Article Text


MRI in the detection of hepatic metastases from high-risk uveal melanoma: a prospective study in 188 patients
  1. Ernie Marshall1,
  2. Christopher Romaniuk1,
  3. Paula Ghaneh2,
  4. Helen Wong1,
  5. Marie McKay1,
  6. Mona Chopra1,
  7. Sarah E Coupland2,
  8. Bertil E Damato2
  1. 1Clatterbridge Cancer Centre, UK
  2. 2Department of Molecular and Clinical Cancer Medicine, University of Liverpool, Liverpool, UK
  1. Correspondence to Dr Ernie Marshall, Department of Medical Oncology, Clatterbridge Centre for Oncology, Bebington, Wirral CH63 3JY, UK; emarshall{at}


Background/aims To evaluate MRI in the detection of asymptomatic hepatic metastases from uveal melanoma.

Methods A single-arm prospective cohort study.

Participants We enrolled 188 patients whose predicted 5-year mortality from uveal melanoma exceeded 50%. This prognostication was performed by multivariate analysis of clinical stage, histological grade and genetic type, using our online tool, based on Accelerated Failure Time modelling. These high-risk patients underwent a six-monthly assessment, which included history-taking, clinical examination, hepatic MRI (without contrast, unless suspicious lesions were identified) and biochemical liver function tests.

Results Ninety (48%) of the 188 patients developed detectable metastases, a median of 18 months after ocular treatment. Six-monthly MRI-detected metastases before symptoms in 83 (92%) of 90 patients developing systemic disease, with 49% of these having less than five hepatic lesions all measuring less than 2 cm in diameter. Of these 90 patients, 12 (14%) underwent hepatic resection, all surviving for at least a year afterwards.

Conclusions Six-monthly MRI detects metastases from high-risk uveal melanoma before the onset of symptoms, enhancing any opportunities for early treatment of metastatic disease and clinical trial participation. Whether these actually result in prolongation of life, after taking lead-time bias into account, requires further investigation.

Statistics from

Uveal melanoma is fatal in almost 50% of patients.1 The cause of death in such patients is almost invariably metastatic disease, which usually involves the liver. Such disease is apparent in less than 1% of patients when the primary tumour is diagnosed.2

Metastatic death occurs almost exclusively in patients whose tumour shows chromosome 3 loss and/or a class 2 gene expression profile.3 ,4 In such patients, the survival time correlates inversely with clinical stage and with histologic grade of malignancy.5 We have developed an online tool that estimates the survival prognosis of individual patients according to these predictors, also taking age and sex into account.6

There is no convincing evidence that screening for metastases from uveal melanoma prolongs life.7 ,8 The scope of screening has, nevertheless, increased recently, thanks to advances such as partial hepatectomy, radiofrequency ablation, selective internal radiotherapy, intrahepatic chemotherapy and molecular therapies, such as ipilimumab.9–11 Screening also enhances opportunities for evaluating novel treatments.

There is currently no consensus as to which patients with uveal melanoma should be screened for metastatic disease, what investigations should be performed, how often patients should be examined, and for how long.

The aim of this study was to evaluate MRI in detecting aymptomatic hepatic metastases from uveal melanoma.

Patients and methods

Patients were included in this study if treated for uveal melanoma at the Liverpool Ocular Oncology Centre, and if their risk of metastatic death at 5 years exceeded 50%. They were excluded if they were unable or unwilling to attend Clatterbridge Centre for Oncology, where the screening was performed.

The survival probability was estimated according to clinical, histological and genetic predictors, using the Liverpool Uveal Melanoma Prognosticator Online, which deploys Accelerated Failure Time modelling.6 Race was not recorded, as this is not a recognised prognostic factor.

The clinical stage was determined by ophthalmoscopy and ocular ultrasonography, recording largest basal tumour diameter, tumour thickness, ciliary body involvement and extraocular spread.

The histological grade of malignancy was assessed by light microscopy, assessing the melanoma cell cytomorphology (ie, presence or absence of epithelioid cells), using the Periodic Acid Schiff stain to assess extravascular matrix patterns and counting mitoses in 40 high-power fields.

Chromosome 3 loss was identified using fluorescence in situ hybridisation until 2006, when we replaced this with multiplex ligation-dependent probe amplification.12 ,13

Initially, histological and genetic analyses were performed only if tumour tissue became available after enucleation or local resection. Since 2007, we have routinely performed biopsy on tumours treated by radiotherapy, unless the patient declined, or if there were any contraindications (eg, anticoagulation).

