This study investigated the hypothesis that the insertion/deletion (4G/5G) polymorphism of the plasminogen activator inhibitor type-1 gene affects the risk for ischemic stroke, since results concerning this association have been controversial. We therefore performed a meta-analysis of published data regarding this issue. A comprehensive electronic search was carried out until January 2006. The analysis was performed using random-effects models and meta-regression. Eighteen eligible studies were retrieved (15 case-control studies and three cohort studies). The case-control studies included 3104 cases and 4870 control individuals concerning the contrast of 4G/4G versus remaining genotypes. The 4G pooled allele frequencies in cases and controls were 54.21 and 54.75%, respectively. Overall, the per-allele odds ratio of the 4G allele was 0.98 (95% confidence interval, 0.858-1.121). Regarding genotypes, we derived nonsignificant odds ratios in all contrasts. The subanalysis including the three studies with a prospective design in the 4G/4G versus 5G/5G contrast derived a significant result (relative risk, 0.523; 95% confidence interval, 0.353-0.775), but the estimated effect size was insignificant when cohort and case-control studies were analyzed together (relative risk, 0.848; 95% confidence interval, 0.662-1.087). We failed to demonstrate a significant association between the 4G/5G polymorphism and ischemic stroke under basal conditions. Determination of plasminogen activator inhibitor type-1 function seems of much higher clinical value than determination of the 4G/5G polymorphism. The effect of this genotype on risk of ischemic stroke in acute stressful diseases and the role of cohort studies in genetic epidemiology, however, warrant further investigation.