Efficacy and safety of anti-PD-1-based therapies in advanced melanoma across acral and mucosal subtypes: a nationwide retrospective cohort study Authors: Yuan Qiao, Xiao Fu, Iltaf Hussain, Weihong Ge, Wei Yang, Renbin Jiang, Xiaoyan Dai, Haisheng Chen, Jinyi Zhao, Meng Tang, Fei Mu, Hao Wang, Qiuhui Wu, Xing Xia, Yunnan Zhang, Suzhi Ji, Ying Xu, Xinrui Cao, Jiexin Wang, Le Gao, Jingwen Wang, Yu Yao, Yu Fang Published: 2026 ABSTRACT BackgroundAcral melanoma (AM) and mucosal melanoma (MM) respond poorly to anti-PD-1 monotherapy. Therefore, there is an urgent need to explore safe and effective anti-PD-1-based combined therapies.MethodsA nationwide real-world cohort study consecutively included patients with advanced melanoma treated with anti-PD-1 in China between July 1, 2018, and June 30, 2023. The Research Electronic Data Capture (REDCap) system was used across all centers. The primary endpoints were progression-free survival (PFS), overall survival (OS), and immune-related adverse events (irAEs), with a secondary endpoints of objective response rate (ORR). Multivariable Cox and logistic regression models were conducted to compare the efficacy and safety of different anti-PD-1-based combination therapies with anti-PD-1 monotherapy. Stratified analyses were performed to evaluate the effect of therapies in cutaneous, acral, mucosal, and unknown primary subtypes.ResultsA total of 431 advanced melanoma patients comprising 22% cutaneous, 39% acral, 32% mucosal, and 7% unknown primary subtypes were included. Multivariable Cox analysis indicated that compared with anti-PD-1 monotherapy, anti-PD-1 + anti-VEGF + chemotherapy (HR, 0.4 [0.2, 0.7], P = 0.005) and anti-PD-1 + anti-VEGF therapy (HR, 0.4 [0.2, 0.9], P = 0.018) showed significant PFS benefit in AM and MM, respectively. Notably, anti-PD-1 + IFN-α1b showed a nonsignificantly longer PFS and OS in AM. A multivariate logistic regression model found that anti-PD-1-based combination therapies were significantly associated with grade 1–2 irAEs, while only anti-PD-1 + IFN-α1b + anti-VEGF therapy was significantly associated with high risk of grade 3–5 irAEs (OR, 4.4 [1.3–14.7], p = 0.017), without a significant survival benefit.ConclusionsOur results provide evidence for anti-PD-1 + anti-VEGF + chemotherapy and anti-PD-1 + IFN-α1b in AM and anti-PD-1 + anti-VEGF therapy in MM, as the long-term benefit and acceptable toxicity. However, for the patients who received anti-PD-1 + IFN-α1b + anti-VEGF, cautionary notes and awareness of drug safety should be raised, as the triple therapy with multiple mechanisms may increase the potential risk of toxicity.Electronic supplementary materialThe online version of this article (https://doi.org/10.1007/s00262-026-04460-6) contains supplementary material, which is available to authorized users. Introduction Melanoma is a significant global health concern due to its poor prognosis and high mortality rate [1]. The traditional systemic therapy of metastatic melanoma was ineffective [2], with a 5-year survival rate of < 10% [3] and a median survival time of 6–9 months [4] in the past. Although targeted therapy with BRAF and MEK inhibitors has obvious therapeutic effects[5], its application is limited to patients with BRAF gene mutations (The proportion of Chinese melanoma patients with BRAF gene mutations is about 25% [6]). Fortunately, programmed cell death protein 1 inhibitor (anti-PD-1) has become a significant breakthrough [7] and brought the promise of prolonged life for melanoma, which is an immunologically “hot” tumor [8] and was the earliest indication approved for immunotherapy. The 5-year survival rate was raised to 34–39% [3, 9], with prolonged overall survival (OS) 32.0–36.9 months [3, 10]. However, although immunotherapy had demonstrated encouraging therapeutic effects, only 10.9–33.9% of patients responded [11]. So many clinical trials have tried combinations of anti-PD-1 with other immune checkpoint inhibitors (ICIs) to improve response rate and combat drug resistance [12, 13]. The most reported combination was anti-PD-1 and cytotoxic T lymphocyte antigen 4 (CTLA-4). However, the increased toxicity of grade 3–5 immune-related adverse events (irAEs) was a significant concern, which may be life-threatening [14]. Previous clinical trial reports mostly focused on cutaneous melanoma (CM). However, acral melanoma (AM) and mucosal melanoma (MM) showed worse prognoses than CM [15]. Even after anti-PD-1 treatment, the response rates of AM and MM (0–13.3%) subtype were low [16, 17], because they have a different genetic profiles [18, 19] (such as BRAF, NRAS, NF1, SF3B1, CDKN2A, and TP53), a lower tumor mutational burden (TMB), lower tumor infiltrating lymphocytes (TILs), and lower programmed cell death-ligand 1(PD-L1) expression than CM. So it is a challenge to explore other effective and safety anti-PD-1-based combination therapies with different mechanisms of action. In addition to BRAF and MEK inhibitors, there are three most widely used promising anti-PD-1-based