Long-Term Outcomes Support Utilization of 23-GEP in Clinically Ambiguous Melanocytic Neoplasms

Supplemental Digital Content is Available in the Text.

Abstract:

Difficult-to-diagnose melanocytic neoplasms often require additional diagnostic tools beyond histopathological assessment to reach a definitive diagnosis. The 23-gene expression profile (GEP) test provides additional diagnostic information, returning a result of suggestive of benign, intermediate, or suggestive of malignant. Here, we present a real-world cohort of ambiguous melanocytic lesions where the 23-GEP test was used to arrive at a definitive diagnosis with long-term follow-up. Melanocytic lesions were included in this study if 23-GEP testing was performed as part of their original diagnostic workup (n = 267). Long-term follow-up to determine disease recurrence and/or metastasis (ie, an event) was obtained (median, 6 years). Following 23-GEP testing, 89.5% of difficult-to-diagnose lesions were provided definitive diagnoses. Event rates were 11.5% for lesions with malignant 23-GEP results, 2.9% for lesions with intermediate results, and 1.3% for lesions with benign results (Fisher exact, P = 0.002). Lesions with at least 5 years of clinical follow-up and/or an event (n = 163) had similar results to the overall cohort (P = 0.007). A benign 23-GEP result is associated with a low risk of recurrence/metastasis, whereas a malignant 23-GEP result is associated with a significantly higher risk of recurrence/metastasis, supporting the value of 23-GEP in guiding clinical management decisions for ambiguous melanocytic lesions.

INTRODUCTION

Most benign nevi are regarded as clinically insignificant lesions and, as a result, typically do not require further therapeutic intervention. Management strategies are significantly intensified, however, when a diagnosis of melanoma is established. The National Comprehensive Cancer Network and American Academy of Dermatology provide clear and detailed guidance for patients with melanoma, including primary treatment, consideration of sentinel lymph node biopsy, and structured clinical follow-up that can involve multidisciplinary care teams, consideration of surveillance imaging, adjuvant therapy, and treatment of recurrence and metastatic disease based on a patient's risk of disease progression. Given the divergence of management pathways for benign nevi and melanomas, as well as the impactful implications of a misdiagnoses, accurate classification of histologically ambiguous lesions is clinically critical.^1,2

Ambiguous melanocytic neoplasms are difficult to diagnose and subject to variability in interpretation between pathologists.^3–5 Without a definitive diagnosis, ambiguous lesions have varied surgical treatment and subsequent management, which may have negative implications for the patient, their family, and the healthcare system.⁶ Moreover, a missed melanoma can have devastating consequences, placing patients at risk of poor outcomes due to the likelihood of a more advanced stage of disease by the time an accurate diagnosis is rendered.

Ancillary tools such as gene expression profiling (GEP) can be used by dermatopathologists to assist in classifying ambiguous lesions as benign or malignant⁷ and are recommended by several patient management guideline entities.^1,8–10 The 23-GEP uses the differential expression of 14 discriminant genes and 9 control genes to differentiate nevi from melanoma and provides 1 of 3 test results: suggestive of benign, intermediate, or suggestive of malignant.^11,12 Accuracy metrics of 90.0%–96.3% sensitivity and 87.3%–96.2% specificity have been described across many studies.^11,13–17 The test performance of the 23-GEP has been described by the Appropriate Use Committee of the American Society of Dermatopathology as ‘Majority appropriate use’ in 6 categories including “Distinction of nevus from primary melanoma in an adult patient when the morphologic findings are ambiguous by light microscopic parameters.”⁸

In ambiguous samples clinically tested with the 23-GEP, performance of 92.1% (95% confidence interval: 84.2%–100%) sensitivity and 94.7% (95% confidence interval: 86.8%–100%) specificity compared with histopathology has been reported (n = 369; intermediate cases n = 44), with a median follow-up of 3.3 years.¹⁸ However, there is a lack of information related to long-term follow-up of patients with 23-GEP test use during the diagnostic process. The 23-GEP has been available commercially since 2013 and was used at Southern California Kaiser Permanente as part of the diagnostic workup of ambiguous melanocytic lesions. The aim of this study was to further confirm the high performance of the 23-GEP in difficult-to-diagnose lesions in a real-world community practice cohort with long-term outcome data.

