We have retrospectively evaluated our results after aortic surgery in adults using deep hypothermia with circulatory arrest to determine the patient predictors of early death and postoperative stroke by logistic regression analysis. Of the 656 patients operated on between July 7, 1979, and January 30, 1991, 43% (n = 283) were female, the median age of the patients was 64 years (range 10 to 88 years), 12% (n = 77) had acute dissection, 26% (n = 173) had previously undergone either cardiac or ascending aortic operations, and 13% (n = 85) had a history of cerebrovascular disease. Eighty-four patients underwent elephant trunk procedures. The median circulatory arrest time was 31 minutes (range 7 to 120 minutes). The univariable predictors of transient or permanent stroke, defined as clinical evidence of neurologic injury, either global or hemiparetic, which occurred in 44 patients (7%), were as follows (p < 0.05): increased age; a history of cerebrovascular disease; circulatory arrest time (7 to 29 minutes = 12/298 [4%], 30 to 44 minutes = 15/201 [7.5%], 45 to 59 minutes 9/84 [10.7%], 60 to 120 minutes 7/48 [14.6%]; cardiopulmonary bypass time; and concurrent descending thoracic aorta repair. The multivariably determined predictors were as follows (p < 0.05): a history of cerebrovascular disease; previous aortic surgery distal to the left subclavian artery; and cardiopulmonary bypass time. A history of aortic valve incompetence, however, was associated with a lower risk of stroke (adjusted odds ratio 0.42, p = 0.015). The multivariably determined predictors for increased risk of early death (p < 0.05), which occurred in 66 (10%) patients, were as follows: increased age; Marfan syndrome; concurrent distal aortic aneurysm; previous ascending aortic operation; cardiopulmonary bypass time; cardiac complications; renal complications; and stroke. In this study, the occurrence of stroke was observed to increase after 40 minutes of circulatory arrest; furthermore, the mortality rate increased markedly after 65 minutes of circulatory arrest. Thus the "safe" period for strokes not developing appeared to be limited to approximately 40 minutes. We conclude that deep hypothermia with circulatory arrest is a safe technique for the repair of complex aortic problems provided both the circulatory arrest and the cardiopulmonary bypass times are not excessive. In addition, the clinical characteristics of the patients are important determinants of stroke and death.