15.4 to 31.1) vs 44.6% (38.1 to 51.2)) and rash (21.1% (14.5 to 29.5) versus 45.5% (38.9 to 52.1)).Table 3Adverse events among NIVO+RELA and ATEZO+VEM+COBI, before and after matchingBefore matchingAfter matchingAEs, % (95% CI)AEs, % (95% CI)NIVO+RELA(n=132)ATEZO+VEM+COBI(n=231)NIVO+RELA(ESS=106)ATEZO+VEM+COBI(n=231)Any AE98.5 (95.4 to 99.7)100.0 (100.0 to 100.0)98.8 (94.5 to 99.8)100.0 (100.0 to 100.0)Any grade 3/4 AE39.4 (31.3 to 47.9)74.9 (68.8 to 80.3)40.0 (31.1 to 49.7)74.9 (68.8 to 80.3)AE leading to discontinuation30.3 (22.9 to 38.5)38.1 (31.8 to 44.7)29.3 (21.4 to 38.5)38.1 (31.8 to 44.7)Specific AEs (any grade) Rash22.0 (15.5 to 29.5)45.5 (38.9 to 52.1)21.1 (14.5 to 29.5)45.5 (38.9 to 52.1) Diarrhoea26.5 (19.5 to 34.4)50.2 (43.6 to 56.8)28.9 (20.9 to 38.3)50.2 (43.6 to 56.8) Fatigue31.8 (24.3 to 40.1)30.3 (24.4 to 36.7)31.3 (23.2 to 40.8)30.3 (24.4 to 36.7) Arthralgia22.7 (16.1 to 30.4)44.6 (38.1 to 51.2)22.3 (15.4 to 31.1)44.6 (38.1 to 51.2) Pruritus19.7 (13.5 to 27.0)28.1 (22.4 to 34.4)20.5 (13.7 to 29.4)28.1 (22.4 to 34.4)AE, adverse event; ATEZO, atezolizumab; COBI, cobimetinib; ESS, effective sample size; NIVO, nivolumab; RELA, relatlimab; VEM, vemurafenib.Discussion Although the results from the DREAMseq trial have indicated a survival benefit for first-line NIVO+IPI over BRAF/MEK inhibitors and solidified IO as the preferred option over targeted therapy for most patients with BRAF-mutant advanced melanoma, there are currently no head-to-head trials similarly comparing these outcomes to those of patients receiving NIVO+RELA in this setting. The present MAICs leveraged individual patient-level data from the RELATIVITY-047 trial of NIVO+RELA and published aggregate data from the trials of four BRAF/MEK inhibitor combinations to compare their efficacy as first-line treatment for BRAF-mutant advanced melanoma. After matching on baseline characteristics, NIVO+RELA was associated with greater OS benefits compared with DAB+TRAM, ENCO+BINI, VEM+COBI and ATEZO+VEM+COBI, particularly over longer observation times (ie, more than 12 or 24 months). Additionally, NIVO+RELA demonstrated longer-term PFS benefits (after 12 months) over DAB+TRAM and VEM+COBI, or after 24 months versus ATEZO+VEM+COBI. NIVO+RELA had lower ORR per investigator than the comparators. Finally, the safety profile of NIVO+RELA was generally similar to or more favourable than that of the BRAF/MEK inhibitors.To our knowledge, this is the first indirect comparison of NIVO+RELA with BRAF/MEK inhibitors in the BRAF-mutated patient population. Boutros et al35 analysed PFS, ORR and grade ≥3 AEs via a network meta-analysis of data from 18 trials of first-line treatments for advanced melanoma (irrespective of BRAF mutation status), including NIVO+RELA and BRAF/MEK inhibitors.35 They found that, compared with NIVO+RELA, NIVO+IPI had similar PFS and ORR, while BRAF/MEK inhibitors had similar PFS and significantly higher ORR; NIVO+RELA also showed a trend of fewer grade ≥3 AEs than NIVO+IPI. However, as that meta-analysis included patients with or without BRAF mutation, their findings are not directly comparable to the present results only among the BRAF-mutated patient population.The current results are similar to those of a prior MAIC by Tarhini et al demonstrating durable OS and PFS benefits for patients with BRAF-mutant advanced melanoma treated with NIVO+IPI versus BRAF/MEK inhibitors, with the greatest benefits realised after 12 months.34 This observation may be explained by considering the results of an indirect treatment comparison using patient-level data that recently reported similar efficacy for NIVO+RELA as compared with NIVO+IPI as first-line treatment in patients with untreated advanced melanoma.36 Further, a network meta-analysis reported that, across all-comers and a PD-L1 <1% subgroup, NIVO+RELA showed efficacy (OS and PFS) broadly comparable to NIVO+IPI, with lower odds of grade 