BACKGROUND: The DiRECT trial assessed remission of type 2 diabetes during a primary care-led weight-management programme. At 1 year, 68 (46%) of 149 intervention participants were in remission and 36 (24%) had achieved at least 15 kg weight loss. The aim of this 2-year analysis is to assess the durability of the intervention effect. METHODS: DiRECT is an open-label, cluster-randomised, controlled trial done at primary care practices in the UK. Practices were randomly assigned (1:1) via a computer-generated list to provide an integrated structured weight-management programme (intervention) or best-practice care in accordance with guidelines (control), with stratification for study site (Tyneside or Scotland) and practice list size (>5700 or ≤5700 people). Allocation was concealed from the study statisticians; participants, carers, and study research assistants were aware of allocation. We recruited individuals aged 20-65 years, with less than 6 years' duration of type 2 diabetes, BMI 27-45 kg/m2, and not receiving insulin between July 25, 2014, and Aug 5, 2016. The intervention consisted of withdrawal of antidiabetes and antihypertensive drugs, total diet replacement (825-853 kcal per day formula diet for 12-20 weeks), stepped food reintroduction (2-8 weeks), and then structured support for weight-loss maintenance. The coprimary outcomes, analysed hierarchically in the intention-to-treat population at 24 months, were weight loss of at least 15 kg, and remission of diabetes, defined as HbA1c less than 6·5% (48 mmol/mol) after withdrawal of antidiabetes drugs at baseline (remission was determined independently at 12 and 24 months). The trial is registered with the ISRCTN registry, number 03267836, and follow-up is ongoing. FINDINGS: The intention-to-treat population consisted of 149 participants per group. At 24 months, 17 (11%) intervention participants and three (2%) control participants had weight loss of at least 15 kg (adjusted odds ratio [aOR] 7·49, 95% CI 2·05 to 27·32; p=0·0023) and 53 (36%) intervention participants and five (3%) control participants had remission of diabetes (aOR 25·82, 8·25 to 80·84; p<0·0001). The adjusted mean difference between the control and intervention groups in change in bodyweight was -5·4 kg (95% CI -6·9 to -4·0; p<0·0001) and in HbA1c was -4·8 mmol/mol (-8·3 to -1·4 [-0·44% (-0·76 to -0·13)]; p=0·0063), despite only 51 (40%) of 129 patients in the intervention group using anti-diabetes medication compared with 120 (84%) of 143 in the control group. In a post-hoc analysis of the whole study population, of those participants who maintained at least 10 kg weight loss (45 of 272 with data), 29 (64%) achieved remission; 36 (24%) of 149 participants in the intervention group maintained at least 10 kg weight loss. Serious adverse events were similar to those reported at 12 months, but were fewer in the intervention group than in the control group in the second year of the study (nine vs 22). INTERPRETATION: The DiRECT programme sustained remissions at 24 months for more than a third of people with type 2 diabetes. Sustained remission was linked to the extent of sustained weight loss. FUNDING: Diabetes UK.