The long-term impact of tyrosine kinase inhibitor (TKI) discontinuation on resistance and survival in patients with advanced gastrointestinal stromal tumours (GIST) is unclear. We report the exploratory long-term outcomes of patients with advanced GIST stopping imatinib in the BFR14 trial. BFR14, an open-label, randomised, phase 3 trial, was done in 17 comprehensive cancer centres or hospitals across France. Patients with advanced GIST aged 18 years or older with an Eastern Cooperative Oncology Group performance status of 0-3, no previous treatment with imatinib, and no previous malignancy were eligible. Patients were treated with oral imatinib 400 mg daily. Patients with a complete or partial response, or stable disease, according to Response Evaluation Criteria in Solid Tumours (1.0) at 1 year, 3 years, and 5 years from the start of treatment were randomly assigned (1:1) to treatment discontinuation until progression (interruption group) or treatment continuation until progression (continuation group). Randomisation was done centrally with computer-generated permuted blocks of two and six patients stratified by participating centre and presence or absence of residual disease on CT scan. The primary endpoint was progression-free survival. Secondary endpoints included time to imatinib resistance and overall survival. Analyses were conducted on an intention-to-treat basis in all randomly assigned patients who were not lost to follow-up. This trial is registered with ClinicalTrial.gov, NCT00367861. Between May 12, 2003, and March 16, 2004, after 1 year of imatinib, 32 patients were randomly assigned to the interruption group and 26 to the continuation group. Between June 13, 2005, and May 30, 2007, after 3 years of imatinib, 25 patients were randomly assigned to the interruption group and 25 to the continuation group. Between Nov 9, 2007, and July 12, 2010, after 5 years of imatinib, 14 patients were randomly assigned to the interruption group and 13 to the continuation group. Median follow-up was 235·2 months (IQR 128·8-236·6) after the 1-year randomisation, 200·9 months (190·2-208·4) after the 3-year randomisation, and 164·5 months (134·4-176·4) after the 5-year randomisation. Median progression-free survival in the interruption group versus the continuation group after 1 year of imatinib was 6·1 months (95% CI 2·5-10·1) versus 27·8 months (19·5-37·9; hazard ratio [HR] 0·36 [95% CI 0·20-0·64], log-rank p=0·0003), after 3 years of imatinib was 7·0 months (3·5-11·7) versus 67·0 months (48·8-85·6; 0·15 [0·07-0·32], log-rank p<0·0001), and after 5 years of imatinib was 12·0 months (9·0-16·6) versus not reached (NR; NR-NR; 0·13 [0·03-0·58], log-rank p=0·0016). The median time to imatinib resistance after 1 year of imatinib was 28·7 months (95% CI 18·1-39·1) versus 90·6 months (25·3-156·1; HR 0·93 [95% CI 0·51-1·71], log-rank p=0·82), after 3 years was 66·2 months (43·0-89·6) versus 127·3 months (15·0-239·7; 0·35 [0·17-0·72, log-rank p=0·0028), and after 5 years was 58·6 months (0·0-167·4) versus NR (NR-NR; 0·24 [0·05-1·12], log-rank p=0·049). Median overall survival after 1 year of imatinib was 56·0 months (95% CI 30·3-82·9) versus 105·0 months (20·6-189·6; HR 0·84 [95% CI 0·46-1·54], log-rank p=0·57), after 3 years was 104·0 months (90·7-118·7) versus 134·0 months (89·7-178·3; 0·40 [0·20-0·82], log-rank p=0·0096), and after 5 years was NR (NR-NR) versus 110·4 months (82·7-154·1; 1·28 [0·41-3·99]; log-rank p=0·67), INTERPRETATION: Imatinib interruption in patients with GIST without progressive disease is not recommended. Imatinib interruption in non-progressing patients with GIST was associated with rapid progression, faster resistance to imatinib, and shorter overall survival in the long-term follow-up when compared with imatinib continuation in patients after 3 years and 5 years of imatinib. Centre Léon Bérard, INCa, CONTICANET, Ligue Contre le Cancer, and Novartis.