Abstract Introduction Previous studies have provided inconsistent results on whether variants in the MBL2 gene, coding for the complement-activating mannan-binding lectin (MBL) protein, associate with rheumatoid arthritis (RA). We re-evaluated this in context of the main environmental and genetic risk factors (smoking, HLA-DRB1 'shared epitope' (SE), PTPN22 *620W), which predispose to rheumatoid factor (RF) and/or anti-citrullinated-protein antibody (ACPA)-positive RA. Methods In this population-based EIRA study, rheumatoid factor (RF), ACPA, smoking, SE and PTPN22 *620W status was determined in incident RA cases and matched controls. MBL-high ( n = 1330) and MBL-low ( n = 1257) genotypes predicting MBL levels were constructed from four promoter and exon-1 polymorphisms in the MBL2 gene. Odds ratios with 95% confidence interval (OR, 95% CI) were calculated by logistic regression. In extended families ( n = 316), previously reported data were re-analyzed, considering RF and smoking. Results MBL-high genotypes tended to be associated with RF-negative (OR = 1.20, 95% CI 0.96-1.51) but not RF-positive (OR = 1.00, 95% CI 0.83-1.20) RA. Results divided by ACPA status did not differ. When stratified for smoking, MBL-high genotype was strongly associated with RF-negative RA in never smokers (OR = 1.82, 95% CI 1.24-2.69) but not in ever smokers (OR = 0.96, 95% CI 0.73-1.30). In never smokers, the association was observed in both the RF-negative/ACPA-negative (OR = 1.67, 95% CI 1.10-2.55) and RF-negative/ACPA-positive subgroups (OR = 3.07, 95% CI 1.37-6.89), and remained on an SE/ PTPN22 *620W negative background. In the extended families, the reported association between high MBL and RA was in fact confined to never smokers. Conclusions High MBL may predispose to RF-negative RA but only in individuals who have never smoked. This illustrates the importance of phenotypic subgrouping in genetic studies.