We aimed to investigate the potential association between vitamin D and serum leptin levels by pooling together the results from observational studies and clinical trials. A systematic literature search of PubMed, Scopus and Google Scholar was conducted up to March 2015. The analysis of observational studies was conducted on six papers that reported nine correlation coefficients using Fisher's Z and its standard error. Then, effect sizes of eligible trials were pooled using random-effects models (the DerSimonian-Laird estimator). Results of observational studies showed an inverse association between leptin and 25-hydroxyvitamin D (25(OH)D) (Fisher's Z = -0.93, 95% CI: - 0.95, - 0.91). After combining trials, pooled mean difference (PMD) for 25(OH)D was 24.06 ng/ml (95% CI, 17.27–30.85; P<0.001) with significant heterogeneity among studies (P<0.001; I2 = 89.1%). Raising 25(OH)D was associated with significant increase in leptin level (PMD = 4.60 ng/ml, 95% CI, 0.55–8.66, P = 0.026) with significant heterogeneity (P<0.001; I2 = 96.4%). Population with diabetes (PMD: 13.63 ng/ml), age younger than 50 years (PMD: 1.884 ng/ml), doses less than 1000 IU/day (PMD: 1.53 ng/ml), duration less than 24 weeks (PMD: 14.668 ng/ml) and baseline 25(OH)D <50 nmol/l (PMD: 13.483 ng/ml) were sources of heterogeneity. Current evidence indicates that inverse association between leptin level and 25(OH)D concentration was observed in observational studies, which was not demonstrated in intervention studies with high heterogeneity. Clearly, there is a need for properly designed and large prospective dose-response trials with long-term follow-up to assess the sources of heterogeneity.