Patients estimated to have a similar survival probability to the matched general population were reassured and were not screened for metastasis. Our current protocol is for the ocular oncologist to inform patients of their laboratory results, survival probability and management options. Patients are referred either to an oncology centre with a special interest in uveal melanoma, their local hospital, or the Nurse Ocular Oncology Clinic (NOOC) at the Royal Liverpool University Hospital (RLUH). At RLUH, patients have an MRI scan (without contrast) in the morning and attend the NOOC in the afternoon, when they are informed of the MRI result.14 A health psychologist and an ophthalmologist are on hand. The nurse also takes a history, examines the eye, collects blood for liver function tests, completes a quality-of-life questionnaire and provides counselling.

Patients were invited to participate in this study if their estimated 5-year risk of metastatic death exceeded 50%. Investigations consisted of: history-taking (eg, weight loss, abdominal pain, symptoms inducing visit to general practitioner), clinical examination, biochemical liver function tests and hepatic MRI imaging without contrast. The MRI protocol incorporated axial T2-weighted turbo spin echo, T1-weighted gradient echo and axial stir images with 8 mm slice thickness, 2 mm gap and 400 mm field of view. These were performed six-monthly, unless highly suspicious lesions were identified, in which case further characterisation was sought with contrast enhancement. Equivocal abnormalities were investigated by repeating the MRI without contrast 3 months later. Screening was continued until metastases were detected, and for a minimum of 5 years.

Patients were referred to the hepatobiliary multidisciplinary team for consideration of partial hepatectomy if the MRI scan suggested resectable disease (ie, limited tumour burden with not more than four deposits). Inoperable patients were enrolled in a clinical trial, or offered palliative chemotherapy.

The study followed the tenets of the Declaration of Helsinki, with informed patient consent and ethical committee approval (Wirral local research ethics committee reference number 99/01).


Baseline characteristics

Between January 2000 and November 2010, 279 patients with high-risk uveal melanoma were referred for screening. Of these, 188 (67%) underwent at least one hepatic MRI scan. The main reasons for not screening at our hospital included long travel time in 48 (53%), preference for care close to home in 15 (17%), claustrophobia/reluctance to have MRI in 7 (8%), comorbidity in 4 (4%) and other/uncertain in 17 (19%) patients. Seventy-three percent of patients lived more than 50 miles (80 km) from our centre.

The 188 patients (86 male, 102 female) whom we screened had a median age of 63 years (range 24–84). The median tumour diameter and thickness were 16.6 mm (range, 5.5–23.6) and 8.0 (range, 1.6–17.5), respectively. The tumour extended anterior to ora in 105 (56%) patients, and involved the optic disc in 31 (17%), with 41 (22%) spreading extraocularly. Histology showed epithelioid cells in 148 (79%) tumours, closed loops in 95 (51%) and a mitotic count exceeding 7 per 40 high-power fields in 48 (26%). Chromosome 3 loss was detected in 172 (92%) patients. The ocular treatment consisted of enucleation in 140 (74%), proton beam radiotherapy in 16 (9%), brachytherapy in 16 (9%), scleral resection in 15 (8%), endoresection in 1 (0.5%) and photodynamic (soon followed by proton beam radiotherapy) in 1 (0.5%) patients.


The 188 patients had a median follow-up of 28.8 months (range, 1–118 months). Four presented with detectable hepatic metastases at the time of the first MRI screening.

Detection of metastases

The actuarial survival rates at 3, 5 and 8 years were 47.8% (95% CI 39.2 to 56.4), 34.8% (95% CI 26.0 to 43.6) and 24.5% (95% CI 9.0 to 40.0) (figure 1). Screening detected metastases in 90/188 patients (48%) (figure 2) (table 1).

Table 1

Number and size of detected metastases

Figure 1

Kaplan–Meier survival curve showing relapse-free survival in all 188 patients enrolled in the study.

Figure 2

Flow diagram showing outcomes in patients enrolled into the study.

Of these, 66% had metastases measuring less than 2 cm, and 68% had less than 5 lesions detected, with 44 (49%) having both these features. There were 5 patients (6%) having at least 1 metastasis exceeding 10 cm in diameter, and 10% with more than 10 metastases, with 6% showing both these adverse features. Of these 90 patients, 87 were known to have a melanoma with chromosome 3 loss, the other three being included because of histological and clinical risk factors. At the time of detection, these metastases were confined to the liver in 82 (91%) patients, with extrahepatic metastases being found in 8 patients, either alone, in 2 (2%) patients, or together with hepatic lesions in 6 (7%). These extrahepatic metastases were located in skin in 3 patients, lung in 2 and bone in 3. The mode of detection of metastatic disease was MRI in 83/90 (92%) patients. In the remaining 7 patients, 5 presented with symptoms of liver dysfunction (ie, pain, anorexia, fatigue) and 2 with asymptomatic skin metastases.

