The current focus of the screening for individuals at risk for type 1 diabetes has moved from first-degree relatives to the general population. However, there is a shortage of data on the predictive utility of various risk markers in the background population in various countries, and predictive strategies for the general population remains open. We studied the frequency of diabetes-associated autoantibodies in a series of 3,652 unaffected Finnish school-children, and determined the relationships between autoantibodies and HLA-DQB1 risk markers. In addition, all subjects were observed for progression to type 1 diabetes.
The reported frequencies of autoantibodies in the general population have varied widely. In the present study, the frequencies of ICA (islet cell antibodies), GADA (GAD antibodies), IA-2A (antibodies to IA-2 protein), and IAA (insulin autoantibodies) were 2.8, 0.5, 0.6, and 0.9%, respectively. Multiple antibodies (i.e., two or more) were detected in 21 children (0.6%),and 9 children (0.25%) had three or four antibody specificities in their initial blood sample. The present cut-off limits for antibody-positivity(determined as the 99th percentile in >370 healthy control subjects) are close to the 99.5th percentile for IA-2A and GADA and the 99.0th percentile for IAA in the present series of 3,652 children. The use of different approaches for the definition of cut-off limits(1,2)resulted in somewhat higher frequencies of autoantibodies, although this did not improve the diagnostic sensitivity of any autoantibody. These data illustrate the difficulties in defining borderline positivity and in directly comparing of results from different studies, and emphasize the need for common international standards to be used in these autoantibody assays.
Our current knowledge of the relation between HLA-DQB1 risk markers and autoantibodies is mainly based on family surveys and studies on patients with type 1 diabetes. We genotyped ∼600 healthy schoolchildren, including all but one of the 141 antibody-positive children. The subjects with GADA (>6.6 relative units [RU]) carried the DQB1*0302 allele [58% (CI 33-80)vs. 24% (20-29), P = 0.002] and the DQB1*02/0302 genotype[21% (6-46) vs. 4% (2-6), P = 0.007] more frequently than the antibody-negative control subjects, and carried the DQB1*0602 or *0603 allele [16% (3-40) vs. 41% (36-46), P = 0.031] less frequently. In addition, the subjects with moderate or high titers of GADA(≥20 RU) carried the DQA1*05-DQB1*02 haplotype more frequently than the antibody-negative control subjects [47% (CI 21-73) vs. 17%(14-21), P = 0.010]. The children with detectable levels of ICA carried the DQA1*05-DQB1*02 haplotype more frequently than the ICA- children [27% (18-36) vs. 17% (14-21), P =0.036]. No specific associations were observed between IA-2A or IAA and DQB1 alleles or genotypes, except for a weak positive association between moderate or high levels of IA-2A (≥10 RU) and the DQB1*0302 allele [5 of 9 (56%, CI 21-86) vs. 106 of 436 (24%, CI 20-29), P = 0.047]. None of the nine children with three or four antibody specificities had a high-risk DQB1 genotype, whereas seven of them (78%) carried either the DQB1*0302 or the DQB1*02 risk allele. Surprisingly,three of these nine subjects had a protective genotype. Accordingly, the ongoing β-cell autoimmunity in these three subjects is probably initiated by strong environmental factors and/or is related to genetic factors other than those residing in the HLA region.
Four subjects (0.11%) progressed to type 1 diabetes over a median follow-up of 5.3 years (range 5.2-5.5). The intervals from the initial blood sampling to the diagnosis ranged from 0.9 to 4.4 years, and the age at diagnosis varied from 7.8 to 13.4 years. All progressors had multiple (more than two)antibodies in their initial blood sample, whereas none of them carried the high-risk DQB1*0302 allele. One of them carried the protective DQB1*0602, and another subject carried the DQB1*0603 allele. An intravenous glucose tolerance test was performed in these two latter subjects, and both of them had a markedly reduced first-phase insulin response for a long time before the diagnosis, as we have reported previously(3). Two subjects carried the DQA1*05-DQB1*02 haplotype, one of them being homozygous and the other also having the protective DQB1*0602 allele. One subject carried the DQB1*x/x genotype (x = other than *02, *0301, *0302, or *0602), and one carried the protective DQB1*0301/0603 genotype. These results suggest that, at least in some individuals representing the general population, HLA-DQB1 high-risk markers are not indispensable for progression to over type 1 diabetes, and the “protective” alleles do not provide full protection against the disease. In the present series, all of the children were older than 6 years of age when initially screened, and all but one of the progressors were older than 10 years of age at diagnosis, which must be taken into account when considering the DQB1 genotype distribution in the progressors, because we previously observed an association between age and the prevalence of DQB1 risk markers in patients with type 1 diabetes(4). In addition, the number of children who have so far presented with clinical diabetes is still low, and accordingly, the genotype distribution may be a matter of chance. Furthermore,17 (0.4%) children from the initial target population of 4,280 schoolchildren were initially excluded from the study, because they had previously been diagnosed with clinical type 1 diabetes(3). HLA-DQB1 data was available for 15 of these subjects, and only one of them carried the protective DQB1*0602 allele, whereas 6 (40%) had the DQB1*02/0302 high-risk genotype(3).
Two recent studies have proposed a two-step screening strategy for prediction of type 1 diabetes in the general population(1,5). Although autoantibodies were 100% sensitive to predict type 1 diabetes in the present series of schoolchildren, the positive predictive value of any screening strategy did not exceed 50%. The HLA-DQB1 risk markers complicated the prediction, because the two progressors with a “protective”DQB1*0602 or *0603 allele would have been classified as subjects with a low risk of progression to type 1 diabetes. The positive predictive value of ICA ≥20 Juvenile Diabetes Foundation units was 29%, and that of IA-2A and multiple antibodies was 19%, all having a disease sensitivity of 100%. The positive predictive value and sensitivity of GADA were lower (11 and 50%), because only two GADA+ subjects have progressed to type 1 diabetes. On the other hand, GADA have been suggested to be related to a slower progression to clinical disease; therefore, a number of GADA+ children in the present study may later progress to over diabetes. Environmental factors and/or genetic factors other than HLA-DQB1 may also be involved in the appearance of β-cell autoimmunity in the general population, at least after the age of 6 years. With the currently available methods, the accurate assessment of the individual risk of type 1 diabetes is complicated. The present data strongly support the use of combined screening for autoantibodies, whereas further studies are needed to establish the value of current genetic risk markers in the prediction of type 1 diabetes in populations of schoolchildren.