To evaluate the consistency, strength, and independence of the relation of carotid atherosclerosis to coronary atherosclerosis, we quantified coronary artery disease risk factors and extent of carotid atherosclerosis (B-mode score) in 343 coronary artery disease patients and 167 disease-free control patients. In univariable analyses, there was a strong association between coronary status and extent of carotid artery disease in men and women older than and younger than 50 years (p less than 0.001 for men and women greater than 50 years, p less than 0.001 for women less than or equal to 50 years, p = 0.045 for men less than or equal to 50). The relation remained strong after control for age in men and women older than 50 years and in women younger than 50 (p less than 0.001 for men and women greater than 50 years, p = 0.003 for women less than or equal to 50) but did not persist after control for age in men younger than 50. Logistic models that included coronary disease risk factors, with or without B-mode score, as independent variables and presence or absence of coronary disease as the outcome variable indicated that the extent of carotid atherosclerosis was a strong, statistically significant independent variable in models for men and women older than 50 years of age. Next, we examined the usefulness of B-mode score as an aid in screening for coronary artery disease in men and women older than 50 years. Classification rules, both including and excluding B-mode score, were developed based on logistic regression and, for comparison, recursive partitioning (decision trees). The performance of these rules and the bias of their performance statistics were estimated. The improved classification of the study sample when B-mode score was incorporated in the rule was statistically significant only for men (p = 0.015). However, the addition of B-mode score was found to 1) increase the median discrimination score for both sex groups based on the logistic model, and 2) yield better sensitivities and specificities for rules based on recursive partitioning. Thus B-mode score is strongly, consistently, and independently associated with coronary artery disease in patients older than 50 and is at least as useful as well-known risk factors for identifying patients with coronary artery disease.