The aim of this study was to investigate the potential role of brain-derived neurotrophic factor (BDNF) in children with autism spectrum disorders (ASD) by measuring serum circulating levels of BDNF as well as calcium and comparing them with age-matched and sex-matched normal controls. The study included 75 drug-naive ASD children and 75 age-sex-matched healthy children. The concentration of serum BDNF was determined using the enzyme-linked immunosorbent assay method at baseline. Clinical information was collected. The severity of ASD was assessed at admission using the Childhood Autism Rating Scale total score. The results indicated that the mean serum BDNF levels were significantly (P<0.0001) higher in children with ASD compared with the control cases (17.59±5.55 vs. 11.21±2.79 ng/ml; t=8.902). On the basis of the receiver operating characteristic curve, the optimal cutoff value of serum BDNF levels as an indicator for auxiliary diagnosis of autism was projected to be 12.65 ng/ml, which yielded a sensitivity of 80.8% and a specificity of 70.2%; the area under the curve was 0.840 [95% confidence interval (CI), 0.777-0.904]. In univariate logistic regression analysis, with an unadjusted odds ratio of 9.42 (95% CI, 4.33-25.95; P<0.0001), BDNF of 12.65 ng/ml or more had an association with a diagnosis of ASD. After adjusting for possible covariates, BDNF of 12.65 ng/ml or more is still an independent indicator of ASD, with an adjusted odds ratio of 7.33 (95% CI, 2.98-16.55; P<0.0001). Serum BDNF levels may be associated independently with the severity of ASD, and higher BDNF levels could be considered an independent diagnostic marker of ASD.