ADAPT-M: A tool to predict individualized benefit from adjuvant anti–PD-1 (PD-1) in resected high-risk melanoma.

Abstract

9567 Background: Resected stage IIB/C and microscopic stage IIIA/B/C/D melanoma presents a significant risk of recurrence risk of 37 to 62% at 5 years (yrs). While 1 year of adjuvant PD1 therapy reduces this risk by approximately half, the individual patient’s baseline risk without treatment and their specific benefit from adjuvant PD1 remain unknown. We aimed to generate two separate predictive models for recurrence – one for untreated and one for treated pts - to calculate the individualised benefit of adjuvant PD1. Methods: Pts with resected stage IIB/C or microscopic stage IIIA/B/C/D melanoma, either treated with adjuvant PD1 or untreated at 13 major melanoma centres, and with at least 2 years of follow-up from surgery were included. We analysed pts demographics, disease characteristics, blood parameters, pathological and imaging data at baseline, and clinical outcomes. Propensity scores were estimated using logistic regression with covariates associated with treatment in univariate screening (p<0.05). Separate penalised multivariable logistic regression models were built to predict recurrence for treated and untreated pts. A tool based on these models was created to calculate the individual patient benefit from adjuvant PD1. Results: A total of 3560 pts was included, divided into untreated (discovery n=2384 and validation n=597) and treated (discovery n=404 and validation n=175) cohorts. The model for untreated pts included age, gender, primary site, histological subtype, Breslow thickness, ulceration, mitosis, number and site of regional lymph nodes (LN), mutation status and LDH (AUC 0.70 in discovery, 0.69 in validation). The model for treated pts included age, gender, Breslow thickness, site of regional LN, and mutation status (AUC 0.69 in discovery, 0.61 in validation). ADAPT-M (ADjuvant Anti-PD-1 Tool for Melanoma) calculates the individualized benefit of adjuvant PD1. For example, patient A, a woman ≤45 years old with a primary head & neck melanoma (superficial spreading, thickness >1-2 mm, no ulceration, 1 mitosis/mm 2 ), 2 clinically occult LN metastases in the neck, NRAS mutant, and normal LDH, has a 12-month recurrence risk of 21.9% untreated vs 2.3% treated (benefit 19.6%), and a 24-month risk of 41.1% untreated vs 6.4% treated (benefit 34.7%). While, patient B, a man ≤45 years old with a primary melanoma in the lower limb (nodular, thickness >1-2 mm, no ulceration, 1 mitosis/mm 2 ), 2 clinically occult LN metastases in the groin, BRAF V600 mutant, and normal LDH, has a 12-month recurrence risk of 14.2% untreated vs 11.8% treated (benefit 2.5%), and a 24-month risk of 28.0% untreated vs 30.1% treated (no benefit). Conclusions: ADAPT-M, a clinical tool that estimate individualized recurrence risks with and without adjuvant anti-PD1 therapy, quantifies the absolute benefit of treatment for pts with high-risk resected melanoma, facilitating personalized adjuvant therapy decisions.