The incidence of malignant melanoma (MM) increases consistently (39% between 2006 and 2016) with a current incidence of over 132.000 estimated cases worldwide each year (1, 2). Low risk MM (stages I-IIB) can be effectively treated with surgical excision only (3, 4). In contrast, high-risk MM (stages IIC-IV) with no evidence of disease (NED) after excision is associated with a worse survival rate (5) and therefore an efficient and tolerable adjuvant therapy is needed (4, 6). Interferon alpha (IFN-α) has lost its relevance in the wake of new therapeutic options due to its inconsistent impact on overall survival (OS) and high toxicity (7–9). After IFN-α, ipilimumab was the first agent to be approved for stage III (10). However, due to its unfavorable side effect profile, it was soon replaced by nivolumab and pembrolizumab (11). A recent phase II trial demonstrated the superior efficacy of the nivolumab/ipilimumab combination versus nivolumab or placebo in stage IV MM with NED (12). For patients with BRAF-V600 mutant (BRAFmut) MM, targeted adjuvant therapy is another option. While initial results with the BRAFi vemurafenib were not encouraging, the combination of the BRAFi dabrafenib with the MEKi trametinib demonstrated a clear benefit versus placebo in stage III disease (13, 14). Despite these significant developments, there is still no standard of care for the adjuvant therapy of high-risk MM, especially in the presence of a BRAF driver mutation (15).

Previous meta-analyses of adjuvant therapy for MM either included IFN-α, did not include subgroup-specific data or lacked a direct comparison of nivolumab versus placebo within a randomized placebo-controlled trials (RCT) (8, 16).

The RFS-benefit of modern adjuvant therapy (HR 0.57, 95% CI= 0.45–0.71) is higher than that shown for IFN-α in previous meta-analyses (HR 0.82, 95% CI = 0.77–0.87) (23).

Immune checkpoint inhibitor (ICi) monotherapy with pembrolizumab or nivolumab is similarly effective in improving RFS and both agents seem to be superior to ipilimumab while being less toxic, as previously demonstrated for nivolumab in the Checkmate-238 trial (24). This trial was not included in the current meta-analysis because of the lack of a placebo arm. An indirect analysis of adjuvant nivolumab versus placebo in stage III MM based on the Checkmate-238 and EORC-18071 trials calculated a HR for RFS of 0.53 (95% CI = 0.41–0.68) which is similar to the HR for RFS in the IMMUNED trial in stage IV (25).

The BRAF/MEKi combination therapy in the COMBI AD trial was associated with a clinical benefit across all subgroups with a tolerable adverse effect profile. In fact, the combination therapy demonstrated the highest numerical RFS-benefit in stage III melanoma (HR 0.49; 95% CI= 0.40–0.59). In stage IV melanoma with NED, nivolumab/ipilimumab demonstrated an impressive RFS benefit (HR 0.23; 97.5% CI= 0.12–0.45). The superiority of the combination versus ipilimumab or nivolumab monotherapy has also been demonstrated in metastatic melanoma, although at the cost of more grades 3–4 AE [24, 32]. However, only interim results from the rather small IMMUNED trial are currently published, and thus they must be interpreted with caution. Furthermore, stage IV patients with NED are per se at a higher risk of relapse, thus RFS benefits with adjuvant therapy between stages III and IV are not comparable.

