Clinical determinants of long-term survival in metastatic uveal melanoma

Uveal melanoma (UM) is the most common tumor of the eye in adults but still represents a rare subtype of melanoma. The mean age-adjusted incidence is 5.2 per million [1] and shows a north-to-south decreasing gradient in Europe [2], with the highest current mean incidence of 9.5 per million in Ireland [3]. The treatment approach of the primary tumor depends on the tumor size, patient preference, and tumor localization, most commonly by brachytherapy or enucleation [4, 5]. The most important risk factor for the development of metastases is monosomy 3, and at least 40–50% of patients will develop metastases, predominantly to the liver [6, 7]. Since there is no standardized and effective treatment for advanced UM, the prognosis is bleak once metastases develop [8]. A meta-analysis of studies published between 1980 and 2017 including 2494 patients calculated a median overall survival (OS) across all treatment modalities of 1.07 years [9]. However, the population of this meta-analysis was treated mainly in the time before immune checkpoint blockade (ICB) was available. In both settings, some patients show a more favorable disease course with longer OS. Thus, this study aimed to identify parameters that are linked to a more favorable clinical outcome in patients with metastatic UM.

Considering that there are no standardized treatment options and only a few high-quality trials for metastatic UM, data from a real-world setting and registries are of particular value for this orphan cancer condition. In this study, we followed a two-step strategy to identify clinical parameters that are associated with a favorable outcome. First, we assembled a population from multiple centers within Germany and registry data and performed univariate analyses to explore possible prognostic factors. Second, we further validated these parameters after the population was divided into two cohorts with survival of less and more than 2 years as a proxy for short-term and long-term survival, respectively. Thus, patients with an unknown death status or ongoing treatment within two years after entry into stage IV were excluded. This selection of the population may explain why the OS was relatively high compared to previous studies [10‐12]. The median OS for the entire population was 22.3 months which is by far longer than the proposed benchmark of 10.2 months in a recent meta-analysis of 912 patients from clinical trials but comparable to monocentric data other sites [13, 14]. Furthermore, this difference could be explained by a higher proportion of patients treated with ICB. The benchmark study included patients from numerous clinical trials before ICB was available. Accordingly, the proportion of patients receiving immunotherapy was also only 15%, while 34% were still treated with chemotherapy [14]. In our study, a large proportion received ICB at least once in stage IV disease. The PR rate to single PD1 inhibition and combined ICB were 10% and 21%, respectively. These results are slightly higher than in two recently published phase II trials [15, 16]. Piulats et al. achieved a PR rate of 10% in 52 treatment-naïve patients in the Spanish GEM-1402 trial. A higher PR rate of 15% was reported by Pelster et al. in 33 patients with any prior treatment for UM. One CR occurred in each of these two prospective studies. Similarly, Pelster et al. reported more favorable OS and PFS values than in the Spanish trial (5.5 vs. 3 months and 19.1 vs. 12.7 months, respectively). These data are in line with our results from a real-world setting and clearly beat the benchmark values [17]. Although the response rates to ICB are by far lower than in cutaneous melanoma, we conclude that the overall prognosis of metastatic UM has been improved significantly by the introduction of ICB.

