Several studies have evaluated the association between mannose-binding lectin (MBL) polymorphisms and sepsis. However, the results are inconclusive and conflicting. To better understand the roles of MBL polymorphisms in sepsis, we conducted a comprehensive meta-analysis. All relevant studies were searched from PubMed, EMBASE and Web of Knowledge databases, with the last report up to 7 May 2013. Twenty-nine studies addressing four MBL polymorphisms (–550G/C, –221G/C, structure variant A/O, Gly54Asp) were analysed for susceptibility to sepsis and one study for sepsis-related mortality. Overall, significant associations between structure variant A/O and susceptibility to sepsis were observed for AO + OO vs. AA [odds ratio (OR) 1·27, 95% confidence interval (CI) 1·05–1·52, P = 0·01] and O vs. A (OR 1·19, 95% CI 1·02–1·40, P = 0·03). In subgroup analysis based on age group, increased risk was found in the paediatric group in the dominant model (OR 1·72, 95% CI 1·16–2·56, P = 0·007). Moreover, there was a slight association between the +54A/B polymorphism and susceptibility to sepsis in Caucasians (recessive model: OR 10·64, 95% CI 1·24–91·65, P = 0·03). However, no association was observed for –550G/C and –221G/C polymorphisms both overall and in subgroup analysis. For sepsis-related mortality, only one study suggested AO/OO was associated with in-hospital mortality in pneumococcal sepsis patients after controlling for confounding variables. Our meta-analysis indicated that MBL structure variants might be associated with susceptibility to sepsis but further studies with a large sample size should be conducted to confirm these findings.