Article Text

Download PDFPDF

Risk-stratified systemic surveillance in uveal melanoma
  1. Jacquelyn M Davanzo,
  2. Elaine M Binkley,
  3. James F Bena,
  4. Arun D Singh
  1. Cole Eye Institute, Cleveland, Ohio, USA
  1. Correspondence to Dr Arun D Singh, Cole Eye Institute, Cleveland Clinic, Cleveland, Ohio 44195, USA; singha{at}


Background/Aim Molecular prognostication provides clinically applicable prognostic information for patients with uveal melanoma. Most ocular oncologists recommend intensive metastatic surveillance for patients with high-risk tumours. However, socioeconomic variables may limit a patient’s ability to adhere to recommended surveillance. We aim to analyse socioeconomic data from patients with uveal melanoma who underwent molecular prognostication to determine which variables influence adherence.

Methods This was a retrospective review of 107 consecutive patients who were diagnosed and treated for uveal melanoma from January 2014 to June 2015. Patients were categorised into low/unknown risk and high risk for metastasis. The low-risk group was followed with hepatic ultrasonography every 6 months. The high-risk group was followed with more frequent hepatic imaging or incorporation of hepatic CT/MRI into the surveillance protocol. Adherence to surveillance recommendations was recorded for the first 2 years following primary treatment. Socioeconomic data including age at diagnosis, baseline systemic staging, gene expression profile status, marital status, insurance, distance of primary residence, median household income and Charlson Comorbidity Index score were recorded. Frequency/modality of imaging and metastatic status were also recorded.

Results High-risk patients were more likely to develop metastasis than low-risk/unknown-risk patients (p<0.001). High-risk patients were more likely to have scans at baseline (p=0.008) and to have expected scans relative to low-risk/unknown-risk patients (p<0.001). There was no significant relationship between the likelihood of adhering to recommended surveillance and the other variables analysed.

Conclusions Prognostic risk level is a significant predictor of surveillance and remains significant after adjustment for socioeconomic variables. Adherence to surveillance recommendations for high-risk patients may translate into improved survival.

  • uveal melanoma
  • metastatic surveillance
  • gene expression profiling

Statistics from

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.


Despite adequate local tumour control, patients with uveal melanoma have a 10%, 25% and 34% risk of developing metastatic disease at 2, 5 and 10 years following treatment, respectively.1 However, not all tumours are the same in regard to metastatic risk, with clinical, histopathological and molecular features placing some tumours at higher risk for metastasising than others.2 3 The advent of molecular prognostication has provided ocular oncologists with a useful tool, particularly for those patients undergoing treatment with brachytherapy who would not have been able to obtain prognostic information from histopathological analysis. Using one of the commercially available tests, gene expression profiling (GEP, Castle Biosciences, Friendswood, Texas), tumours can be classified into 1A or 1B with a relatively low risk of metastasis (2% and 21% risk of metastasis, respectively, at 5 years), or a class 2 tumour with higher risk for metastasis (72% risk of metastasis at 5 years).2 4 Given that there is no effective adjuvant therapy for metastasis of uveal melanoma at present, management generally focuses on intensive metastatic surveillance in the hope of identifying metastatic disease early, so that treatment modalities such as traditional or targeted systemic chemotherapy or local hepatic treatment with ablation or resection can be initiated.5 6

Given the propensity for uveal melanoma to metastasise to the liver, most ocular oncologists focus on hepatic surveillance with imaging modalities including hepatic ultrasound, CT and MRI.7 8 Hepatic ultrasound is non-invasive and useful given its wide availability.7 8 CT provides a sensitive, reproducible test, and MRI is both highly specific and able to detect lesions less than 1 cm in size.7 8 For patients with high-risk tumours, oncologists often recommend either more frequent (greater than every 6 months) and/or more intensive surveillance such as inclusion of hepatic CT/MRI in addition to hepatic ultrasonography (USG).9 10 Patient adherence to these surveillance exams is crucial to early, presymptomatic detection of metastatic disease. Literature on metastatic surveillance from patients with a variety of malignancies, including prostate, breast and colon cancer, suggests that the patient’s psychological, socioeconomic and demographic variables play an important part in a patient’s desire and ability to adhere to surveillance recommendations.11–13 Patient age, comorbidities, ethnicity, comprehension of their disease and anxiety can have an effect on adherence to surveillance protocols.11–13

Frequent surveillance exams can place financial, time and psychological burden on individuals, and the ‘real world’ compliance with surveillance may deviate from the recommendations of the oncologist. In fact, in one study of patients with uveal melanoma, many patients were dissatisfied with insufficient financial and psychological counselling, and these feelings could translate into poor compliance with follow-up and surveillance.14 We therefore aim to study adherence to recommended surveillance protocols in a population of patients diagnosed with uveal melanoma and analyse the influence of socioeconomic factors.


