31-Genetic Expression Profile Test

The quantitative reverse transcription polymerase chain reaction–based 31-GEP test is obtained from samples that are collected from formalin-fixed paraffin-embedded CMM tissue and arranged in 5-μm sections on microscope slides. RNA isolation is performed followed by an assessment of its quality and quantity. The RNA is then converted to complementary DNA and undergoes amplification before being loaded to microfluidics gene cards containing primers specific for the 31 gene targets. The gene expression assay is performed in triplicate. Radial basis machine predictive modeling is performed, which is a nonlinear classification based on the normalized values for each gene. The modeling transforms the gene measurements using a kernel function to find an optimal hyperplane in multivariate dimension, thus providing a predicted classification of high and low risk tumor biology.

Initial Development and Validation

For the development of the 31-GEP, Gerami and colleagues used published genomic analysis of CMM tumors to determine a unique prognostic genetic signature for metastatic risk. Genes were selected on the basis of significant genetic expression variation in metastatic and nonmetastatic CMM across several published studies. Of 54 identified genes, the investigators selected 20 based on chromosomal location. Genes from a similar uveal melanoma panel were added in addition to specific BAP1 gene probes. A signature comprising 28 prognostic genetic targets and 3 control genes was developed from the expression data. The 31-GEP was applied to 268 primary CMM cases (collected from 7 independent centers) with clinical follow-up of at least 5 years unless there was a well-documented metastatic event, including positive SLNB.

The study initially reported the use of the test to predict metastasis in patients diagnosed with stage I or II CMM using an independent validation set consisting of 104 cases. Of these cases, 35 had developed metastatic disease, and there was median follow-up of 7.3 years for the cases that did not. The 5-year disease-free survival was 97% among the 61 cases with a class 1 “low-risk” signature and 31% for the 43 cases with a class 2 “high-risk” signature. Negative predictive value and positive predictive value were 93% and 72%, respectively. The receiver operating characteristic curve was 0.91 for the validation set and 0.93 for the original training set, which is consistent with a clinically relevant predictive model.

For stage I and II cases in the validation set that had either a metastatic event or more than 5 years of follow-up without metastasis, class 1 disease-free survival was 98% compared with class 2 with a rate of 37%. Median follow-up for cases in this cohort was 7.6 years. When combined, the validation and training cohorts consisted of 220 stage I and II CMM cases. Overall, the 31-GEP accurately identified 120 of 134 (90%) stage I/IIA cases without documented evidence of metastasis as class 1 (low risk) and 24 of 30 (80%) stage I/IIA cases with documented metastasis as class 2 (high risk) (sensitivity, 90%; specificity, 84%; positive predictive value = 72%; negative predictive value = 95%).

When compared with other AJCC staging criteria, regression analysis characterized the 31-GEP as a strong independent predictor of metastatic risk along with Breslow thickness, ulceration, age, and higher level of staging. The initial 31-GEP development and validation study revealed the test to be an independent and powerful prognosticator of metastasis in stage I and II CMM patients.

31-Genetic Expression Profile and American Joint Committee on Cancer Online Prognostic Tool

The AJCC Individualized Melanoma Outcome Prediction Tool Web-based prognostic tool is often used by clinicians to attain survival rate estimates and appropriate management recommendations. In a separate follow-up study, Ferris and colleagues compared the accuracy of the 31-GEP test with the accuracy of the AJCC Web-based prognostic tool. For the evaluation, two 5-year overall survival rate cutoffs (79% and 68%) were used because they reflect the 5-year survival rates for CMM stage IIA (79%) and stage IIB (68%). These stages were chosen because patients in these cohorts can receive significantly different surveillance and treatment based upon national guidelines. These rates were used as cutoff scores to establish low- and high-risk groups based on the AJCC-predicted outcomes, which were then compared with the 31-GEP results.

Two hundred five stage I and II CMMs were used in the comparison. Univariate analysis revealed significant risk assessment for all 3 predictors (31-GEP, AJCC 79%, and AJCC 68%). Determination of risk was significantly correlated with recurrence, distant metastasis, and death. Multivariate regression indicated that the 31-GEP is more significantly associated with the endpoints of distant metastasis and death. The 31-GEP also had a significantly higher sensitivity, but lower specificity when compared with the AJCC predictor at both cutoff scores.

The 31-GEP classification combined with the AJCC prognostic tool enhanced sensitivity, reflecting a more accurate identification of high-risk stage I and II melanoma patients. The combination predictor resulted in accurate identification of 90% of cases with recurrences, 88% of distant metastases, and 82% of deaths. Multivariate analysis comparing the combined 31-GEP and AJCC tools also outperformed risk prediction by the AJCC tool alone.

31-Genetic Expression Profile and Sentinel Lymph Node Biopsy

The predictive accuracy of the 31-GEP and SLNB was evaluated individually and in combination in order to assess disease-free survival, distant metastasis-free survival, and overall survival. Two hundred seventeen invasive CCM samples from patients who also underwent the SLNB procedure were analyzed using the 31-GEP test in a multicenter prospective study. Of the 58 patients with a positive SLNB, 37 (64%) experienced a metastatic event, 32 (55%) developed distant metastasis, and 18 (31%) died of all causes. The 31-GEP test identified 141 cases as class 2 (high risk), of which 91 (65%) progressed to metastatic disease. Seventy-one (50%) developed distant metastasis, and 53 (38%) died of all causes. Although both SLNB and the 31-GEP were found to be significant predictors on Cox regression, for each end point the 31-GEP had a higher hazard ratio compared with SLNB ( P <.0001).

