FC19 Poster Castle Biosciences 3 CM_01_011.FCD 2019.accuracy metrics.V7 Recurrence prediction in cutaneous melanoma with 31-gene expression profiling and sentinel lymph node biopsy Graham Litchman DO1, MS, Hillary Caruso PhD2, Kyle Covington PhD2, Robert Cook PhD2, Darrell Rigel MD, MS3 1. National Center for Cutaneous Melanoma 2. Castle Biosciences, Inc. 3. New York University School of Medicine BACKGROUND •The prognostic 31-GEP test is an accurate predictor of metastatic risk for patients with cutaneous melanoma, stratifying patients into low (Class 1A), intermediate (Class 1B/2A), and high-risk (Class 2B) 1-2. •Previous prospective and retrospective studies have validated the 31- GEP as a strong predictor of risk of recurrence and distant metastasis, independent of other clinical and pathological features 2-7. •The test has demonstrated clinical utility in guiding patient management, including frequency of follow-up, use of imaging for surveillance, and referral for interdisciplinary care 8-12. •Use of the 31-GEP to guide sentinel lymph node biopsy (SLNB) decisions has recently been validated, such that patients with T1-T2 melanoma over the age of 55 with a GEP Class 1A result have a very low probability of SLN positivity and favorable outcomes11. To perform a comparative evaluation of the prognostic accuracy of the 31-GEP and sentinel lymph node biopsy. OBJECTIVE REFERENCES & ACKNOWLEDGMENTS CONCLUSIONS AND SIGNIFICANCE • These results show that the 31-GEP detects more patients with high-risk melanoma than SLNB and combining both achieves best sensitivity. • The high NPV of the 31-GEP test assures that patients with low-risk results have low probability of metastasis or death from melanoma. • Use of 31-GEP to guide SLNB decisions does not alter the sensitivity of combined 31-GEP and SLN status on prognosis. FUNDING & DISCLOSURES This study was sponsored by Castle Biosciences, Inc. HC, KC, and RC are employees and options holders of Castle Biosciences, Inc. GL is a fellow of the National Society for Cutaneous Medicine which receives grants from Castle Biosciences. DSR served as a consultant to Castle Biosciences, Inc.. METHODS RESULTS • We evaluated the prognostic accuracy of the 31-GEP and SLNB in a cohort of 1479 patients comprised of 3 previously described cohorts4-6 and a novel cohort of stage I-III cutaneous melanoma cases collected from 7 institutions under an IRB-approved protocol. • We also compared the sensitivity and NPV of 31-GEP and SLNB in a model which assumed use of the 31-GEP to triage patients for SLNB. GEP SLN GEP and SLN RFS n=1479 n=867 n=1479 Sensitivity 76% (71-80%) 57% (51-63%) 86% (82-90%) Specificity 76% (73-78%) 74% (70-77%) 69% (66-72%) PPV 46% (42-50%) 50% (44-56%) 43% (41-45%) NPV 92% (90-94%) 79% (75-82%) 95% (93-96%) DMFS n=1223 n=867 n=1223 Sensitivity 76% (70-82%) 61% (55-68%) 88% (83-92%) Specificity 69% (66-72%) 72% (68-75%) 60% (57-63%) PPV 35% (31-39%) 39% (34-44%) 32% (30-34%) NPV 93% (91-95%) 86% (83-89%) 96% (94-97%) MSS n=1157 n=630 n=1157 Sensitivity 86% (77-92%) 72% (60-80%) 94% (87-98%) Specificity 65% (62-68%) 67% (64-70%) 57% (54-60%) PPV 17% (13-20%) 16% (13-21%) 15% (14-16%) NPV 98% (97-99%) 96% (95-98%) 99% (98-100%) Variable All Cases (n=342) Age, median (range) 61 (18-94) Breslow thickness, mm median (range) 1.2 (0.1-29) Ulceration present 346 (23.4%) AJCC stage I 852 (57.6%) II 313 (21.1%) III 312 (21.1%) Table 1. Demographics of the 1479 patients in study Table 2. Accuracy metrics of GEP and SLN Only patients with reported end points were included in analysis. Patients from Greenhaw, et. al. did not have DMFS or SLN status reported. Patients from Hsueh, et. al. did not have MSS reported. Accuracy metrics of combined tests determined for patients who had positive SLN OR GEP Class 2 result. Study Design n Novel cohort Archival, multi-center 211 Gastman et al.4 Archival, multi-center 690 Greenhaw et al.6 Prospective, single center 256 Hsueh et al.5 Prospective, multi-center 322 1. Gerami et al. Clin Cancer Res 2015;21:175-83. 2. Gerami et al. J Am Acad Dermatol 2015;72:780-5 e783. 3. Zager et al. BMC Cancer 2018;18(1):130. 4. Gastman et al. JAAD 2019;80(1):149-157. 5. Hsueh et al. J Hematol Oncol 2017;10:152. 6. Greenhaw et al. Dermatol Surg 2018; Dec;44(12):1494-1500. 7. Keller et al. Cancer Med 2019. Apr.5 doi:10.1002/cam4.2128. 8. Berger et al. Curr Med Res Opin 2016;32(9):1599-1604 9. Dillon et al. J Cut Med 2018;2(3):111-121 10. Schuitevoerder et al. J Drugs Dermatol 2018;17(2)196-199 11. Vetto et al. Future Oncol. 2019 Apr;15(11):1207-1217. 12. Burman, et al. SKIN. 2019 Sept; 3(5): 291-306. T1-T2 T3-T4 Age <55 Age ≥ 55 and Class 1B, 2A, 2B Age ≥ 55 and Class 1A GEP Result SLN status GEP Result GEP Result SLN status All Ages Table 2. Sensitivity and NPV of 31-GEP and SLNB in a model assuming use of 31-GEP to guide SLNB decisions GEP SLN GEP and SLN RFS n=1479 n=842 n=1479 Sensitivity 76% (71-80%) 56% (50-62%) 84% (79-88%) NPV 92% (90-94%) 79% (75-82%) 94% (93-95%) DMFS n=1223 n=842 n=1223 Sensitivity 76% (70-82%) 61% (53-68%) 86% (80-90%) NPV 93% (91-95%) 86% (83-89%) 95% (94-97%) MSS n=1157 n=609 n=1157 Sensitivity 86% (77-92%) 72% (60-82%) 94% (97-98%) NPV 98% (97-99%) 96% (95-98%) 99% (98-100%)