Performance of the 23-gene Expression Profile (23-GEP) Test by Histopathological Evaluation in an Independent, Multi-Center Performance Cohort of Cutaneous Melanocytic Neoplasms

Abstract

Histopathologic evaluation can effectively diagnose most melanocytic neoplasms; however, lesions considered to be difficult-to-diagnose pose challenges for accurate classification of malignant potential, which can lead to over- or under-treatment. Ancillary testing to provide additional information is available for such cases. The validated diagnostic 23-GEP test provides an objective result of benign, malignant, or intermediate. In this large, independent cohort, the performance of the 23-GEP in its current laboratory is presented. Lesions from patients ≥18 years old were enrolled from eight centers or from melanoma cases clinically submitted for prognostic 31-GEP testing. Lesions were independently reviewed by 3–5 dermatopathologists and included in the study if they were fully concordant or non-concordant without conflicting diagnoses (i.e., both benign and malignant designations) or majority of unknown malignant potential designations, resulting in a cohort (n=2512) of benign nevi (n=1140) and malignant melanomas (n=1372). Accuracy metrics and two-tailed 95% confidence intervals (CIs) were calculated using resampling x10,000 iterations to establish a balanced number of benign versus malignant samples. The 23-GEP performance within this cohort was 91.3% (95% CI, 89.2–93.2%) sensitivity, 91.9% (89.8–93.8%) specificity, 92.2% (9.30–94.0%) positive predictive value, and 91.0% (89.0–92.9%) negative predictive value; 7.8% of lesions received an intermediate result. These metrics do not deviate significantly from previously published studies. These data demonstrate that the 23-GEP has an overall accuracy of 91.6% (90.1–93.0%), further supporting its use as an ancillary test that can be integrated with clinical and histopathologic information to guide final diagnosis.