Stroke is a polygenic or multifactorial disease, and each single susceptibility gene has modest effects. We hypothesize that combined effects of multiple genes might confer a higher stroke risk than a single susceptibility gene. To test our hypothesis we initially recruited 2000 stroke patients (44.3% thrombosis, 28.3% lacunar infarction and 27.4% intracerebral hemorrhage) and 2000 controls, and examined 6 polymorphisms in 5 candidate genes for stroke. Plasma lipoprotein(a) [Lp(a)] level was defined as a categorical variable and also included. Interactions between genetic risk factors were detected by the multifactor dimensionality reduction (MDR) method and further evaluated by multivariate logistic regression analyses. A significant combined effect on stroke due to the C677T polymorphism of methylenetetrahydrofolate reductase (MTHFR), the T2354A polymorphism of 5-lipoxygenase activating protein (ALOX5AP), and Lp(a) level, was detected using the MDR method. Furthermore, the combination of MTHFR 677TT, ALOX5AP 2354AA and Lp(a) elevation (Lp(a) concentration > or = 30 mg/dL) was found to be strongly associated with thrombotic stroke in males (OR, 10.419; 95%CI, 2.602 to 41.749; P= 0.001) using the multivariate logistic regression model. In conclusion, our results show that a combination of genetic risk factors can confer a higher risk for stroke than a single risk factor, indicating that people with multiple genetic risk factors have a higher risk of stroke and should be targets for prevention of this disease.