combination therapies in AM/MM. One of the combination therapy was anti-PD-1 combined with avascular endothelial growth factor inhibitor (anti-PD-1 + anti-VEGF) in MM, the objective response rate (ORR) was raised from 0–13.3% [16, 17] with anti-PD-1 monotherapy to 24.5–48.3% with combination therapy. Another combined therapy was anti-PD-1 combination with anti-VEGF and chemotherapy (anti-PD-1 + anti-VEGF + chemotherapy) in AM, and recent CAP 03 phase Ⅱ clinical trial (single-arm, n = 50) showed amazing ORR of 64.0% [22] in AM. Based on the results, anti-PD-1 + anti-VEGF + chemotherapy was recommended in the updated 2024 Chinese Society of Clinical Oncology (CSCO) guidelines for melanoma [23]. The third one was anti-PD-1 combined with interferon-α1b (anti-PD-1 + IFN-α1b). Although interferon Alfa-2b (IFN-α2b) was approved and recommended for adjuvant therapy, its application on advanced melanoma was limited by the significant toxicity [24]. IFN-α1b, which was the first genetic engineered drug of China, showed lower toxicity than IFN-α2b [25]. However, clinicians are still unable to answer patients’ safety-related questions due to the lack of large-scale clinical experience and high-level evidence from the literature. Furthermore, the clinical studies [16, 17, 20, 26] on AM/MM were mostly single-arm reports with small sample sizes (n = 50–200) and shorter follow-up time. There are major concerns that these anti-PD-1-based combination therapies may potentially increase toxicity risks with uncertain long-term benefits. Therefore, we conducted a national, multicenter cohort study to obtain large-scale safety and efficacy data across melanoma subtypes in China, with extended follow-up. This study aims to compare the risk–benefit profiles of different anti-PD-1-based combination therapies versus anti-PD-1 monotherapy, which will provide evidence-based guidance for rational immunotherapy selection across biological subtypes. Methods Patients and study design This nationwide multicenter retrospective cohort study was conducted across 7 institutions in the eastern, western, and central regions of China, all of which are grade A tertiary hospitals with the largest sample size of melanoma cases. Patients diagnosed with unresectable or metastatic stage III/Ⅳ advanced melanoma between July 1, 2018, and June 30, 2023, and who received at least one dose of anti-PD-1 therapy (either anti-PD-1 monotherapy or anti-PD-1 combined with chemotherapy, anti-angiogenic agents, or cytokine therapy) were consecutively enrolled. Patients who received anti-PD-1 as adjuvant therapy were excluded. Patients who participated in clinical trials, such as anti-PD-1 combined with BRAF ± MEK inhibitor, CTLA-4, or cell transfusion therapy, were excluded. Patients with uveal melanoma or non-measurable lesions were also excluded. Data were collected through the Research Electronic Data Capture (REDCap) system, following centralized training for all centers to ensure standardized data entry procedures. Quality control was implemented throughout the data collection process to ensure the accuracy and authenticity of the data. Patients were enrolled and numbered continuously, and unjustified deletions were not allowed. This cohort study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines. The study was approved by the Biomedical Ethics Committee of the Medical Department, Xi'an Jiaotong University. It was registered with the Chinese Clinical Trial Registry (ChiCTR2300077379). The ethical approval numbers of each participating center are listed in Supplementary Table S1. Efficacy assessment The primary endpoints were progression-free survival (PFS), overall survival (OS), and immune-related adverse events (irAEs), and the secondary endpoint was objective response rate (ORR). The irAEs were classified and graded according to the Common Terminology Criteria for Adverse Events (CTCAE) version 5.0. Tumor response and progression were assessed according to the Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1. ORR was defined as the proportion of patients who achieved a best overall response of complete response (CR) or partial response (PR). PFS was defined as the time from the initiation of anti-PD-1 administration to disease progression or any-cause death, whichever occurred first. Patients who were alive without progression at the last follow-up were censored. OS was defined as the time from the initiation of anti-PD-1 administration to any-cause death, with patients otherwise censored at the last follow-up. The cutoff date for the last follow-up was September 30, 2024. Statistical analyses Categorical variables were presented as frequencies and percentages. Normally distributed continuous variables were presented as mean ± standard deviation, while non-normally distributed variables were expressed as median (interquartile range). Group comparisons were performed using Pearson’s chi-square test or Fisher’s exact test for categorical variables and t tests or Mann–Whitney U tests for