1264MO IMpower010: Exploratory analysis of disease-free survival (DFS) by TGFβ cancer-associated fibroblast (CAF) gene signature expression in patients (pts) with resected NSCLC treated with atezolizumab (atezo) or best supportive care (BSC)

CAFs are an important component of the tumour microenvironment. TGFβ-induced CAFs have been associated with poor prognosis in several indications. In IMpower010, adjuvant atezo showed significant DFS benefit vs BSC in resected early-stage NSCLC after chemotherapy (chemo), leading to approval of atezo after platinum-based chemo for resected (R0) PD-L1 TC ≥1% or ≥50% (SP263) stage II-IIIA NSCLC. We report results from exploratory analyses to identify predictive biomarkers for atezo outcomes using IMpower010 RNA sequencing (RNAseq) data. The IMpower010 study design and primary data have been reported ( Felip, Lancet 2021 ). The predictive effects of features including cell-type and -state gene signatures for DFS (atezo vs BSC) were evaluated from baseline (resection) tumour samples. Relevant features were identified based on variable importance scores calculated by generalised random forests method. Median DFS was estimated using Kaplan-Meier curves; HRs were estimated by Cox regressions. Of the 1005 randomised pts (ITT), 500 comprised the RNAseq biomarker-evaluable population. The TGFβ CAF gene signature had the highest variable importance. DFS HRs (95% CI) were 0.54 (0.37, 0.80) and 0.94 (0.63, 1.41) in the TGFβ CAF-high (≥ median) and -low (< median) populations (Table). Improved outcomes with atezo vs BSC in the TGFβ CAF-high population were driven by shorter DFS in the BSC arm. Baseline characteristics were generally balanced across populations. Exploratory analyses of IMpower010 RNAseq data showed that the TGFβ CAF-high population had improved DFS with atezo vs BSC, regardless of histology or PD-L1 expression level. These data suggest that after surgical resection, pts with enrichment of the TGFβ CAF gene signature may have improved DFS after adjuvant atezo vs BSC.Table: 1264MOAtezo median DFS, moBSC median DFS, moDFS HR (vs BSC) 95% CIITTn=507 NRn=498 37.20.81 0.66, 0.99RNAseq BEPn=248 36.1n=252 30.40.71 0.54, 0.94TGFβ CAF-lown=123 NRn=127 37.30.94 0.63, 1.41TGFβ CAF-highn=125 NRn=125 29.70.54 0.37, 0.80TGFβ CAF-high, squamousn=61 NRn=62 41.40.48 0.26, 0.89TGFβ CAF-high, nonsquamousn=64 36.1n=63 24.70.60 0.36, 0.99TGFβ CAF-high, PD-L1 TC ≥1%n=74 NRn=69 35.30.49 0.29, 0.84TGFβ CAF-high, PD-L1 TC <1%n=51 NRn=56 28.60.64 0.36, 1.14BEP, biomarker evaluable population; NR, not reached; TGFβ, transforming growth factor beta. Open table in a new tab