Certain HLA-DQ alleles are known to contribute to predisposition to coeliac disease (CD). The existence of additional independent risk-modifying loci in the HLA complex is still being debated. The DR3-DQ2 haplotype has been studied most, but the evidence is conflicting. The discrepancies may stem from the absence of such an effect, insufficient statistical power to detect an effect (i.e. small studies) and/or incomplete control of linkage disequilibrium (LD) to the neighbouring DQ-loci, known to elicit a strong effect. In the present study, we aimed to undertake a statistically high-powered family-based analysis, fully controlling effects of LD between the major DQ-risk haplotypes and neighbouring candidate loci. We investigated five markers on DR3-DQ2, DR5-DQ7 and DR7-DQ2 haplotypes in 327 Norwegian and Swedish families. Our primary finding was that TNF-308A ( TNF2) was significantly associated on the DR3-DQ2 haplotype [stratum specific odds ratio (OR) = 2.40 (1.25-4.48), Pc = 0.009, where P(c) = Pn and n = number of tests performed]. Furthermore, we confirmed earlier indications that LD between TNF2 and DQA1*05-DQB1*02 on the DR3 haplotype is more strongly maintained in family-based cases than family-based controls. In conclusion, we confirmed in this study, the largest of its kind, that additional CD risk factors independent of DQ2 alleles do exist on the DR3 haplotype.