continuous variables. The Kaplan–Meier method was used to estimate median PFS and OS with two-sided 95% confidence interval. And the log-rank test was used to compare differences among groups. Univariate and multivariable Cox regression analyses were performed to evaluate the associations between baseline characteristics and long-term survival outcome (PFS and OS). Stratified analyses were conducted to evaluate the treatment effect across cutaneous, acral, mucosal, and unknown primary subtypes. Univariate and multivariable logistic regression analyses were conducted to identify factors associated with irAEs. Multivariable models were adjusted for age, sex, biological subtypes, combination therapies, and variables with P < 0.1 in univariate analysis. Interaction and stratified analyses were conducted according to sex, age, Eastern Cooperative Oncology Group Performance Status (ECOG PS), disease stage, number of metastatic sites, liver/kidney dysfunction, lactate dehydrogenase (LDH), and BRAF mutation status. All statistical analyses were performed using R statistical software (version 4.2.0) and Empower Stats software (version 4.2.0). A p value < 0.05 was considered statistically significant. Results Baseline patient characteristics Initially, 554 patients and their associated data were extracted and collated using REDCap. After applying the inclusion and exclusion criteria, a final cohort of 431 patients was ultimately included for analysis (Supplementary Figure S1). Baseline characteristics are listed in Table 1. The study cohort included 94 (22%) cutaneous, 169 (39%) acral, 138 (32%) mucosal, and 30 (7%) melanoma of unknown primary (UPM). There were no significant differences in most baseline characteristics. As expected, the frequency of BRAF mutations differed significantly across biological subtypes. BRAF mutations were most frequent in CM (27%) and UPM (23%), but rare in AM (4%) and MM(6%). The baseline characteristics, including primary sites for each subtype, are presented in Supplementary Table S2-S5.Table 1Patient baseline characteristicsNumber of patients (%)PBaseline characteristicsTotal n = 431CutaneousAcralMucosalUnknownn = 94n = 169n = 138n = 30Sex0.051Male211 (49)44 (47)94 (56)56 (41)17 (57)Female220 (51)50 (53)75 (44)82 (59)13 (43)Age0.327 ≤ 60215 (50)52 (55)82 (49)63 (46)18 (60) > 60216 (50)42 (45)87 (51)75 (54)12 (40)ECOG PS0.3140–1399 (93)87 (93)161 (95)124 (90)27 (90) ≥ 232 (7)7 (7)8 (5)14 (10)3 (10)Stage0.001Unresectable Ⅲ39 (9)6 (6)12 (7)18 (13)3 (10)Stage Ⅳ (M1a)74 (17)17 (18)28 (17)24 (17)5 (17)Stage Ⅳ (M1b)110 (26)28 (30)57 (34)20 (14)5 (17)Stage Ⅳ (M1c)158 (37)28 (30)53 (31)67 (49)10 (33)Stage Ⅳ (M1d)50 (12)15 (16)19 (11)9 (7)7 (23)Number of metastatic sites0.131 < 3287 (67)62 (66)108 (64)101 (73)16 (53) ≥ 3144 (33)32 (34)61 (36)37 (27)14 (47)Liver/kidney dysfunction0.934No391 (91)85 (90)152 (90)126 (91)28 (93)Yes40 (9)9 (10)17 (10)12 (9)2 (7)Autoimmune disease0.835No417 (97)90 (96)163 (96)135 (98)29 (97)Yes14 (3)4 (4)6 (4)3 (2)1 (3)LDH0.293 ≤ ULN261 (65)60 (71)104 (65)81 (62)16 (53) > ULN142 (35)24 (29)55 (35)49 (38)14 (47)BRAF mutation < 0.001Wild-type173 (40)35 (37)75 (44)57 (41)6 (20)Mutation47 (11)25 (27)7 (4)8 (6)7 (23)Not investigated211 (49)34 (36)87 (52)73 (53)17 (57)Line of therapy0.1201 line295 (68)63 (67)122 (72)87 (63)23 (77)2 line105 (24)26 (28)40 (24)38 (28)3 (10) ≥ 3 line31 (7)5 (5)7 (5)13 (9)4 (13)Anti-PD-10.005Toripalimab359 (83)82 (87)148 (88)102 (74)27 (90)Pembrolizumab72 (17)12 (13)21 (12)36 (26)3 (10)Radiotherapy0.366No367 (85)84 (89)145 (86)112 (81)26 (87)Yes64 (15)10 (11)24 (14)26 (19)4 (13)Combination therapy < 0.001Anti-PD-1 monotherapy84 (19)26 (28)36 (21)13 (9)9 (30)Anti-PD-1 + chemotherapy60 (14)11 (12)26 (15)17 (12)6 (20)Anti-PD-1 + anti-VEGF102 (24)13 (14)17 (10)64 (46)8 (27)Anti-PD-1 + anti-VEGF + chemotherapy65 (15)11 (12)36 (21)18 (13)0 (0)Anti-PD-1 + IFN-α1b84 (19)23 (24)39 (23)17 (12)5 (17)Anti-PD-1 + IFN-α1b + anti-VEGF36 (8)10 (11)15 (9)9 (7)2 (7) Table 1Patient baseline characteristicsNumber of patients (%)PBaseline characteristicsTotal n = 431CutaneousAcralMucosalUnknownn = 94n = 169n = 138n = 30Sex0.051Male211 (49)44 (47)94 (56)56 (41)17 (57)Female220 (51)50 (53)75 (44)82 (59)13 (43)Age0.327 ≤ 60215 (50)52 (55)82 (49)63 (46)18 (60) > 60216 (50)42 (45)87 (51)75 (54)12 (40)ECOG PS0.3140–1399 (93)87 (93)161 (95)124 (90)27 (90) ≥ 232 (7)7 (7)8 (5)14 (10)3 (10)Stage0.001Unresectable Ⅲ39 (9)6 (6)12 (7)18 (13)3 (10)Stage Ⅳ (M1a)74 (17)17 (18)28 (17)24 (17)5 (17)Stage Ⅳ (M1b)110 (26)28 (30)57 (34)20 (14)5 (17)Stage Ⅳ (M1c)158 (37)28 (30)53 (31)67 (49)10 (33)Stage Ⅳ (M1d)50 (12)15 (16)19 (11)9 (7)7 (23)Number of metastatic sites0.131 < 3287 (67)62 (66)108 (64)101 (73)16 (53) ≥ 3144 (33)32 (34)61 (36)37 (27)14 (47)Liver/kidney dysfunction0.934No391 (91)85 (90)152 (90)126 (91)28 (93)Yes40 (9)9 (10)17 (10)12 (9)2 (7)Autoimmune disease0.835No417 (97)90 (96)163 (96)135 (98)29 (97)Yes14 (3)4 (4)6 (4)3 (2)1 (3)LDH0.293 ≤ ULN261 (65)60 (71)104 (65)81 (62)16 (53) > ULN142 (35)24 (29)55 (35)49 (38)14 (47)BRAF mutation < 0.001Wild-type173 (40)35 (37)75 (44)57 (41)6 (20)Mutation47 (11)25 (27)7 (4)8 (6)7 (23)Not investigated211 (49)34 (36)87 (52)73 (53)17 (57)Line of therapy0.1201 line295 (68)63 (67)122 (72)87 (63)23 (77)2 line105 (24)26 (28)40 (24)38 (28)3 (10) ≥ 3 line31 (7)5 (5)7 (5)13 (9)4 (13)Anti-PD-10.005Toripalimab359 (83)82 (87)148 (88)102 (74)27 (90)Pembrolizumab72 (17)12 (13)21 (12)36 (26)3 (10)Radiotherapy0.366No367 (85)84 (89)145 (86)112 (81)26 (87)Yes64 (15)10 (11)24 (14)26 (19)4 (13)Combination therapy < 0.001Anti-PD-1 monotherapy84 (19)26 (28)36 (21)13 (9)9 (30)Anti-PD-1 + chemotherapy60 (14)11 (12)26 (15)17 (12)6 (20)Anti-PD-1 + anti-VEGF102 (24)13 (14)17 (10)64 (46)8 (27)Anti-PD-1 + anti-VEGF + chemotherapy65 (15)11 (12)36 (21)18 (13)0 (0)Anti-PD-1 + IFN-α1b84 (19)23 (24)39 (23)17 (12)5 (17)Anti-PD-1 + IFN-α1b + anti-VEGF36 (8)10 (11)15 (9)9 (7)2 (7) Patient baseline characteristics Treatment characteristics, prior therapy, and post-progression therapy As shown in Table 1, the numbers of patients treated with anti-PD-1 monotherapy, anti-PD-1 + chemotherapy, anti-PD-1 + anti-VEGF, anti-PD-1 + anti-VEGF + chemotherapy, anti-PD-1 + IFN-α1b, and anti-PD-1 + IFN-α1b + anti-VEGF were 84 (19%), 60 (14%), 102 (24%), 65 (15%), 84 (19%), and 36 (8%), respectively. There were significant differences in anti-PD-1-based combination therapies across biological subtypes. Anti-PD-1 monotherapy was relatively commonly used for CM (28%) and UPM (30%). Patients with MM were more likely to be treated with anti-PD-1 + anti-VEGF (46%), such as anti-PD-1 combined with axitinib. Patients with AM received a higher proportion of anti-PD-1 + IFN-α1b (23%), anti-PD-1 + anti-VEGF + chemotherapy (21%), and anti-PD-1 monotherapy (21%). No significant difference in prior ICIs therapy was observed across biological subtypes. In contrast, CM and UPM were more likely to receive BRAF ± MEK inhibitor (Supplementary Tables S6-S7). Following disease progression and discontinuation of anti-PD-1-based therapy, post-progression therapies were administered in 298 of 431 patients (69%). No significant difference in post-progression therapy was observed across biological subtypes, except for BRAF ± MEK inhibitors, which were more frequently used in CM and UPM (Supplementary Tables S8-S9). Objective response according to biological subtypes and treatment Of the 431 patients, 23 were unable to determine the best overall response after the first dose of anti-PD-1. Thus, 408 patients were evaluable for ORR. The overall ORR was 19.85% in the cohort (including anti-PD-1 monotherapy and anti-PD-1-based combination therapies), with a subtype-specific rate of 22.22% in CM, 17.83% in AM, 19.85% in MM, and 23.33% in UPM (Table 2). By treatment modality, ORR for anti-PD-1 monotherapy, anti-PD-1 + chemotherapy, anti-PD-1 + anti-VEGF, anti-PD-1 + anti-VEGF + chemotherapy, anti-PD-1 + IFN-α1b, and anti-PD-1 + IFN-α1b + anti-VEGF were 14.46%, 18.97%, 21.05%, 26.56%, 16.44%, and 25.71%, respectively (Supplementary Table S10).Table 2Objective response rate (ORR)Treatment responseTotal no. of patients (%)Overalln = 431Cutaneousn = 94Acraln = 169Mucosaln = 138Unknownn = 30Best overall responseComplete response7223Partial response741826237Stable disease19139805913Progressive disease13631494610Unable to determine2341270ORR19.85%22.22%17.83%19.85%23.33% Table 2Objective response rate (ORR)Treatment responseTotal no. of patients (%)Overalln = 431Cutaneousn = 94Acraln = 169Mucosaln = 138Unknownn = 30Best overall responseComplete response7223Partial response741826237Stable disease19139805913Progressive disease13631494610Unable to determine2341270ORR19.85%22.22%17.83%19.85%23.33% Objective response rate (ORR) We further performed stratified analyses of ORR for different anti-PD-1 combination therapies in each subtype (Supplementary Tables S11–S14). The ORR of anti-PD-1 monotherapy was 19.23% in CM, 8.57% in AM, 7.69% in MM, and 33.33% in UPM, respectively. The ORR of anti-PD-1-based combination therapies ranged from 0 to 40% across biological subtypes. Notably, the ORR of each combination therapy differed across subtypes. For example, anti-PD-1 + IFN-α1b was associated with a higher ORR in AM, but a lower ORR in MM (24.24% vs. 0%). Interestingly, anti-PD-1 + anti-VEGF showed a higher ORR in MM compared with AM (26.23% vs 0%). Progression-free survival (PFS) and overall survival (OS) according to subtypes The median PFS and OS for CM, AM, MM, and UPM were 5.4, 4.4, 4.6, 5.2 months (P = 0.73), and 39.8, 21.8, 19.8 and 27.1 months (P = 0.045), respectively (Fig. 1). There was no significant difference in PFS among biological subtypes. However, OS was significantly longer in CM and UPM than in AM and MM, regardless of the therapies used in combination. To eliminate the potential survival bias associated with biological subtypes and analyze the efficacy of anti-PD-1-based therapies, we further performed stratified analyses of PFS and OS according to different therapies in each subtype (Fig. 2). Kaplan–Meier analysis with the log-rank test showed no significant differences in PFS and OS across treatment modalities in CM (Fig. 2A, B) or UPM (Fig. 2G, H).Fig. 1Kaplan–Meier curves of (A) progression-free survival (PFS) and (B) overall survival (OS) among patients with advanced melanoma stratified by biologic subtypes. P values were determined using log-rank testFig. 