We conducted a retrospective study of 107 consecutive patients diagnosed with primary uveal melanoma at our institution from January 2014 to June 2015. Patients were treated with primary therapy including enucleation, plaque brachytherapy or primary resection. Each patient was offered molecular prognostication with GEP using the commercially available GEP test (Castle Biosciences). We studied adherence to recommendations regarding systemic surveillance for the first 2 years after primary ocular therapy. Patients were categorised into two groups: low risk for metastasis (class 1A or 1B) or unknown risk for metastasis (prognostication not performed—small tumour, patient refusal, technical failure) and high risk for metastases (class 2). Prognostic class was determined by GEP testing of either trans-scleral or transvitreal fine-needle aspiration biopsy specimens from patients who elected to undergo prognostic testing. For patients undergoing treatment with plaque brachytherapy, fine-needle aspiration biopsy was performed at the time of plaque insertion.

The low-risk/unknown-risk group of patients was advised our standard surveillance protocol that included hepatic USG every 6 months (expected number of scans over 2 years=4 or less).10 Patients identified to be at high risk for metastasis (class 2) were referred to a medical oncologist and advised intensive protocol which included either more frequent systemic surveillance (every 3-month hepatic USG) or incorporation of hepatic CT/MRI in the surveillance protocol. These patients were expected to have four or more scans in the initial 2 years following primary therapy. In addition, baseline systemic staging that included a CT scan of the chest, abdomen and pelvis with and without contrast was advised in all cases prior to primary ocular therapy and prognostication assessment.

In order to identify socioeconomic factors that could pose barriers to receiving recommended care, we analysed several factors. The socioeconomic data were obtained from review of electronic medical records. Data points included age at diagnosis, baseline systemic staging, genetic expression profiling prognostication status, marital status, insurance type, distance of primary residence to the Cole Eye Institute at the Cleveland Clinic, home zip code, median household income, Charlson Comorbidity Index (CCI) score,15 frequency of imaging and modality of imaging (USG, CT, MRI) for systemic surveillance, and metastatic status of the patient.

The age at the time of diagnosis was calculated by subtracting the birth year from the first encounter date. Family support was determined by the marital status or the patient’s emergency contact person which was listed as spouse or another.

The type of insurance at the time of diagnosis was categorised into three separate groups: Medicare/Veteran’s Association (Veterans Administration)/military, private/other, or Medicaid/none/self.321

Distance from the Cleveland Clinic was estimated using Google Maps to calculate the number of miles the patient’s primary address zip code was from Cleveland Clinic’s zip code (44195), and then divided into three groups: 0-160 Km, 161–320 Knm miles and 321+ Km away from Cole Eye Institute, Cleveland Clinic.

The median household income derived from primary address zip code listed in the medical chart from the website, www.incomebyzipcode.comwhich, which provides current census bureau income statistics for the USA.

The CCI score was calculated using an online CCI calculator ( -cci), which gives a numerical score to predict 10-year survival in patients with multiple comorbidities, with lower scores indicating longer length of survival.

Categorical variables were described using frequencies and percentages, while continuous variables were described using means and SD. The relationship between risk and categorical variables was assessed using χ2 tests for nominal measures and Kruskal-Wallis tests for ordered data, while the relationship between risk and continuous variables was assessed using analysis of variance models. A multivariable logistic regression was run to predict expected surveillance from predictor variables. Models maintained at least five events per variable.16 In the first model, age, support type, distance from the Clevland Clinic Foundation (CCF), CCI score and median household income were used as adjustment factors, while in the second model age and median household income were replaced with insurance type. Analyses were performed using SAS Software V.9.4.


Thirty-seven patients had high-risk GEP results, 31 patients had low-risk GEP results and 39 patients had unknown risk. Of the 39 patients with unknown risk, GEP testing was not performed in 20 patients due to small tumour size, patient age or medical comorbidities. Of the remaining unknown-risk patients, nine patients declined testing, four patients had testing denied by insurance and four patients did not have testing for unknown reasons.

There was a statistically significant difference in the risk of metastasis between the low-risk/unknown-risk (n=70) and the high-risk group of patients (n=37) (p<0.001), with only eight patients developing detectable metastasis, all of them in the high-risk group. There was no statistically significant difference in age, marital status, spouse support, type of insurance, distance from the Cleveland Clinic, median household income or CCI score, and between patients in the low-risk/unknown-risk group and the high-risk group (table 1).

Table 1

Relationship between metastasis risk and demographic variables

High-risk patients were more likely to have scans done at baseline (p=0.008). High-risk patients had more scans performed (p=0.011) and were also more likely to have any scans (p<0.001), intensive imaging protocol scans (p<0.001) and expected surveillance (4 scans for low/unknown risk, >4 for high risk) (p<0.001) compared with unknown-risk/low-risk patients (table 2).

Table 2

Relationship between metastasis risk and baseline systemic staging and surveillance scans

Two statistical models were designed to predict expected surveillance. Predictor variables included risk group, support, distance from the Cleveland Clinic, CCI score, income and age. In the first model, the odds of high-risk patients having expected surveillance were 9.8 times higher as compared with low-risk patients (p<0.001). No other factors were significant. In the second model, the predictor variables were risk group, support, distance from the Cleveland Clinic, CCI score and insurance. The odds of high-risk patients having expected surveillance were 9.5 times higher as compared with low-risk patients (p<0.001). Again, no other prognostic factors were significant. Based on these findings, risk level is a significant predictor of surveillance adherence and remains significant after adjustment for other factors.

