NMAs were assessed using the Preferred Reporting Items for Systematic Reviews and Network Meta-Analyses (PRISMA) guidelines (19). We searched PubMed, Embase, WHO International Clinical Trials Registry Platform, and Clinical Trials.gov for phase II or III RCT from database inception to December 24, 2021 ( Supplementary Table S1 provides the search strategy).

The inclusion criteria of this research were as follows: (a) a phase II or III RCT with parallel assignment; (b) patients aged 18 years or older; (c) patients had been histologically confirmed with unresectable or metastatic BRAF V600-mutant melanoma; and (d) the patient’s disease scored 0 or 1 with the Eastern Cooperative Oncology Group performance evaluation (20). The exclusion criteria were as follows: If an RCT used the intervention model of the crossover assignment, it should be excluded. However, if the RCT provided usable data from the first period of the randomized crossover trial, this can be viewed as a parallel-group trial (21, 22). In addition, women who were pregnant or breastfeeding and patients with active malignancy other than melanoma were excluded. In addition, patients receiving vaccines, traditional Chinese medicines, or other nontargeted or nonspecific immunotherapies will also be excluded. Three investigators independently reviewed study abstracts and full text. Where the investigators are unable to determine whether an RCT shall be included, a discussion with and vote-counting by all authors are deemed necessary to resolve the issue.

Data were extracted using a standardized collection form in Excel. The following data were extracted: publication details (the year of publication and first author), trial details (intervention, comparator, and the number of patients), and acceptability outcomes (the cumulative incidence of any-grade AEs and severe AEs). All included studies contained the most updated data, such as the case-extended follow-up data. We assessed individual trials according to the Cochrane Collaboration’s Risk of Bias tool by Review Manager v5.2 (23).

The classification of any-grade AEs (grades 1–5) and severe AEs (grades 3–5) that occurs in patients should be in accordance with the National Cancer Institute Common Terminology Criteria for Adverse Events, version 4.0 (24). If the data remained unavailable, studies would be excluded from the NMAs. For acceptability, the outcome measure was relative risk ratio (RR) along with its 95% credible intervals (CrI).

Conventional meta-analysis is based on pairwise head-to-head direct comparison, but pairwise head-to-head comparisons in oncology treatment are relatively limited (25). In contrast, the need for indirect comparisons by NMA has attracted significant attention (26). Statistical analysis and graph generation were performed using the networking commands of StataSE 16 (27). Firstly, we need to install the network meta-package, including st0411 and st0410, by typing the command <help network>. In addition, the mvmeta and metareg packages also should be installed for subsequent analysis. Secondly, we can enter the extracted data into the data editor, where <id> represents study, <t> means treatment, <sd> is the standard deviation, and <n> is the sample size. Noteworthy, each row represents one arm of one study. Thirdly, we can process our data into a specific format that can be used for network analysis using the command statement <network setup mean sd n, study(id) trt(t) format(augment)>. This contains the premise that our effect size must be a SD value. If our effect size is SMD, we can enter the command statement <network setup mean sd n, study(id) trt(t) format(augment) smd>. We can then enter <network map, improve> in the command window to obtain the network map. The circular nodes indicate the treatment regimens, the size of each circle corresponds with the number of participants, and the width of the lines indicates the number of studies. Next, we can type the command <network meta i> to check for inconsistency hypothesis, which is only considered when the network forms a closed-loop (28–30). If the p-value is less than 0.05, the inconsistency model is significant, indicating that the consistency model cannot be used for analysis, and therefore, the random-effects model should be selected instead (31). We then need to examine the source of heterogeneity, including the sensitivity analysis (29), loop heterogeneity (25), and node-splitting (32). Furthermore, we can judge the best treatment option based on the surface under the cumulative ranking (SUCRA) for each treatment provided by STATA (33). The SUCRA can be transformed into acceptability with specific settings, which means a larger SUCRA indicates higher acceptability of a treatment regimen. Finally, the most important part of NMA is a league table, which can be generated by <netleague, lab(Ate+vem+cob vem+cob Niv+ipi Niv ipi Dab+Tra vem enc+bin enc Dab Dab+Tra+pem Dac+sel Dac) sort(ipi Niv Dac Dac+sel Dab+Tra enc+bin Dab Dab+Tra+pem enc vem Niv+ipi Ate+vem+cob vem+cob) eform>. Based on the league table, we can directly or indirectly analyze all included treatments.

