BACKGROUND: All studies of treatment-free remission (TFR) in patients with chronic myeloid leukaemia have discontinued tyrosine kinase inhibitor (TKI) treatment abruptly and have focussed on patients with stable MR4 (BCR-ABL to ABL ratio ≤0·01%). We aimed to examine the effects of gradual treatment withdrawal and whether TFR is feasible for patients with less deep but stable remission. METHODS: The De-Escalation and Stopping Treatment with Imatinib, Nilotinib, or sprYcel (DESTINY) study is a non-randomised, phase 2 trial undertaken at 20 UK hospitals. We recruited patients (aged ≥18 years) with chronic myeloid leukaemia in first chronic phase, who had received TKI therapy for 3 years or more, with three or more BCR-ABL quantitative PCR transcript measurements (BCR-ABL to ABL1 ratio) less than 0·1% (major molecular response [MMR]) in the 12 months before entry. Patients with all PCR measurements less than 0·01% were assigned to the MR4 group. Patients with results between 0·1% and 0·01% were allocated to the MMR group. TKI treatment was de-escalated to half the standard dose for 12 months, then stopped for a further 24 months, with frequent PCR monitoring. Recurrence was defined as the first of two consecutive samples with PCR measurement greater than 0·1%, which required treatment recommencement at full dose. The primary endpoint was the proportion of patients who could first de-escalate their treatment for 12 months, and then stop treatment completely for a further 2 years, without losing MMR. Analysis was by intention to treat. This study is registered with ClinicalTrials.gov, number NCT01804985. FINDINGS: Treatment at entry was imatinib (n=148), nilotinib (n=16), or dasatinib (n=10), for a median of 6·9 years (IQR 4·8-10·2). Between Dec 16, 2013, and May 6, 2015, we enrolled 49 patients into the MMR group and 125 into the MR4 group. In the MR4 group, 84 (67%) patients reached the 36-month trial completion point and recurrence-free survival was 72% (95% CI 64-80). In the MMR group, 16 (33%) entrants completed the study and recurrence-free survival was 36% (25-53). No disease progression was seen and two deaths occurred due to unrelated causes. All recurrences regained MMR within 5 months of treatment resumption. INTERPRETATION: Initial de-escalation before discontinuation might improve the success of TFR protocols, although the mechanism of its benefit is not yet clear. The findings also suggest that TFR merits further study in patients with stable MMR. FUNDING: Newcastle University and Bloodwise.