Baseline carcinoembryonic antigen (CEA) serum levels to predict bevacizumab-based treatment response in patients with KRAS exon wild-type metastatic colorectal cancer (mCRC) receiving 1st-line therapy with FOLFIRI plus cetuximab or bevacizumab (AIO KRK0306, FIRE3 trial).

3581 Background: Recently, Prager et al. proposed the hypothesis that carcinoembryonic antigen (CEA) is involved in tumor angiogenesis and therefore might predict bevacizumab-based treatment response in metastatic colorectal cancer (mCRC) (Cancer Research, 2013; Cancer Science, 2014). The present study aimed to evaluate this hypothesis in the FIRE-3 trial where patients (pts) with KRAS exon 2 wild-type mCRC were randomized to receive first-line chemotherapy with either FOLFIRI + cetuximab (cet) or FOLFIRI + bevacizumab (bev). Methods: Baseline CEA levels were analyzed either as continuous variable (ln) or categorized in two groups ( ≤ / > 25% quartile; ≤ / > 6.2ng/ml) with regard to their predictive impact on PFS and OS in the two different treatment arms. Correlations relied on Cox regression analysis. Survival was estimated by Kaplan-Meier. Results: For analysis of CEA, 522/592 pts (cet: 257 pts; bev: 265 pts) were eligible. When CEA baseline level was evaluated as a continuous variable, CEA (ln) was inversely correlated with PFS (p= 0.02) and OS (p= 0.001) in the bev arm, but not in the cet arm. Using the 25% quartile of CEA as a cut-off, only in the bev arm an effect on OS (24.2 mo vs. 33.3 mo; HR 0.60; Cl95%:0.41-0.88; p= 0.009) and PFS (10.4 mo vs. 11.7 mo; HR 0.67; Cl95%:0.48-0.94; p= 0.02) was observed, but not in the cet arm. Pts with high CEA baseline levels ( > 6.2ng/ml) showed a shorter OS in the bev arm compared to the cet arm (bev: 24.2 mo vs. cet: 30.0 mo; HR 1.52; Cl95%:1.16-2.00; p= 0.002). In pts with low CEA baseline level ( ≤ 6.2ng/ml) no significant survival difference was observed between arms (HR 0.92; Cl95%:0.58-1.44; p= 0.70). Univariate analysis identified CEA as an independent predictor of OS in the bev arm, but not in the cet arm. Conclusions: The present analysis supports the hypothesis that CEA acts as a predictor of bev activity, while outcome parameters in the cet arm are not affected by baseline CEA. Patients with elevated CEA baseline levels appeared to benefit from anti-EGFR directed therapy, while anti-VEGF based treatment was inferior.