Uveal Melanoma Metastatic to the Liver: Treatment Trends and Outcomes

Uveal melanoma is the most common intraocular malignant tumor in adults [1]. Although local tumor control treatments have been highly effective with brachytherapy and enucleation, up to 50% of patients develop distant metastasis [2]. Five-year cumulative rate of metastasis is approximately 25%, with the liver being the most common site of metastasis (89% of sites) [3]. Median overall survival (OS) in patients with uveal melanoma liver metastasis ranges between 14 and 28 months by recent reports, depending on the method of treatment [4-6].

Treatment options for metastatic liver lesions include systemic therapy, and local therapy to the liver [7]. No therapy has been consistently proven to be superior to others in the treatment of metastatic liver lesions [8, 9]. Systemic chemotherapy has generally shown poor response with high toxicity rates [10, 11]. A retrospective study demonstrated that local therapy of the liver tumor had a significant effect on survival (26 months) compared to those who received systemic therapy (9.1 months) [7]. In a study comparing 16 patients who underwent laparoscopic resection or RFA compared to 28 who had systemic treatment, the OS of the surgery group was significantly longer than the systemic group (35 vs. 15 months) [12]. However, not all patients are candidates for local therapy due to their overall performance status or diffuse disease burden. The advent of immunotherapy in recent years has provided promising results in the survival of cancer patients, including those with cutaneous melanoma [13, 14]. However, it is unclear if this benefit clearly translates to improved survival in uveal melanoma patients with liver metastasis [2].

Our study investigated uveal melanoma liver metastasis treatment in a single-center tertiary referral center. Treatment trends and outcomes of different treatment modalities, including immunotherapy and local therapy, are reported.

A retrospective case series of patients with uveal melanoma metastatic to the liver was performed through chart review. The study was approved by the IRB of Cleveland Clinic. All patients were managed at Cleveland Clinic, Cleveland, OH, and were diagnosed with liver metastasis between 2004 and 2016. First-line treatment was clarified as no treatment, systemic therapy, not including checkpoint inhibitors (carboplatin with paclitaxel, dacarbazine, temozolomide, sunitinib, and interferon with stibogluconate), checkpoint inhibitor immunotherapy (ipilimumab, pembrolizumab and nivolumab), or local therapy (surgical resection, radiofrequency ablation, radioembolization, chemoembolization, or isolated hepatic perfusion). Patients were excluded if there was insufficient information in the chart to determine their treatment or follow-up.

Baseline characteristics were collected, including gender, age at diagnosis of uveal melanoma, type and treatment of uveal melanoma. The time from uveal melanoma diagnosis to liver metastasis diagnosis was collected as well as whether the metastasis was detected by symptoms, or by imaging, baseline staging or laboratory studies in asymptomatic patients. Presence of extrahepatic metastasis, number of liver lesions, and lactate dehydrogenase (LDH) levels were also recorded [10]. Time periods for trend analysis were divided between 2004–2011 (pre-checkpoint inhibitors) and 2012–2016 (post-checkpoint inhibitors).

Treatment type and date(s) of treatment for the metastasis were recorded. Time to progression from first treatment (as noted by progression on imaging and/or clinical assessment) was recorded as progression-free survival (PFS). PFS was not calculated for patients who had no treatment since they did not have follow up imaging. Date of death was obtained through chart review and/or http://www.ancestry.com (which links through Social Security Death Index, death certificates, and obituaries).