2Survival according to biologic subtypes and anti-PD-1-based combination therapies PFS (A) and OS (B) for cutaneous melanoma, PFS (C) and OS (D) for acral melanoma, PFS (E) and OS (F) for mucosal melanoma, and PFS (G) and OS (H) for melanoma of unknown primary Kaplan–Meier curves of (A) progression-free survival (PFS) and (B) overall survival (OS) among patients with advanced melanoma stratified by biologic subtypes. P values were determined using log-rank test Survival according to biologic subtypes and anti-PD-1-based combination therapies PFS (A) and OS (B) for cutaneous melanoma, PFS (C) and OS (D) for acral melanoma, PFS (E) and OS (F) for mucosal melanoma, and PFS (G) and OS (H) for melanoma of unknown primary However, there was a nonsignificant trend toward prolonged PFS with anti-PD-1 + anti-VEGF + chemotherapy (median PFS 7.6 months, Fig. 2C) and prolonged OS with anti-PD-1 + IFN-α1b (median OS 27.2 months, Fig. 2D) in AM. In addition, anti-PD-1 + anti-VEGF was associated with significantly longer median PFS (6.3 months, Fig. 2E) and nonsignificant trend toward prolonged OS (25.5 months, Fig. 2F) in MM. Besides biological subtypes and treatment regimen, BRAF mutation status was associated with longer OS (supplementary Figure S2). Furthermore, stage Ⅳ (M1d), ECOG PS ≥ 2, ≥ 3 metastatic sites, hepatic or renal dysfunction, and LDH > upper limit of normal (ULN) were associated with significantly shorter PFS or OS (Supplementary Figure S2). Toxicity Of the 431 patients, 293 (68%) experienced any grade irAEs and 62 (14%) experienced 3–4 irAEs. No treatment-related death occurred (Fig. 3, Supplementary Tables S15-S16). The most common event was thyroid dysfunction (181, 42%), followed by skin toxicity (67, 16%), hypertriglyceridemia (66, 15%), hepatobiliary disorders (53, 12%), and gastrointestinal disorders (34, 8%). The rates of steroid use, discontinuation of anti-PD-1, and hospitalization were 8%, 8%, and 5%, respectively. There was no significant difference in the incidence, severity, or management of irAEs across different subtypes (Table S15).Fig. 3Incidence of irAEs according to spectrum and management in different combination therapies Incidence of irAEs according to spectrum and management in different combination therapies However, the incidence of irAEs differed across different treatment regimens. Compared with anti-PD-1 monotherapy, anti-PD-1-based combination therapies significantly increased the incidence of irAEs, steroid use, and discontinuation of anti-PD-1, especially in the anti-PD-1 + IFN-α1b + anti-VEGF group (Fig. 3). Considering the specific types of irAEs induced by anti-PD-1 + anti-VEGF and anti-PD-1 + IFN-α1b therapies, these regimens mainly increased the incidence of thyroid dysfunction, skin toxicity, hypertriglyceridemia, hepatobiliary, and gastrointestinal disorders, which were tolerable and not life-threatening. Nevertheless, the triple therapy of anti-PD-1 + IFN-α1b + anti-VEGF significantly increased grade 3–5 irAEs, which led to higher hospitalization and permanent discontinuation rate (Table S16). Univariable and multivariable analysis for PFS, OS, and irAEs Both univariable and multivariable Cox regression analyses consistently indicated that ECOG PS ≥ 2, AM, and MM subtypes were significantly associated with worse PFS and OS (Supplementary TableS17). Anti-PD-1 + anti-VEGF and anti-PD-1 + anti-VEGF + chemotherapy were significantly associated with longer PFS, but not OS. Stratified analyses revealed subtype-specific treatment effects. In CM subtype, no significant differences were observed across different anti-PD-1-based therapies, and results from univariate and multivariate Cox regression analyses were consistent. In AM subtype, compared with anti-PD-1 monotherapy, anti-PD-1 + anti-VEGF + chemotherapy was associated with significantly longer PFS in both univariate (HR, 0.6 [0.3, 1.0], P = 0.047) and multivariate Cox regression analyses (HR, 0.4 [0.2, 0.7], P = 0.005). In MM subtype, anti-PD-1 + anti-VEGF was significantly associated with longer PFS in multivariable Cox regression analysis (HR, 0.4 [0.2, 0.9], P = 0.018) (Supplementary Table S18). In addition, anti-PD-1 + anti-VEGF + chemotherapy (HR, 0.6 [0.3, 1.5], P = 0.288), anti-PD-1 + IFN-α1b (HR, 0.5 [0.2, 1.1], P = 0.077) and anti-PD-1 + IFN-α1b + anti-VEGF (HR, 0.4 [0.1, 2.1], P = 0.295) showed nonsignificant trends toward improved OS in AM, but not in MM. Anti-PD-1 + chemotherapy (HR, 0.8 [0.2, 2.9], P = 0.739) and anti-PD-1 + anti-VEGF (HR, 0.6 [0.2, 1.7], P = 0.392) showed nonsignificant trends toward improved OS in MM, but not in AM (Supplementary Table S19). To further confirm