Background: Studies of a modified XELIRI (mXELIRI; capecitabine plus irinotecan) regimen suggest promising efficacy and tolerability profiles in the first-line and second-line settings. Therefore, we aimed to compare the efficacy and safety of the mXELIRI regimen with that of standard FOLFIRI (leucovorin, fluorouracil, and irinotecan), with or without bevacizumab in both regimens, as a second-line therapy for metastatic colorectal cancer. Methods: We did a multicentre, open-label, randomised, non-inferiority, phase 3 trial. We enrolled patients from 98 hospitals in Japan, China, and South Korea who were aged 20 years or older with histologically confirmed and unresectable colorectal adenocarcinoma, and who had withdrawn from first-line chemotherapy for metastatic colorectal cancer. We randomly assigned patients (1:1) to receive either mXELIRI with or without bevacizumab (irinotecan 200 mg/m 2 intravenously on day 1 plus oral capecitabine 800 mg/m 2 twice daily on days 1–14, repeated every 21 days, with or without bevacizumab 7·5 mg/kg intravenously on day 1) or FOLFIRI with or without bevacizumab (irinotecan 180 mg/m 2 intravenously on day 1, leucovorin 200 mg/m 2 intravenously on day 1, fluorouracil 400 mg/m 2 intravenously on day 1, and a 46-h continuous intravenous infusion of fluorouracil [2400 mg/m 2 ], repeated every 14 days, with or without the addition of bevacizumab 5 mg/kg intravenously on day 1) via a centralised electronic system. We used the minimisation method to stratify randomisation by country, Eastern Cooperative Oncology Group performance status, number of metastatic sites, previous oxaliplatin treatment, and concomitant bevacizumab treatment. Patients and clinicians were not masked to the allocated treatment. The primary endpoint was overall survival analysed on an intention-to-treat basis with a non-inferiority upper margin of 1·30 for the hazard ratio (HR). This study is registered with ClinicalTrials.gov, number NCT01996306, and is ongoing but no longer recruiting participants. Findings: Between Dec 2, 2013, and Aug 13, 2015, 650 patients were enrolled and randomly assigned to receive mXELIRI with or without bevacizumab (n=326) or FOLFIRI with or without bevacizumab (n=324). After a median follow-up of 15·8 months (IQR 8·7–24·9), a total of 490 patients had died (242 in the mXELIRI with or without bevacizumab group and 248 in the FOLFIRI with or without bevacizumab group) and the median overall survival was 16·8 months (95% CI 15·3–19·1) in the mXELIRI group and 15·4 months (13·0–17·7) in the FOLFIRI group (HR 0·85, 95% CI 0·71–1·02; p non-inferiority <0·0001). In the per-protocol safety population, the most common grade 3–4 adverse event was neutropenia (affecting 52 [17%] of 310 patients in the mXELIRI group and 133 [43%] of 310 in the FOLFIRI group). Incidences of grade 3–4 diarrhoea were higher in the mXELIRI group (22 [7%]) than in the FOLFIRI group (ten [3%]). Serious adverse events were reported in 46 (15%) of 310 patients in the mXELIRI group and 63 (20%) of 310 in the FOLFIRI group. Two treatment-related deaths (one pneumonitis and one lung infection) were observed in the mXELIRI group and there was one treatment-related death (lung infection) in the FOLFIRI group. Interpretation: mXELIRI with or without bevacizumab is well tolerated and non-inferior to FOLFIRI with or without bevacizumab in terms of overall survival. mXELIRI could be an alternative to FOLFIRI as a standard second-line backbone treatment for metastatic colorectal cancer, at least for Asian patient populations. Funding: Chugai Pharmaceutical and F Hoffmann-La Roche.