Hepatic metastasis surveillance in uveal melanoma: a retrospective cohort study from a UK tertiary centre (2006–2022)

Liver surveillance imaging is essential for detecting early asymptomatic metastases in uveal melanoma, which predominantly involves the liver. Early detection may improve treatment opportunities, but variability in imaging protocols and a lack of consensus on surveillance duration present challenges. This study aimed to evaluate our systemic surveillance protocol, optimise pathways, and assess risk factors for metastasis.

We retrospectively analysed patients diagnosed with uveal melanoma between 2006 and 2021 who underwent hepatic imaging surveillance at Sheffield Teaching Hospitals NHS Foundation Trust. Demographics, tumour characteristics, treatments, disease status, and survival outcomes were collected.

Among 1086 patients (45% female, 79% White; median age 68 years), 315 (29%) developed metastases, with 293 (93%) detected within five years of ocular treatment. The number needed to scan (NNS) increased substantially after five years, indicating reduced detection efficiency. Higher T stage and ciliary body involvement were significantly associated with increased metastatic risk ( P < 0.01).

Most metastases from uveal melanoma occur within five years of treatment. Personalised, risk-based surveillance strategies considering tumour stage and location may improve efficiency and optimise healthcare resource use.

In the UK, between 700 to 800 individuals are diagnosed with uveal melanoma (UM) each year [ 1 ]. Almost 50% of patients develop metastatic disease, which most commonly involves the liver (93%), lungs (24%), and bones (16%) [ 2 , 3 ]. The median time from ocular treatment to liver metastasis is 27 months (IQR, 13–46) [ 4 ]. Metastatic disease is associated with a poor prognosis, with a median survival of 14–28 months, with only 20% surviving one year or more [ 2 , 5 , 6 ].

Early detection of metastases (i.e., before the onset of symptoms) enhances any opportunities for treatment, identifies patients for possible clinical trial inclusion, and enables earlier palliative care, if needed [ 7 ]. Ultrasound is widely used because it offers high spatial resolution enabling identification of small lesions, is quick, inexpensive and widely available. However, it is operator dependent, and sensitivity may be limited in patients with a high BMI or when bowel gas interferes with imaging [ 8 ]. Due to these patient factors, it may miss smaller lesions with a quoted specificity of 85% [ 8 ]. At our centre, good sonographer expertise and same-day reporting enhance its value, especially when paired with same-day ophthalmic assessment. MRI has a reported sensitivity of 95% and is often used to characterise lesions identified on ultrasound, but this sensitivity decreases with sub centimetre metastases and significantly in metastases 9 ]. A percutaneous biopsy can be obtained if confirmatory histopathology is required [ 5 , 10 , 11 ]. 18-FDG PET is an alternative modality with high sensitivity, but it is not utilised in our centre as the high radiation exposure that would result from repeated examinations limits its applicability as a surveillance tool [ 12 ].

The approach to liver surveillance imaging is highly variable both internationally and within the UK, with no established consensus on the optimal imaging technique, frequency or duration due to lack of evidence within the literature [ 13 ]. The European Society for Medical Oncology (ESMO) has not established specific surveillance guidelines for uveal melanoma [ 14 ]. In the United States (USA) The National Comprehensive Cancer Network (NCCN) provides guidelines for uveal melanoma surveillance utilising risk stratification based on the tumour size, location, histological features, genetic factors and liver function blood test [ 15 ]. Intensive imaging is reserved for patients with high-risk features, typically every 3–6 months for the first five years [ 15 ]. In the UK, Melanoma Focus recommended a total of 10 years surveillance with six monthly intervals in the first 5 years and annually thereafter [ 13 ]. However, surveillance has not been recommended for patients who have small tumours with favourable genetic markers [ 13 ]. In Scotland, patients with uveal melanoma are stratified into high or low/medium risk, with high-risk patients undergoing one liver MRI examination every 6 months, for 10 years [ 16 ].

Sheffield Teaching Hospitals NHS Foundation Trust (STH) is one of the four National Centres of Excellence for uveal melanoma in the UK [ 17 ]. Similar to the Melanoma Focus guidelines, patients undergo hepatic surveillance every six months for the first five years using ultrasound. This is then followed up by annual ultrasound surveillance for up to ten years. Patients with lesions detected on ultrasound, undergo MRI to characterise the lesion, and then undergo temporaneous MRI surveillance. If the patient remains free of metastatic disease on subsequent MR imaging, they preferably return to US surveillance. However, in practice patients may remain on MRI surveillance going forward, reflecting the complexity of factoring in individual clinician and patient preferences into surveillance pathways. All patients within the service currently undergo the same surveillance strategy, irrespective of whether they have a high or low risk primary tumour.