Time period and initial treatment groups were compared on categorical predictors using Pearson χ² tests or Kruskal-Wallis tests. PFS and OS were summarized using Kaplan-Meier estimation and compared between groups using Cox proportional hazards models. PFS was plotted out to 1 year because fewer than 10% of patients were actively followed after that. OS is plotted out to 2 years because fewer than 10% of patients were actively followed after that. Results of these models are presented as hazard ratios. An additional analysis using Cox proportional hazards models with repeat treatment as a time-dependent covariate was performed. Results of these models are presented as hazard ratios. Univariable and multivariable Cox proportional hazards models were fit using the six variables: age at diagnosis, gender, time from uveal melanoma diagnosis to metastasis, method of detection, presence of extrahepatic metastasis, number of liver lesions > 1. Multicollinearity among predictors was evaluated using variance inflation factors and condition indices. No multicollinearity was observed. To assess the impact of repeated treatment on OS, repeated treatment was included as a time-dependent covariate in the multivariable model, along with initial treatment type. An interaction between these two factors was considered but dropped when found to be nonsignificant. Hazard ratios with 95% confidence limits from Cox models are presented from univariable and multivariable models. A significance level of 0.05 was assumed for all tests except for post hoc pairwise analysis where a Bonferroni corrected significance level was used. Analysis was performed using SAS software (version 9.4; Cary, NC, USA).

Ninety-two patients were diagnosed with liver metastasis from uveal melanoma. Nineteen patients were excluded due to lack of follow-up or a first-line treatment not in the specified group. A total of 73 patients were included for analysis. The median age at time of uveal melanoma diagnosis was 63.0 years (interquartile range 56–70). Thirty-eight (52%) of the patients were male. The most common location of uveal melanoma was choroid (42 patients, 58%), followed by cilio(choroid) (30, 41%) and unknown (1, 1%). First-line treatment of the uveal melanoma included plaque therapy (38 patients; 52%), enucleation (34; 47%), and proton beam radiation (1; 1%).

The median time from uveal melanoma diagnosis to liver metastasis was 27 months (interquartile range 13–46 months). Median PFS and OS for the entire cohort was 4 months (95% CI 3–5 months) and 15 months (95% CI 11–18 months), respectively.

Ten patients were diagnosed after presenting with symptoms (weight loss, abdominal pain, nausea, back pain). Of the 63 that were asymptomatic, 55 were diagnosed through routine imaging surveillance (48 ultrasound, 5 CT, 1 MRI, 1 lab work), 5 were diagnosed at baseline systemic staging, 2 were diagnosed with biopsy-proven liver metastasis from melanoma while undergoing surveillance for another cancer and 1 was diagnosed incidentally. Of the patients who were not diagnosed at time of staging, 57 patients had surveillance every 6 months, 4 patients had surveillance every 3 months, and the surveillance frequency for 7 patients was unknown. Four patients also had lung metastasis found at the time of liver metastasis diagnosis. Median follow-up duration after liver metastasis was 12 months (interquartile range 7–19 months).

There were 42 patients in the 2004–2011 time period and 31 in the 2012–2016 time period (Table 1). Age, gender, LDH value (elevated or normal), number of liver lesions (single or multiple), extrahepatic metastasis at time of diagnosis (present or absent), and follow-up duration were not statistically different between the 2 time periods. Patients were statistically more likely to receive checkpoint inhibitors in the 2012–2016 time period. No statistically significant difference in PFS (Fig. 1a) and OS (Fig. 1b) was observed across the two time periods. The median PFS for 2004–2011 was 3.4 months, while for 2012–2016 it was 4.0 months. The median OS for 2004–2011 was 12.9 months, and for 2012–2016 it was 17.5 months.

First-line treatments included systemic therapy (14 patients), checkpoint inhibitor therapy (18), local treatment (30), and no treatment (11). Age, gender, LDH value (elevated or normal), and number of liver lesions (single or multiple) were not statistically different between the 4 groups (Table 2). Overall differences in extrahepatic metastasis status were observed, but pairwise comparisons were not significant. Over 90% of patients treated with local therapy had no extrahepatic metastasis, while 64% of patients treated with systemic therapy had no extrahepatic metastasis. Follow-up duration was significantly shorter in the no treatment group relative to those with checkpoint inhibitor therapy and local treatment. Comparison of PFS and OS amongst the different treatment groups is presented in Figure 2a, b, respectively. Patients who received first-line systemic therapy had a median PFS of 2 months and median OS of 10.3 months. The checkpoint inhibitor group had a PFS of 3 months and OS of 15.8 months, which was not significantly different than those of systemic therapy (p = 0.14 and 0.25 respective ly). Patients who received local therapy had a PFS of 4.6 months and OS of 18.7 months, which was significantly different than those receiving systemic therapy.