The search identified 1,243 citations. After removing duplicates, 1,092 citations were retrieved from the specific databases. Eight hundred fifty-four studies were excluded after reading the titles and abstracts. Assessing full text led to the exclusion of another 224 citations. In total, 12 citations describing 11 RCTs were included, and all trials were multicentric. The search result diagram is shown in Figure 1 .

The 11 RCTs involve a total of 5,803 patients with BRAF V600-mutant melanoma. Of the 12 studies, ten were phase III (34–39) (40–43) and two were phase II trials (44, 45). Among them, nine studies are two-arm (34, 36, 39, 41–45) and three are three-arm trials (35, 37, 40). In addition, patients from ten studies were not previously treated with systemic therapy for metastatic melanoma (34, 35, 37–40, 42–45), while patients from the other two studies are treated with immunotherapy (36, 41). Eventually, we demonstrated that the three studies are updated long-term follow-up results (39, 41, 42), and the study is indirect comparisons (39). Table 1 shows the summary characteristics extracted from the RCTs.

The risk of bias graph is shown in Supplementary Figure S1 . The overall risk of bias is relatively low. Of the 12 studies, six used the hierarchical replacement block randomization (34, 35, 38, 41–43, 45), and the remaining six only mentioned random grouping (36, 37, 39–41, 44). About half of the studies allude to allocation concealment (34, 35, 38, 42, 43, 45), and only five studies reported double-blind (34, 35, 38, 39, 45). As the 3 studies were open-label (40–42), their judgments of outcome were measured separately by other researchers, and we considered their results were not skewed.

We compared each treatment intervention and indicated that the 5,803 patients are randomized to receive either conventional chemotherapy (dacarbazine, N = 587) or BRAF inhibitor (including dabrafenib, encorafenib, and vemurafenib N = 1,814), a combination of BRAFi and MEKi (including vemurafenib plus cobimetinib, dabrafenib plus trametinib, and encorafenib plus binimetinib N = 2,257), ipilimumab (N = 311), nivolumab (N = 313), BRAFi+MEKi+anti-PD1 (including atezolizumab plus vemurafenib plus cobimetinib and dabrafenib plus trametinib plus pembrolizumab N = 290), and anti-CTLA4+anti-PD1 (including nivolumab plus ipilimumab N = 437). Of the 11 identified RCTs consisting of 12 studies, 13 treatment options were presented, as follows: (1) atezolizumab plus vemurafenib plus cobimetinib (Ate+Vem+Cob), (2) vemurafenib plus cobimetinib, (3) nivolumab plus ipilimumab (Niv+Ipi), (4) nivolumab, (5) ipilimumab, (6) dabrafenib plus trametinib (Dab+Tra), (7) vemurafenib, (8) encorafenib plus binimetinib, (Enc+Bin), (9) encorafenib, (10) dabrafenib, (11) dabrafenib plus trametinib plus pembrolizumab (Dab+Tra+Pem), (12) dacarbazine plus selumetinib (Dac+Sel), and (13) dacarbazine. The treatment options of the RCTs were connected in the main network ( Figure 2 ).

Relative acceptability concerning any-grade AEs is presented in Table 2 . Differences between treatments within the immunotherapy are statistically significant, suggesting that Niv+Ipi increases any-grade AEs than single-agent Ipi (RR: 0.90; Crl: 0.83, 0.96) or single-agent Niv (RR: 0.90; Crl: 0.84, 0.97). The acceptability of single-agent Niv and Ipi is considered the best treatment agent in the NMA. There are no statistical differences among most of the MEKi+BRAFi treatments, and we only demonstrate that the acceptability of Dab+Tra is better than Vem+Cob (RR: 0.94; Crl: 0.89, 0.98). We cannot compare the incidence of any-grade AEs within BRAFi (Dab, Vem, Enc) because of the lack of statistical differences.

