Background: A common single-nucleotide polymorphism identified in the 5′-untranslated region of the leptin gene (LEP -2548 G/A polymorphism) may be associated with obesity, but the existing research findings are inconsistent, so we conducted this meta-analysis. Methods: Medline, Embase and ISI Web of Science databases were searched to identify relevant studies. Meta-analysis of the total and subgroup populations was conducted using allelic, additive, dominant and recessive models, and odds ratios and their 95% confidence intervals were calculated in a fixed-effect model if no heterogeneity (evaluated as I2 statistic) existed. Otherwise, a random-effects model was adopted. Subgroup analysis was performed by ethnicity. Meta-regression and the HETRED analysis were used to explore the potential sources of between-study heterogeneity. Egger's test and influence analysis were conducted to evaluate the publication bias and study power, respectively. Results: The final selection enrolled 9 studies, including 2,988 subjects (1,372 obese subjects and 1,616 controls). No significant association was identified between the LEP -2548 G/A polymorphism and obesity for all genetic models in the overall population and Caucasians. We found a significant association with allelic, additive and dominant models for subjects of mixed race from South America. Notwithstanding, this significance should be treated cautiously for it is based on a rather small sample (788 involved subjects). Conclusions: In total, the combined analysis of data from current and published studies suggested that the LEP -2548 G/A polymorphism does not contribute to the development of obesity, despite the fact that a significant association exists in a small subgroup from South America. Further studies are needed to elucidate the relationship.