Melanoma brain metastasis (MBM) is a known characteristic for poor prognosis. The median overall survival (mOS) in the era of chemotherapy was 4 months and decreased to 2 months in patients with elevated lactate dehydrogenase (LDH).1 2 The response of MBM to chemotherapy was approximately 5%. This applies to both, drugs that cross the blood-brain barrier, such as temozolomide and fotemustine, and to drugs that do not cross the blood-brain barrier, such as dacarbazine.3 4 The American Joint Committee on Cancer has acknowledged the negative impact of brain metastasis on the prognosis of patients with melanoma in its latest eighth edition staging system by defining this subgroup as M1d.5

With the introduction of targeted treatment (BRAF/MEK inhibitors) and immune checkpoint inhibitors, the prognosis of metastatic melanoma has drastically improved.6–8 In contrast to ample data on the efficacy of novel therapies in stage IV melanoma without MBM, there are only a few small studies on the efficacy of these drugs in patients with cerebral disease. This lack of information is mainly due to the fact that large phase II/III multicenter studies systematically excluded patients with MBM, particularly if symptomatic or previously treated with local therapy, such as stereotactic radiosurgery and surgery (STR/surgery). The first studies investigating targeted therapy and immune checkpoint inhibitors in patients with MBM showed that these therapies were also very effective intracranially.9–12 Currently available data suggest that PD-1-based immunotherapy and particularly combined immunotherapy with nivolumab and ipilimumab (NIVO+IPI) might be more effective than BRAF/MEK inhibitors.8 13 In two retrospective studies with patients with MBM, the authors reported that patients receiving immunotherapy had a mOS between 13 and 14.8 months (95% CI: 8.1 to 17.8 and 9.9 to 19.7, respectively), while in those receiving targeted therapy the mOS was only 7 and 10 months (95% CI: 3.8 to 10.2 and 7.8 to 11.7, respectively).14 15 This difference was also present when these systemic therapies were given in combination with stereotactic radiosurgery, favoring the combination with immunotherapy, which resulted in a mOS between 21–25 months (95% CI: 12.9 to 29.1 and 14.6 to 35.4, respectively) and 12.9–14 months with targeted therapy (95% CI: 12.9 to 29.1 and 9.1 to 16.7, respectively).

Our study provides real-world outcome data from 23 German skin cancer centers, retrospectively assessing the activity of NIVO+IPI alone or in combination with local therapies regardless of treatment line in patients with asymptomatic and symptomatic MBM.

We addressed the following questions: (a) Which prognostic factors for OS can be identified in patients with MBM treated with combined immunotherapy? (b) Does local therapy (STR/surgery) improves survival in patients with MBM treated with NIVO+IPI? (c) Are STR/surgery more effective when given before or after combined immunotherapy? (d) Is there a difference in terms of survival when combined immunotherapy is given as a first-line treatment for MBM or later in the course of the disease? (e) In patients with BRAF V600-mutated melanoma, which first-line systemic therapy for MBM translates into better OS: first-line immunotherapy or first-line targeted therapy? (f) Is there a difference in terms of OS when patients with symptomatic and asymptomatic MBM receive NIVO+IPI? Since a total of 380 patients were included, we were able to perform subgroup analyses with reasonable statistical power.

The present study shows that combined immunotherapy with NIVO+IPI can result in improved survival of patients with MBM, comparable to results in other stage IV patients. This is particularly true if intracranial CR, PR or SD has been achieved. The type of intracranial response is a strong predictor for OS. In our cohort, the 2-year OS rates of patients with SD, PR and CR ranged from 63% to 86%, whereas patients with PD had a 2-year OS rate of only 20% (table 3). Similar favorable results have been reported in the ABC trial, a randomized phase II study of nivolumab or NIVO+IPI in patients with MBM.16 The 3-year intracranial PFS was above 90% for patients with asymptomatic, treatment-naïve MBM achieving an intracranial CR, and above 50% for patients with PR. We have no explanation why in our cohort patients with SD did better than patients with PR.

In our study, the 1-year and 2-year OS rate were 69% and 41%, respectively, in line with previous reports.9 10 In the already mentioned ABC trial, patients who received combined immunotherapy had a 1-year and 2-year OS rate of 63%16 and in the Checkmate-204 trial the reported 1-year OS rate was even higher (81.5%).10 The survival rates in these trials are higher than those reported in our cohort. Compared with the ABC trial and the Checkmate-204 trial, which included patients with asymptomatic MBM and treatment-naïve BRAF wild-type patients, 31% (60/193) of the patients in our trial had symptomatic MBM and 20% of the BRAF wild-type patients were pretreated. In the Checkmate-204 trial, 17% of the patients had received previous systemic therapy for MBM and 52% had only one MBM compared with 34% pretreated patients and 53% patients with more than three MBM in our cohort.

Two studies evaluating pembrolizumab in patients with MBM also reported similar outcomes.17 18 The first study evaluated treatment with pembrolizumab monotherapy in 23 patients with one or more asymptomatic and untreated MBM. With a longer follow-up of 38 months, the mOS time was 17 months (95% CI: 10 months to not reached) and the 2-year OS was 48%. These are in line with our results for patients who did not receive STR/surgery for whom the mOS was 16 months (95% CI: 7.6 to 24.4) and the 2-year OS rate was 41%. However, in this trial, only asymptomatic patients were included and 87% had <3 MBM, a population with potentially better outcome that the one included in our report. In the second study, Anderson et al reported the results of the combination from pembrolizumab and radiation therapy in 21 patients with MBM. Despite the low number of patients included, the percentage of lesions that had a CR (>30%), was higher than previously reported with systemic therapy or STR alone.

