UM is a rare cancer, affecting ~1600-1700 patients per year in the United States, but it is the most common intra-ocular cancer in adults. UM arises in the middle layer of the eye, the uvea tract, which consists of the iris, ciliary body, and choroid. Eye-sparing radiation (brachytherapy or proton beam therapy) is the most common treatment approach, but approximately 10% of patients will undergo enucleation due to large and/or aggressive tumors that cannot be managed with radiation or due to eye pain or vision loss. Local treatment by radiation or enucleation is highly successful at controlling the primary tumor, with only ~5% chance of local recurrence. Most patients present with local disease and no evidence of metastases, however, as many as 50% of patients will ultimately experience distant metastasis, most commonly to the liver.
Clinicopathologic staging cannot reliably identify patients at low or high risk of metastasis, as even early stage patients (AJCC Stage I-II) have a substantial risk of metastasis and mortality. Historically, most UM patients were managed with high intensity surveillance, including frequent imaging and laboratory tests, with the goal of diagnosing early metastasis. Systematic imaging has been shown to be effective at identifying asymptomatic metastases, which is important because treatment of liver metastases with surgical resection or regional therapy is more effective and achieves better outcomes when tumor burden is low. However, since approximately 50% of patients will not experience metastasis, a substantial proportion of patients were subjected to unnecessary imaging, laboratory tests, and clinical visits, resulting in patient burden, undo exposure to radiation and over-utilization of healthcare resources.
An accurate determination of metastatic risk at diagnosis allows for a risk-appropriate surveillance program. Patients at high-risk of metastasis can continue to be followed with a high intensity program as previously prescribed, such as quarterly ultrasound, magnetic resonance imaging (MRI) or computerized tomography (CT) scans alternating withliver function tests (LFTs), and consideration of adjuvant treatment. These patients benefit from early detection of metastatic disease when it can be most effectively treated. Patients with low metastatic risk can be removed from this traditional intensive surveillance and instead followed with a low intensity program, such as yearly exams, imaging, and LFTs.
DecisionDx-UM Test Description and Intended Use
DecisionDx-UM is an ribonucleic acid (RNA) gene expression classifier that is based on the expression levels of 15 mRNA transcripts (3 control and 12 discriminating genes). DecisionDx-UM is performed on tissue from a fresh-frozen fine needle aspirate biopsy (FNAB), formalin-fixed paraffin embedded (FFPE) sections from an enucleated tumor, or, in rare cases, fresh-frozen resection material. Results are reported as a 5-year risk classification for metastasis: low risk (Class 1A), intermediate risk (Class 1B), or high risk (Class 2).
The DecisionDx-UM test is intended for determination of metastatic risk, and to guide surveillance and referral to medical oncology in patients who have a confirmed diagnosis of UM and no evidence of metastatic disease. The test discriminates patients with high risk (class 2) for early distal recurrent disease from those with minimal risk of distal metastasis (class 1A). Identification of high-risk patients allows early referral to a medical oncologist with expertise in the management of UMs, which includes intensified metastatic surveillance and/or metastasis intervention, and stratification for entry into clinical trials with adjuvant therapy. In rare cases where the patient cannot realistically see a medical oncologist due to geographic location (long distance to travel), and/or are among underserved patient populations, if they cannot feasibly see a medical oncologist, surveillance testing for class 2 patients can be directed by an ophthalmologist with specific training in treating patients with UM.
