Serum VEGF-A concentrations as a predictive biomarker of immunotherapy response in refractory metastatic colorectal cancer (mCRC): An exploratory analysis of CCTG CO.26.

3547 Background: Benefits of immune checkpoint inhibitors (ICIs) in unselected proficient mismatch repair (pMMR) mCRC remain unestablished. This exploratory analysis of CCTG CO.26 evaluated whether serum VEGF-A concentrations may influence ICI efficacy. Methods: CCTG CO.26 was a phase II trial that randomized refractory mCRC patients (pts) (2:1) to Durvalumab + Tremelimumab + Best supportive care (DT) vs Best supportive care (BSC) alone and was positive for overall survival (OS). VEGF-A concentrations were determined in baseline serum samples using an ELISA method. OS and progression-free survival (PFS) were compared between high and low groups based on median VEGF-A concentration in each treatment arm. An optimal cutoff was also identified using a minimal p-value approach. VEGFA mRNA expression from tissue-based RNA-seq was analyzed in the INSPIRE study, a separate, single arm, multicohort phase II trial comprising multiple tumor types treated with pembrolizumab for external validation. Results: Baseline VEGF-A concentrations were available for 161 pts (47 BSC, 114 DT). Median VEGF-A concentration was 1166 pg/ml (range: 226 – 4898; IQR: 784-1755). VEGF-A concentrations were significantly higher in ECOG 1 vs 0 pts (p = 0.03), and in pts who had previously received regorafenib (p = 0.03). Grouped by median VEGF-A, median OS was longer in DT-treated patients with low VEGF-A (7.39 months for DT-Low vs. 6.08 DT-high; hazard ratio (HR) 0.62, 95% confidence interval (CI): [0.42–0.90], p = 0.01), but not in BSC treated patients (4.44 months BSC-Low vs. 4.11 BSC-High; HR 1.05 [0.57–1.92], p = 0.88; interaction HR 0.56 [0.28–1.14], p = 0.11). An optimal VEGF-A threshold of 1540 pg/mL was identified using a minimal p-value approach, with 33.5% pts having high VEGF-A concentrations. Median OS was 7.29 vs 4.37 months for DT-Low and DT-High pts (HR 0.53 [0.35–0.80], p < 0.005), and 3.55 vs 7.46 months for BSC-Low and BSC-High pts (HR 1.50 [0.80–2.80], p = 0.21; interaction HR 0.33, [0.16–0.70], p < 0.005). In multivariable analyses incorporating ECOG, presence of liver metastases, plasma tumor mutation burden and arginine levels, low VEGF-A concentrations remained statistically associated with improved OS with DT (HR 0.46 [0.30–0.70], p < 0.001; interaction HR 0.18 [0.083–0.41], p < 0.0001). There was no association between VEGF-A concentrations and PFS in any arm. In the INSPIRE study (n = 66), VEGFA mRNA expression was lower in responders (p < 0.001) and significantly associated with improved PFS (HR 0.45 [0.25–0.79], p = 0.006) and OS (HR 0.47 [0.26–0.85], p = 0.013) across tumor types. Conclusions: Refractory mCRC patients with low VEGF-A concentrations may derive benefit from ICIs. This is the first report to suggest that serum VEGF-A concentration is a potential predictive biomarker for benefit from ICIs. These findings should be prospectively validated.