OBJECTIVE In following patients initially recruited for a cross-sectional study of blood viscosity in ischemic cerebrovascular disease, it was noted that those having a low albumin-globulin ratio appeared to experience the majority of subsequent vascular events. Accordingly, a prospective study in which subjects were assigned to a high or low albumin-globulin cohort was undertaken to examine the relation between a low albumin-globulin ratio, the presence of clinical risk factors for stroke, and the occurrence of subsequent stroke, myocardial infarction, or vascular death. METHODS Three groups of subjects were followed for an average of 1.5 +/- 0.8 years to ascertain vascular end points. Group 1 consisted of 126 patients with acute ischemic stroke; group 2 included 109 subjects matched with group 1 for age, medications, and recognized clinical risk factors for stroke; and group 3 was composed of 84 healthy volunteers, matched for age with groups 1 and 2. The median albumin-globulin ratio for group 1 at enrollment, 1.45, was used to dichotomize patients into two cohorts: all subjects with an albumin-globulin ratio of 1.45 or less were assigned to the "low" albumin-globulin cohort; those whose ratio was greater than 1.45 were assigned to the "high" albumin-globulin cohort. The occurrence of vascular end points was verified during subsequent hospitalizations and outpatient clinic visits and by telephone interviews of patients and providers. RESULTS A total of 51 vascular events occurred, including 39 in group 1, 8 in group 2, and 4 in group 3. Subjects in either group 1 or 2 who were in the low albumin-globulin cohort had at least double the risk for a subsequent vascular event compared with their counterparts in the high albumin-globulin cohort (P < .01 and P < .03, respectively). In comparison with the high albumin-globulin cohort, significantly more patients in the low albumin-globulin cohort in group 1 had a history of prior stroke (P < .03). When groups 1 and 2 were combined, both a low albumin-globulin ratio and diabetes had a significant independent association with increased risk for subsequent vascular events in a Cox proportional-hazards model (P < .01 and P < .03, respectively). CONCLUSIONS The results of this study indicate that significantly increased risk for subsequent vascular events in stroke patients and in subjects with clinical risk factors for stroke is associated with a shift in the concentrations of blood proteins to a prothrombotic environment characterized by lower levels of albumin and an increased concentration of globulins and fibrinogen.