Aims/hypothesis: Controversy surrounds whether the ratio of apolipoprotein B (ApoB) to apolipoprotein A-I (ApoA-I) is the best lipoprotein discriminator of CHD risk in nondiabetic populations, but the issue has never been investigated in type 2 diabetes. Methods: In 2,627 participantswithout known vascular disease in the Collaborative Atorvastatin Diabetes Study, ApoB, ApoA-I, LDL-cholesterol (LDLC) and HDL-cholesterol (HDLC) were assayed at baseline. Results There were 108 CHD and 59 stroke endpoints over 3.9 years. The ApoB:A-I ratio at baseline was the lipoprotein variable most closely predicting CHD risk both by comparison of the hazard ratio for a 1 SD change or tertiles of frequency distribution. The areas under the receiver operator curve for the ApoB:ApoA-I and the LDLC to HDL-HDLC ratios, although not significantly different from each other, were greater (p=0.0005 and p=0.0125 respectively) than that of non-HDLC:HDLC. The 27% decrease in the ApoB:ApoA-I ratio on atorvastatin predicted a 32% (95% CI 5.4-51.2%) risk reduction in CHD, close to the 36% decrease observed. Neither the ApoB:ApoA-I nor any other lipoprotein concentration or ratio predicted the stroke outcome. Conclusions/interpretation: Overall, the ApoB:ApoA-I ratio improved on the non-HDLC:HDLC ratio in predicting CHD, but, depending on the assessment chosen, its superiority over LDLC:HDLC may be marginal. The statin- nduced decrease in stroke risk may not be lipoprotein mediated. Trial registration: ClinicalTrials.gov CT00327418 Funding: The study was supported by unrestricted grants from Diabetes UK, the Department of Health and Pfizer to theUniversity of Manchester and to University College, London. © 2008 Springer-Verlag.