METHODS

Study Design and Cohort Details

Cases from Kaiser Permanente with a valid 23-GEP test result with diagnostic evaluation between 2013 and 2019 were curated. All melanocytic lesions were histopathologically ambiguous on initial review. The 23-GEP test was used during the histopathological review and final diagnosis determination of the cases. Limited patient demographic information was collected along with limited follow-up information related to disease progression. Melanoma recurrences and/or metastases were considered events. Cases with nonprimary tumors were excluded (n = 6). A secondary subset was created including only patients with at least 5 years of follow-up and/or a disease progression event (n = 163 with 12 events).

23-GEP

23-GEP was performed as described previously.^11,12 In brief, RNA was extracted from 9 unstained FFPE slides with at least 10% tumor content by macrodissection, and qRT-PCR was performed. A logistic regression algorithm based on 3 groups of defined gene sets was used to calculate a score and test result of suggestive of benign (−16.7 to −2.1), malignant (0.0–11.1), or intermediate (cannot exclude malignancy) (−2.0 to −0.1).

Diagnostic Binning

Final diagnoses were obtained from electronic medical records and binned as atypical melanocytic proliferation (AMP), atypical Spitz tumor (AST), melanoma in situ (MIS), melanoma, or nevus according to the following parameters. “Favor” or “consistent with” or similar language was used for primary binning. Any lesion that had features of a spitzoid neoplasm was binned as AST. Lesions described as nevi of any kind (other than spitz or spindle cell) were binned in the nevus category. Lesions with verbiage of “Features of” were categorized as an AMP.

Statistical Methods

All analyses were performed using R (v4.3.0) with open-source R packages (stats v4.5.0). Fisher's exact test was used to compare metastatic event rates between 23-GEP result categories. An alpha of 0.05 was used for all statistical comparisons.

RESULTS

Clinically Ambiguous Lesions Tested with 23-GEP

A total of 267 clinically ambiguous lesions from a single center met the study inclusion criteria (Table 1). All neoplasms with biopsies dated from 2013 to 2019 that received the 23-GEP as part of the original histopathological workup to aid in diagnostic clarification were included. The median follow-up time for the full cohort was 6 years (range of 0–10 years). The median age was 36 years, and 54 patients were age 18 years or younger (20.2%). Breslow information was reported for a subset of lesions during the diagnostic work that were favored to represent invasive melanoma (n = 109). T-category for the invasive melanomas (n = 67) is reported in Table 1 with T2a lesions having the highest representation at 28.4%. A small percentage of lesions were ulcerated (4.9%). Lesion subtype descriptions of the full cohort are described in Table 2, and a full list of final diagnoses for the entire cohort is available as Supplemental Digital Content (see Supplemental Table 1, https://links.lww.com/AJDP/A195). Spitzoid lesions were predominant in this study (51.1%), highlighting the level of ambiguity/difficulty in this cohort.

TABLE 1.

Patient Characteristics From Clinically Ambiguous Lesions (n = 267)

Age* (n = 266)
Median (range)	36 (0–89)
Sex (n = 267)
Female, n (%)	151 (56.6%)
Breslow depth†, n (%) (n = 109)
<1 mm	66 (60.6%)
1–2 mm	26 (23.9%)
2–4 mm	10 (9.2%)
>4 mm	7 (6.4%)
Ulceration, n (%) (n = 267)
N/A	114 (42.7%)
No	140 (52.4%)
Yes	13 (4.9%)
T-stage‡, n (%) (n = 67)
T1a	18 (26.9%)
T1b	9 (13.4%)
T2a	19 (28.4%)
T2b	4 (6.0%)
T3a	9 (13.4%)
T3b	1 (1.5%)
T4a	5 (7.5%)
T4b	2 (3.0%)

^*Age was missing in 1 of the patients.

