Gene expression profiles help identify metastasis in primary cutaneous melanoma




In patients presenting with cutaneous melanoma, use of molecular data can help determine whether there is a high likelihood of nodal metastasis that warrants sentinel lymph node biopsy, according to a report published online in the Journal of Clinical Oncology.

A large number of sentinel lymph node (SLN) biopsies could be avoided if nodal metastasis in patients with primary cutaneous melanoma were better identified at the time of diagnosis, according to Dr. Alexander Meves of the department of dermatology at the Mayo Clinic in Rochester, Minn., and colleagues.

“In this study, we found that molecular data in combination with Breslow depth, tumor ulceration, and patient age were useful for discriminating between primary cutaneous melanomas that had or had not metastasized to SLN,” they wrote (J. Clin. Oncol. 2015 July 6 [doi:10.1200/JCO.2014.60.7002]).

Based on samples from 160 patients that included benign nevi and primary skin melanomas with and without SLN metastasis, investigators identified genes differentially expressed between metastatic and nonmetastatic pigmented skin lesions, and found a cluster of genes associated with integrin cell adhesion. Of particular interest, the integrin cell adhesion receptor, beta-3 integrin (ITGB3), was upregulated in regionally metastatic melanoma.

Information from a cohort of 360 patients, including 74 (20.6%) with biopsy-confirmed nodal metastasis, was used to derive prediction models for SLN metastasis based on clinicopathologic factors alone and in combination with molecular data. Younger age, tumor ulceration, and greater Breslow depth contributed to the clinicopathologic model. The combined model added expression data from four genes: ITGB3, cellular tumor antigen p53 (TP53), laminin B1 chain (LAMB1), and tissue-type plasminogen activator (PLAT; protein name, t-PA). Area under the receiver operating characteristic (ROC) curve for the combined clinicopathologic plus molecular model was 0.89, compared with 0.78 for the clinicopathologic model alone (P < .001). Performance of the models on the validation cohort (n = 146) was similar, with the area under the ROC curve at 0.93. Using a 10% cutoff, the false-positive rate was 22% and the false-negative rate was 0%.

Expression data combined with clinicopathologic features can be used to calculate the predicted probability of SLN positivity at the time of primary diagnosis, which has “the potential to improve patient care by avoiding unnecessary SLN procedures,” the authors wrote.

Next Article: