Lipoprotein changes often precede incident type 2 diabetes and may be diabetogenic, but effects of lipid-lowering medications, ethnicity and sex on these associations are unclear. In the Multi-Ethnic Study of Atherosclerosis, among 5636 men and women, ages 45-84, without baseline diabetes or CVD, 694 diabetes cases were identified during mean follow-up of 7.7 years. NMR spectroscopy was used to measure lipoprotein particle concentrations and size, including a calculated lipoprotein-related insulin resistance score (LP-IR), on baseline plasma. Analyses were stratified by baseline lipid-lowering medication (primarily statin) use, as “untreated” (n=4816) or “treated” (n=820), and by glucose <100 vs.100-125 mg/dl (impaired fasting glucose, IFG). Among treated and untreated participants, higher LP-IR, larger mean VLDL particle size, HOMA-IR and (log)insulin were associated with incident diabetes, adjusted for age, ethnicity and sex (HRs per SD >1.5, p<0.05). Associations of LP-IR and VLDL size with incident diabetes remained significant when additionally adjusted for HDL-C, (log) triglycerides, BMI, glucose, family history of diabetes, physical activity and alcohol use and among treated participants with IFG, and were similar by sex and race-ethnicity. In conclusion, lipoprotein particle measures, particularly LP-IR and VLDL size, were robustly associated with incident diabetes among both untreated and treated ethnically diverse adults.