Our results confirm that a good ECOG performance status and normal serum LDH levels are important prognostic factors in stage IV UM, as was outlined by us and others [11] [18]. Male sex and advanced age were not correlated with survival in contrast to the benchmark data. Interestingly, the number of affected organs was significantly higher in cohort B, while patients with liver metastases only were more commonly observed in cohort A. Also, patients with liver metastases only showed significantly worse survival compared to those with a combination of both hepatic and extrahepatic metastases. This distribution is intriguing and was also observed in the two recent phase II studies. Piulats et al. reported that the OS in patients with exclusive liver metastases was shorter than that in patients with metastases in other locations beyond the liver (9.2 vs. 23.5 months) and those with both liver and other metastases (15.5 months) [15, 16]. Pelster et al. achieved a response in 6 patients of that 5 had both liver and extrahepatic metastases. These results imply that exclusive hepatic metastases are a major unfavorable prognostic factor. The liver exerts a particularly immunosuppressive effect, and the hepatic tumor microenvironment may prevent antigen recognition by tumor-infiltrating lymphocytes, even when treated with ICB [19]. However, if extrahepatic metastases are present, this immunosuppressive effect can be bypassed and the resistance to ICB can be overcome more easily. This hypothesis can help to explain the feeble immune response in UM as 95% of the metastases spread to the liver which is, in general, the least responsive metastatic site to ICB [6, 20]. Also, 93% of the patients included in our study had liver metastases with no significant difference between the cohorts (96% vs 89%), whereas the long-term survivors had significantly more frequently pulmonary and “other” metastases. Besides, the number of patients in cohort B with exclusive liver metastases was substantially lower (20 vs. 5 patients). It is noteworthy that the patients with liver metastases only had shorter OS compared to those with several metastatic sites (7.7 vs. 24.8 months) although the overall tumor burden is likely to be higher in the latter group. Consequently, rapidly progressive liver metastases were, counterintuitively, associated with a worse overall survival, while patients with extrahepatic metastasis had a more favorable disease course. We conclude that the presence of multiple affected organ systems with extrahepatic disease is associated with either a more indolent disease course per se or results in better treatment responses. These data support another interesting hypothesis, i.e., that liver metastases cause the elimination of CD8⁺ T-cells via immunosuppressive effects of macrophages [21]. This elimination of T-cells may lead to a poor response to ICB in patients with liver metastases only. Besides, liver metastases of UM have a significantly lower PD-L1 expression compared to metastases of cutaneous melanoma, which might also be a reason for the poor response [22‐24]. Interestingly, the absence of melanin in liver metastases of UM induces stronger immune responses and is associated with an improved ICB response [25]. However, the hepatic tumor microenvironment might be influenced positively by liver-directed treatments. In our population, patients receiving liver-directed treatments showed significant survival advantages. It remains unclear whether this benefit is truly achieved by liver-directed treatment or whether patients with a more isolated disease are selected for this procedure that per se have a better prognosis. Nonetheless, these interesting signals observed by us and others should be further investigated in future studies.

The limitations of our study are its retrospective design and the exclusion of patients with an unknown death status within 2 years after entry into stage IV that may positively bias the survival. Also, the genotype has proven influence on the prognosis but was not considered here due to missing data as no data on the genetic background of the tumors were available.

To date, we have some information on the impact of genetic alterations (1) on development and the pattern of metastasis formation, (2) on the disease course once metastasis has developed, and (3) on the treatment response, in particular to ICB. UM is genetically characterized by chromosomal losses and gains as well as a low mutational burden. Chromosomal alterations like monosomy 3, disomy 3, and gain of chromosome 8q are considered strong prognostic factors. Monosomy 3 is associated with poor disease-free survival (DFS), while tumors with disomy 3 tend to spread rarely and have prolonged DFS. Gain of chromosome 8q and loss of chromosome 3 are associated with a poor OS [26, 27]. Somatic GNAQ and GNA11 mutations are commonly detected leading to consecutive activation of the MAPK pathway which might contribute to the development of UM [28]. However, survival seems not to be affected by these mutations [29]. Further, hemizygous mutations in the BAP-1 gene were found in monosomy 3 tumors resulting in loss or dysfunctional BAP-1 expression associated with poor DFS and OS [30‐32]. Szalai et al. demonstrated a strong association between BAP-1 mutated tumors and a peak of clinical detected metastases 2 years after enucleation [33]. They suggested that this might be an explanation for the observations of the disproved Zimmerman–McLean–Forster effect, who hypothesized that the enucleation causes a spread of tumor cells leading to the peak of metastases [33, 34]. However, BAP-1 mutated tumors might correlate to locally fast progressive liver metastasis observed in our study population. Moreover, SF3B1 and EIF1AX mutations occur in tumors with disomy 3 and are associated with a more favorable disease course [32, 35, 36]. SF3B1 mutations correlate with a long DFS and with a late peak of metastases after 7 years [30, 33]. EIF1AX mutations as well as wild-type BAP-1, SF3B1, and EIF1AX genes are correlated with prolonged survival [30]. Thus, the genotype plays an important role in determining the overall prognosis regardless of treatments. Further, there are signals that a high mutational burden is predictive for response to ICB [37] and a MBD4-related hypermutator phenotype in metastatic uveal melanoma showed an exceptional high sensitivity to anti-PD-1 which is present in up to 2% of UM patients [38].

Altogether, the prognosis of metastatic UM in this population was better than the recently proposed benchmark and may have improved by the more frequent use of ICB. We demonstrate that a good response to ICB, extrahepatic disease, and liver-directed treatments, were associated with long-term survival.