At the time of detection of metastases (by MRI or otherwise), serum alkaline phosphatase was normal, abnormal or not available in 77/90 (86%), 6/90 (7%) and 7/90 (7%) of patients, respectively. Serum lactate dehydrogenase was normal, abnormal or not available in 55/90 (62%), 25/90 (27%) and 10/90 (11%) of patients, respectively.

During the screening period, four patients were found to have a secondary primary cancer: arising in colon (1), bladder (1), kidney (1) and lung (1). One patient who underwent liver resection was found to have only a haemangioma.

The median time to detection of metastases was 18 months (range, 3–97 months).

By the close of the study, 67/188 (35%) uveal melanoma patients at high risk of disseminated disease were still being screened for metastases. In another 31 patients, the screening at our centre had ceased, and this was because the 5-year screening programme had been completed (27 patients), or because the patient asked for follow-up to be continued at the local hospital (four patients).

Management of metastatic disease

The median overall survival for the 90 patients with metastatic disease was 12 months (95% CI 9.4 to 14.6). Thirty-eight (41%) patients were referred for laparoscopy, of whom 13 were found to have potentially operable tumours. One of these patients had radiofrequency ablation of a solitary metastasis, and 12 underwent partial hepatectomy with R0 (ie, complete) resection. The median survival time of these 12 patients was 24 months (95% CI 20.2 to 27.8) with all surviving 1 year, and 5 (40%) still alive at 2 years. Eleven of the 12 patients undergoing partial hepatectomy subsequently relapsed with metastatic disease.

Thirty-seven patients received some form of palliative chemotherapy. In 23 of these, such treatment was administered in the context of either a national or an international multicentre trial. Most patients with proven metastases were discharged to their local hospital for further care, and made an informed decision to continue with best supportive care and/or palliative chemotherapy. The median survival time of these patients was 10 months (95% CI 8.1 to 11.9 months) with 35% alive after 1 year.


Our aim was only to determine how many patients with metastases would be detected before the onset of symptoms if screened with MRI. It was not our objective to determine whether prognostication and screening are cost-effective or justifiable in patients with uveal melanoma. Neither did we seek to evaluate the treatment of metastatic disease in these patients.

Main findings

Prognostication based on clinical, histological and genetic data yielded a cohort of high-risk patients, with the 5-year survival of only 35%. Six-monthly MRI detected the metastases before the onset of symptoms in 92% of patients. Of the 90 patients developing metastases, 49% had less than 5 hepatic lesions, all measuring less than 2 cm in diameter. Twelve (14%) patients were treated by hepatic resection, and one patient received radiofrequency ablation of a solitary lesion. The 12 patients having hepatic resection had a median survival of 24 months (95% CI 20.2 to 27.8 months) as compared with 10 months (95% CI 8.1 to 11.9 months) in patients with inoperable metastases.

Strengths and weaknesses of the study

The main strengths of this study are the prospective design, the large number of patients, and the long follow-up. To our knowledge, no previous studies of this type, with so many patients, have been published.

A weakness of our study is the lack of comparison with other screening modalities, such as CT, and positron emission tomography (PET)/CT. However, such comparisons have been reported elsewhere, as discussed in the next paragraph.

Comparison with previous studies

Orcurto et al15 compared 18F-fluorodeoxyglucose (FDG) PET/CT with MRI in 10 patients. They found 108 liver lesions, which were detected with both types of scan in 31% of lesions, by MRI alone in 65% and by PET/CT alone in 4%. PET-CT detected 79% of 33 lesions exceeding a diameter of 1.1 cm with MRI as compared with 11% of 71 smaller lesions. Servois et al16 compared MRI with FDG-PET in 12 patients, in whom 27 lesions were resected and examined histologically. They found that MRI and FDG-PET, respectively, detected 7/7 and 2/7 lesions measuring 5–10 mm in diameter, and 11/11 and 9/11 lesions exceeding 10 mm in diameter. There were 9 lesions measuring less than 5 mm in diameter, and none of these were detected by either test. Strobel et al17 found that 41% of 27 hepatic metastases from uveal melanoma were FDG-positive as compared with 100% of 43 hepatic metastases from cutaneous melanoma. Feinstein et al reported 91 patients who underwent a CT scan within a month of uveal melanoma diagnosis. Hepatic abnormalities were found in 50 (55%) patients, but metastases were confirmed in only three of these. To reduce this false positive rate, Patel et al,18 from the same institution, investigated CT, reporting metastases to be hypodense, with heterogeneity on enhancement. Interestingly, they reported two patients who seemed to have a solitary lesion with CT but who were found to have multiple metastases when contemporaneous MRI was performed. These studies suggest that MRI is superior to CT and FDG-PET/CT in the detection of small hepatic metastases. Comparison of MRI with contemporaneous, contrast-enhanced CT would seem useful, but must take into account the hazards of radiation exposure when CT scans are repeated many times over a long period.

