Background and Purpose— Cervical artery dissection (CAD) is a recognized cause of ischemic stroke among young and middle-aged individuals. The pathogenesis of dissections is unknown, although numerous constitutional and environmental risk factors have been postulated. To better understand the quality and nature of the research on the pathogenesis of CAD, we performed a systematic review of its risk factors. Methods— PubMed [MEDLINE (1966 to February 22, 2005)] and Embase (1980 to February 22, 2005) were searched to identify studies fulfilling the inclusion criteria. Two reviewers independently assessed methodological quality of the primary studies. Relevant data were extracted, including the risk factor(s) investigated, characteristics of the study population, and strength of the association(s). Results— Thirty-one case-control studies were included for analysis. Selection bias, lack of control for confounding, and inadequate method of data analysis were the most common identified methodological shortcomings. Strong associations were reported from individual studies for the following risk factors: aortic root diameter >34 mm (odds ratio [OR=14.2; 95% confidence interval [CI], 3.2 to 63.6), migraine (OR _adj , 3.6; 95% CI, 1.5 to 8.6), relative diameter change (>11.8%) during the cardiac cycle of the common carotid artery (OR _adj , 10.0; 95% CI, 1.8 to 54.2), and trivial trauma (in the form of manipulative therapy of the neck) (OR _adj , 3.8; 95% CI, 1.3 to 11). A weak association was found for homocysteine (2 studies: OR _crude , unknown; 95% CI, 1.05 to 1.52; OR _crude , 1.3; 95% CI, 1.0 to 1.7), and recent infection (OR _adj , 1.60; 95% CI, 0.67 to 3.80). Two studies had conflicting findings for low levels of α ₁ -antitrypsin, with the methodologically stronger study suggesting no association with CAD. Conclusions— CAD is a multi-factorial disease. Many of the reviewed studies contained 2 or more major sources of bias commonly found in case-control studies. Only one study (of homocysteine) used healthy controls, a robust sample size, and had a low risk of biased results. The relationship between atherosclerosis and CAD has been insufficiently examined.