Systemic treatment of metastatic uveal melanoma: review of literature and future perspectives

Up to 50% of patients with uveal melanoma develop metastatic disease with poor prognosis. Regional, mainly liver-directed, therapies may induce limited tumor responses but do not improve overall survival. Response rates of metastatic uveal melanoma (MUM) to systemic chemotherapy are poor. Insights into the molecular biology of MUM recently led to investigation of new drugs. In this study, to compare response rates of systemic treatment for MUM we searched Pubmed/Web of Knowledge databases and ASCO website (1980–2013) for “metastatic/uveal/melanoma” and “melanoma/eye.” Forty studies (one case series, three phase I, five pilot, 22 nonrandomized, and two randomized phase II, one randomized phase III study, data of three expanded access programs, three retrospective studies) with 841 evaluable patients were included in the numeric outcome analysis. Complete or partial remissions were observed in 39/841 patients (overall response rate [ORR] 4.6%; 95% confidence intervals [CI] 3.3–6.3%), no responses were observed in 22/40 studies. Progression-free survival ranged from 1.8 to 7.2, median overall survival from 5.2 to 19.0 months as reported in 21/40 and 26/40 studies, respectively. Best responses were seen for chemoimmunotherapy (ORR 10.3%; 95% CI 4.8–18.7%) though mainly in first-line patients. Immunotherapy with ipilimumab, antiangiogenetic approaches, and kinase inhibitors have not yet proven to be superior to chemotherapy. MEK inhibitors are currently investigated in a phase II trial with promising preliminary data. Despite new insights into genetic and molecular background of MUM, satisfying systemic treatment approaches are currently lacking. Study results of innovative treatment strategies are urgently awaited. Forty clinical studies on metastatic uveal melanoma were reviewed regarding responses to systemic treatments. New insights into genetic and molecular background led to investigation of new substances but promising in vitro data have not yet been translated into satisfying treatment responses; however, preliminary results of ongoing studies are highly encouraging.


Introduction
Ocular melanoma accounts for 3% of all melanoma cases [1]. Uveal melanoma (UM) is the most common primary intraocular tumor with an incidence of approximately five cases per million individuals [1]. Up to 50% of patients develop metastatic disease with spread of tumor cells to liver (89%), lung (29%), bone (17%), and other organs [1,2]. At this stage UM has a poor prognosis with median overall survival (OS) of 4-15 months [3]. Survival rates in metastatic UM (MUM) have remained almost unchanged in the past 40 years [1].
As far as MUM is restricted to a limited anatomic region, locoregional treatment modalities can be used to control disease, for example, surgical resection, intraarterial chemotherapy, transarterial percutaneous chemoembolization, selective internal radiation therapy, and radiofrequency ablation [4]. Patients in whom surgical resection is feasible show longer OS [5]. Liver-directed therapies may induce remission of single metastases but do not prolong OS [4].
MUM is frequently treated with chemotherapeutics like dacarbazine, fotemustine, or gemcitabine/treosulfan although evidence for these regimens is limited. In clinical practice, responses are rarely seen and the impact of systemic chemotherapy on patients' survival is questionable [3].
Our understanding of molecular genetics and intracellular signaling pathways involved in the pathogenesis of

Cancer Medicine
Open Access MUM has improved over the last decades [6] resulting in the current investigation of targeted therapy approaches. We here review the present status of systemic treatment of MUM and evaluate therapy outcome measured by overall response rate (ORR) (IBM, Ehningen, Germany).

Methods
PubMed search was performed for "metastatic" [and] "uveal" [and] "melanoma" as well as for "melanoma" [and] "eye" [and] "treatment" on 16 May 2013 for the time period between 1980 and May 2013. "Web of Knowledge" and congress abstract search via the American Society of Clinical Oncology homepage was performed (data cut 22 May 2013). The ClinicalTrials.gov website was searched for terms "melanoma" and "eye" on 13 May 2013. All retrieved study summaries were screened and compared to published data.
All titles and abstracts in English language were screened for relevant content by the first author (K. B.).
The selection process was documented according to PRIS-MA criteria (Fig. 1) [7]. Studies on in vitro data, diagnostics, treatment of the primary tumor, single case reports, and clinical trials on locoregional treatment modalities were excluded. Full text versions of all relevant articles in English language were obtained and their references reviewed for additional relevant reports. Studies with less than four MUM patients, ecological design, without description of objective response assessment and studies not reporting ORR were excluded from meta-analysis (Fig. 1). All remaining studies were reviewed for quality aspects including study design, patient population, histological confirmation of disease, and method of staging/ outcome evaluation by first and second author (K. B., A. G.). Patients treated in higher than first-line situations were classified as "non-first-line." Studies were grouped by type of treatment into singleagent or combination chemotherapies, chemoimmunotherapies, immunotherapies, antiangiogenetic therapies, and treatment with kinase inhibitors. In each group, rates of complete (CR) or partial remission (PR) and their exact 95% confidence intervals (95% CI) were computed for each study and overall for the group. In addition, homogeneity of ORR was examined by the exact chisquared test. In case that homogeneity was rejected, the ORR was computed again, excluding the outlier study that caused heterogeneity. An overall summary analysis was carried out equally for all types of treatment. Statistical analysis was performed using SPSS statistics program version 21.0 (IBM, Ehningen, Germany).

