The Paris Prospective Study is a long-term, large-scale study of the factors predicting coronary heart disease. The first follow-up examination included, for subjects not known as having diabetes mellitus, a 75 g oral glucose tolerance test with measurement of plasma insulin and glucose levels, fasting and 2 h post-load. Between 1968 and 1973, 6903 men aged 43-54 years were thus examined. Causes of death were ascertained within this group after 15 years of mean follow-up. The baseline variables were tested as predictors of death from coronary heart disease by a Cox regression analysis. Significant independent predictors of coronary heart disease death were: systolic blood pressure, number of cigarettes per day, plasma cholesterol level, and 2 h post-load plasma insulin level when entered as a categorical variable (below or above 452 pmol/l. i.e. the lower limit of the fifth quintile of the distribution). This dichotomization was performed to account for the non-linear univariate distribution of deaths with increasing post-load insulin values. Fasting plasma insulin level was not an independent predictor of death by coronary heart disease over this long-term follow-up. Levels of blood glucose were not significant independent predictors of death by coronary heart disease when plasma insulin levels were included in the model. The same applied to abnormalities of glucose tolerance when the 125 men with known non-insulin-treated diabetes at baseline were added to the group. Under the assumption that hyperinsulinaemia is a marker of insulin resistance, the results are consistent with the hypothesis that insulin resistance is associated with a higher risk of coronary heart disease mortality. However, it is doubtful that circulating insulin per se is a direct cause of arterial complications.