Apolipoproteins B, A-I and Lp(a) have been proposed as independent predictors of subsequent ischaemic heart disease (IHD) improving on the prediction obtained by routine lipid measurements. In this report we have investigated the relative predictive ability of apolipoproteins and plasma lipids in a prospective study of middle aged men. 2398 men aged 49-65 years from the general population of Caerphilly, South Wales, UK were screened for evidence of IHD. After an overnight fast 2225 men each provided a venous blood sample on which plasma lipids, apolipoproteins B, A-I, A-II, and lipoprotein (a) (Lp(a)) were measured. Over a follow-up period of nearly 9 years, 282 (12%) men developed major IHD. Multiple logistic regression analysis showed that after adjusting for standard cardiovascular risk factors other than lipids there was a strong trend (standardised relative odds (SRO) = 1.20; P = 0.009) for incidence of IHD to increase with apolipoprotein B. However, on further adjusting for total cholesterol this trend largely disappeared (SRO = 1.05; P = 0.57). Similarly, a trend for incidence of IHD to increase with decreasing apolipoprotein A-I (SRO = 1.18; P = 0.02) disappeared when HDL cholesterol was added to the model. Levels of apolipoprotein A-II were not related to risk of subsequent IHD. Incidence of IHD was effectively constant over nearly 90% of the range of Lp(a). Only among the 5% of men with Lp(a) greater than 70 mg dL-1 was the risk of IHD significantly (P = 0.04) greater than among men with Lp(a) less than 10 mg dL-1. Apolipoproteins B and A-I do not improve on the prediction of risk of IHD provided by total and HDL cholesterol, respectively. Apolipoprotein A-II was not related to risk of IHD. Lp(a) may be independently associated with incident IHD among the 5-10% of men with the highest levels.