In a population of 46 children with CD recruited in the Paris area of France, an excess of DRB1*03 and DRB1*07 alleles and of DR3/DR7, DR3/DR3 and DR11(or 12)/DR7 phenotypes was found (RRs of 6.3, 9.3, 24.6, 15, and 15.1, respectively), which is reminiscent of the markers of susceptibility observed in southern rather than in northern European celiac patients. More importantly, the highest association with CD was not found in individuals expressing the DQA1*0501-DQB1*0201 heterodimer in single dosage (RR = 24.9) or in homozygous state, but in people co-expressing one copy of DQA1*0501-DQB1*0201 on one haplotype and a second copy of DQB1*0201 on the second haplotype (RR = 35.7). This suggests that in our population either DQB1*0201 or a gene closely linked to DQB1*0201 influences the susceptibility to CD conferred by the DQA1*0501-DQB1*0201 heterodimer. Significant positive or negative RRs conferred by some TAP2 or DPB1 alleles were found. However, they were moderate compared to the RR conferred by the expression of a second copy of DQB1*0201. Moreover, they were no longer significant when patients were compared with HLA-DR matched controls. This suggests that associations of CD with TAP2 and DPB1 alleles are secondary to linkage disequilibria and argues against the contribution of these alleles in resistance and/or susceptibility to CD. Thus the "raison d'être" of a "DQB1*0201 second haplotype effect" in susceptibility to CD remains to be elucidated.