that the observed findings are robust to potential confounders, we performed stratified analyses by major clinically or statistically significant covariates. The results supported the robustness of the main analyses (Supplementary Tables S20-S21). A multivariate logistic regression model adjusted for all risk factors (Supplementary Table S22) and found that compared with anti-PD-1 monotherapy, anti-PD-1 + anti-VEGF (OR, 2.3[1.1–4.7], p = 0.024), anti-PD-1 + anti-VEGF + chemotherapy (OR, 2.2[1.0–5.0], p = 0.043), anti-PD-1 + IFN-α1b (OR, 4.3[2.0–9.2], p < 0.001), and anti-PD-1 + IFN-α1b + anti-VEGF (OR, 4.4[1.5–12.6], p = 0.006) were significantly associated with any grade of irAEs. While only anti-PD-1 + IFN-α1b + anti-VEGF group was significantly associated with a higher incidence of grade 3–5 irAEs (OR, 4.4[1.3–14.7], p = 0.017). Discussion The high heterogeneity of melanoma determines that melanoma guideline recommendations should differentiate between subtypes. However, most published studies have focused on CM or on single-arm immunotherapy in AM/MM. To the best of our knowledge, this is the first and largest investigation to compare the efficacy and safety of various anti-PD-1-based therapies across different biological subtypes in Asian. The proportions of AM (39%) and MM (32%) in our cohort were much higher than those in Western populations [27] (only 2–3%). In addition, 81% of patients were treated with five anti-PD-1 combination therapies in our national real-world study. Generally, ORR and PFS were improved and showed differences among anti-PD-1-based combination therapies in AM and MM. There was significant difference in OS across biological subtypes, regardless of the combination therapy regimen. Interestingly, the biological subtype difference in clinical benefit was not observed in toxicities. Different anti-PD-1-based combination therapies were strong risk factors for the incidence and spectrum of irAEs. In our study, CM subtype independently predicted superior long-term benefit from anti-PD-1 therapy compared with AM and MM. This finding is consistent with previous genomic analyses, as BRAF mutations occurred more frequently in CM than in AM or MM, which was associated with ultraviolet (UV) light-induced DNA damage [18]. Compared with anti-PD-1 monotherapy, anti-PD-1-based combination therapies were associated with nonsignificantly higher ORR (16.67–37.50%) and longer PFS (3.4–5.9 months), whereas no significant difference in OS was observed among treatment groups. Since combination therapies may increase the incidence of irAEs, but without significant long-term survival benefits in CM, regardless of the line of therapy. Anti-PD-1 therapy remains a mainstay in the treatment of advanced melanoma, and with PD-L1 expression representing a critical predictive biomarker. The phase II POLARIS-01 [16] trial showed that PD-L1-positive Chinese patients (20.5%) benefited more from anti-PD-1 monotherapy, with significantly better ORR (38.5% vs. 11.9%, P = 0.0065) and PFS (7.7 months vs. 2.7 months, P = 0.013) than those with PD-L1-negative. Our findings support that anti-PD-1 monotherapy was still the preferred recommendation for CM, especially for patients with PD-L1-positive, and combination therapies may be considered for subsequent treatment lines. For AM subtype, the ORR and median PFS of patients treated with anti-PD-1 monotherapy were 8.57% and 3.5 months, which were comparable with those reported in single-arm clinical trials of Chinese patients [16] (ORR 14.0%, PFS 3.2 months) and the real-world study on Japanese AM [28] (ORR 16.6%, PFS 3.5 months). Notably, compared with anti-PD-1 monotherapy, the combination therapy of anti-PD-1 + anti-VEGF + chemotherapy showed nonsignificant trends toward prolonged PFS (3.5 months vs. 7.6 months) and improved ORR (8.57 vs. 30.56%) in AM. Our results were consistent with a recent single-arm real-world study published in 2025 [29] (ORR 38.1%, PFS 8.5 months), but lower than those reported in the CAP 03 phase Ⅱ clinical trial (ORR 64%, PFS 17.5 months) [22]. It can be explained that the baseline characteristics of AM in our real-world study were worse than those in the CAP 03 clinical trial, such as a lower percent of first-line therapy (67% vs. 100%), a higher percent of patients with brain metastasis (22.2% vs. 0%), and baseline hepatic or renal dysfunction (11% vs. 0%). The amazing effect of anti-PD-1 + anti-VEGF + chemotherapy may be due to the synergistic effect of chemotherapy with anti-angiogenesis agent and anti-PD-1 in AM [30] and has been validated in phase Ⅱ/Ⅲ trials of other solid tumors [31, 32]. Nevertheless, although anti-PD-1 + anti-VEGF + chemotherapy has been explored for melanoma in recent years, long-term follow-up results are still needed in the future. In the MM subtype analysis, compared with anti-PD-1 monotherapy, anti-PD-1 + anti-VEGF showed significantly higher ORR (26.23% vs. 7.69%) and longer PFS (6.3 months vs. 3.1 months) in our study. The result was slightly lower