Peripheral blood brain-derived neurotrophic factor (BDNF) has been proposed as a potential biomarker related to disease activity and neuroprogression in bipolar disorder, speculated to mirror alterations in brain expression of BDNF. The research area is rapidly evolving; however, recent investigations have yielded conflicting results with substantial variation in outcomes, highlighting the need to critically assess the state of current evidence. The aims of the study were to investigate differences in peripheral blood BDNF concentrations between bipolar disorder patients and healthy control subjects and between affective states in bipolar disorder patients, including assessment of the effect of treatment of acute episodes on BDNF levels. A systematic review of English language studies without considering publication status was conducted in PubMed (January 1950-November 2014), Embase (1974-November 2014) and PsycINFO (1806-November 2014), and 35 studies comprising a total of 3798 participants were included in the meta-analysis. The results indicated that crude peripheral blood BDNF levels may be lower in bipolar disorder patients overall (Hedges' g = -0.28, 95% CI: -0.51 to -0.04, P = 0.02) and in serum of manic (g = -0.77, 95% CI: -1.36 to -0.18, P = 0.01) and depressed (g = -0.87, 95% CI: - 1.42 to -0.32, P = 0.002) bipolar disorder patients compared with healthy control subjects. No differences in peripheral BDNF levels were observed between affective states overall. Longer illness duration was associated with higher BDNF levels in bipolar disorder patients. Relatively low study quality, substantial unexplained between-study heterogeneity, potential bias in individual studies and indications of publication bias, was observed and studies were overall underpowered. It could thus not be excluded that identified differences between groups were due to factors not related to bipolar disorder. In conclusion, limitations in the evidence base prompt tempered conclusions regarding the role of peripheral BDNF as a biomarker in bipolar disorder and substantially improving the quality of further research is warranted.