3/4 AEs versus NIVO+IPI and generally better PFS than IO monotherapies.37 The phase III DREAMseq trial, in which patients with treatment-naïve BRAF-mutant metastatic melanoma received front-line NIVO+IPI or DAB+TRAM at enrolment and the alternate therapy at disease progression, confirmed that first-line combination therapy with NIVO+IPI followed by BRAF/MEK inhibitors (if necessary) should be the preferred treatment sequence for the majority of patients.23 Additionally, the current safety results are in line with prior literature reporting moderate treatment-related AEs associated with BRAF/MEK inhibitors that may require dose reduction or treatment discontinuation.38 Immunotherapies may present with differing AEs (ie, immune-related),39 and the risks and benefits of the different treatment classes should be considered when selecting among therapeutic options.The observed differences in the results of comparisons across BRAF/MEK inhibitors may be influenced by variations in treatment and residual confounding factors. The efficacy of NIVO+RELA appears most favourable against VEM+COBI, while comparisons with ENCO+BINI were less conclusive, potentially reflecting ENCO+BINI’s greater activity relative to other BRAF/MEK inhibitors and more favourable clinical outcomes as reported in the COLUMBUS trial. As a second-generation BRAF inhibitor, ENCO was reported to have higher potency against wild-type BRAF compared with DAB;40 in preclinical models, ENCO demonstrated more potent inhibition of BRAF V600 mutant cell lines than DAB or VEM.41 Furthermore, ENCO+BINI generally has a lower risk for fever than DAB+TRAM and a lower rate of photosensitivity than VEM+COBI,42 43 which may contribute to longer duration of on-treatment follow-up. Finally, there was limited follow-up for PFS per investigator for ENCO+BINI (median follow-up, 16.7 months; maximum follow-up, approximately 26 months), which limited the relative HR calculation between ENCO+BINI and NIVO+RELA.Although ATEZO+VEM+COBI is not a recommended first-line treatment for BRAF-mutant advanced melanoma per the National Comprehensive Cancer Network Clinical Practice Guidelines in Oncology due to excessive toxicity concerns and minimal added benefit,44 a study showed that approximately 7% of oncologists who prescribe immunotherapy and/or BRAF/MEK inhibitors to patients with advanced melanoma in community or academic settings prescribed ATEZO+VEM+COBI.45 This MAIC indicated both improved survival and potentially reduced toxicity with NIVO+RELA versus ATEZO+VEM+COBI, suggesting that the PD-L1 plus BRAF/MEK triplet therapy may not be necessary.The long-term OS and PFS advantage of first-line NIVO+RELA, paired with a better safety profile, has implications for clinical practice and management of patients with BRAF-mutant advanced melanoma. The findings are consistent with a growing body of literature suggesting that NIVO+RELA may rival or challenge the dominance of NIVO+IPI as the standard first-line treatment choice. For example, a prior analysis indirectly comparing NIVO+RELA and NIVO+IPI as first-line treatment for advanced melanoma in the BRAF-mutant subgroup found that the treatments had similar efficacy, but that NIVO+IPI had a better ORR; NIVO+RELA had a better safety profile overall, although this was not evaluated in the BRAF-mutant subgroup.36 Subgroup analyses were not conducted in the current study due to lack of data availability within subgroups. Our findings, along with prior evidence that BRAF/MEK inhibitors may be beneficial for patients with symptomatic disease,46–48 underscore the value of a personalised approach to treatment selection. For example, considering that BRAF/MEK inhibitors were associated with similar short-term PFS and more favourable ORR in this study, patients with disease characteristics that favour early response or disease control may be candidates for targeted therapy.MAICs can