Recent developments of gene expression profiling as a prognostic biomarker in melanoma

May 2021 EADO 2021 Tom Feys

Michael Marchetti, MD, Dermatologist, Memorial Sloan Kettering Cancer Center, New York, United States

The use of gene expression profiling (GEP) as prognostic biomarkers is an emerging area of research. Specifically for melanoma, three commercial GEP assays are available. Importantly, the validation data for these three assays vary and they all test for different set of genes. Moreover, some clinical utility studies dispute their applicability to clinical practice. 

Commercially available gene expression profiling tests

Prognostic biomarkers enable better clinical decision making through the prediction of patient outcomes. For this reason, prognostic biomarkers have become an essential part of clinical practice. Unfortunately, the AJCC cancer staging system proved to be inadequate in this respect, with comparable survival rates across disease staging. As a result, when using this model, physicians are both under- and overtreating patients. As an alternative, gene expression profiling could be used in a prognostic manner by measuring the activity of certain oncogenes. For melanoma, three GEP tests are commercially available: DecisionDx Melanoma (Castle Biosciences), Melagenix (NeraCare GmbH) and SkylineDx. Remarkably, there is no overlap between discriminant genes in the three GEP tests. However, there is some functional overlap between the discriminant genes.

Unclear which GEP is superior for sentinel lymph node biopsy selection

GEP can be used to assess the need for a sentinel lymph node biopsy (SLNB) or the risk of recurrence. Currently, the NCCN guidelines recommend against a SLNB if the likelihood of a positive node is below 5%. DecisionDx, which assesses 31 genes, has been evaluated in this setting. A prospective study of 1,421 T1/2 melanoma patients from Vetto JT, et al., found a SLNB positivity rate of 2% in patients over 65, meaning that DecisionDx could potentially reduce SLNB by 52% in this population.1. In another study, CP-GEP (a model using eight discriminative genes combined with age and thickness) was compared against clinicopathology alone in an archival cohort of 745 patients with t1b-t3b melanoma. This model was shown to have a higher specificity compared to the clinicopathological model (49% vs. 32%). In addition, CP-GEP was found to have a negative predictive value of 96%, resulting in a reduction of SLNB by 42%.2 Unfortunately, it is currently not possible to determine which GEP is superior, as direct comparisons are statistically problematic and uninformative.3,4

Disputed clinical utility

Data also support the use of GEPs as a prognostic indicator for recurrence, metastasis or death. The MelaGenix 11-GEP test was evaluated in 245 stage II melanoma patients, finding a 5-year disease-free survival of 76% in patients with a low-risk GEP score, and 58% in high-risk patients.5 The CP-GEP model has also been applied to 831 patients who received a SLNB, finding that biopsy-negative patients who had a low or high CP-GEP risk had significantly different survival outcomes (HR: 2.40[1.53-3.74], p< 0.001).6

The 31-GEP test has the largest body of validation data, with more than 10 studies reporting a statistically significant discrimination between GEP high and low risk patients in relation to survival outcomes. A recent meta-analysis of 1,479 melanoma patients found the 31-GEP assay to be a significant prognostic indicator of survival (HR[95%CI]: 2.90[2.01-4.19]).7 However, contradictory evidence also exists, with one single-centre study finding that the 31-GEP test was not associated with recurrence free survival in a multivariate analysis.8 Marchetti et al, performed a meta-analysis, assessing stage-specific performance, in 623 patients with stage I disease. Of the 21 patients who experienced recurrence, only 6 (29%) were high-risk GEP patients. Conversely, of the 602 patients who had no recurrence, 541 (90%) turned out to be GEP low-risk. Results were different for the 212 patients with stage II disease. In these patients, the majority of patients who had a recurrence, were also classified as being high-risk (82%). On the flipside, however, the majority of patients without a recurrence where incorrectly told to have a high-risk. As such, sensitivity and specificity seems to reverse between stage I and stage II disease.9 As a result, 31-GEP may be more applicable in the selection of stage II patients who require more aggressive interventions. These results may also be explained by the fact that the 31-GEP training set included 67 metastases overall, with only 9% being <1mm thick, 13% being stage I and 58% of cases being stage II.10 The clinical utility of the 31-GEP test has also been established in a study by Dillon et al, For 269 melanoma patients, pre- and post- GEP test recommendations were collected to determine changes in management resulting from the addition of GEP testing. In 49% of cases, management plans changed after results of 31-GEP.11 Decision curve analysis is another way to quantify clinical benefit, and has been applied to the CP-GEP model. At a 5% threshold risk for SLNB, there was seemingly no benefit over a CP-only model.12 Applying this analysis to disease staging, there was a low probability of net benefit in stage I disease with 31-GEP, whereas the probability for benefit in stage II disease was far greater.

Conclusion

Gene expression profiling can provide valuable insights into the risk of recurrence and need for SLNB in melanoma patients. Furthermore, these tests can be incorporated into prognostic models to predict patient outcome. However, contradictory evidence disputes the additional clinical utility these tests provide. Ultimately, GEP tests should not be used blinded to gain additional prognostic information. Instead, testing should be based on the pre-test probability of the outcome. Finally, other prognostic markers, such as T-cell fraction and 121-GEP testing may offer additional clinical utility in the future.

References

  1. Vetto JT et al., Guidance of sentinel lymph node biopsy decisions in patients with T1-T2 melanoma using gene expression profiling. Future Oncol. 2019; 15(11).
  2. Bellomo D et al., Model Combining Tumour Molecular and Clinicopathologic Risk Factors Predicts Sentinel Lymph Node Metastasis in Primary Cutaneous Melanoma. JCO Precis Oncol. 2020; 4: 319-334.
  3. Baker SG. Putting Risk Prediction in Perspective: Relative Utility Curves. J Natl Cancer Inst. 2009; 101(22): 1538-1542.
  4. Vickers AJ, Elkin EB., Decision curve analysis: a novel method for evaluating prediction models. 2006; 26(6): 565-74.
  5. Amaral TMS et al., Clinical validation of a prognostic 11-gene expression profiling score in prospectively collected FFPE tissue of patients with AJCC v8 stage II cutaneous melanoma. Eur J Canc. 2020; 125: 38-45.
  6. Eggermont AMM et al., Identification of stage I/IIA melanoma patients at high risk of disease relapse using a clinicopathologic and gene expression model. Eur J Canc. 2020; 140: 11-18.
  7. Greenhaw B et al., Molecular risk prediction in cutaneous melanoma: A meta-analysis of the 31-gene expression profile prognostic test in 1,479 patients. JAAD. 2020; 83(3): 745-753.
  8. Kangas-Dick AW et al., Evaluation of a Gene Expression Profiling Assay in Primary Cutaneous Melanoma. Ann Surg Oncol. 2021.
  9. Marchetti M et al., Performance of Gene Expression Profile Tests for Prognosis in Patients With Localized Cutaneous Melanom: A Systematic Review and Meta-analysis. JAMA Dermatol. 2020; 156(9): 953-962.
  10. Gerami P et al., Development of a Prognostic Genetic Signature to Predict the Metastatic Risk Associated with Cutaneous Melanoma. Clin Cancer Res. 2015; 21(1): 175-183.
  11.   Dillon LD et al., Prospective, Multicenter Clinical Impact Evaluation of a 31-Gene Expression Profile Test for Management of Melanoma Patients. SKIN. 2018; 2(2): 111-121.
  12. Bartlett EK et al., Gene Expression Profile-Based Risk Modeling to Select Patients With Melanoma Who Can Avoid Sentinel Lymph Node Biopsy: Are We There Yet?. JCO Precis Oncol. 2020; 4: 988-989.