In the interferon era, BRAFmut MM has been independently associated with a worse overall survival with HR of 1.7 (95% CI= 1.37–2.12) (26). Another, more recent meta-analysis of 52 trials also found that the presence of a BRAF mutation was associated with a reduced OS (HR 1.23, 95% CI= 1.09–1.38) (27). Most trials however included neither BRAF/MEKi nor ICi therapy. Prognosis of BRAFmut MM is expected to be crucially influenced by modern therapeutic agents. ICi have dramatically improved outcomes in the adjuvant and metastatic setting and additionally, patients with a BRAF mutation have now the option of targeted therapy. Thus, their prognosis can be expected to improve in the context of modern therapy. Interestingly, in our analysis, BRAFmut resectable MM was associated with higher RFS in trials which reported outcomes according to BRAF mutational status (IMMUNED and EORTC-1325). In contrast, in advanced melanoma, Puzanov and colleagues found in a pooled analysis of three RCTs with ICi (pembrolizumab) that BRAFmut patients had similar OS as patients with BRAF wild-type MM (PFS; 19.8% and 22.9% and OS; 35.1% and 37.5%). Patients with BRAFmut MM who did not receive BRAFi +/- MEKi therapy had a worse prognosis than those who did (28). This contradiction could be explained through the fact that, while BRAF mutations are early events in their evolution (29), metastatic melanomas accumulate further genomic alterations such as whole-genome duplication over time, which may account for resistance to treatment (30). Furthermore, the higher tumor burden of unresected melanomas might lead to increased potential of developing resistant clones under BRAF/MEKi compared to completely resected melanoma. These two factors could account for the discordance in the prognostic influence of BRAF mutations between completely resected stage III/IV in our meta-analysis and advanced/unresectable melanoma in the meta-analysis by Puzanov et al.

Age does not influence outcomes after adjuvant therapy. Specifically, the HR for RFS in patients ≥65 years old is even numerically lower than that of their younger counterparts. Therefore, advanced age alone should not discourage administration of adjuvant therapy. Recently published data on elderly patients with MM receiving ICi also demonstrated good clinical outcomes without increased toxicity (31).

Ulceration status of the primary tumor may be predictive of RFS when ICi are used. Several studies have shown that ulcerated melanomas have distinct biologic characteristics (32, 33). In our study, ipilimumab and pembrolizumab are both associated with a significant RFS benefit in patients with ulcerated melanoma, but not in those with primary tumors without ulceration. In contrast, dabrafenib/trametinib showed benefit regardless of ulceration status, while vemurafenib is also associated with superior RFS in non-ulcerated stage IIIC MM. A post hoc meta-analysis of the EORTC-18952 (IFN α-2b versus observation in stages IIB–III) and -18991 trials (pegylated-IFN versus observation in stage III) also demonstrated that the absence of ulceration was predictive for inefficacy of adjuvant treatment with IFN-α (34). Therefore, it would be reasonable to prefer a BRAF/MEKi combination in non-ulcerated BRAFmut MM.

Adjuvant treatment in stage IIIA is associated with a similar RFS-benefit as in stages IIIB/C. Thus, our meta-analysis supports administration of adjuvant therapy in stage IIIA. However, all the above trials are powered for DFS/RFS and OS data have only been reported for the first interim analysis of the COMBI-AD trial and for the EORC-18071 trial, where a benefit could be demonstrated in favor of treatment. The recently reported update of the Checkmate-238 trial (35) showed no difference in OS between ipilimumab and nivolumab despite a significant RFS benefit. This and the general lack of OS data pose the issue of early vs late treatment, particularly in stages II/IIIA. Further data on OS are needed to guide treatment decisions. The influence of toxicity in decision making in earlier stages is discussed below.

As IFN-α has been associated with substantial AE and drug related fatalities, modern adjuvant therapeutics have to meet high expectations (23, 36, 37). The highest toxicity was seen with the nivolumab/ipilimumab combination in stage IV MM with NED. Pembrolizumab toxicity in the adjuvant setting (31.6% grade ≥3 AE) was lower compared to the AE rate of nivolumab (41%). However, data on nivolumab toxicity in our meta-analysis are derived from stage IV MM, while pembrolizumab was tested in stage III patients. The Checkmate238 trial, which also included stage III patients demonstrated a 25.4% rate of grades 3–4 AE in patients treated with nivolumab (11). Monotherapy with pembrolizumab or nivolumab have been shown to have a considerably better tolerability profile than ipilimumab (10, 24, 38). The dabrafenib/trametinib combination demonstrated similar grades 3–4 AE rates as pembrolizumab and nivolumab (26%) and less than half compared to vemurafenib (59%), as in previous trials comparing BRAF/MEKi combinations to BRAFi monotherapy (39, 40). The ICi combination and ipilimumab monotherapy were associated with the highest toxicity. In the context of the curative adjuvant setting, potentially permanent toxicities involved with ICi become particularly relevant. The incidence is higher with ICi combination than with PD-1 monotherapy (hypophysitis: 8.0% vs. 1.1, hypothyroidism: 13.2% vs. 7.0%). Primary adrenal insufficiency and insulin dependent diabetes are rare events (cumulative incidence after ICi: 0.7% and 0.2% respectively) (41–43). Because of this potentially long-lasting toxicity and the lack of consistent data on OS as discussed above we generally prefer BRAF/MEKi as adjuvant treatment in the context of BRAF-mutant MM, especially in stages IIIA/B.

