Background – Predictive biomarkers for first-line (1L) ICI therapy in advanced melanoma are lacking. aLAG-3/PD1 is approved regardless of LAG3 expression despite evidence of poorer RR and PFS (Relativity-047). The relationship between LAG3expression and clinical benefit to 1L non-aLAG-3 ICIs is unknown.

Methods – We evaluated LAG3 expression as a biomarker for benefit towards 1L non-aLAG-3 ICIs in a real-world (rw) advanced melanoma cohort.De-identified NGS data from patients (pts; N=367) with stage III-IV (advanced) cutaneous melanoma was analyzed in the Tempus Database. Pts were treated with 1L ICIs, aCTLA4/PD1 (n=222) or aPD1 (n=145). Pts treated with 1L aLAG-3/PD1 were limited and not included. Tissue was collected prior to 1L and sequenced with xT DNA (648-gene panel) and xR RNA assays. LAG3 high (LAG3-H) and low (LAG3-L) expressors were defined by median LAG3 RNA transcripts per million. PD-L1 was determined by IHC. QuanTIseq was used to estimate immune cell proportions. Rw overall survival (rwOS) was calculated from treatment (tx) start to death from any cause. Rw objective response rate (rwORR) was defined as the proportion of pts with a documented complete or partial response within 90 days of tx start. Hazard ratio (HR) was calculated using a Cox proportional hazards (CoxPH) model and p-values using the Wald test.

Results – LAG3-H was associated with PD-L1+ status (LAG3-H 52% vs LAG3-L 5.5%) and higher proportions of adaptive immune cells, including CD8+ (2.7% vs 0.1%), B (4.3% vs 3.1%), and inhibitory Treg cells (4.5% vs 2.9%; p<0.001 for all). In the aPD1 cohort, LAG3-H (n=77) had higher rwORR vs LAG3-L (n=68; 67% vs 15%, p=0.002). In the aCTLA4/PD1 cohort, LAG3-H (n=109) had similar rwORR vs LAG3-L (n=113; 46% vs 37%, p=0.42). CoxPH analysis found a significant interaction between LAG3 and 1L ICIs on rwOS (p=0.02), suggesting the ICI effect on rwOS differs between LAG3-L and LAG3-H (p=0.02). In the LAG3-L group, pts treated with aPD1 had reduced rwOS compared to the aCTLA4/PD1 cohort (HR, 1.88, p=0.03), while in the LAG3-H group pts treated with aPD1 had higher rwOS compared to the aCTLA4/PD1 cohort (HR, 0.67, p=0.2). LAG3 was associated with immune microenvironment changes and modulated 1L aPD1 but not aCTLA4/PD1 efficacy in pts with advanced melanoma. Similar to the Relativity-047 trial that showed LAG3 alters aLAG-3/PD1 efficacy, pts with LAG3-H treated with aPD1 had improved outcomes vs pts with LAG3-L. However, pts with LAG3-H treated with aCTLA4/PD1 had similar outcomes vs pts with LAG3-L.

Conclusions – Differences in rwOS observed between aCTLA4/PD1 and aPD1-treated pts within the LAG3 expression subsets indicates potential predictive value of this biomarker. Future studies will prospectively validate these novel findings in pts with advanced melanoma to confirm the finding that aCTLA4/PD1 is optimal for pts with LAG3-L and alternative 1L ICIs are best for pts with LAG3-H.

VIEW THE PUBLICATION