Immune phenotype-driven treatment outcome of IO-only versus chemo-IO in PD-L1-high, first-line, advanced non-small cell lung cancer (NSCLC).

Abstract

e21179 Background: PD-L1 tumor proportion score (TPS) is the only biomarker for clinical decision-making regarding immunotherapy (IO) in advanced NSCLC. Currently, there is uncertainty when selecting treatment options between IO monotherapy (IO-only) and in combination with chemotherapy (Chemo-IO) when solely based on PD-L1, in patients with PD-L1 TPS ≥ 50%. Here, we explored a novel biomarker, immune phenotype as assessed by artificial intelligence (AI)-powered spatial tumor infiltrating lymphocytes (TIL) analysis to provide additional information in determining first-line treatment for advanced NSCLC, using a real-world dataset. Methods: A total of 349 whole-slide images (WSIs) of H&E-stained slides for advanced/metastatic NSCLC patients without actionable EGFR mutation and ALK translocation treated with IO-only or Chemo-IO as their first-line were retrospectively collected from Samsung Medical Center. An AI-powered spatial TIL analyzer, Lunit SCOPE IO classified inflamed immune phenotype (IIP, the proportion of inflamed area ≥ 33.3% in tumor microenvironment) versus non-inflamed IP (non-IIP). PD-L1 TPS was assessed based on PD-L1 pharmDx 22C3 staining. Progression-free survival (PFS) was measured by the investigators per RECIST v1.1. Results: In the analysis set, all IO-only group (n = 84) had TPS ≥ 50%, but Chemo-IO group (n = 265) consisted of 21.1%, 29.8%, and 49.1% of TPS ≥ 50%, 1-49%, and < 1%, respectively. Proportions of squamous cell carcinoma were not significantly different between IO-only and Chemo-IO (32.1% vs 27.5%, p = 0.5008). Proportions of IIP were significantly correlated with the TPS group, as they were 33.6%, 19.0%, and 13.1% in TPS ≥ 50%, 1-49%, and < 1%, respectively (p = 0.0002). In the merged dataset (n = 349), PFS was significantly increased in IIP (22.6%) compared to non-IIP (median PFS [mPFS] 14.6 vs 6.0 m, 6-m PFS rate 72.1% vs 49.9%, hazard ratio [HR, 95% confidence interval] 0.57 [0.39-0.82], p = 0.0014). In TPS ≥ 50% group, IIP was significantly correlated with favorable PFS in IO-only group (IIP vs non-IIP; mPFS 14.6 vs 4.6 m, 6-m rate 65.5 vs 45.2%, HR 0.55 [0.31-0.98], p = 0.0375), however, it was not statistically different in Chemo-IO (n = 56, mPFS not reached [NR] vs 8.3 m, 6-m rate 88.5 vs 65.3%, HR 0.45 [0.15-1.33], p = 0.1185). Interestingly, in the subgroup of TPS ≥ 50% with IIP (n = 47), PFS was not significantly different based on treatment regimen (Chemo-IO vs IO-only, mPFS NR vs 14.6 m, 6-m rate 88.5 vs 65.5%, HR 0.52 [0.17-1.59], p = 0.2263), but Chemo-IO was significantly superior to IO-only in TPS ≥ 50% with non-IIP group (n = 93, mPFS 8.3 vs 4.6 m, 6-m rate 65.3 vs 45.2%, HR 0.53 [0.30-0.94], p = 0.0255). Conclusions: Immune Phenotype based on AI-powered spatial TIL analysis may provide complementary information for clinical decision-making between IO-only and Chemo-IO in advanced NSCLC patients with TPS ≥ 50%.