The role of liver imaging was reported by Eskelin et al,19 with the detection of metastases at a presymptomatic stage in 59% of 46 patients by means of annual liver ultrasonography and biochemical liver function tests. The authors concluded that the rate of asymptomatic detection may be optimised with semiannual imaging, which they estimated would detect metastases before symptoms in more than 95% of patients; however, patient outcomes were not reported.

Discussion of methods

We selected the detection of asymptomatic metastases as an outcome for this study because the likelihood of a response to treatment is greater in this group than in patients who are detected with advanced disease.

The cost-effectiveness of the screening programme was enhanced by selecting high-risk patients, which was possible because of the prognostic methods we developed. Without such selection, the number of patients needing to be enrolled would have been so high as to make this study unfeasible. Maeda et al20 screened 159 patients with uveal melanoma with MRI, detecting hepatic metastases in 9% of patients after a mean follow-up of 5.7 years. They reported five patients with false positive results, caused by vascular anomalies and haemorrhagic cysts. A full evaluation of our policy of screening only high-risk patients is beyond the scope of this study and would require another study design, with different inclusion and exclusion criteria. Nevertheless, our selection process for screening was successful, as evidenced by the fact that only 35% of our patients survived more than 5 years.

It was not our intention to evaluate the efficacy of surgical resection of liver metastases. This would require a different study, which would ideally be a randomised trial so as to take lead-time bias into account.

Clinical implications

This study indicates that six-monthly hepatic MRI detects almost all patients with hepatic metastasis from uveal melanoma before the onset of symptoms. Hepatic MRI has made it possible to evaluate treatments for metastases at an earlier stage of disease. We hope to evaluate systemic adjuvant therapies for high-risk patients with uveal melanoma, and when this becomes possible, we anticipate that MRI screening will be a useful end-point, in addition to overall survival.21


In conclusion, screening with hepatic MRI without contrast detects metastases from uveal melanoma before the onset of symptomatic disease in more than 90% of patients. The number of scans needed to detect each patient with metastasis is reduced by targeting high-risk patients by multivariate analysis that includes clinical, histological and genetic risk factors.


We are grateful to the National Specialist Commissioning Team of the National Health Service of England for funding our ocular oncology service.


View Abstract


  • Contributors BD made a substantial contribution to the conception and design of this study. He performed the ocular treatments and prognostication, and was critically involved in data acquisition. He was also involved in the drafting of the article, as well as critically revising it and approving the final version to be published. EM made a substantial contribution to the conception and design of this study. He was the oncologist managing the systemic care of the patients included in the study, and was critically involved in data acquisition. He was also involved in the drafting of the article, as well as critically revising it and approving the final version to be published. CR contributed to the conception and design of this study, particularly with regard to the MRI aspects. He was critically involved in MRI data acquisition. He was also involved in the drafting of the article, as well as critically revising it and approving the final version to be published. SC contributed to the design of this study. She performed pathology-based analyses of the hepatic metastases, and was critically involved in data acquisition. She was also involved in the drafting of the article, as well as critically revising it and approving the final version to be published. MC contributed to the design of this study, and was critically involved in data acquisition. She was also involved in the drafting of the article, as well as critically revising it and approving the final version to be published. PG, as the hepatic surgeon, was involved in the conception and design of this study. She was critically involved in data acquisition. She was also involved in the drafting of the article, as well as critically revising it and approving the final version to be published. HW, as the statistician, was involved in the design of this study. She was critically involved in data analysis. She was also involved in the drafting of the article, as well as critically revising it and approving the final version to be published. MM was involved in the design of this study. She was critically involved in data analysis. She was also involved in the drafting of the article, as well as critically revising it and approving the final version to be published.

  • Financial support Clatterbridge Cancer Research Trust provided funding for MRI sessions. The funding organisation had no role in the design or conduct of this research.

  • Funding None.

  • Competing interests None.

  • Ethics approval Wirral local research ethics committee reference number 99/01.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement Data are available on request from the corresponding author, Dr E Marshall.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.