Real-world impact of vascular endothelial growth factor (VEGF) on (chemo)immunotherapy (IO) response in non-small cell lung cancer (NSCLC)

Introduction
VEGF family signaling shapes angiogenesis and immune regulation in NSCLC; however, ligand/isoform-specific effects on (chemo)IO outcomes remain unclear.

 

Methods
TEMPUS clinicogenomic data (2011-2025) were evaluated for patients with metastatic NSCLC treated with frontline chemoIO, IO, or chemo. Pre-treatment tumor transcriptomics were assessed for VEGF ligand, receptor, and splice isoform log2 transcripts per million. Real-world progression-free survival (rwPFS) and overall survival (rwOS) were assessed using multivariable Cox proportional hazards models adjusted for clinical covariates and inferred cellular fractions. Gene expression tertiles visualized effect sizes with Kaplan–Meier estimates compared using log-rank tests. Tumor microenvironment (TME) composition used EPIC/MCP-counter deconvolution algorithms from bulk RNA sequencing data. Feature stability was confirmed using 5-fold nested cross-validated Cox elastic-net regression.

 

Results
2,852 cases were included, 1799 received chemoIO, 538 IO, and 535 chemo. Multivariable Cox models demonstrated distinct and opposing prognostic patterns among VEGF family ligands and isoforms. Higher VEGF-C expression associated with inferior survival (HR 1.09, p<0.001), alongside VEGF-A-165 (HR 1.07, p=0.004) and FLT4-long (HR 1.08, p=0.004) whereas VEGF-B (HR 0.95, p=0.038) and VEGF-A-121 (HR 0.95, p=0.034) were associated with improved outcomes. These findings were concordant in tertile-based analyses, demonstrating clinically meaningful separation in rwPFS/OS with up to 4.8-month differences (Table). VEGF-C expression strongly correlated with cancer-associated fibroblast enrichment (R= 0.50, p<0.001), supporting a stromal-driven immunosuppressive TME. Results were consistent across treatment subgroups and remained stable in cross-validated models.

 

Conclusion
VEGF ligand and splice isoforms demonstrate divergent and biologically coherent associations with (chemo)immunotherapy outcomes in NSCLC, supporting an opportunity for ligand-specific strategies in IO optimization.

 

Table. VEGF expression tertiles and survival outcomes in metastatic NSCLC treated with (chemo)immunotherapy.

 

Gene	rwPFS (months) High	rwPFS (months) Low	rwPFS p-value	rwOS (months) High	rwOS (months) Low	rwOS p-value
VEGF-A	7.2	6.2	0.076	19.4	15.0	0.007*
VEGF-A 121 (↑)	7.0	6.9	0.430	18.5	16.7	0.027*
VEGF-A 165 (↓)	6.0	7.9	0.025*	15.2	18.6	<0.001**
VEGF-B (↑)	7.7	6.7	0.016*	20.4	16.0	<0.001**
VEGF-C (↓)	5.9	7.7	0.015*	15.2	20.0	<0.001**
FLT4	6.7	7.7	0.023*	16.7	18.7	0.280
FLT4-long (↓)	6.3	7.3	0.026*	15.2	19.1	0.012*
FLT4-short	6.2	7.1	0.200	16.7	19.2	0.057

 

*p<0.05; **p<0.001; (↑) improved outcomes with higher expression; (↓) worse outcomes with higher expression