PD-L1 tumor proportion score and clinical benefit from first-line pembrolizumab in patients with advanced nonsquamous versus squamous non-small cell lung cancer (NSCLC).

Abstract

9539 Background: The predictive value of PD-L1 tumor proportion score (TPS) on NSCLC tumor cells as a biomarker for response to PD-(L)1 inhibitors is well established. However, its histology specific value in advanced (a) squamous (Sq) versus nonsquamous (NS) cancers remains unclear. Here, we used real world data to assess the differential value of PD-L1 TPS as a predictive biomarker for overall survival (OS) after first-line pembrolizumab (P) in patients (pts) with Sq versus NS NSCLC. Methods: Inclusion criteria for this analysis of the Flatiron Health EHR-derived de-identified database required that pts were diagnosed with aNSCLC, tested for PD-L1 TPS and received a numerical result, had no alteration in EGFR, ALK, or ROS, and received first-line, single agent P. The primary endpoint was overall survival (OS) from the first dose of P in patients with TPS ≥ 50% (H) compared to patients with TPS < 50% (L). Due to the violation of the proportional hazards assumption, a generalized gamma model of OS was used, adjusting for demographic variables and estimated median OS and their confidence intervals with the bootstrap method. The PD-L1-histology interaction was examined by comparing the differences in median OS (H vs. L) between Sq and NS patients. Results: Of 1560 pts with NSCLC treated from 1/2011 – 5/2019, 1055 had NS and 405 Sq. No meaningful differences in age, gender, or smoking history were seen between PD-L1 H and L pts with either histology. Among NS pts, H had significantly longer OS than L, with unadjusted hazard ratio (HR) of 0.71 (95% CI: 0.53 - 0.94; p = 0.018). Among Sq pts, H was not associated with longer OS than L, with unadjusted HR 0.89 (95% CI: 0.64 - 1.25; p = 0.514). Based on the generalized gamma model, PD-L1 H in Sq patients was associated with a 0.19 year improvement in median OS (95% CI: -0.22-0.49, P = 0.283), whereas PD-L1 H in the NS group was associated with a 0.70 year improvement in OS (95% CI: 0.34-1.05, P < 0.001). The median improvements between the Sq and NS patients were significantly different (P = 0.034), after adjusting for demographic variables. Conclusions: PD-L1 TPS of ≥ 50% predicted longer OS in pts with NS NSCLC treated with first-line P compared to pts whose tumors had a TPS of < 50%. However, no relationship between PD-L1 TPS and OS after first-line P was seen in patients with Sq NSCLC. These data suggest that PD-L1 may not be an appropriate predictive biomarker for checkpoint inhibitor use in NSCLC with squamous histology.