Although the use of arteriography to define coronary artery disease is considered the optimal method for determining associations of risk factors with atherosclerosis, the strengths of associations found have varied greatly between arteriographic studies. To assess whether such associations are affected by changes in the arteriographic definition of absence of disease, the authors evaluated one sample of patients undergoing coronary arteriography. The prevalence of specific risk factors by the extent of atherosclerosis (0-24, 25-49, and greater than or equal to 50% maximal stenosis) was determined, as were their associations with coronary artery disease (greater than or equal to 50% stenosis), using persons with 0-24 or 0-49% maximal stenosis, separately, as control groups. Among both males and females, analysis showed increased prevalence of risk factors with increasing atherosclerosis, with greater similarity of those who had moderate atherosclerosis to the cases than to those who had minimal atherosclerosis. The strength of association of arteriographically defined coronary disease with age, hypertension, diabetes, low and high density lipoprotein cholesterol, smoking, and family history generally varied inversely with the extent of atherosclerosis permitted within the control group. These findings indicate that epidemiologic understanding of associations of risk factors with atherosclerosis may be weakened by including persons with moderate, subclinical atherosclerosis within the control group.