Genetic prediction of type 1 diabetes in a population with low frequency of HLA risk genotypes and low incidence of the disease (the DIABFIN study)

Abstract
Background To develop a sensitive, specific screening strategy for predicting genetic risk for type 1 diabetes mellitus (T1DM) in the low‐incidence continental Italian population, and to define with this tool, a cohort of high‐to‐moderate risk infants for an immunological follow‐up study aimed at identifying environmental risk factors for T1DM. Methods 4855 newborns in three regions of continental Italy were screened for T1DM HLA‐DRB1‐DQB1 risk genotypes using a reverse line blot typing method. Risk classification was based on odds ratios (OR) found in a preliminary case–control study (356 T1DM patients, 412 controls). Screening efficiency was optimized by allele subtyping. Results Screening for well‐known T1DM susceptibility genotypes [DRB1*03/*04‐DQB1*0302; DRB1*03/*03; DRB1*04/*04‐DQB1*0302; DRB1*04‐DQB1*0302/X where X ≠ DRB1*03, DRB1*04‐DQB1*0302, DQB1*0602 or DQB1*0603] was associated with <60% sensitivity due to their low frequencies in the general Italian population. Inclusion of an additional genotype from which protective DRB1 and DQB1 alleles had been excluded [DRB1*03/X° where DQB1 ≠ *0301, *0503, *0602, or *0603 and X° ≠ DRB1*03, DRB1*04‐DQB1*0302 or DRB1*07] increased screening sensitivity to 75% (specificity: 85%). Among 4855 newborns, we have found the high‐risk genotype [DRB1*03/*04‐DQB1*0302; estimated absolute risk (AR) 1/23] to be present in only 0.9%. The moderate‐risk genotypes were found in 13.8% of newborns (estimated AR 1/177). Conclusions Risk classification must be tailored to the characteristics of the individual population, in particular, the allelic frequencies in the background population and T1DM prevalence. We have developed a screening strategy with good levels of sensitivity that should prove effective for use throughout the Italian peninsula. Copyright © 2004 John Wiley & Sons, Ltd.