Post–PD-1 failure in advanced melanoma: Real-world outcomes with lenvatinib-based versus non-lenvatinib-based therapy.

Abstract

e21520 Background: Post–anti–PD-1 advanced melanoma has few effective salvage options and CHEMO outcomes are modest. We evaluated real-world comparative effectiveness (incl. OS) of LENVA+Pembro versus non-Lenva-based therapy after anti–PD-1 progression, given encouraging activity in LEAP-004 despite negative 1L data in LEAP-003. Methods: Single-center retrospective RWD study at N.N. Blokhin NMRCO using the full EMR (01/2022–09/2025; 37,942 records; 7,061 pts). Advanced cutaneous melanoma pts progressing after anti–PD-1 were identified (ICD-10 C43 + free-text; non-cutaneous primaries manually excluded). Results: Treatment assignment was available for 235 patients (chemo without lenva, n = 117; lenva-based regimen, n = 118). Baseline characteristics were broadly similar: median age 59 (32–95) vs 61 (25–86) years, men 46.2% vs 37.3%, BRAF-mutant 30.8% vs 28.0%, and metastatic stage at treatment start (M1d 26.5% vs 22.9%; M1c 32.5% vs 32.2%). Prior ipilimumab was more frequent in the lenva group (66.1% vs 56.4%). The main imbalance was the therapy line: among patients with documented lines (70/117 and 80/118), the median line was 2 (1–6) vs 2.5 (1–6) and ≥3rd line used 42.9% vs 50.0%. Median FU was 16.2 mo (95% CI 11.2–22.2) with chemo and 14.4 mo (95% CI 10.8–22.6) with lenva. In OS Kaplan–Meier analysis (N = 235; 82 deaths), median OS was 77.2 mo (95% CI 24.5–NR) with chemo (37 deaths) and 24.6 mo (95% CI 13.4–NR) with lenva-based therapy (45 deaths) (log-rank p = 0.183). A multivariable Cox model with time-varying lenva effect was fitted in pts with known BRAF (WT/mutant) and complete covariates (n = 129; 46 deaths), adjusting for age, BRAF, metastatic stage at treatment start, prior ipilimumab, and line of therapy. Line of therapy was independently associated with OS (HR 1.36 per one-line increase; 95% CI 1.03–1.79). To account for non-proportional effects, lenva exposure was modeled piecewise (0–6, 6–12, ≥12 months), yielding interval-specific hazard ratios vs chemotherapy-only: 0–6 mo HR 0.82 (95% CI 0.35–1.92), 6–12 mo HR 3.13 (0.65–14.99), and ≥12 mo HR 3.11 (0.80–12.09). The proportional hazards test for the interval-specific lenva effect was borderline (p = 0.063), while the global test was not significant (p = 0.27). Exploratory effect-modification analyses did not identify heterogeneity of the time-varying lenva association by therapy line across pre-specified cutoffs (likelihood ratio tests for interaction p = 0.20 for 1 vs ≥2, p = 0.81 for 1–2 vs ≥3, and p = 0.13 for 1–3 vs ≥4), noting limited power and sparse events in early-line strata. Conclusions: LENVA-based salvage was not associated with improved OS vs CHEMO in aPD-1–refractory melanoma and effect estimates were time-dependent/uncertain amid strong confounding by therapy line. These data do not justify prioritizing a LENVA-vs-CHEMO prospective trial and highlight the need for new strategies in PD-1–resistant disease.