At our centre, the cost of an MRI scan for uveal melanoma surveillance is approximately £150, almost four times the cost of an abdominal ultrasound scan, and the current waiting times for urgent and routine MRI scans are 10 or 16 weeks, respectively. Therefore, a ten-year MRI-focused surveillance programme places significant demands on medical imaging and financial resources [ 18 ].

The aim of this study was to evaluate the frequency and timing of liver metastases in patients with uveal melanoma undergoing surveillance imaging in Sheffield and assess the appropriateness of current practice. It also aimed to examine trends in detection efficiency over time and explore clinical factors associated with metastatic risk, including tumour location and T-stage.

All methods were performed in accordance with the relevant guidelines and regulations. This single-centre retrospective study was approved by the local clinical research office, the local research ethics committee (Sheffield 3D Lab; reference 17/YH/0142), and the information governance team. As this was a retrospective study using non-identifiable data, the requirement for informed consent was waived by the clinical research office and ethics committee.

Patients were identified retrospectively by searching the Sheffield Ocular Oncology Service database. They were included if more than 16 years old, diagnosed with uveal melanoma between January 2006 and October 2021 and if they underwent hepatic surveillance with ultrasound or MRI at our hospital. Patients were excluded if no hepatic imaging was performed at our hospital after the initial staging scan at STH, or if they had synchronous (cancers diagnosed within 6 months of each other) or metachronous (cancers diagnosed more than 6 months apart) cancers with liver metastases without histological confirmation of the metastatic origin of the liver lesions, or if TNM data was not available. Patients were not pre-selected based on their primary characteristics, including the risk profile of the tumour.

Data was copied to a Microsoft Excel spreadsheet from a combination of the STH CRIS (Computerised Radiology Information System) and PACS (Picture archiving and communication system). For each eligible patient, the following variables were recorded: demographic information (including current age, sex, and self-reported ethnicity); details of initial diagnosis (tumour subtype, TNM staging, and primary treatment); disease status (disease-free or metastatic) and date of death; the date and available findings of the most recent liver surveillance imaging (ultrasound or MRI), noting that no second image interpretation was performed; and the date of the initial documented diagnosis of liver metastasis. With patients who had initial surveillance imaging at STH, with subsequent surveillance imaging at their local centre, the date and report of the latest STH imaging were documented.

Where necessary BR acted as an arbitrator to resolve any discrepancies or conflicts which had been flagged during data collection. In cases of missing data, patients’ electronic health records and letters were reviewed to fill in any gaps. For patients outside of our integrated care system, missing data was collected by contacting their local General Practitioner.

Simplification of TNM staging focused on consolidation of subcategories based on T staging alone; for instance, T1a, T1b, and T1c cases were aggregated under T1 [ 19 ]. This was done to focus on size as the most important factor for liver metastasis, in line with previous literature, and to strengthen the results given the small sample size [ 20 ]. Additionally, patients’ ages were grouped into six distinct categories: under 40, 40–49, 50–59, 60–69, 70–79, and 80 and above.

Tumour subtype was determined by clinical examination. Ciliochoroidal tumours were defined as those involving both the ciliary body and the choroid, while iridociliary tumours were defined as those involving the iris and the ciliary body.

Statistical analysis and the generation of graphs were conducted using RStudio (2022.07.1 running R 4.2.1.) running the following packages: ‘data.table’, ‘dplyr’, ‘tidyverse’, and ‘ggplot2’ and Prism (version 9.4.1; San Diego, CA, USA). Kaplan-Meier survival analysis and Cox proportional hazards models were employed to evaluate liver metastasis and mortality rates, using the ‘survival’ package with a significance threshold of p < 0.05. Factors influencing liver metastasis risk were analysed according to age at treatment, TNM T category, and tumour location.

For mortality, we analysed the time from diagnosis to either the occurrence of death or the last known follow-up (censoring). For metastasis evaluation, we measured the time from diagnosis to either the first detection of liver metastasis or the most recent surveillance scan date. Median times, including time to metastasis and survival intervals, were calculated using Kaplan–Meier estimates to account for censoring. The likelihood ratio for each year was calculated by dividing the proportion of patients without metastasis in that year by the proportion of patients with metastasis in the same year. The number needed to scan (NNS) for each year was calculated as the inverse of the annual risk of developing metastasis, defined as the number of patients at risk at the start of the year divided by the number who developed metastasis during that year. For overall NNS estimates within subgroups, including T-stage and tumour location, the total number of patient-years at risk was divided by the total number of metastases within each subgroup as per study by Hagström et al. [ 21 ].