Figure 3 shows the number of treatment modalities used, stratified by initial therapy. In patients who started with checkpoint inhibitor therapy, 66% went on to receive secondary, tertiary, or even quaternary treatment modalities; in those who started with local therapy, 57% went on to receive multiple treatments. Twenty-nine percent of patients who started with systemic therapy went on to receive different treatments. A Cox proportional hazards model with switched treatment modality as a time-dependent covariate was performed to determine the effect of multiple treatments. This analysis was performed on 61 patients after excluding those with no treatment and those with a second treatment but unknown timing of this therapy. The hazard ratio was 1.4 (95% CI 0.76, 2.59) for death in the multiple treatment group; however, this was not statistically significant (p = 0.28). An interaction between initial treatment type and multiple treatment modalities also did not show statistical significance.

Univariate analysis (Table 3) demonstrated that patients who were diagnosed with metastasis by symptoms and those with extrahepatic metastasis had significantly greater risk of mortality (HR 2.72 and 2.28, p < 0.005 and p < 0.033, respectively). However, with multivariable analysis, the only variable that remained a significant risk of mortality was diagnosis by symptoms (HR 2.39 p < 0.017). Of the 10 patients who were diagnosed by symptoms, 2 received systemic therapy first-line, 1 received checkpoint inhibitor therapy, 2 received local treatment, and 5 received no treatment. OS of asymptomatic patients was 16 months compared to 7 months for those who were diagnosed by symptoms (p = 0.005).

A patient-level plot of OS time, stratified by initial treatment, is shown in Figure 4. Except for 6 patients who are all in the local treatment group, no patient lived longer than 3 years. Two of these 6 exceptionally long survivors are still alive, and one of them is still living 129 months (10.75 years) after liver metastasis diagnosis. Three out of the 6 patients were female, and 3 patients were less than 60 years old at time of metastasis diagnosis. Four patients had choroidal melanomas and 2 had ciliochoroidal melanomas. The median time from uveal melanoma diagnosis to liver metastasis diagnosis for those 6 patients was 37.5 months, and they were all asymptomatic at time of diagnosis. Three patients had more than 1 liver lesion at the time of metastasis.

Our study described the trends and outcomes of uveal melanoma liver metastasis treatment in a single tertiary care center across 12 years. Median PFS and OS for the entire cohort was 4 and 15 months, respectively.

In our study, the use of checkpoint inhibitors was the only first-line treatment that significantly changed across time, with increased use in the 2012–2016 time period, which reflects its introduction to the market [15, 16]. Survival outcomes have not improved for patients across time as PFS and OS did not significantly change across the two time periods of 2004–2011 and 2012–2016.

Patients who received first-line local treatment had the longest progression-free and OS, as was shown previously [12]. This could be secondary to overall functional status as we observed differences in extrahepatic metastasis status across first-line treatment groups. Over 90% of patients initially treated with local therapy had no extrahepatic metastasis, while 64% of patients treated with systemic therapy and checkpoint inhibitors had no extrahepatic metastasis. These pairwise comparisons were not significant possibly due to low sample numbers in each of the groups. The longer survival of local therapy groups compared to systemic treatment groups have been documented in previous studies but our OS for local therapy (18 months) is lower than the previously reported survivals of 26 and 35 months [7, 12]. This may be due to differences in treatment practices or the order of multiple therapies since patients who received local therapy first-line may have different characteristics than patients who received local therapy at any time. Also, local therapies are a heterogeneous group with specific treatments proportionally and differently represented in each study.

Those who received first-line checkpoint inhibitor therapy did not have better survival than those who received first-line systemic therapy. This study’s reported PFS for checkpoint inhibitors is comparable to previous findings [15, 16]. However, this study’s OS of 15.8 months for patients treated with checkpoint inhibitors is longer than previous reports of 6–11 months [16-18]. This may be due to the fact that we included patients who received any of the checkpoint inhibitors, as well as combinations of checkpoint inhibitors, while previous studies mainly looked at one of the drugs as monotherapy. The OS in our patients with systemic therapy not including checkpoint inhibitors (10.3 months) is comparable to previously reported OS [8]. Not surprisingly, patients with no treatment had the shortest OS.

