Objectives. The Val66Met polymorphism (rs6265) of the BDNF gene is a non-synonymous polymorphism, previously associated with anorexia nervosa (AN). Methods. We genotyped rs6265 in 235 patients with AN and 643 controls. Furthermore, we performed a systematic review of all case-control and family-based studies testing this SNP in AN, and combined the results in a meta-analysis. Results. The results of the case-control study were non-significant. For the meta-analysis, nine studies were identified (ncases = 2,767; ncontrols = 3,322, ntrios = 53) and included. Primarily, the analyses indicated an association with OR of 1.11 (P = 0.024) in the allelic contrast, and OR of 1.14 (P = 0.025) for the dominant effect of the Met allele. However, additional analyses revealed that the first published study (from those included in the meta-analysis) overly influenced the pooled effect size (possibly due to a phenomenon known as a winner's curse). When this case-control study was replaced by a trio study (ntrios = 293) performed on a largely overlapping sample, the effect size became smaller and non-significant, both for the allelic contrast (OR = 1.07, P = 0.156) and the dominant effect (OR = 1.07, P = 0.319). The quality of included studies was good and there was no significant heterogeneity across the effect sizes. Conclusions. Our analyses indicate that the BDNF Val66Met variant is not associated with AN at detectable levels. © 2013 Informa Healthcare.