Results
The selection process is outlined in Figure 1. Of 59 retrieved articles including 11 congress abstracts, four were excluded because of small patient numbers (n < 4) [8][9][10] or ecological design [2]. Nine were excluded because ORR was not reported [11][12][13][14][15] or mixture study design did not permit separate analysis of MUM data [16][17][18][19]. Forty-six studies were included in review discussion, that is, one case series, five pilot, three phase I, one phase I/II, 29 phase II, and one phase III study, data from three expanded access programs, and four retrospective data analyses. In six of the studies response criteria were insufficiently described. The first authors of these reports were contacted by e-mail to comment on response criteria. In case of authors' response studies were included in numeric analysis [20][21][22] while studies for which response criteria could not be elucidated were excluded [23][24][25]. Two publications were excluded because of presentationdriven interim analyses of incomplete clinical trials, one of them reported within a review publication [26,27] (NCT00338130, NCT01143402). One publication was excluded from numeric analysis because the drug could not be assigned to one treatment subgroup [28].
The numeric analysis included 40 publications with a total of 841 patients (Table 1). Patient numbers ranged from four in a pilot study [29] to 171 patients in a randomized multicenter study [30]. Histological confirmation of metastatic disease was reported in 19/40 studies. Immunohistochemical stains of c-kit were performed in one study [31], mutational analysis of c-kit in another study [32] and GNAQ sequencing in a limited number of patients in two studies [33,34]. Mean patients' age was 59 years; 546 patients were treated in first-line situation whereas 229 patients had received prior therapies including surgery, liver-directed treatment, chemotherapy, or immunotherapy. Response was evaluated according to WHO response criteria of 1979 [35] in 12 and according to RECIST 1.0/1.1 [36] in 27 studies.
Response, including CR and PR, was achieved in 39 of 841 patients; ORR was 4.6% (95% CI 3.3-6.3%). No responses were observed in 22/40 studies. Stable (SD) versus progressive disease (PD) was reported over all studies for 184 versus 379 patients (ratio 1:2) while nine studies did not provide information on SD/PD numbers. Median OS was reported in 26/40 studies ranging from 5.2 months in pretreated, predominantly end-stage patients [37] to 19.0 months in selected first-line patients [38]. Progression-free survival (PFS) was reported in 21/40 studies ranging from 1.8 to 7.1 months.

