ObjectivePostprandial hyperglycemia drives insulin resistance and inflammation, leading to metabolic dysfunction-associated fatty liver disease (MAFLD). Prediction of postprandial glycemic responses by digital twin (DT) technology can fashion a personalized nutrition, activity, and sleep to treat type 2 diabetes (T2D) and MAFLD. This study examines the effects of DT-enabled personalized nutrition, activity, and sleep on glycemic status, surrogate markers of MAFLD, and magnetic resonance imaging-derived proton density fat fraction (MRI-PDFF) in patients with T2D. MethodsIn an open-label randomized trial (2:1), 319 people with T2D were eligible to intervention (DT) or standard care (SC). DT patients followed personalized meal plans with foods suggested by artificial intelligence with least predicted postprandial glycemic response. The primary end point was to compare change in hemoglobin A1c (HbA1c) and medicine reduction between the DT and SC groups. Key secondary end points included remission to compare liver function test scores and visceral adiposity using MRI. ResultsHbA1C was significantly better for DT than for SC (−2.9 [1.8] vs −0.3 [1.2]; P < .001) at 1 year with 72.7% remission of T2D. In patients with abnormal baseline values, significant improvements were seen in DT vs SC patients from baseline to 1 year in nonalcoholic fatty liver disease liver fat score (mean [SD]; −2.5 [2.0] vs −0.1 [1.5]; P < .001) and nonalcoholic fatty liver disease fibrosis score (−1.20 [0.9] vs −0.1 [1.0]; P < .001), respectively. Improvements are seen with DT compared with SC in other liver fat, fibrosis score, and %liver fat by MRI-PDFF. ConclusionAt 1 year, DT-enabled personalized treatment significantly improved hyperglycemia and surrogate markers of MAFLD and MRI-PDFF.