New paradigm for stage III melanoma: from surgery to adjuvant treatment

The American Joint Committee on Cancer (AJCC) classification has represented the reference staging system for melanoma for many decades. Recently, the AJCC 8th version has been introduced, and it has attempted to define a more accurate and precise definition of prognosis in line with the major progresses in understanding the biology and pathogenesis of melanoma [1].

However, this new staging system has been largely criticized [2]. For instance, it uses the same histological factors of the AJCC 7th version, and although with some refinements, it does not introduce any new prognostic biomarkers; moreover, in stages I–III melanoma, the AJCC 8th version is still based on 5- and 10-year melanoma-specific survival, and does not take into account the increased survival in stage IV melanoma associated with the introduction of new therapies. Even more importantly, the AJCC 8th version introduces major changes in the stage III staging system [2]. Indeed, surgical practice is changing in stage III patients, since two studies have shown that there is no survival benefit in radical lymph node dissection after a positive sentinel lymph node dissection [3, 4]. Therefore, some patients currently staged IIIB–C after dissection could be downgraded to IIIA (as in the case of patients with metastatic non-sentinel lymph nodes) since many completion lymph node dissections (CLND) will no longer be performed. Moreover, new and effective targeted and immune strategies are being introduced in the pharmacological armamentarium in the adjuvant setting, showing major efficacy [5‐7]. Other trials are ongoing (Table 1), and others will start, likely leading to complete changes in the therapeutic approach for stage III melanoma.

This article provides the authors’ personal view on the above-mentioned topics.

In a recent landmark international trial, Faries et al. randomly assigned patients with sentinel node metastases to lymph node dissection (dissection group; n = 824) or nodal observation with ultrasonography (observation group; n = 931) [4]. Patients in the dissection group did not show improved melanoma-specific survival compared with those in the observation group (86 ± 1.3% vs 86 ± 1.2%) at a median follow-up of 43 months (Fig. 1) [4]. However, the rate of disease-free survival was slightly higher in the dissection group (68 ± 1.7% vs 63 ± 1.7%; p = 0.05), given an increased rate of disease control in the regional nodes at 3 years (92 ± 1.0% vs 77 ± 1.5%; p < 0.001). Statistical analysis revealed that non-sentinel node metastases (reported in 11.5% of the patients in the dissection group) represent an independent prognostic factor for recurrence (hazard ratio [HR]: 1.78; p = 0.005). Overall, these findings demonstrate that immediate completion of lymph node dissection can increase the rate of regional disease control and provide prognostic information, without actually improving melanoma-specific survival. However, the patients in this study were a highly selected group with an extremely low burden of sentinel node disease (median tumor burden < 1 mm and only one node involved in 70% of patients). Similar results were reported in another pivotal multicenter, phase III trial by Leiter et al. [3]. In this latter study, which was terminated early due to difficulties in enrollment and low event rate, patients were randomly assigned to either CLND (n = 243) or observation (n = 233) [3]. Distant metastasis-free survival at 3 years was 74.9% (90% CI 69.5–80.3) following CLND and 77.0% (90% CI 71.9–82.1) in the observation group, similarly no differences in overall survival (OS) were found between the two groups.

According to the above-mentioned bases, there appears to be no survival benefit associated with CLND, although this strategy can lead to more refined staging and regional node control. For instance, a recent retrospective study from the European Organisation for Research and Treatment of Cancer (EORTC), including 1015 patients, showed that CLND led to upstaging in N-category in 19% and in AJCC stage in 5–6% [8]. Similar findings were reported by Madu et al. [9].

Active surveillance for patients who do not undergo CLND following a positive sentinel lymph node should include nodal ultrasound as a component of the follow-up strategy. CLND should be discussed with patients clarifying the risks, benefits and alternatives of the procedure, including their overall risk of harboring metastatic non-sentinel lymph nodes and the impact of dissection on staging, regional control and survival. In this regard, the Italian Melanoma Intergroup has developed a nomogram to predict the risk of non-sentinel node positivity, which can help clinicians to discuss with patients the opportunity of CLND [10].

However, with the advent of the new adjuvant treatments, there is probably less need for indicating a CLND.

In light of the evolving landscape of adjuvant therapy in melanoma and the lack of survival benefit of CLND, it becomes important to explore possible consequences of omitting CLND, and whether it is possible to stratify positive sentinel node patients on the basis of information retrieved from sentinel lymph node biopsy [8]. Verver et al. in a retrospective analysis from nine EORTC melanoma group centers, showed that adequate stratification of patients with sentinel lymph nodes was possible based on ulceration and tumor burden category, with 1 mm being the threshold to distinguish between low-/intermediate- and high-risk patients [8]. In particular, the identification of low-, intermediate- and high-risk patients could help select adjuvant therapy in clinical practice.