adjust for differences between trial populations that are potential treatment effect modifiers and prognostic factors that might otherwise bias the comparison. While the MAICs did not substantially alter the clinical interpretation in this study, they provide reassurance in the results by ensuring that comparisons are robust and not influenced by baseline population differences, enhancing the reliability of indirect comparisons across trials.This study is subject to several limitations, many of which are common to all MAICs. First, only known baseline factors that were consistently reported across trials for a given comparison were included among the matching covariates in the MAICs. Specific to the CoBRIM and IMspire150 trials, the comparisons of NIVO+RELA versus VEM+COBI and NIVO+RELA versus ATEZO+VEM+COBI could be affected if any baseline characteristics that were not balanced post-match (ie, race and region) represent treatment modifiers. Second, as with any comparison of non-randomised treatment groups, such comparisons are subject to potential bias due to unobserved or unmeasurable confounding factors. Not all baseline characteristics were consistently available across comparator trials. For example, the number of disease sites was reported only for DAB+TRAM, while history of brain metastases was reported only for VEM+COBI and ATEZO+VEM+COBI. Other potential markers of disease severity, such as PD-L1 expression or presence of liver metastasis, were also not available across comparators. However, the direction of any potential bias cannot be determined without information on how these characteristics were distributed in each trial relative to RELATIVITY-047. Third, patterns of subsequent systemic therapy after progression differed across trials. Approximately half of the patients in each trial received a subsequent therapy, but immunotherapy was more common after BRAF/MEK inhibitor combinations (~32–35%) than after NIVO+RELA (14%), while targeted therapy was more frequent after NIVO+RELA (35% versus ~8%–19% after BRAF/MEK inhibitor doublets and ~27% after the BRAF/MEK triplet). These differences may have had some impact on OS that cannot be adjusted via MAIC. Fourth, the RELATIVITY-047 trial used the American Joint Committee on Cancer (AJCC) 8th edition, while the comparator trials used the AJCC 7th edition; pooling the M1c and M1d categories together for disease stage mitigated the impact of this discrepancy, but some remaining differences might persist between populations. Fifth, PFS was assessed per investigator, as this was a consistent metric across trials, but could potentially indicate a degree of subjectivity. The limited follow-up for investigator-assessed PFS in the COLUMBUS trial also restricted the relative HR calculation between ENCO+BINI and NIVO+RELA. Sixth, the results of the study may not be generalisable beyond the study sample. Finally, due to data limitations, matching could not be conducted to the all-treated population in IMspire150 (among which safety outcomes were collected); thus, the results from the safety comparisons versus ATEZO+VEM+COBI should be interpreted with caution.In the absence of head-to-head randomised clinical trials, this MAIC provides insights into the comparative efficacy and safety of first-line treatment options for BRAF-mutant advanced melanoma. The results of these MAICs suggest that, although NIVO+RELA is associated with lower ORRs, it is associated with long-term (after 12 months) OS advantage over DAB+TRAM, ENCO+BINI and VEM+COBI (or at any time vs ATEZO+VEM+COBI) as first-line treatment for BRAF-mutant advanced melanoma. In addition, these comparisons demonstrated better longer-term PFS (after 12 months) for NIVO+RELA over DAB+TRAM and VEM+COBI (or after 24 months vs ATEZO+VEM+COBI). These findings can help inform decision-making when selecting the first treatment for BRAF-mutant advanced melanoma.