Background: FOLFOXIRI and bev achieve high response rate/R0 resections and encouraging PFS in patients with unresectable CLM. Pathological response, a surrogate endpoint of survival, is increased by bev. COI-B (capecitabine, oxaliplatin, irinotecan and bev) is a feasible regimen in pts with advanced colorectal cancer. In this phase II study, we aimed at assessing the perioperative use of this combination in pts with borderline resectable CLM (NCT2086656). Methods: Inclusion criteria: borderline resectability due to technical (need of 2-stage hepatectomy, involvement of > 1 hepatic vein or > 4 segments) and/or biological reasons ( > 4 metastases, CEA > 200, synchronicity). Limited resectable extraepatic disease and in situ primary allowed. Primary endpoint: pathological response according to Rubbia-Brandt et al., with a Simon 2-stage design (first step 22 pts; target 46 pts); secondary endpoints: objective response rate (ORR); R0 resection; safety; progression-free and overall survival. Pts received biweekly irinotecan (180 mg/mq) and bev (5 mg/kg) day 1, oxaliplatin (85 mg/mq) day 2 and capecitabine 1000 mg/mq day b.i.d.) days 2-6; 5 cycles pre-operatively (the last without bev) and 4 post-operatively. Results: we present preliminary data on the first 29 pts (26 resected, 3 still awaiting). M/F: 17/12, median age 57 years (38-76), synchronous disease 79%, multiple nodules 55%, N+ primary tumor (69%), CEA > 200 14%, extrahepatic disease 0.03%, RAS mutation 59%/no BRAF mutation. Two-stage hepatectomy 0.04%, ≥ 4 segments involved 34%, > 1 hepatic veins 14%. ORR was 89% (24/27 restaged patients), with 2 SD and 1 PD. R0 resection in 22 (85%) resected pts. TRG 1-2 was observed in 8 (31%), while TRG 3 in 12 (46%), i.e. pathological response rate 77% (first step reached). Grade ≥ 3 toxicities: diarrhea 2, neutropenia 2, thrombosis 2, neuropathy 1, fatigue 1. At a median follow up of 16 months, only 7 relapses and 2 deaths were observed. Conclusions: COI-B regimen is a feasible neoadjuvant strategy for borderline resectable CLM. This is the first trial to select pathological response as primary endpoint with encouraging activity. Clinical trial information: NCT2086656