An important unanswered clinical question is adjuvant therapy for patients with stage II MM, where rates of distant recurrence after resection can reach 44% (44, 45). Vemurafenib monotherapy is not approved in the adjuvant therapy of MM, however, BRIM8 was the only trial to include patients with stage IIC. In this subgroup, no events occurred in the vemurafenib arm (0/15) whereas six patients suffered a relapse in the placebo arm (6/12) (13). Although IFN-α remains an adjuvant therapeutic option for patients with stages IIB and IIC melanoma, it is rarely used in daily practice due to its significant toxicity (9). Currently ongoing trials are comparing pembrolizumab and nivolumab vs. placebo in resected stage II MM (45, 46).

Neoadjuvant approaches with both BRAF/MEKi and ICi in high-risk resectable MM are also currently under investigation (47, 48).

Although based on well-designed trials with robust results, our meta-analysis still has some limitations. First, it does not address the contemporary question of a comparison between adjuvant and neoadjuvant therapy. Second, it was restricted to placebo-controlled studies and thus forced the exclusion of relevant trials like those comparing two agents (e.g., ipilimumab versus nivolumab). Moreover, due to the significant trial variability regarding endpoint reporting and lacking consistent OS data reporting/availability, our meta-analysis is based on RFS and not on OS data. On this matter, significant inter-trial heterogeneity is also noted in the stages included. Furthermore, comparisons between subgroups were not possible for all the included RCTs, as subgroup-definition as well as data availability for each subgroup varied across the trials. Another limitation is the inconsistent representation of BRAFmut MM across trials. In addition, treatment and definition of stage III within the included RCTs does not correspond completely to current standards. Moreover, complete lymph node dissection was required for trial enrollment in the EORTC-1325, EORTC-18071 and COMBI-AD trials. This practice has been meanwhile replaced by sentinel lymph node biopsy according to results from RCTs (49, 50). Additionally, it must be kept in mind, that the current definition of stage III disease according to the 8^th edition of AJCC is different than the one used in the RCTs above. Stages IIIA/B/C as defined in the 7^th edition carry a worse prognosis, and may therefore benefit more from adjuvant therapy (51).

In conclusion, contemporary adjuvant therapy in the post interferon-alpha era for patients with high-risk completely resected MM is effective and tolerable and should be recommended in all patients in the absence of contraindications. BRAFmut MM was associated with higher RFS. Furthermore, some subgroups may benefit more from specific treatments and this can guide treatment choice. Advanced age and stage IIIA should not discourage adjuvant treatment. Options in BRAF wildtype melanoma are limited to ICi. In BRAFmut MM, BRAF/MEKi should be preferred, especially in the absence of ulceration and stage IIIA. Adjuvant treatment should be adapted to patient preference like the intake schedule or pre-existing conditions.

The data sets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: “The Open Science Framework—Center for OpenScience” (https://osf.io/m9vr5).

KC, KY, and DK-M performed the systematic review. KC, KY, MB, LT, and TR performed the data analysis. KC, CP, CM, TV, SS, and DK-M evaluated the data. KC, KY, and DK-M drafted the manuscript. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.