There was no statistically significant relationship between age, marital status, spouse support, type of insurance, distance from the Cleveland Clinic, median household income or CCI score, and adherence to recommended surveillance (table 3).

Table 3

Relationship between expected surveillance and predictor variables for all patients


While there are clinical trials under way to assess the effectiveness of adjuvant therapy for micrometastatic disease in patients with high-risk tumours, management typically focuses on intensive metastatic surveillance with hepatic imaging.9 17–19 While previous work has shown that early detection of metastatic disease does not improve overall survival, more recent work has suggested that survival is better in patients with smaller hepatic metastases at diagnosis and in patients who have not yet developed symptoms from metastasis.20–22 Newer treatments for metastatic disease including small molecule, biologic and targeted therapy may one day lead to improved survival.5 As adjuvant therapies and effective treatment of subclinical metastatic disease become available, developing appropriate surveillance protocols and understanding patient barriers to compliance with such protocols will be of increasing significance. However, a patient’s decision to comply with surveillance recommendations is complex, and may reflect the individual’s perception of their disease, resources, comorbidities and socioeconomic background.11 13

The present study found no statistically significant relationship between socioeconomic variables and adherence to recommended surveillance (table 3). This is different from previous work examining adherence to surveillance protocols for other types of malignancy in which variables such as race, age and comorbid conditions were associated with adherence.11–13 While this could represent a specific difference in the population of patients with uveal melanoma, one important consideration is that the group of individuals who elect to proceed with GEP testing may be more highly motivated to obtain prognostic and follow-up information about their disease and in essence ‘self-select’ regardless of socioeconomic background. Additionally, a previous study of patients who have had GEP testing showed that many patients experience anxiety and depression regarding their diagnosis that often improves over time.23 One consideration for future study is whether anxiety regarding the high-risk tumour diagnosis drives these patients to adhere to surveillance. One limitation of this study was the inclusion of patients with low and unknown risk in one group. While it is possible that patients with low and unknown risk may behave differently, the recommended surveillance protocol was the same for each of these groups. Low-risk and unknown-risk patients were therefore analysed together in order to compare adherence relative to the intensive screening regimen used for the high-risk patients.

Previous work has shown that 95% of patients with uveal melanoma in our practice elected to proceed with molecular prognostication when offered.24 While individuals with uveal melanoma who choose to undergo GEP testing may be of a different socioeconomic background than those who decline testing, the number of individuals who decline testing is so small that this makes this difficult to study. The lack of an association of socioeconomic variables such as median household income and insurance coverage with surveillance adherence may represent a true lack of association, but one limitation of the present study is the small number of patients (n=7) from the Medicaid/self-pay/or uninsured group. However, a previous study has shown that in the mid-southern USA, higher area-based socioeconomic measures are associated with a higher rate of uveal melanoma diagnosis.25 Some populations of patients diagnosed with uveal melanoma may be of an overall higher socioeconomic class due to better access to eye care.25

Our finding that patients with high-risk tumours were more likely to adhere to surveillance protocols than patients with low-risk or unknown-risk tumours is important. At this time, patient survival after the development of metastatic disease remains poor.26 27 While there are a number of strategies under investigation, there is no effective adjuvant therapy for metastatic disease.6 28 However, the increased adherence to intensive surveillance in patients at the highest risk for developing metastasis is promising in that this could ultimately result in earlier detection of disease at a time when targeted hepatic therapies could be employed and ultimately improve survival outcomes. The fact that high-risk patients in this study were referred to medical oncology may have additionally influenced the improved surveillance adherence in this group and supports the benefit of referring high-risk patients. Likewise, the decreased adherence to surveillance recommendations in the low-risk/unknown-risk group needs to be considered when managing patients. While the risk of metastatic disease is lower in this cohort, the risk of metastasis is not completely absent, and patients should be encouraged to adhere to recommended surveillance.2 4 A recent survey of patients with uveal melanoma found that many patients were not satisfied with the counselling provided regarding genetic testing.14 Communication with patients is key to conveying the need for surveillance, and while we may be optimistic with patients with low-risk tumours regarding their decreased risk for metastasis we also need to emphasise the importance of surveillance and identify barriers to adherence.

In summary, prognostic risk level is a significant predictor of surveillance adherence irrespective of age, type of insurance, marital status, spousal support, distance from the institution or median household income. Patients with high-risk tumours were more likely to undergo recommended surveillance than those with low-risk or unknown-risk tumours. Adherence to surveillance recommendations in patients with high-risk tumours may translate into improved survival outcomes.



  • Contributors JMD collected the data. EMB wrote the manuscript. JFB performed the statistical analysis. ADS conceived, edited and approved the final manuscript.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient consent for publication Not required.

  • Ethics approval Approval was obtained from the Institutional Review Board at the Cleveland Clinic Foundation and the project adhered to the tenets of the Declaration of Helsinki.

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

Linked Articles

  • At a glance
    Keith Barton Jost B Jonas James Chodosh