Future perspectives
Advances in knowledge about genetics and signaling pathways led to initiation of clinical trials with innovative therapeutics. Screening the ClinicalTrials.gov website for ongoing clinical trials on MUM revealed 15 studies, two of them with randomized design ( Table 2). Only two of the phase II studies evaluate chemotherapies (albuminbound paclitaxel [25], liposomal vincristine [29]).
The occurrence of UMs in an immunologically privileged site makes immunotherapy a promising treatment approach. Current data on ipilimumab were gained from retrospective analyses only. One was published at the time of manuscript revision and showed, in line with the previously published studies, an ORR of 5.1% (2/39); SD was observed in 44% (week 12) and 25% (week 23) of patients [64]. Anti-CTLA4 antibodies are further assessed in three prospective trials. While two of them are expected to report outcomes soon, another large trial on ipilimumab will not be finished before 2017 (Table 2).  PD-1 and PD-L1 have become important targets in cutaneous melanoma. To our knowledge, MUM patients have not been included in trials with PD-1 or PD-L1 antibodies yet. However, as PD-L1 expression is found in MUM cells [65] and probably in the tumor environment further investigation of treatment strategies targeting PD-1/PD-L1 in MUM are warranted. Activating somatic mutations in GNAQ/GNA11, two members of the guanine nucleotide-binding protein family (G-proteins), were found in 83% of UMs [66]. Both mutations result in the constitutive activation of the mitogen-activated protein kinase (MAPK) pathway thereby inducing proliferation in the absence of external growth stimuli [67]. Hence, blocking this pathway by specific inhibitors may be an effective therapeutic approach for MUM [68][69][70]. Several kinase inhibitors are currently studied in five phase I/II and II studies. A phase II study presently conducted in the US compares selumetinib versus temozolomide/DTIC with a much noticed interim analysis on PFS; ORR was 15% in the selumetinib-group (7/46) compared to 0% in the temozolomide-group (0/46) and 0% in the cross-over group [26]. However, tumor regression without reaching RECIST-defined response was seen in 50% in the selumetinib-group versus 11% in the temozolomide group and 23% in the crossover group. PFS in week 16 was 43.1% for selumetinib versus 8.5% for temozolomide. Interestingly, responses were also seen in GNAQ/GNA11 (Q209, exon 5) wildtype patients. However, retrospective assessment of codon R183, exon 4 revealed a mutation in the patient with objective response according to RECIST. These promising but preliminary data on MEK inhibition had to be excluded from our numeric analysis as final study outcomes should be awaited [26,27].
GNAQ/GNA11 signaling induces activation of phospholipase C (PLC) and protein kinase C (PKC) further downstream of PLC with subsequent MAPK pathway activation [71]. There are two trials under way investigating PKC inhibition alone and in combination with MEK inhibition. GNAQ/GNA11 signaling is also linked to the PI3K-AKT pathway in UM, usually in an activating manner resulting in increased cell proliferation and survival [71]. Hence inhibition of PI3K or AKT, possibly in combination with MAPK pathway inhibition, appears to be another attractive treatment strategy.
On the basis of promising data on the multikinase inhibitor sorafenib in small case series, a placebo-controlled phase II study is currently conducted in Germany investigating sorafenib versus placebo. Preliminary data on cabozantinib, a c-Met/VEGFR2 inhibitor currently under investigation [12], prompted investigators to initiate a randomized phase II study on cabozantinib versus dacarbacine or temozolomide. Search on the ClinicalTri- als.gov website additionally revealed results on one terminated, yet unpublished study analyzing the combination of sunitinib/lenalidomide/cyclophosphamide, which showed no response in 12 patients (NCT00482911). Mutations in BAP1, a deubiquitinating enzyme located on chromosome 3p, are seen in 85% of high-risk ("class-2") UMs and correlate with development of metastatic disease [72]. One substrate of BAP1 is histone H2A; histone-deacetylase inhibitors were shown to reverse the H2A hyperubiquitination caused by BAP1 knock-down in vitro [73] and might therefore be a therapeutic strategy [74]. The histone-deacetylase inhibitor vorinostat is currently studied in MUM.
Antiapoptotic bcl-2, which is (over)expressed in more than 95% of UMs [72], provides another potential target. The bcl-2-antisense oligonucleotide oblimersen is currently under investigation. Upregulation of insulin-like growth factor (IGF)-1 and IGF-1R receptor in UM [72] potentially offers the possibility of treatment with the anti-IGF-1R-antibody cixutumumab. Further compounds currently under investigation in phase II studies include the HSP90 inhibitor ganetespib, and the somatostatin receptor analog pasireotide in combination with everolimus. Other treatment approaches such as targeting of somatostatin receptors by octreotid [24] and a phase I/II study on pegylated arginine were disappointing [28].
Altogether, immunotherapeutics and kinase inhibitors are currently the most investigated agents with encouraging interim results on MEK inhibition.

