Transcatheter arterial chemoembolisation (TACE) has been reported to be an efficient treatment of liver metastases of endocrine tumours in short series of patients. However, several factors seem to affect its results. The aim of this work is to identify predictors of response to TACE for liver metastases of endocrine tumours. A total of 163 TACE procedures were performed in 67 patients between 1994 and 2004. Forty-four patients were treated with streptozotocin and 23 with doxorubicin. Primary tumour was located in the pancreas for 19 patients, and had been removed in 43. Thirty-eight tumours were functioning. Response rate was 37% (confidence interval [CI] 95%: 28-49%). Median time to progression (TTP) was 14.5 months (CI 95%: 9-41). In multivariate analysis (n=43), predictors of tumour response were body mass index (BMI) (odds ratio [OR]: 1.3; CI 95%: 1.04-1.63; P=0.022), functioning type of tumour (OR: 7.31; CI 95%: 1.26-42.5; P=0.027), arterial phase enhancement on abdominal computed tomography (CT) (OR: 8.11; CI 95%:1.06-62; P=0.044) and use of streptozotocin for cytotoxic agent (OR: 21.3; CI 95%: 1.48-306; P=0.025). Analysis of TTP predictors showed that BMI (hazard ratio [HR]: 0.85; CI 95%: 0.76-0.86; P=0.01) and arterial phase enhancement (HR: 0.3; CI 95%: 0.12-0.73; P=0.008) were associated with delayed progression. This large study confirms the previously reported results of TACE regarding its efficacy for the treatment of liver metastases of endocrine tumours. Arterial phase enhancement on abdominal CT and BMI are predictors of treatment's efficacy. Streptozotocin should be the preferred cytotoxic agent in order to save anthracycline for systemic chemotherapy.