^†Breslow information was reported for a subset of lesions during the diagnostic work that were difficult to distinguish from invasive melanoma.

^‡T-stage was not available for 8 patients with a final diagnosis of invasive melanoma.

N/A, not available.

TABLE 2.

Lesion Growth Patterns of the Clinically Ambiguous Lesions (n = 267)

Growth Pattern, n (%)
Acral nevus/acral lentiginous	2 (0.7%)
Blue nevus/blue	21 (7.9%)
Combined nevus	20 (7.5%)
Common/superficial spreading	18 (6.7%)
Compound/nevoid	41 (15.4%)
Deep penetrating/nodular	10 (3.7%)
Desmoplastic	7 (2.6%)
Dysplastic/lentigo maligna	8 (3.0%)
Spitz/spitzoid	136 (50.9%)

Growth patterns were not available for 4 lesions.

23-GEP Aids in the Classification of Ambiguous Lesions

The 23-GEP contributed to the number of definitive diagnoses from presumed zero to 89.5% (Table 3). Overall, 33.7% of patients received a diagnosis of nevus, and 35.2% received diagnosis for melanoma/MIS, both of which have defined treatment guidelines and clear management pathways. An AST diagnosis was provided for 20.6% of the cohort, further evidencing the level of case complexity included in this cohort. Importantly, the number of lesions that received a nondefinitive diagnosis of AMP or a similar variation decreased to 10.5%.

TABLE 3.

Final Diagnoses in the Clinically Ambiguous Lesions (n = 267)

Diagnosis, n (%)
AMP	28 (10.5%)
AST	55 (20.6%)
Melanoma	75 (28.1%)
MIS	19 (7.1%)
Nevus	90 (33.7%)

Clinically ambiguous lesions were provided a nondefinitive diagnosis of atypical melanocytic proliferation (AMP), or a final, definitive diagnosis of atypical Spitz tumor (AST), melanoma, nevus, or melanoma in situ (MIS) post-23-GEP testing.

Long-term Outcomes of Lesions with 23-GEP Testing

Long-term outcomes for patients with 23-GEP test results are described in Table 4. A significant difference in event rate (ie, recurrence and/or metastases) was detected between the three 23-GEP result categories (Fisher exact test, P = 0.002). Lesions with a 23-GEP result had event rates of 1.3% for benign, 2.9% for intermediate, and 11.5% for malignant. The negative predictive value of the 23-GEP was calculated to be 98.7%. Two 23-GEP benign lesions had disease progression. Lesions with a 23-GEP malignant result had a 9× higher event rate (11.5%) than lesions with a 23-GEP benign result (1.3%).

TABLE 4.

Performance of 23-GEP Clinical Outcomes

	No Event (%)	Event (%)
23-GEP benign	153 (98.7)	2 (1.3)
23-GEP intermediate	33 (97.1)	1 (2.9)
23-GEP malignant	69 (88.5)	9 (11.5)

Fisher exact test, P-value = 0.002.

Analyses of lesions with ≥5 years of follow-up time and/or an event (ie, recurrence and/or metastases) were also considered (≥5 years cohort) (n = 163). The median follow-up time for the ≥5 years cohort was 7.5 years. Long-term outcomes are described in Table 5. Similar to the full cohort, statistical differences were detected in event prevalence between the 23-GEP result categories (Fisher exact test, P = 0.007). Lesions with 23-GEP benign results had an event rate of 2.2%, intermediate results had an event rate of 5.3%, and malignant results had an event rate of 16.4%.

TABLE 5.