Although we investigated PFS and OS across different treatment groups, the groups were stratified based on their first-line treatment, and many patients had multiple treatment modalities. Only for the patients who did not receive treatment could we say that the OS is a true reflection of the no treatment effect. However, using multiple treatment modalities as a time-dependent covariate, we found no statistically significant survival advantage of multiple treatments.

Previous studies with limited number of patients suggest prolonged survival may be associated with age less than 60, longer disease-free interval from primary tumor to metastasis (median 94 months in one study), female sex, choroidal melanomas, and fewer number of liver lesions at diagnosis [11, 19, 20]. A large study of 152 patients with metastatic uveal melanoma from one center and 102 from another center demonstrated that liver involvement, elevated LDH, and WHO performance status ≤1 were associated with worse prognosis, while those with longer disease-free interval from diagnosis of uveal melanoma to metastasis had a survival advantage [21]. Additionally, other studies have validated an updated staging system for metastatic melanoma that includes performance index, largest diameter of largest metastasis, and alkaline phosphatase levels for prognostication [22, 23]. In our multivariable analysis, we did not find significant survival advantage by gender, age, number of liver lesions, presence of extrahepatic metastasis or disease-free interval from primary tumor to metastasis. However, we found that patients who were diagnosed by symptoms did have a greater mortality risk than those who were asymptomatic. Patients who were symptomatic at the time of metastasis diagnosis had shorter OS than those who were asymptomatic, but this statistical significance should be taken cautiously due to the small number of symptomatic patients and possible lead time bias (although almost all patients were screened every 6 months). This result supports the need for screening of asymptomatic patients [24].

We identified 6 patients who had unusually long survival (more than 3 years) after diagnosis of liver metastasis. They were all in the local treatment group. We also have one patient who had experimental immunotherapy (interferon with tremelimumab) as part of a clinical trial as first-line treatment (and was thus excluded from our analysis) and who was diagnosed with liver metastasis 10 years ago and is still alive. There is a previous report of a metastatic uveal melanoma patient treated with ipilimumab who was alive 3 years after treatment [25]. In our group of long-term survivors, we did not find an increased incidence of factors that predicted survival in previous studies or our multivariable analysis. Interestingly, all our patients were treated with first-line local therapy, and one study also showed surgery and intrahepatic chemotherapy were correlated with prolonged survival [11]. Another study of long-term survivors did not specify the first-line therapy but noted 8/9 of their patients received chemotherapy, 6/9 had surgery, and 2/9 had intrahepatic perfusion (classified as local therapy in our cohort) [19]. Further studies are needed to identify and characterize patients who have unusually long survival despite liver metastasis.

The study protocol has been approved by the research institute’s committee on human research.

This study was supported in part by an unrestricted grant from The Research to Prevent Blindness Inc., awarded to the Cole Eye Institute. Ahmad Tarhini, MD: BMS (consultant), Novartis (consultant), Array Biofarma (consultant), NewLink Genetics (consultant), Merck (consultant), HUYA (consultant). Eren Berber, MD: Ethicon (honoraria) and Aesculap (honoraria). Arun D. Singh, MD: Aura Biosciences (stock interest), Isoaid LLC (honoraria), Eckert and Zeigler (honoraria). The other authors have no financial disclosures.

L.T.X.: conception and design of the study, data collection, writing the manuscript, analysis and interpretation, critical revision of the manuscript; P.F.F.: conception and design of the study, analysis and interpretation, critical revision of the manuscript; J.F.B.: analysis and interpretation, statistical expertise, critical revision of the manuscript; M.L.: analysis and interpretation, statistical expertise, critical revision of the manuscript; A.T.: analysis and interpretation, critical revision of the manuscript; E.B.: analysis and interpretation, critical revision of the manuscript; A.D.S.: conception and design of the study, analysis and interpretation, critical revision of the manuscript.