than that from single-arm phase IB clinical trial [20] (ORR 48.3%; PFS 7.5 months), but similar to the single-center real-world study [21] (ORR 24.5%; time to treatment failure 5.2 months). In addition, anti-PD-1 + anti-VEGF also showed long-term benefit (median OS 25.5 months), comparable with the 3-year survival update of the clinical trial (OS 20.7 months) [33]. Vascular endothelial growth factor (VEGF)-mediated immunosuppression in the tumor microenvironment [34] plays a vital role in tumor metastasis. Studies showed that the use of anti-VEGF and anti-PD-1 can synergistically enhance T cell infiltration and suppress tumor growth [35]. Patients treated with anti-PD-1 monotherapy in our study also showed a longer OS (19.8 months) in MM than previous report (11.3 months) [36]. It may benefit from the high proportion of anti-PD-1 + anti-VEGF (29%) used in the post-progression therapy in our study (Table S8). For UM subtype, the ORR and median PFS treated with anti-PD-1 monotherapy were 33% and 3.9 months, which were higher than those in AM and MM, but lower than those in CM. Previous reports demonstrated that the prognosis of UM was better than melanoma of known primary [37]. However, the median PFS of UPM in our result was shorter than that reported in the POLARIS-01 clinical trial [16] (7.0 months). The possible reason may be that baseline characteristics of UPM in our study were worse than those in POLARIS-01 trial (Table 1), especially the higher percent of patients with brain metastasis (23% vs. 0%) and ECOG PS ≥ 2 (10% vs. 0%). Genetic analyses revealed that UPM resembles the genotype of CM [38], which further supports a better prognosis of UPM than AM and MM. National Comprehensive Cancer Network (NCCN) [39] and American Society of Clinical Oncology (ASCO) [40] guidelines for Western and White populations mainly focus on CM and recommend combination therapy with two ICIs, such as anti-PD-1 + CTLA-4 or anti-PD-1 + relatlimab. Although the combination therapy with anti-PD-1 and CTLA-4 demonstrated higher ORR than anti-PD-1 monotherapy, it did not show significant long-term survival benefit in AM [41, 42] or MM [43]. In addition, the combination therapy of anti-PD-1 and CTLA-4 is associated with higher toxicity and poorer tolerance in Asian populations, as 50–80% of patients developed grade 3–5 treatment-related adverse events and a treatment discontinuation rate of 30–60% [44, 45]. The uncertain long-term benefits and high risk of irAEs have limited the widespread application of this combination in AM/MM in China. In our initial REDCap-based data collection, all patients receiving anti-PD-1-based therapies were consecutively enrolled from multiple centers in China, regardless of the combination therapy regimens. It can be seen from Figure S1 that among the 554 initially enrolled patients, only 3 received this combination therapy. These real-world data reflected that the combination therapy of anti-PD-1 and CTLA-4 has not been used in advanced melanoma in China. The sample size of the three patients was too small to conduct comparative analyses across different therapies, so they were excluded. Although published clinical research on anti-PD-1 + anti-VEGF + chemotherapy in AM and anti-PD-1 + anti-VEGF therapy in MM is limited to single-arm study, the recommendations for this two combination regimens in the CSCO guidelines [23] have promoted their real-world use in China. Our nationwide comparative study confirmed significant PFS benefits and tolerable toxicity of anti-PD-1 + anti-VEGF + chemotherapy in AM and anti-PD-1 + anti-VEGF therapy in MM. But these two combination therapies did not demonstrate significantly prolonged OS benefits compared with anti-PD-1 monotherapy. This may be due to post-progression therapy, as patients in the anti-PD-1 monotherapy group had a higher likelihood of receiving anti-PD-1 + anti-VEGF therapy for MM or anti-PD-1 + anti-VEGF + chemotherapy for AM after disease progression. Notably, although anti-PD-1 + IFN-α1b was not recommended in CSCO guidelines and did not show significant PFS benefit in comparative analyses, it demonstrated a nonsignificant trend toward improved OS compared with anti-PD-1 monotherapy in AM subtype (27.2 vs 21.1 months; HR, 0.5 [0.2, 1.1], P = 0.077). What is more, the median OS of anti-PD-1 + IFN-α1b group in our research (27.2 months) was longer than previous reports in Japanese AM [28] and white AM [46] treated with anti-PD-1 monotherapy (18.1–22.8 months) or anti-PD-1 + CTLA-4 (7.6–15.6 months). Type I IFN plays an important role in the generation and activity of cytotoxic T lymphocytes, and activate CD4+, CD8+ T cells [47]. Anti-PD-1 + IFN-α1b may potentially enhance the immunogenicity of AM. To our knowledge, our study is the first comparative report demonstrating the long-term benefit and acceptable toxicity of anti-PD-1 + IFN-α1b, which provides a promising therapeutic option for refractory melanoma. However, anti-PD-1 + IFN-α1b + anti-VEGF group was associated with significantly