OBJECTIVE We examined the impact of acarbose, an α-glucosidase inhibitor, on incident diabetes and regression to normoglycemia in 6,522 Acarbose Cardiovascular Evaluation (ACE) trial participants in China who had impaired glucose tolerance (IGT) and coronary heart disease (CHD). RESEARCH DESIGN AND METHODS Participants were randomly assigned to acarbose or placebo and followed with four monthly fasting plasma glucose (FPG) tests and annual oral glucose tolerance tests. Incident diabetes was defined as two successive diagnostic FPG levels ≥7 mmol/L or 2-h plasma glucose (PG) levels ≥11.1 mmol/L while taking study medication or a masked adjudicated confirmation of this diagnosis. Regression to normoglycemia was defined as FPG <6.1 mmol/L and 2-h PG <7.8 mmol/L. Intention-to-treat and on-treatment analyses were conducted using Poisson regression models, overall and for subgroups (age, sex, CHD type, HbA1c, FPG, 2-h PG, BMI, estimated glomerular filtration rate, for IGT alone, for IGT + impaired fasting glucose, and for use of thiazides, ACE inhibitors [ACEis]/angiotensin receptor blockers [ARBs], β-blockers, calcium channel blockers, or statins). RESULTS Incident diabetes was less frequent with acarbose compared with placebo (3.2 and 3.8 per 100 person-years, respectively; rate ratio 0.82 [95% CI 0.71, 0.94], P = 0.005), with no evidence of differential effects within the predefined subgroups after accounting for multiple testing. Regression to normoglycemia occurred more frequently in those randomized to acarbose compared with placebo (16.3 and 14.1 per 100 person-years, respectively; 1.16 [1.08, 1.25], P < 0.0001). This effect was greater in participants not taking an ACEi or ARB (1.36 [1.21, 1.53], Pinteraction = 0.0006). The likelihood of remaining in normoglycemic regression did not differ between the acarbose and placebo groups (P = 0.41). CONCLUSIONS Acarbose reduced the incidence of diabetes and promoted regression to normoglycemia in Chinese people with IGT and CHD.