Clinical Validation
In both prospective and retrospective multicenter studies, DecisionDX-UM has been shown to be a more accurate prognostic indicator of metastasis compared to any other factor.1-4
Onken et al1 reported the migration of the RNA expression profile from a hybridization-based microarray platform to a polymerase chain reaction (PCR)-based 15 gene assay and analyzed the technical performance of the assay in a prospective study of FNAB tumor samples from multiple centers. The gene expression profile distinguished between low metastatic risk (class 1 signature) and high metastatic risk (class 2 signature). The role of RNA quality and tumor heterogeneity was evaluated. A clinically annotated training set of 28 UMs was used to support the vector machine algorithm for classification. One hundred seventy-two patients from a single center with a median follow-up of 16 months were utilized to evaluate prognostic performance which demonstrated technical performance of 94.8%. Kaplan–Meier analysis showed an accuracy of risk-classification with 5-year metastatic-free survival (MFS) rates of 98% and 24% for predicted Class 1 and 2 cases, respectively (P < 0.0001).1
The Collaborative Ocular Oncology Group study was a prospective, multi-center, blinded study to assess clinical validity of the DecisionDx-UM test.2 Comparison with other genetic and clinicopathologic variables was evaluated. Of 494 patients, 446 were considered evaluable. The 50-month metastasis-free survival was 97% vs 20% for Class 1 and 2 respectively (p<0.0001). By Cox multivariate proportional hazards analysis, Class 2 identified metastasis better than any other prognostic factor (p<0.006). The Net Reclassification Improvement study showed improvement of gene expression profiling over TNM (T describes size of primary tumor, N describes regional lymph nodes status, M describes distant metastasis) classification of 37% at 2 years (p=0.008) and 43% at 3 years (p=0.001). When compared to chromosome 3 status, the improvement of gene expression profiling over TNM was 36% at 2 years (p= 0.006) and 38% at 3 years (p=0.004).2
In a retrospective, single-center clinical study designed to assess clinical validity of the DecisionDx-UM test, in 187 patients, Chappell, et al3 showed disease specific survival was predicted with high accuracy. Kaplan-Meier analysis for 5-year disease specific survival was 93% and 38% for Class 1 vs 2 cases, respectively (p<0.0001). By multivariate Cox modeling, the DecisionDx-UM class was the only independent significant predictor of outcome for both metastasis-free survival (HR=8.4, p<0.0001) and disease-specific survival (HR=12.3, p<0.0001).
Another prospective, single-center clinical study evaluated the clinical validity of the DecisionDx-UM test in 299 UM patients. In this study, Cox multivariate analyses confirmed that the 15-gene expression profile was the only significant predictor of metastatic risk (p=0.0013).4
A step-down algorithm analysis of two genes in Class 1 patients has since been performed to identify those patients with Class I classification at risk for late metastasis. Due to this refinement, Class 1 includes low-risk Class 1A patients and a small number of intermediate-risk Class 1B subjects with late relapse.
Clinical Utility
A retrospective chart review study showed that the DecisionDx-UM test results direct appropriate surveillance and treatment plans by matching an individual patient’s risk for metastasis to informed medical management decisions. Aaberg et al6 reported on 88 Medicare beneficiaries in which all Class 1 patients received low-intensity surveillance while Class 2 patients received high intensity surveillance plans (imaging and/or liver function testing every 3-6 months). Test results also influenced referral decisions with 29% of Class 2 patients being referred to medical oncology for follow-up and 10% recommended for adjuvant therapy consideration whereas no Class 1 patients were referred.
In a prospective, multi-center study of 70 patients, the majority (81%) of Class 1 patients had low-intensity surveillance and all (100%) Class 2 patients received high-intensity surveillance (p<0.0001); four Class 2 patients were recommended for systemic adjuvant therapy.
A decision tree analysis was performed to model the impact of DecisionDx-UM on resource utilization, comparing the previous framework in which all patients received high-intensity surveillance regimens with one in which the surveillance regimen is guided by DecisionDx-UM. Strict compliance with DecisionDx-UM results was associated with a 50% reduction in the number of surveillance procedures performed at two years compared to the previous framework, and a 63% reduction at five years. These results indicate that use of DecisionDx-UM can help avoid high intensity, imaging-based surveillance in patients with a low risk of metastasis, thereby reducing resource utilization in the management of UM patients, which is associated with overall cost savings.