higher risk of grade 3–5 irAEs (OR, 4.4[1.3–14.7], p = 0.017), without superior benefit to that of anti-PD-1 + IFN-α1b group. Recent research identified that anti-VEGF may enrich IFN expression by increasing senescence marker expression and promoting type I IFN and STING signaling [48]. So we hypothesized that the reason for the higher risk of grade 3–5 irAEs in the anti-PD-1 + IFN-α1b + anti-VEGF group may be due to synergistic overactivation of IFN expression. Thus, larger prospective trials are warranted to validate this safety observation. There are some limitations in our study. First, the study was retrospective, so unmeasured confounding and bias might exist. In order to minimize possible bias, we consecutively extracted data from the REDCap system and then strictly applied the inclusion and exclusion criteria. To reduce possible confounding, we collected and adjusted for all potential confounders. Nevertheless, some unmeasured confounding may remain. For instance, although PD-L1 expression may affect clinical responses to anti-PD-1-based therapies, few patients in our real-world Chinese cohort underwent PD-L1 testing, largely owing to the high associated costs. In addition, clinicians tend to administer anti-PD-1 combination therapy to patients with more severe disease in real-world clinical practice, and this inherent selection bias may lead to the underestimation of survival benefits associated with combination treatments. Second, the incidence of grade 1–2 irAEs may be underestimated, since we only reported irAEs clearly confirmed and documented in medical records or by clinical physicians. Adverse events most likely caused by chemotherapy and difficult to distinguish from irAEs, such as nausea, vomiting, fatigue, and blood toxicity, were not recorded. So the incidence of hematologic toxicities maybe lower than the actual rate. However, the comparison results between different groups may not be affected, as the standards were consistent across groups. Third, although we are currently the largest melanoma cohort study in Asia, the sample size of each subgroup was still small after stratification by treatment regimen and melanoma subtype, due to the high heterogeneity of melanoma. The lack of standardized clinical treatment protocols led to treatment heterogeneity across different melanoma subtypes. The limited number of patients receiving different treatment regimens in each subgroup may reduce the statistical power of analyses, making it difficult to detect potential significant differences or associations between treatment regimens. In addition, there may be differences in baseline characteristics among patients in different group. As biological subtypes are strong influencing factors associated with long-term benefit, we conducted univariable and multivariable Cox regression analyses to explore factors associated with PFS and OS within each biological subtype. To confirm that the observed findings are robust to potential confounders, we further performed stratified analyses by major covariates with clinical or statistical significance. Although the results support the robustness of the analysis, we acknowledge that inadequate sample sizes remain an issue in some subgroups. Fourth, the impact of post-progression treatment on the study results was not fully analyzed. Due to the high malignancy characteristics of advanced melanoma in China, most patients received multiple lines of subsequent therapies after disease progression, which may influence the key endpoint of OS. Nevertheless, owing to the retrospective nature of the study, detailed information on the first-, second-, and third-line post-progression treatments was incomplete. Accordingly, we could not fully adjust for these factors in the statistical analyses. This limitation may introduce confounding effects when interpreting the long-term clinical benefits of initial treatment. So further prospective studies with larger sample sizes, standardized treatment strategies, and comprehensive collection of post-progression treatment data are needed to further confirm and supplement our findings. Conclusion The study results support that both biological subtypes and anti-PD-1-based combination therapy regimens are associated with clinical benefit in melanoma. CM subtype independently predicted better long-term benefit than AM/MM subtypes, regardless of the combination therapy regimens. In addition, the differences in clinical benefit across biological subtypes were not observed in toxicity profiles. Different anti-PD-1-based combination therapies were strong risk factors of incidence and spectrum of irAEs. In conclusion, our results provide more reliable evidence for anti-PD-1 + anti-VEGF + chemotherapy and anti-PD-1 + IFN-α1b in AM and anti-PD-1 + anti-VEGF therapy in MM, as the long-term benefit and acceptable toxicity. Electronic supplementary material Below is the link to the electronic supplementary material.Supplementary file 1 (DOCX 1131 kb) Supplementary file 1 (DOCX 1131 kb)