Epidemiological studies have evaluated the associations between brain-derived neurotrophic factor (BDNF) 196A/G gene polymorphism and Alzheimer's disease (AD) risk. However, the results remain inconclusive. Sexually dimorphic effect of the polymorphism of BDNF 196A/G in AD patients had been proposed previously, specifically in female group. As more cases were reported, therefore, we performed a meta-analysis of published case-control studies to better understand these results. We systematically searched online databases of Embase, PubMed and Web of Science, as well as hand searching of the references of identified articles and meeting abstracts. Review Manager (Version 5.2.4) and Stata software (Version 12.0) were used for statistical analyses. The pooled odds ratios (ORs) with 95% confidence intervals (95% CIs) were calculated. A total of 23 publications including 25 studies were identified and entered the analysis. No significant association was observed in overall population, as well as subgroups stratified by ethnicity (Caucasian and Asian). However, when stratified by gender, significant association was observed just in female subgroup (A allele vs. G allele: OR = 1.15, 95% CI = 1.06-1.25; A/A vs. G/G: OR = 1.29, 95% CI = 1.06-1.57; A/A + A/G vs. G/G: OR = 1.30, 95% CI = 1.11-1.53). This meta-analysis confirmed the gender-related association between BDNF 196A/G polymorphism and AD risk, which may indicate a certain effect of female hormone on progression of the disease. Larger sample size and more studies with homogeneous AD patients and well-matched controls are needed in future.