One of the excluded studies appeared to examine the GP 1a association of pre-eclampsia with several polymorphisms including PAI-1. The report stated that a statistically significant association was not found but did not provide any data. This form of selective reporting bias may also exist undetected in other studies. The prevalence of the PAI-1 polymorphism varies between ethnic groups and ethnicity affects the risk of pre-eclampsia. Therefore, different ethnic distributions of cases and controls may distort any gene-disease associations. Although most of the studies in this review recruited cases and controls from ethnically-similar populations in a defined geographical region, the use of broad ethnic group categorisations and inconsistencies in how ethnicity data were defined and reported may have caused residual confounding. The possibility that population admixture has distorted the strength of the pooled estimate of association remains. Family-based studies using transmission disequilibrium testing to account for population admixture have not yet examined the association of the PAI-1 polymorphism with pre-eclampsia. A major limiting factor in determining the validity and applicability of the findings of any systematic review and meta-analysis is the quality of the included studies. However, unlike the quality assessment of randomised controlled trials, empirical evidence to determine the degree to which different genetic association study characteristics introduce bias is lacking. Furthermore, the methods used in many genetic association studies are reported variably and incompletely. The potential methodological design weakness in genetic association studies include imprecise phenotype definition, inappropriate control selection, and lack of blinding of laboratory genotyping staff to the clinical status of cases and controls. We specifically excluded studies in which participants included women with gestational hypertension in order to improve the homogeneity of phenotype between studies. The absence of HEPES statistical heterogeneity in the metaanalyses is reassuring and may reflect the a priori requirement for studies to have used an internationally-accepted case definition. Furthermore, only studies in which controls were women who had a pregnancy uncomplicated by pre-eclampsia were eligible for inclusion. This inclusion criterion reduces the risk that any observed association is due to another unrelated or unknown factor. Only two studies included in this review reported blinding of genotyping. Some of the older genotyping techniques are more subjective than the modern methods and knowing the clinical status of the case or the control may have influenced the interpretation of the genotyping result. We have managed this problem to some extent by excluding studies in which the genotype distribution within the controls deviated substantially form Hardy Weinberg equilibrium. The underlying pathophysiology in most cases of ischemic stroke involves thrombotic or thromboembolic arterial occlusion.