To this end, the therapeutic efficacy of adjuvant treatment with IFN-α for the treatment of patients with AJCC stage II–III cutaneous melanoma in terms of both disease-free survival and, to a lower extent, OS was already shown in two pivotal meta-analyses [11, 12]. More recently, a number of studies have investigated adjuvant therapy with newly introduced treatments, and therefore since 2011 the treatment of advanced or metastatic melanoma has undergone a sort of revolution with the introduction, which has also been recently evaluated in an adjuvant setting in a number of randomized clinical trials (Table 2) [13].

The new therapies introduced for stage IV melanoma do retain their efficacy also in the adjuvant setting: indeed, the best performers in stage IV are best performers in the adjuvant setting. In particular, recent developments in immunotherapy have prolonged OS in metastatic melanoma with the possibility to reach a long-term benefit [18]. Targeted therapies based on the combined BRAF and MEK inhibition also exert a long-term beneficial effect, which is more evident in patients with favorable baseline characteristics, namely normal levels of lactate dehydrogenase, with no brain metastases, and low tumor burden while elevated LDH, high tumor burden and brain metastasis should be considered as negative predictive factors [18]. This long-term benefit of targeted therapies may be related to an immune modulation: indeed, BRAF and MEK inhibitors affect tumor microenvironment and immune surveillance, and patients with complete response to targeted treatment have a pre-existing favorable immunologic signature [18].

What can we learn from the recent trials on surgery for stage III melanoma and trials in the adjuvant setting? Perhaps, the most relevant finding with immediate relevance to clinical practice comes from the Combi-AD update, presented by Georgina Long at the latest ESMO meeting. From these results it clearly stems that, at least for target therapy, adjuvant treatment prevents mainly loco-regional recurrence (− 11% with respect to placebo). Therefore, although in other recent adjuvant studies the enrolled patients underwent to an elective lymph nodes dissection before treatment, we believe that surgery could be postponed in case of loco-regional recurrence, and not performed immediately after sentinel node biopsy. This suggestion is supported by the results of the MSLT-II trial, which showed no impact of lymphadenectomy on melanoma-specific survival [4]. Of course, this applies only in case of micrometastases, as discovered by histology; in case of clinical overt recurrence, elective lymph node dissection should still be considered the first option. Moreover, patients eligible to adjuvant treatment in real life are markedly different from those included in clinical trials, as they are classified according to the new AJCC 8th classification instead of AJCC 7th classification, and also because the majority of patients in clinical practice would not have undergone CLND.

While for stage IIIB, C and D is well recognized the role of adjuvant, the role of this strategy in stage IIIA is more debated. Indeed, according to recent evidence we feel that in these patients the tumor burden within metastatic sentinel nodes should drive the decision on addressing or not them to the adjuvant treatment. In particular, a tumor burden of the sentinel nodes metastases > 1 mm allows to identify a high-risk patient, eligible to adjuvant treatment, as first suggested by van Akkooiin a landmark paper published in 2008 (Fig. 2) [10, 19]. On the other hand, in case of IIIA disease with a sentinel node metastasis < 1 mm, adjuvant should not be considered, since the risk of side effects—which can be also permanent in case of immunotherapy—is too high given the ultimate expected prognosis. Remarkably, most trials on stage IIIA enroll patients with sentinel nodes metastases > 1 mm. Overall, we suggest that T1a/T1b/T2a and N1a/N2a can be considered as selection criteria for adjuvant treatment. Moreover, more extended investigation on mutational analysis at stage III (e.g., by next-generation sequencing [20, 21]) will likely pave the way to new classification approaches.

Last, preliminary evidence had suggested that checkpoint inhibitors in the neoadjuvant setting (i.e., patients with surgically palpable and resectable lymph nodes metastasis) may be superior over adjuvant therapy [22‐25]. In a pilot study, 20 patients with palpable stage III melanoma were randomly assigned to ipilimumab and nivolumab, as either four courses after surgery (adjuvant arm) or two courses before surgery and two courses postsurgery (neoadjuvant arm) [21]. Neoadjuvant therapy was feasible: all patients underwent surgery at the preplanned time point, and pathological response was achieved in 7/9 patients. No relapses were reported at the time of analysis; however, 9/10 patients experienced 3/4 adverse events. In another study, 8/27 patients on neoadjuvant therapy experiences a complete or major pathological response after a single dose of anti-PD-1 therapy, and all of them remained disease-free [23]. In another randomized, phase II study treatment with combined ipilimumab and nivolumab was associated with high response rates (overall response rate: 73%, pathological complete response:, 45%) but substantial toxicity (73% grade III adverse events), while nivolumab monotherapy led to modest responses (ORR 25%, pathological complete response: 25%) and low toxicity (8% grade 3 treatment-related adverse events) [24]. Last, a randomized phase II trial in 21 patients showed that neoadjuvant plus adjuvant dabrafenib and trametinib prolonged event-free survival versus standard of care (19.7 months vs 2.9 months) in patients with high-risk, surgically resectable, clinical stage III–IV melanoma [25]. Neoadjuvant therapy appears to be feasible, although associated with a high burden of toxicity and future studies should clarify when this strategy should be preferred over adjuvant treatment [26].