Aims/hypothesis <p>Recent studies advocate prediabetes remission as a goal in diabetes prevention, but the optimal dietary composition for prediabetes remission over the long term is unknown. We aimed to examine the long-term effects on prediabetes remission of a prudent diet with moderate protein and a moderate glycaemic index (GI) (akin to general dietary guidelines) vs a high-protein, low-GI diet.</p> Methods <p>This study is a secondary analysis of PREVIEW, which is a 3 year, multicentre, parallel, randomised trial. Adults with overweight/obesity (BMI ≥ 25 and &lt;30 kg/m² or BMI ≥ 30 kg/m², respectively) and prediabetes (fasting glucose 5.6–6.9 mmol/l and/or 2&#xa0;h glucose 7.8–11.0 mmol/l determined using an OGTT) were recruited. Eligible participants underwent an 8-week rapid weight loss programme comprising a low-energy diet, followed by a 3 year weight maintenance phase comprising lifestyle intervention. At baseline, participants were randomly assigned to a high-protein (25% of energy from protein), low-GI (GI&lt;50) diet, or a prudent diet with moderate protein (15% of energy from protein) and moderate GI (GI&gt;56). The primary outcome of the current analysis was the number of participants who achieved prediabetes remission (i.e. a return to normal fasting glucose and normal glucose tolerance) at 1 year or 3 years. Secondary outcomes were changes in body weight and composition over 3 years (continuous variables) and maintenance of a ≥8% weight loss target (binary variable). Modified intention-to-treat analyses were performed on all participants who received the dietary intervention (<i>n</i>=1856). Risk ratios and 95% CI for prediabetes remission and maintaining the weight loss target in each diet group were estimated using multilevel modified Poisson regression adjusted for age and sex. Linear mixed models were used to estimate the dietary effects on changes in body weight and composition.</p> Results <p>The moderate-protein, moderate-GI group (<i>n</i>=923) had a higher rate of remission than the high-protein, low-GI group (<i>n</i>=933) at both 1 year (rate of remission 26.3% vs 20.7%; RR 1.26; 95% CI 1.04, 1.53; <i>p</i>=0.025) and 3 years (20.6% vs 15.5%; RR 1.26; 95% CI 1.06, 1.50; <i>p</i>=0.015). However, body weight and composition changes were similar for participants on the moderate-protein, moderate-GI vs high-protein, low-GI diet at 1 year (54.0% vs 57.3% of participants met the weight loss maintenance target [≥8% of initial body weight]; <i>p</i>=0.215) and 3 years (31.4% vs 30.4%, respectively; <i>p</i>=0.793). The differences in remission rates of the two dietary patterns were independent of body weight and composition changes.</p> Conclusions/interpretation <p>Following rapid weight loss, a prudent diet with moderate protein and moderate GI was more effective for long-term prediabetes remission than a high-protein, low-GI diet, irrespective of weight change.</p> Trial registration <p>ClinicalTrials.gov NCT01777893</p> Graphical Abstract <p></p>