Though there were many studies assessing the relationship between maternal thyroxine levels during pregnancy and cognitive development in children, there was still lack of evidence for the association from a comprehensive assessment of published data. To get a more comprehensive estimate of the influence of low maternal thyroxine levels on cognitive function, a meta-analysis of prospective cohort studies was performed. Two electronic databases, MEDLINE and EMBASE, were searched for relevant prospective cohort studies. Relative risks (RR) with 95% confidence intervals (95% CI) were pooled using random-effect model of meta-analysis to assess the risk of delayed cognitive development in children. Seven prospective cohort studies with a total of 8273 mother-child pairs were included into the meta-analysis. There was obvious between-study heterogeneity in the meta-analysis (I(2) = 69.6%). Meta-analysis of using random-effect model showed that low maternal thyroxine level was significantly associated with a threefold risk of delayed cognitive development in children (random RR = 3.08, 95% CI 1.83-5.18, P < 0.001). When excluding the study with largest weight, there was no obvious between-study heterogeneity in the left studies (I(2) = 47.6%), and meta-analysis using random-effect model showed that low maternal thyroxine level was still significantly associated with increased risk of delayed cognitive development in children (random RR = 3.76, 95% CI 2.14-6.58, P < 0.001). Sensitivity analysis by omitting other studies by turns showed that there was no obvious change in the pooled risk estimates, and all pooled RRs were statistically significant. Therefore, the findings from the meta-analysis provide strong evidence for the association between maternal thyroxine levels during pregnancy and cognitive development in children. Low maternal thyroxine level is significantly associated with a threefold risk of delayed cognitive development in children.