Background Liraglutide 3·0 mg was shown to reduce bodyweight and improve glucose metabolism after the 56-week period of this trial, one of four trials in the SCALE programme. In the 3-year assessment of the SCALE Obesity and Prediabetes trial we aimed to evaluate the proportion of individuals with prediabetes who were diagnosed with type 2 diabetes. Methods In this randomised, double-blind, placebo-controlled trial, adults with prediabetes and a body-mass index of at least 30 kg/m2, or at least 27 kg/m2 with comorbidities, were randomised 2:1, using a telephone or web-based system, to once-daily subcutaneous liraglutide 3·0 mg or matched placebo, as an adjunct to a reduced-calorie diet and increased physical activity. Time to diabetes onset by 160 weeks was the primary outcome, evaluated in all randomised treated individuals with at least one post-baseline assessment. The trial was conducted at 191 clinical research sites in 27 countries and is registered with ClinicalTrials.gov, number NCT01272219. Findings The study ran between June 1, 2011, and March 2, 2015. We randomly assigned 2254 patients to receive liraglutide (n=1505) or placebo (n=749). 1128 (50%) participants completed the study up to week 160, after withdrawal of 714 (47%) participants in the liraglutide group and 412 (55%) participants in the placebo group. By week 160, 26 (2%) of 1472 individuals in the liraglutide group versus 46 (6%) of 738 in the placebo group were diagnosed with diabetes while on treatment. The mean time from randomisation to diagnosis was 99 (SD 47) weeks for the 26 individuals in the liraglutide group versus 87 (47) weeks for the 46 individuals in the placebo group. Taking the different diagnosis frequencies between the treatment groups into account, the time to onset of diabetes over 160 weeks among all randomised individuals was 2·7 times longer with liraglutide than with placebo (95% CI 1·9 to 3·9, p<0·0001), corresponding with a hazard ratio of 0·21 (95% CI 0·13–0·34). Liraglutide induced greater weight loss than placebo at week 160 (–6·1 [SD 7·3] vs −1·9% [6·3]; estimated treatment difference −4·3%, 95% CI −4·9 to −3·7, p<0·0001). Serious adverse events were reported by 227 (15%) of 1501 randomised treated individuals in the liraglutide group versus 96 (13%) of 747 individuals in the placebo group. Interpretation In this trial, we provide results for 3 years of treatment, with the limitation that withdrawn individuals were not followed up after discontinuation. Liraglutide 3·0 mg might provide health benefits in terms of reduced risk of diabetes in individuals with obesity and prediabetes. Funding Novo Nordisk, Denmark