There is an unmet need for predictive biomarkers of the clinical benefit of anti-angiogenic drugs.The aim of the present study was to prospectively evaluate the value of FDG-PET/CT performed during and after preoperative radiochemotherapy (CRT) with bevacizumab for the prediction of complete pathologic tumor regression and survival in magnetic resonance imaging (MRI)-defined high-risk locally advanced rectal cancer (LARC) patients Methods: Sixty-one patients treated in a non-randomized phase II study (BRANCH) with concomitant or sequential (4 days before CRT) administration of bevacizumab with preoperative CRT were included. 18F-FDG PET/CT was performed at baseline, 11 days after the beginning of the CRT (early) and before surgery (late). Metabolic changes were compared with pathological complete (TRG1) vs incomplete (TRG2-5) tumor regression, progression-free survival (PFS), cancer specific survival (CSS) and overall survival (OS). Receiver operating characteristic (ROC) curves were calculated for those FDG-PET/CT parameters that significantly correlated with TRG1. Results: Early total lesion glycolysis (TLG-early) and its percent change compared to baseline (ΔTLG-early %) could discriminate TRG1 from TRG2-5. Only ROC analysis of ΔTLG-early % showed an AUC > 0.7 (0.76), with an optimal cutoff at 59.5% (80% sensitivity, 71.4 specificity) for identifying TRG1. Late metabolic assessment could not discriminate between the two groups. After a median follow-up of 98 months (range 77-132) metabolic responders (ΔTLG-early % ≥ 59.5 %) demonstrated a significantly higher 10-year PFS (89.3 vs 63.6 %, P = 0.02) and CSS (92.9 vs 72.6 %, P = 0.04) compared to incomplete metabolic responders . Conclusion: Our results suggest that early metabolic response can act as a surrogate marker of the benefit of antiangiogenic therapy and provide further support for the use of early -FDG-PET/CT evaluation to predict pathological response and survival in the preoperative treatment of patients with LARC. ΔTLG-early% showed the best accuracy in predicting TRG and may be particularly useful in guiding treatment modifying decisions during preoperative CRT based on expected response.