Abstract In order to assess the effect of glycaemic improvement on lipoprotein concentrations, we studied 73 type 2 diabetic subjects before (HbA1c 10.1 (6.2–16)%) and after (HbA1c 6.6 (3.8–8.0)%) glycaemic improvement. Total triglyceride and cholesterol (c), LDLc, HDLc, non-HDLc and apolipoproteins AI (apoAI) and B (apoB) were measured. Bivariate correlations and step-wise, multivariate analysis were performed to find predictors of change in the different components of diabetic dyslipidaemia. Changes in HDLc ( r = −0.358, P = 0.001), apoAI ( r = −0.355, P = 0.003), apoAI/apoB ratio ( r = −0.333, P = 0.005), weight ( r = −0.245, P = 0.046) and BMI ( r = −0.253, P = 0.039) correlated with that of HbA1c, but, in multivariate analysis, only changes in HDLc, apoAI and apoAI/apoB ratio were predicted by the decrease in HbA1c. For the median observed change in HbA1c (−3.3 percentage-points), the estimated changes were +0.14 mmol/l, +0.12 g/l and +0.20 for HDLc, apoAI and apoAI/apoB ratio, respectively, accounting for 81, 92 and 80% of the observed changes. In conclusion, for the component of diabetic dyslipidaemia for which less therapeutic tools are available, glycaemic improvement is most effective.