Performance of 23-GEP by Clinical Outcome in Lesions With 5+ yrs Follow-up/Event

	No Event (%)	Event (%)
23-GEP benign	87 (97.8)	2 (2.2)
23-GEP intermediate	18 (94.7)	1 (5.3)
23-GEP malignant	46 (83.6)	9 (16.4)

Fisher exact test, P-value = 0.007.

DISCUSSION

In a large cohort of histopathologically ambiguous lesions for which the 23-GEP test was used during the diagnostic process from 2013 to 2019, accuracy was determined by analyzing final diagnoses and long-term outcomes. The majority of lesions in the study were definitively classified into guideline-directed treatment pathways, indicating appropriate clinical utility and risk-aligned care for patients.^1,9 Long-term follow-up of patients with benign 23-GEP results revealed an excellent NPV of 98.7%, and these patients had statistically significantly lower event rates (1.3%) than patients with intermediate (2.9%) or malignant (11.5%) 23-GEP results. This evidence demonstrates that the 23-GEP test provides accurate diagnostic information that can be useful in establishing risk of disease progression in patients with ambiguous lesions.

Receiving a definitive diagnosis that aligns with guideline-associated treatment pathways has impactful implications for patient health.² A definitive diagnosis can ensure that patients with melanoma (∼35% of this real-world cohort) receive appropriate, risk-aligned care and consideration of radiologic imaging and laboratory studies, sentinel lymph node biopsy, collaboration with medical and surgical oncology, increased clinical follow-up for recurrence detection, genetic counseling for the patient and their family, and prognostic testing.^1,9 For nonmelanoma definitive diagnoses (nevus and AST), which made up ∼65% of this real-world cohort, patients avoid the implications of a melanoma overdiagnosis, which can impact the patient, their family, and the associated healthcare system.^19,20 ASTs represent a subset of very challenging-to-interpret melanocytic neoplasms, which are generally recommended to be treated by definitive surgery to ensure complete removal of the lesion.²¹ Fifty-five lesions (20.7%) in this cohort were provided with a final diagnosis of AST, and 1 AST patient had disease progression. Owing to the retrospective nature of this study design, pretest diagnoses were not always documented, and therefore, if they were not available, were presumed ambiguous.^22,23

The minimal events observed in the patients who received a benign 23-GEP test result are consistent with prior performance data where detection of melanoma in lesions with disease progression was 90.4%–96.3% across 3 studies.^15–17 In fact, in the study by Tschen et al,²⁴ no patients with a 23-GEP benign result had an event (n = 25). Patients with a 23-GEP malignant test result in this study were more likely to have disease progression when compared with the other test results (ie, benign or intermediate); these trends were also demonstrated in the cohort of patients with ≥5 years of follow-up. Overall, event rates were 9× more likely for patients with malignant 23-GEP test results compared with patients with benign results and 8× more likely for patients with more than 5 years of follow-up. Interestingly, lesions with intermediate results showed an incrementally increased risk of metastatic melanoma.

The data described herein are from patients from a single center and from a single board-certified dermatopathologist, which may not reflect scenarios in other settings. However, the data were from a large cohort representing real-world use and included long-term outcomes, enabling assessment of event rates between the 23-GEP results classes. Lesion body location and details related to long-term treatment were not collected as part of this study.

In summary, diagnostic ambiguity was resolved in ∼90% of cases when the 23-GEP was used during histopathological workup, ensuring most patients received risk-aligned management. The 23-GEP provided results of suggestive of benign, intermediate, or suggestive of malignant, with statistically different disease progression event rates as determined by long-term follow-up. Most importantly, in this study, the 23-GEP benign result was obtained for patients that experienced a very low rate of disease progression after more than 5 years of follow-up. Therefore, when the distinction of a nevus from primary melanoma in an adult patient cannot be made definitively, the 23-GEP test can contribute toward resolving diagnostic ambiguity, and patients with lesions that return the suggestive of benign result are unlikely to experience disease progression.

Supplementary Material

ajd-48-878-s001.docx ^{(32KB, docx)}