Meta-analyses, published in 2008-2010, have confirmed abnormally low serum brain-derived neurotrophic factor (BDNF) concentrations in depressed patients and normalization of this by antidepressant treatment. These findings are believed to reflect peripheral manifestations of the neurotrophin hypothesis, which states that depression is secondary to an altered expression of BDNF in the brain. Since the publication of these meta-analyses, the field has seen a huge increase in studies on these topics. This motivated us to update the evidence on the aforementioned associations and, in addition, to compile the data on serum BDNF concentrations in relation to the symptom severity of depression. Using a manifold of data as compared with earlier meta-analyses, we find low serum BDNF concentrations in 2384 antidepressant-free depressed patients relative to 2982 healthy controls and to 1249 antidepressant-treated depressed patients (Cohen's d=-0.71 and -0.56, P-values <0.0000001). When publication bias is accounted for, these effect-sizes become substantially smaller (d=-0.47 and -0.34, respectively, P-values<0.0001). We detect between-study heterogeneity in outcomes for which only year of publication and sample size are significant moderators, with more recent papers and larger samples sizes in general being associated with smaller between-group differences. Finally, the aggregated data negate consistent associations between serum BDNF concentrations and the symptom severity of depression. Our findings corroborate the claim that altered serum BDNF concentrations are peripheral manifestations of depression. However, here we highlight that the evidence for this claim is slimmer as was initially thought and amidst a lot of noise.