Discussion
Depending on the genetic signatures of the primary tumor [6], up to 50% of UM patients develop metastatic disease. Once metastases occur prognosis is bad and therapeutic options are limited with ORR being considerably low.
The only randomized controlled phase III trial on treatment of MUM (intravenous vs. intraarterial fotemustine) showed improved ORR of liver metastases and prolonged PFS in intraarterially treated patients but similar OS in both groups [30]. Response to intravenous fotemustine was as low as 2.4%. Only two phase II studies have up to now been published that were designed as randomized trials with two subgroups [22,44]. One of these showed 8% ORR in the dacarbazine group. Relatively high ORRs reported for single-agent or combination chemotherapy in small studies [29] are possibly due to selection bias in small patient numbers. Phase II trials on liposomal vincristine and albumin-bound paclitaxel are ongoing but uncertain to reproduce promising results of previous smaller studies [29,40].
In our pooled analysis, chemoimmunotherapy shows slightly better tumor responses than chemotherapy. This observation has to be interpreted with caution as our analysis considered first-line and higher line studies as well as studies that did not differentiate the outcome of first-and second-line treated patients. Better OS in the chemoimmunotherapy studies might thus partially be due to a first-line treatment situation in the majority of trials.
New insights into tumor biology led to investigation of immunotherapies, antiangiogenetic agents, and targeted therapies. While ipilimumab is effective in metastatic cutaneous melanoma [75], it did not yet appear to be superior to chemotherapy regimens in MUM, possibly because published data have mainly been generated from expanded access programs in non-first-line situations. However, OS of 5.2-10.3 months in pretreated patients might still be promising [37,54,56]. Final conclusions can only be drawn from randomized studies, preferably in first-line situations, which are still lacking.
Although VEGF plays a major role in MUM [6], treatment regimens focusing on antiangiogenetic agents did not reveal responses in first-line treatment. Nevertheless, pooled OS of 12.7 months appear promising.
Kinase inhibitors including sorafenib, sunitinib, and imatinib did not show any responses in six of nine studies. Promising results from a small case series on sorafenib combined with fotemustine [20] led to initiation of a large phase II study of sorafenib the results of which are still pending. The ORR for sunitinib was 5% in a pilot trial [63], which, however, could not be confirmed in a subsequent randomized phase II study [22].
GNAQ/GNA11 mutations in over 80% of MUM leading to aberrant activation of the MAPK pathway especially makes MEK an attractive therapeutic target [6]. Patients recruited in phase I/II studies, however, did not show objective responses upon MEK inhibitor treatment [33,34]. Falchook et al. [33] did not observe a correlation between the mutational status of GNAQ/GNA11 and clinical response to MEK inhibition but the analyzed exons were not specified in the publication. A phase II study is currently conducted on selumetinib with a promising interim analysis but pending final results [26]. According to preliminary data, there is no proven correlation of ORR or PFS with GNAQ/GNA11 mutational status. OS was not significantly improved compared to chemotherapy.
Increasing insight into the biology of MUM has not yet translated into higher ORR. Unexpectedly, a correlation of treatment response to mutational/expression status of molecular targets has not been found in smaller trials [32][33][34] and ongoing clinical studies [26]. So far, there is no evidence of a clinically meaningful survival benefit due to novel targeted agents.
With respect to appallingly low ORR, the question is whether disease stabilizations are treatment related or simply reflecting the natural course of disease [13]. None of the currently available therapies has shown prolongation of patients' OS. Survival data were reported in 75% of the analyzed studies but cannot be compared due to inhomogeneous patients' characteristics throughout the studies. Only 7/40 publications reported the lengths of metastases-free intervals as primary diagnosis of UM and, if reported, a wide range was seen within and among these studies (0-25 years) [39,49,50,53,63,76,77]. As metastases may develop 10 or more years after primary UM, this "dormancy" phenomenon has a high impact on patients' prognosis [78,79]. Moreover, other prognostic parameters such as lactate dehydrogenase, sites of metastases, and patients' performance status would need to be equally distributed in the studies to allow comparison of survival data.
According to available study data, survival appears to depend on patient-and tumor-related characteristics rather than on the actual treatment [3]; it therefore can only be analyzed in randomized studies recruiting patients with comparable characteristics. Given a poor response rate in most of the studies, determining PFS at a certain time point might be a more suitable endpoint. This would require defined staging intervals, which, however, were inhomogeneous throughout the analyzed studies here and therefore not considered in this review.

Conclusion
This review analyzes data of studies on systemic treatment of MUM published between January 1980 and May 2013. Altogether, published data mainly provided low-level evidence. The limited efficacy of current treatment approaches illustrates the high medical need for more effective treatment options in MUM.
To date, no chemotherapeutic, immunotherapeutic, or targeted drug has shown reproducible ORR >10% or prolonged OS in MUM. Targeted therapeutics as well as immunotherapies might be promising strategies, but need evaluation in prospective trials. Investigation of chemoimmunotherapy-based strategies appeared to become less important, probably due to toxicity profiles although ORR has been superior to all other therapeutic approaches. Most promising preliminary data are available for MEK inhibition. However, these therapeutic regimens should be judged after final data analyses become available. A future goal should be careful design of randomized clinical trials.

Conflict of Interest
K. Buder received educational/travel grants and honoraria for oral presentations from TEVA GmbH, Roche Pharma, and Bristol-Myers Squibb. A. Gesierich received travel grants for congress participation, and was an advisory board member for Bristol-Myers Squibb and Roche Pharma. No conflicts of interest declared for G. Gelbrich. M. Goebeler was an advisory board member for MSD SHARP and DOHME GmbH.