Relative acceptability on severe AEs is presented in Table 3 . Differences between treatments within the immunotherapy are statistically significant, indicating that Niv+Ipi also increases severe AEs than single-agent Niv (RR: 0.38; Crl: 0.27, 0.52) or Ipi (RR: 0.47; Crl: 0.34, 0.64). Dac has better acceptability than other options in the NMA. There is also no statistical difference between most treatments of MEKi+BRAFi; nevertheless, Dab+Tra presents a lower incidence than Vem+Cob (RR: 0.57; Crl: 0.45, 0.71) in severe AEs. In BRAFi, single-agent Dab exhibits superior acceptability than Vem (RR: 0.66; Crl: 0.50, 0.87) and Enc (RR: 0.64; Crl: 0.43, 0.94). Although we cannot contrast the incidence of severe AEs within the triple therapies (Ate+Vem+Cob and Dab+Tra+Pem), it is associated with the worst acceptability.

For any-grade AEs in Supplementary Table S2 , Ipi (SUCRA: 0.87) and Niv (SUCRA: 0.86) are associated with the best safety profile in NMA, followed by single-agent Dac (SUCRA: 0.82). In the single-agent BRAFi, the acceptability of Dab (SUCRA: 0.46) is better than Enc (SUCRA: 0.38) and Vem (SUCRA: 0.34). Moreover, the ranking of Dab+Tra (SUCRA: 0.59) is superior to Enc+Bin (SUCRA: 0.47) and Vem+Cob (SUCRA: 0.10). Finally, the treatments of Vem+Cob+Ate (SUCRA: 0.15) and Vem+Cob (SUCRA: 0.10) rank the last.

For severe AEs in Supplementary Table S3 , the ranking of Niv (SUCRA: 0.95) is ahead of Dac (SUCRA: 0.90), followed by Ipi (SUCRA: 0.82). The ranking of Vem (SUCRA: 0.44) and Enc (SUCRA: 0.39) here is opposite to our observation in any-grade AEs’ ranking, and the acceptability of Dab (SUCRA: 0.75) is higher than Vem or Enc. In MEKi+BRAFi, the ranking of Dab+Tra (SUCRA: 0.66) is superior to Enc+Bin (SUCRA: 0.39) and Vem+Cob (SUCRA: 0.18). Dab+Tra+Pem (SUCRA: 0.06) and Ate+Vem+Cob (SUCRA: 0.11) rank the last in this study.

Previous studies have compared the effectiveness of allowing treatments to compete against one another, but an overall ranking of safety or acceptability remains unknown. The NMA yielded three important findings regarding the risk of AEs among patients with BRAF V600-mutant melanoma. Firstly, immunotherapy (Niv and Ipi) has better acceptability than most targeted therapies in severe AEs, and triplet therapies (Ate+Vem+Cob, Dab+Tra+Pem) have the worst acceptability. Second, Dab can be used as the first choice in single-agent BRAFi, and the treatment of Dab+Tra is commonly preferred in BRAFi+MEKi. Eventually, the combination of immunotherapy drugs (Niv and Ipi) increases anygrade and severe AEs than a single agent, whereas the condition of targeted therapy cannot be simply generalized.

In this paper, the patients receiving anti-PD1 (Niv) or anti-CTLA4 (Ipi) have better acceptability than patients receiving targeted therapies, and there is no statistical difference between Niv and Ipi. On the contrary, the study by Sandro Pasquali et al. showed that the acceptability of anti-CTLA4 is worse than targeted therapies, and there are significant statistical differences between anti-PD1 and anti-CTLA4 in severe AEs (46). The probable reason causing this difference is that the patients selected in his study include BRAF-mutant-type and wild-type, which may be led to a certain amount of heterogeneity (47). Another possible reason is that we only analyzed a specific drug but not a class of drugs, so we did not find the statistical difference between Niv and Ipi. Besides the differences, we have a common cognition that the patients receiving anti-PD1 has the best acceptability in severe AEs, which is also proved by Devji et al. (15).

