From 1977 to 1982 screening for cardiovascular disease was performed in three Norwegian counties. All those aged between 40 and 54 years were invited, of whom 23,690 men and 23,425 women (90%) attended. Smoking habits and previous cardiovascular disease were recorded; total cholesterol, high-density lipoprotein cholesterol (HDL cholesterol), triglycerides and blood pressure were measured. During subsequent follow-up (mean 6.8 years) 422 men and 54 women died from coronary heart disease, 514 and 114 from all cardiovascular diseases and 983 and 404 from all causes, men and women respectively. For men, mortality decreased with increasing HDL cholesterol, to a minimum of around 1.5 mmol.l-1 (58 mg.dl-1), whereafter mortality increased. This applies to coronary, cardiovascular and all causes of death, as well as to men with and without a history of disease. The association between mortality and HDL cholesterol in healthy men disappeared when total cholesterol was below 6.5 mmol.l-1 (251 mg.dl-1). The inverse association between mortality and HDL cholesterol in women was somewhat stronger than in men, both for coronary and cardiovascular diseases. The relative risks of coronary death, associated with an increase in HDL cholesterol of 0.5 mmol.l-1 (19 mg.dl-1), from the Cox proportional hazards regression, with other major cardiovascular risk factors as covariates, were 0.8 (95% confidence interval: 0.6, 1.0) and 0.8 (0.7, 1.0) for men with and without history of disease, respectively. Corresponding figures for women were 0.5 (0.3, 0.9) and 0.7 (0.4, 1.3).