The combination of immunotherapy and local therapy with stereotactic irradiation or surgery improved patients’ survival compared with patients who only received NIVO+IPI. This benefit might be related to a synergic effect between radiotherapy and immunotherapy that has been demonstrated both in preclinical and clinical studies.19–23 The combination of radiation and immune checkpoint inhibitors seems to be effective both in the irradiated and non-irradiated lesions, and this effect might be associated with the activation of cytotoxic T-cells and reduction of myeloid-derived suppressor cells.18 24 25

The benefit of combining local and systemic therapy in MBM has been previously shown by our group and others, with mOS that range from 14 to 25 months and 1-year OS rates between 58% and 78% in the groups that received local and systemic therapy, clearly superior to the outcomes of patients receiving only systemic therapy (mOS between 6 and 13 months and 1-year OS rates ranging from 34% to 53%).14 15 26–33

In our study, the time point at which the patients received local therapy did not seem to play a significant role in OS: local therapy performed upfront or after initiation of NIVO+IPI resulted in similar OS rates, with a trend benefiting local therapy upfront (mOS 26 months vs 16 months). Different retrospective studies have also addressed this question, and, similar to our cohort, upfront local therapy seems to have better outcomes (mOS of 11–23 months in the group receiving local therapy upfront and 3–9 months in patients receiving local therapy after systemic therapy).34 35

There is still an ongoing debate whether some patients might be better served with systemic therapy alone, as we see very positive outcomes.9–11 36 Not applying local therapy reduces local complications, potential cognitive impairment and might be particularly adequate for patients with a low number of asymptomatic MBM. This question along with the best sequence regarding local therapy is being addressed in ongoing clinical trials, and in the future, we might be better equipped to decide which patients to treat with the different modalities.37 38

In this study, there was a high proportion of patients with BRAFV600-mutated melanoma (63%), but similar to other publications where this subgroup represents between 52% and 65% of the patients.14 15 26 28 Previously, it has been postulated that even in patients with BRAFV600-mutated MBM, first-line systemic treatment should consist of combined immunotherapy. Our analysis showed that there was no difference in OS of patients receiving first-line NIVO+IPI or first-line targeted therapy followed by combined immunotherapy (p=0.085). The two subgroups did not differ significantly (online supplementary table S4), except for the number of MBM, where a higher proportion of patients with >3 MBM received first-line targeted therapy (p=0.002). Our results in this subgroup need to be interpreted with caution since we have not included patients with BRAFV600 mutation who only received targeted therapy.

In the multivariate Cox regression analysis, we identified LDH, S-100B, ECOG-PS and number of MBM as independent prognostic factors. These prognostic factors have already been described in previous analyses,8 14 39–41 but to the best of our knowledge, S100B has only been described as independent prognostic factor for checkpoint inhibitor immunotherapy in one monocentric study.42 It is interesting, however, that both tumor markers, LDH and S100B, remained independent prognostic factors in the multivariate analysis, suggesting that these non-invasive and easy to determine blood parameters can and should be used early in the course of the disease to inform about patients’ prognosis.

Regarding the presence of symptomatic MBM, there was no OS differences between patients with and without symptoms (p=0.065), but a trend can be seem showing that patients with symptomatic MBM have worse prognosis that those who are asymptomatic (1-year OS rate 46% and 63%, respectively). In other prospective studies investigating similar cohorts, the OS rate ranged from 66% at 6 months43 to 31% at 12 months.16 Unfortunately, information regarding the presence of symptomatic MBM is missing in approximately 50% of the patients in our study, and therefore, definitive conclusions cannot be drawn from our data.

Strengths of this investigation are that data from 23 German-certified skin cancer centers with high standards for data quality were included. Three-hundred and eighty patients were analyzed which is thus far the largest published cohort of patients with MBM managed in a routine clinical setting. This high number of patients allowed us to perform subgroup analyses, with results of reasonable sensitivity. Furthermore, this study provides long-term follow-up data of patients with MBM covering a period of up to 18 months.

The study limitations are related to its retrospective design. Patients were included regardless of previous systemic and local therapies prior to the combined immunotherapy and thus some heterogeneity of the study population might have contributed to differences in survival outcomes observed in our cohort. The decision to offer local therapy or not was probably influenced by the number and size of MBM. Additionally, the maximum number of MBM considered to be treated individually by STR/surgery might vary between different centers. We have not evaluated intracranial toxicities. However, this aspect might have been considered when planning local therapy and targeted therapy in patients with BRAFV600-mutated melanoma, influencing the systemic therapy offered as well as the therapy sequence in this subgroup.

In conclusion, our study shows that treatment with NIVO+IPI, particularly in combination with STR/surgery improves survival of patients with MBM. Results presented herein also suggest that local therapy with STR/surgery either before or after starting combined immunotherapy might be advantageous to prolonging OS.