Combining the 31-GEP test and SLNB status further improved metastatic risk prognostication. This combination is particularly valuable because the 2 tests may identify patients at high risk for recurrence who otherwise would have been classified as lower risk. The study showed this for the 42% of patients who had a 31-GEP class 2 (high risk) but a negative SLNB. For this group, the 5-year disease-free, metastasis-free, and overall survival were 35%, 49%, and 54%, respectively. These rates were not different from the subset of patients who were class 2 with a positive SLNB. When used in combination with SLNB, the 31-GEP may help clinicians identify high-risk SLNB-negative patients with aggressive disease.

A major limitation of this study was that the overall risk of metastatic events was about 30% higher in the SLNB-negative cohort of patients than is usually found in the general CMM population.

Clinical Utility

Although the 31-GEP has demonstrated reproducibility and clinical validity in assessing recurrence risk, another important aspect is clinical utility: the impact of the test results on clinical decision making. Three unique prospective studies have evaluated the management changes of clinicians using the 31-GEP test.

One study reviewed 156 CMM patients who were consecutively 31-GEP tested from 6 institutions (3 dermatology and 3 surgical oncology practices) to evaluate the clinical impact and influences on physicians directing management. Frequency of clinical visits, frequency and modality of imaging, SLNB procedure utilization, referrals, and frequency of blood laboratory work was compared. Changes in management were observed in 82 (53%) patients with most class 2 patients (77%) undergoing management changes compared with only 37% of class 1 patients ( P <.0001). The majority (77/82, 94%) of these changes were concordant with the risk indicated by the 31-GEP test result with increased management intensity for class 2 patients and reduced intensity for class 1 patients. Similar significant risk-appropriate changes were found in frequency of office visits and use of advanced imaging (PET/computed tomography).

An important finding of this utilization study was the minimal effect of the 31-GEP on recommendation and use of SLNB. Only 2 of the 156 cases had a change in SLNB utilization for which the procedure was performed. However, both cases were 31-GEP class 1, which suggests other factors prompted the modified management. For patients that underwent SLNB (112), the 31-GEP test results appeared impactful on subsequent management. Ninety-nine patients (89%) had a negative result. Within this group, 55 patients were class 1; 27 (49%) patients had a reduction in the intensity of their management, whereas none had an increase. Forty-four of the patients were class 2; 35 (80%) of these patients were given increased management intensity. Although the utility of the 31-GEP test in combination with SLNB status had shown value in prognostication, the test now has shown physicians are able to individualize patient management by increasing or decreasing surveillance according to each patient’s clinical presentation.

Two separate studies also evaluated the 31-GEP clinical utility by dermatologists using patient vignettes. Dermatology resident physicians (n = 169) responded to a series of questions asking to identify the Breslow thickness at which decisions about SLNB, imaging, and oncology referrals would be made. Additional responses were obtained about willingness to use SLNB procedure or imaging based on patient vignettes with variable clinical characteristics. Answers for all questions were compared when class 1, class 2, or no 31-GEP test result information was provided.

When the 31-GEP test results were not provided, most respondents adhered to current management guidelines for recommending SLNB. The addition of the 31-GEP test information led to significant risk-appropriate changes in management decisions. A class 2 result had an impact on SLNB, imaging, and oncology referral in 47%, 50%, and 47% of physician respondents where as a class 1 result led to a change in only 23%, 18%, and 19%. Similar results were found for the vignettes, where a class 1 designation resulted in a significant decrease in recommendations for SLNB and imaging, but an appropriate increase for class 2 results.

Although the resident physician study showed the 31-GEP may appropriately impact clinical decision making, the test was also evaluated by board-certified dermatologists who may be more cautious in adopting newer technology. Similar results were found for the effects of the 31-GEP in clinical management in their study of 214 dermatologists (Glazer AM, et al. Submitted for publication). In their inflection point analysis, a class 2 result caused 69% and 61% of respondents to choose a thinner Breslow thickness of 0.7 mm for recommending SLNB and imaging. The Breslow thickness inflection point selected to guide SLNB and imaging was changed by a class 1 result (41% and 32%, respectively) and class 2 result (66% and 67%, respectively). Interestingly, the experienced dermatologists were more willing to modify patient management than the resident physicians when incorporating the results of the 31-GEP test.

Limitations and Technical Issues

Molecular testing has been used for prognostic purposes, but the clinical use of these methods has been limited by their susceptibility to sampling error resulting from tumor heterogeneity, limited clinical validation, lack of standardized testing, and high technical failure rates. The development and validation set included 268 patients followed by multiple subsequent validations studies. These results have allowed dermatologists to augment management of CMM patients, but given the toxicity of adjuvant therapy, further validation may be needed before oncologists responsible for treating patients with advanced disease use the test in a significant manner. Furthermore, the test does not currently have a role in determining which patients are candidates for adjuvant immunotherapy, either as a standard of care or as part of clinical trials.

The impact of the 31-GEP results on health outcomes may also be further evaluated. Although the endpoint of the utilization study was to analyze changes in clinical management resulting from class assignments, the study was limited by the absence of follow-up data required to correlate the 31-GEP with outcomes. Future studies would benefit from the collection of follow-up data to show the impact of clinical practice adjustments on patient outcomes.

Although there is great promise, the 31-GEP has not been included in any national or professional association guideline recommendations. If the 31-GEP continues to prove to be an accurate predictor of metastasis, the test may guide decision making regarding staging procedures and adjuvant therapy.