Objective To explore the relationship between selenium and childhood autism spectrum disorder. Methods PubMed， Embase， Cochrane Library， Web of Science， VIP， Wanfang Data， CNKI， and CBM were searched to collect studies of the relationship between Se and childhood autism from the establishment of the databases up to Octuoer 2021. Newcastle-Ottawa scale （NOS） was used to evaluate the quality of methodology. RevMan 5.3 and Stata 16 softwares were used to conduct a Meta-analysis of the extracted data. Results Totally 1 225 literatures were retrieved by computer， and including 20 case-control studies involving 2 119 subjects. The level of selenium from serum， nail and urine in the autism group were much lower than those in the control group ［SMDserum=-1.08， 95%CI（-1.61， -0.54）， P＜ 0.01； SMDnails=-1.28， 95%CI（-2.41， -0.14）， P=0.03； SMDurine=-0.24， 95%CI（-0.46， -0.03）， P=0.03］， and the differences were statistically significant. The level of selenium from hair， red cell and plasma in the autism group were not significantly different from those of the control group ［SMDhair=0.46，95%CI （-0.49， 1.40）， P=0.34； SMDred cell=-0.39， 95%CI （-1.76， 0.98）， P=0.58； SMDplasma=-0.09， 95%CI （-1.39， 1.20）， P=0.89］. Conclusions Selenium levels in serum， urine and nails are associated with autism spectrum disorders in children. Therefore， helping children develop good eating habits has a certain reference significance for the prevention of autism spectrum disorders.