AIMS--To assess whether the ability of lipoprotein related variables to discriminate between individuals with or without premature clinical ischaemic heart disease (IHD) was improved using data on high density lipoprotein-lipoprotein AI (HDL-LpAI) fractions, alone or in combination with data on Lp(a). METHODS--Lipid and apolipoprotein concentrations were measured in 26 middle-aged men (mean age 50.3 years) with early onset IHD and coronary artery bypass grafting prior to sampling, and in 26 matched lipaemic and 26 normolipaemic asymptomatic controls. RESULTS--Triglyceride and Lp(a) concentrations were higher, while HDL cholesterol and apolipoprotein A-I (apoA-I) concentrations were lower in patients than in controls. LpAI concentrations were also lower in IHD patients and were correlated with HDL and apoA-I in both IHD and control groups. Lp(a) was not correlated with any other lipid or apolipoprotein measured in either patients or controls. Univariate discriminant function analysis showed that the proportion correctly classified as patients or controls was marginally greater using LpAI concentrations as the discriminator, which was not increased in combination with Lp(a). Serum triglycerides, HDL cholesterol, apoA-I and Lp(a) alone all had similar, but weaker, discriminant power, which increased in various combinations with LpAI. CONCLUSIONS--LpAI particle measurement may be useful in research to define mechanisms of cardiovascular protection by HDL but the discriminating power for IHD was only marginally superior to measuring total apoA-I or Lp(a) concentrations. Little further advantage arose through combining LpAI data with other variables.