According to National Institutes of Health estimates, more than 90,000 new cases of melanoma were diagnosed in 2018.1 Overall 5-year survival for patients with melanoma exceeds 90%, but individual survival estimates are highly dependent on stage at diagnosis, and survival decreases markedly with metastasis. Therefore, early and accurate diagnosis is critical.
Diagnosis of melanocytic neoplasms usually is performed by dermatopathologists through microscopic examination of stained tissue biopsy sections, a technically simple and effective method that enables a definitive diagnosis of benign nevus or malignant melanoma to be made in most cases. However, approximately 15% of all biopsied melanocytic lesions will exhibit some degree of histopathologic ambiguity,2-4 meaning that some of their microscopic features will be characteristic of a benign nevus while others will suggest the possibility of malignant melanoma. Diagnostic interpretations often vary in these cases, even among experts, and a definitive diagnosis of benign or malignant may be difficult to achieve by microscopy alone.2-4 Because of the marked reduction in survival once a melanoma has metastasized, these diagnostically ambiguous lesions often are treated as possible malignant melanomas with complete surgical excision (or re-excision). However, some experts suggest that many histopathologically ambiguous melanocytic neoplasms are, in fact, benign,5 a notion supported by epidemiologic evidence.6,7 Therefore, excision of many ambiguous melanocytic neoplasms might be avoided if definitive diagnosis could be achieved.
A gene expression signature was developed and validated for use as an adjunct to traditional methods of differentiating malignant melanocytic neoplasms from their benign counterparts.8-11 This test quantifies the RNA transcripts produced by 14 genes known to be overexpressed in malignant melanomas by comparison to benign nevi. These values are then combined algorithmically with measurements of 9 reference genes to produce an objective numerical score that is classified as benign, malignant, or indeterminate. When used by board-certified dermatopathologists and dermatologists confronting ambiguous melanocytic lesions, the test produces substantial increases in definitive diagnoses and prompts changes in treatment recommendations.12,13 However, the long-term consequences of foregoing surgical excision of melanocytic neoplasms that are diagnostically ambiguous but classified as benign by this test have not yet been formally assessed. In the current study, prospectively tested patients whose ambiguous melanocytic neoplasms were classified as benign by the gene expression signature were followed for up to 4.5 years to evaluate the long-term safety of treatment decisions aligned with benign test results.
As part of a prior study,12 US-based dermatopathologists submitted tissue sections from biopsied melanocytic neoplasms determined to be diagnostically ambiguous by histopathology for analysis with the gene expression signature (Myriad Genetics, Inc). Diagnostically ambiguous lesions were those lesions that were described as ambiguous, uncertain, equivocal, indeterminate, or other synonymous terms by the submitting dermatopathologist and therefore lacked a confident diagnosis of benign or malignant prior to testing. Patients initially were tested between May 2014 and August 2014, with samples submitted through a prospective clinical experience study designed to assess the impact of the test on diagnosis and treatment decisions. This study was performed under an institutional review board waiver of consent (Quorum #33403/1).
Patients were eligible for inclusion in the current study if their biopsy specimens (1) had an uncertain preliminary diagnosis according to the submitting dermatopathologist (pretest diagnosis of indeterminate); (2) received a negative (benign) score from the gene expression test; (3) were treated as benign by the dermatologist(s) involved in follow-up care; and (4) were submitted by a single site (St. Joseph Medical Center, Houston, Texas). Although a single dermatopathology site was used for this study, multiple dermatologists were involved in the final treatment of these patients. Patients with benign scores who received additional intervention were excluded, as they may have a lower rate of adverse events (ie, metastasis) than those who did not receive intervention and would therefore skew the analysis population. A total of 25 patients from the prior study met these inclusion criteria. The previously collected12 pretest and posttest de-identified data were compiled from the commercial laboratory databases, and the patients were followed from the time of testing via medical record review performed by the dermatology providers at participating sites. Clinical follow-up data were collected using study-specific case report forms (CRFs) that captured the following: (1) the dates and results of clinical follow-up visits; (2) the type(s) of treatment and interventions (if any) performed at those visits; (3) the specific indication for any intervention performed; (4) any evidence of persistent, locally recurrent, and/or distant melanocytic neoplasia (whether definitively attributable to the tested lesion or not); and (5) death from any cause. The CRF assigned interventions to 1 of 5 categories: excision, excision with sentinel lymph node biopsy, referral to dermatologic or other surgeon, examination only (without surgical intervention), and other. Selection of other required a free-text description of the treatment and indications. Pertinent information not otherwise captured by the CRF also was recordable as free text.
Gene Expression Testing
Gene expression testing was carried out at the time of specimen submission in the prior study12 as described previously.14 Briefly, formalin-fixed, paraffin-embedded, unstained tissue sections and/or tissue blocks were submitted for testing along with a single hematoxylin and eosin–stained slide used to identify and designate the representative portion(s) of the lesion to be tested. These areas were macrodissected from unstained tissue sections and pooled for RNA extraction. Expression of 14 biomarker genes and 9 reference genes was measured via quantitative reverse transcription–polymerase chain reaction, performed in triplicate for each individual gene. The assay score was generated through application of a weighted algorithm to the expression values generated through quantitative reverse transcription–polymerase chain reaction. Scores were plotted on a scale ranging from −16.7 to 11.1, with scores from 0.0 to 11.1 classified as malignant, scores from −16.7 to −2.1 as benign, and scores from −2.1 to −0.1 as indeterminate.
Demographic and other baseline characteristics of the patient population were summarized. Follow-up time was calculated as the interval between the date a patient’s gene expression test result was first issued to the provider and the date of the patient’s last recorded visit during the study period. All patient dermatology office visits within the designated follow-up period were documented, with a nonstandard number of visits and follow-up time across all study patients. Statistical analyses were conducted using SAS software (SAS Institute Inc), R software version 3.5.0 (R Foundation for Statistical Computing), and IBM SPSS Statistics software (IBM SPSS Statistics for Windows, Version 25).