Patients receiving the triplet therapies (Ate+Vem+Cob, Dab+Tra+Pem) have increased incidence of any-grade or severe AEs. In a previous study, the combination therapy of BRAFi (Vem) and anti-CTLA4 (Ipi) has been discontinued due to severe hepatotoxicity (48). Based on this, the researchers proposed that triplet therapies may increase the patient’s immune system’s sensitivity and block BRAF and MEK genes (49). Several triplet therapies that have been evaluated in early-phase clinical trials with promising anti-tumor effects existed, but they also showed obvious toxicity (50, 51). Therefore, taking the triplet therapies’ better effectiveness and higher risk of AEs together, whether they present any competitive advantage over the anti-PD1 or BRAFi+MEKi combinations remains to be seen.

The results are generally consistent with the previous studies in terms of MEKi+BRAFi comparison (52). In our data, Dab+Tra is associated with a lower incidence of any-grade or severe AEs than Vem+Cob, which is the same as in Daud et al. (53). Similarly, there are no statistical differences between Enc+Bin and Dab+Tra, as described by Consoli et al. (54). Although we cannot reveal statistical differences between Vem+Cob and Enc+Bin, we can sort them according to SUCRA. Clinicians can use this ranking to implement individualized medication treatment for patients to reduce patient resistance and suffering. The Dab+Tra has the lowest incidence of any-grade or severe AEs, followed by Enc+Bin, and the last is Vem+Cob. In addition, single-agent Dab has higher acceptability than single-agent Vem or Enc in severe AEs; thus, Dab can be used as the first choice for patients in single-agent BRAFi. We found that Enc was associated with high any-grade AEs and Vem was accompanied with high severe AEs, which was consistent with the report of Ascierto et al. (37). Therefore, when patients have to use Vem or Enc, clinicians should pay more attention to physiological indicators that can easily induce adverse reactions.

The present study has compared the combination of different targeted therapies versus monotherapy. The results of this research present some distinctions from the previous findings in several aspects. Whether the combination of BRAFi+MEKi increases the toxicity of single-agent BRAFi depends on the specific circumstances. Previous studies concluded that Dab+Tra did not raise any-grade or severe AEs than single-agent Dab (36, 38, 55). Our results have confirmed some of the previous findings, and also compared the incidence of AEs between combination and single-agent. Our study showed that compared with single-agent Dab, Enc+Bin and Vem+Cob have significantly increased severe AEs but not Dab+Tra. As for single-agent Vem, Vem Cob has increased severe AEs but not Enc+Bin and Dab+Tra, which is aligned with the report by Robert et al. (37). Finally, it should be noted that Dab+Tra decreases the severe AEs of single-agent Vem; the conclusion is different from Robert’s conclusion that there are no statistical differences between Dab+Tra and single-agent Vem. The possible reason for the distinction is that it is a single RCT and only included limited clinical samples.

Immunotherapy and targeted therapy have become the first-line treatment options for malignant melanoma, but some ambiguities remain (56). The primary purpose of this research is to provide a better drug of choice for patients with BRAF V600-mutant melanoma. This paper offers an acceptability ranking of monotherapy, which provides a new tool for drug selection in clinical practice. It is clear that the therapeutic effect of monotherapy is limited, but it has a few side effects and can be the considered first when managing mild cases. In addition, our article also mentioned the steps in choosing between monotherapy and combination therapy. Although the combination therapy will increase some AEs, its efficacy is affirmed. It is necessary to compare combination therapy with monotherapy and select combination therapy with limited side effects but apparent curative effects. Clinicians should clinically avoid unnecessary combination therapy in case of increased unexpected adverse reactions.