Summary of Analytical and Clinical Performance
General
| Intended Use |
The DecisionDx-UM test is intended for the determination of metastatic risk, and to guide surveillance and referral to medical oncology in patients who have a confirmed diagnosis of uveal melanoma (UM) and no evidence of metastatic disease. |
| Validated Specimen Type(s) |
Fresh frozen fine needle aspirate biopsies (FNAB), frozen resections, and formalin-fixed, paraffin-embedded (FFPE) specimens |
Analytical Performance
| Description |
Results (with 95% Confidence Intervals if applicable)1
|
Repeatability (within run precision)
4 samples (in triplicate) twice on a single PCR card, 1 instrument, 1 operator, 1 run, 1 day, 1 manufacturing reagent lot; repeated on a separate card. |
100% (63.1-100%) |
Intermediate precision (between run precision)
Inter-operator/instrument: 28 samples, 2 instruments, 2 operators, 2 runs, 1 day, 1 manufacturing reagent lot (3 discordances: 1 class 2 vs 1A; 2 class 1A-1B)
Inter-assay: 16 samples, 1 instrument, 2 operators, 3 runs, 3 consecutive days, 1 manufacturing lot (0 discordances) |
89.3% (71.8-97.7%)
100% (79.4-100%) |
| Reproducibility (between sites) |
Not applicable |
| Minimum input cDNA quantity |
125 ng |
| Minimum tumor content (for FFPE specimens) |
80% by histomorphology |
| Limit of blank (LOB) |
CT undetermined for blank |
| Limit of detection (LOD) |
Not applicable |
| Limit of quantitation (LOQ) |
Not applicable |
| Linearity |
Not applicable |
| Interfering substances |
Not applicable2 |
| Specimen stability, primary |
FNAB 96 hours at -80 °C (Onken et al., 2010)
FFPE 4 years at RT (Onken et al., 2012) |
| Specimen stability, intermediate (extracted RNA) |
96 hours when stored at -80 °C per manufacturer and literature |
| Specimen stability, intermediate (cDNA) |
24 hours when stored at 4 °C per manufacturer
30 days when stored at -20 °C |
| Critical reagent closed/shelf-life stability |
Applied Biosystems TaqMan® Low Density Array, 24 months at -20 7deg;C per manufacturer
Arcturus PicoPure® RNA Extraction Kit, 10 months at RT per manufacturer
Applied Biosystems High-Capacity cDNA Reverse Transcription Kit, 8 months at -20 °C per manufacturer
Applied Biosystems TaqMan® PreAmp Master Mix Kit, 9 months at 4 7deg; C per manufacturer
Applied Biosystems TaqMan® Gene Expression Master Mix, 12 months at -20 °C per manufacturer
Applied Biosystems RNAse Inhibitor, 42 months at -20 °C per manufacturer |
| Critical reagent open/in use stability |
Per manufacturer’s specifications |
1Using Clopper-Pearson method
2Since the gene expression profile is based on ratios of gene to controls, rather than an absolute value, the effect of an interfering agent is expected to affect all genes equally and result in failed amplification.
Clinical Performance: Validity
| Description |
5-year metastasis-free (a) or disease-specific (b) survival rates1
(Non-censored recurrence rate; 95% Confidence Intervals of event rates)2 |
| |
Class 1A |
Class 1B |
Class 2 |
| Onken et al2 (n=514)* |
98%a
(0.8%; 0.1-3%; 2/241) |
79%a
(10.4%; 4.3-20.3%; 7/67) |
28%a
(29.6%; 23.5-36.4%; 61/206) |
| Chappell et al3 (n=187)* |
93%b
(2.5%; 0.5-7.3%; 3/118) |
38%b
(28.9%; 18.7-41.2%; 20/69) |
| Correa et al4 (n = 158)† |
92%b
(4.58%; 1.5-10.4%; 5/109) |
55%b |
| Demirci, et al. 2015 (n = 203)†** |
95%a |
90%a |
27%a |
| Correa et al5 (n = 299)† |
92%b |
55%b |
1Survival rates according to Kaplan-Meier analysis
2Overall non-censored recurrence rates and 95% Confidence Intervals (Coppler-Pearson method) not accounting for censored patients.
*Tests performed at Washington University (Wash U);
†Tests performed at Wash U and Castle Biosciences Inc.
**Published survival rates but not event numbers so cannot calculate confidence intervals
Clinical Performance: Utility
| Description |
Clinical Use Outcomes
(with 95% Confidence Intervals if applicable)1 |
| |
Class 1A |
Class 1B |
Class 2 |
Aaberg et al6
(n = 88 with documentation)
(Note: Retrospective decision impact study of Medicare beneficiaries) |
100% (92.6-100.0%; 48/48) received low intensity surveillance. None referred to medical oncology or adjuvant trials.1 |
100.0%
(91.2-100.0%; 40/40) received high intensity surveillance, referral to medical oncology or adjuvant trials. |
| Plasseraud et al7 (n = 70) |
(65.3-94.4%; 25/30)
received low intensity management |
(29-96.3%; 5/7)
received low intensity management |
(89.4-100.0%; 33/33) received high intensity management |
16.7%
(5.6-34.7%; 5/30) received high intensity management |
28.6%
(3.7-71.0%; 2/7) received high intensity management |
10.0%
(2.1-26.5%; 3/30) received referral to medical oncology |
|
33.3%
(18.0-51.8%; 11/33) received referral to medical oncology |
1Low intensity management is defined LFTs and/or imaging studies annually. High intensity management is defined LFTs and/or imaging studies every 3-6 months.
