Risk of Type 1 Diabetes Development in Children With Incidental Hyperglycemia: A multicenter Italian study

In November 1991, the Italian Society of Pediatric Endocrinology and Diabetology (Parma, Italy) promoted the creation of the Screening and Treatment of Prediabetes Study Group. To date, three projects have been carried out: the first project set guidelines for the screening of prediabetes in children (13,14); the second established normal values for first-phase insulin response (FPIR) in normal subjects (15); and the third kept a registry of Italian subjects at risk of type 1 diabetes (16).

According to the guidelines, there were four categories of individuals to be screened: 1) first-degree relatives of type 1 diabetic probands (siblings or offspring); 2) subjects with fasting plasma glucose >5.6 mmol/l (confirmed on two occasions), with no history of type 1 diabetes, without obesity, and not receiving drugs that cause hyperglycemia; 3) subjects with autoimmune endocrinopathies; and 4) subjects with congenital rubella.

A total of 748 of 2,467 subjects fitting into the four previously mentioned categories were found to have incidental elevated glycemia without any family history of type 1 diabetes (17). These subjects, 1–18 years of age (9.04 ± 3.62, mean ± SD) (480 males and 268 females) have been screened by 31 Italian pediatric centers. In the majority of cases, elevated glycemia was observed during a routine blood test. In a few subjects, high blood glucose levels were found during acute illnesses but needed to be confirmed after recovery, before enrollment into the study. Obese subjects (BMI >30 kg/m²) were excluded.

Screening procedures, according to the flow chart established by the Italian Diabetes Study Group, included immunological, metabolic, and immunogenetic investigations. Moreover, to exclude the diagnosis of type 1 diabetes, an oral glucose tolerance test (OGTT) was recommended, according to World Health Organization (WHO) criteria (18). The following immunological markers were determined: ICAs, IAAs, and GADAs. Since 1996, IA-2As were also determined. FPIR to intravenous glucose tolerance testing (IVGTT) was used as a metabolic marker. The results were compared with centile values according to pubertal stage in 138 normal subjects (47 females and 91 males, 3–20 years of age), established by the Italian Prediabetes Study Group (15). Regarding immunogenetic markers, at least in the subjects with immunological and/or metabolic markers, serological HLA typing for class I and class II was done, and molecular analysis of HLA-DQA1 and -DQB1 genes was performed to identify the HLA-DQ genotypes that confer susceptibility to type 1 diabetes and to calculate the number of DQα and DQβ heterodimers.

Regarding the criteria for the follow-up, OGTT was performed almost every 6 or 12 months, according to the previous results (i.e., every 6 months in cases of impaired glucose tolerance and every 12 months in cases of normal glucose tolerance). A normal diet was allowed. Immunological markers and IVGTT were repeated in intervals of 6–12 months in the cases of abnormal results in initial studies. Insulin therapy was instituted according to WHO criteria.

Blood glucose concentration was measured in venous blood at the time of the venipuncture for routine evaluation. Each blood glucose value for hyperglycemic patients was confirmed by a second evaluation.

The presence of ICAs was determined by indirect immunofluorescence on unfixed snap-frozen human pancreas. A sample was considered to be positive if the undiluted serum staining was similar or more intense than the laboratory standard serum sample that had been calibrated to ∼5 Juvenile Diabetes Foundation (JDF) units (19).

IAAs were measured by protein A/G microradiobinding assay, as previously described. Results for each assay were expressed as arbitrary units, and the threshold for positivity was 5 U, corresponding to the 99th centile of normal control subjects (20).

GADAs and IA-2As were measured by radiobinding assay, using in vitro transcribed and translated recombinant human protein, as previously described (21). The thresholds for positivity were 3 U (GADA) and 1 U (IA-2A). Using these thresholds, the assays gave a sensitivity, specificity, and reproducibility of 88, 98, and 100% for GADA and of 70, 99, and 100% for IA-2A in the combined autoantibody workshop (22).

A total of 409 subjects underwent IVGTT according to National Diabetes Data Group recommendations for OGTT, i.e., after 3 days of unrestricted diet (at least 150 g carbohydrates), normal physical activity, and absence of acute illness and administration of drugs that cause hyperglycemia (18). After an overnight fast for 12 ± 1 h, the test started between 7:30 a.m. and 10:00 a.m. The procedure followed the recommendations of the Italian Study Group (15). The sum of 1- and 3-min serum insulin values (FPIR) was calculated and expressed as microunits per milliliter. Blood samples for insulin assay were centrifuged at 4°C, and then plasma was kept at −20°C until analysis. All plasma samples were then sent to the Parma University Laboratory for insulinemia measurement. Insulin levels were measured by radioimmunoassay (Radim; Rome). The intra- and interassay coefficients of variation were 8.2 and 6.9% for low values and 7.7 and 6.0% for high values, respectively. The results were compared with the percentile established in healthy subjects (15).

In all subjects, an OGTT was performed using a dose of 1.75 g glucose/kg body wt. Samples for blood glucose determination were obtained at baseline and 2 h after glucose administration, according to the criteria of the National Diabetes Data Group (18). Subjects underwent OGTT, as recommended by the National Diabetes Data Group, i.e., after 3 days of unrestricted diet (at least 150 g carbohydrates), normal physical activity, and absence of acute illness and administration of drugs that cause hyperglycemia (18). After an overnight fast of 12 ± 1 h, the test began between 7:30 a.m. and 10:00 a.m. The results were considered according to WHO criteria (18).

HLA typing was performed according to the microlymphocytotoxicity technique (23). The analysis of HLA-DQA1 and -DQB1 polymorphisms at genomic levels was performed by the polymerase chain reaction/sequence-specific primers technique (24). The results were compared with those of 40,071 Italian bone marrow donors enrolled in the Italian Bone Marrow Donor Registry and representative of the Italian population (25). Mean age of control subjects was 26 years and the male-to-female sex ratio was 0.98.

Descriptive statistics were computed for all variables: frequency distribution was reported for categorical variables, and mean and SD or median and quartiles were reported for continuous variables, if skewed. Cumulative event-free probability was computed by Kaplan-Meier estimates for the whole case series and after stratifying by risk factor. Event-free survival curves were drawn. The prognostic value for the occurrence of diabetes during follow-up of a series of covariates was assessed by univariate Cox models. Hazard ratios and 95% CIs were calculated for each tested variable. A χ² test was computed to compare HLA distribution between hyperglycemic patients and control subjects. Reported P values were not corrected for multiple test bias. Multivariate analysis could not be performed because of the low number of events. P < 5% was considered statistically significant. Stata 6 (StatCorp, College Station, TX) was used for computation.

ICA levels >5 JDF U were present in 50 of 498 (10%) tested subjects with hyperglycemia, and IAA levels were elevated in 19 of 408 (4.6%) subjects, GADA in 23 of 465 (4.9%) subjects, and IA-2A in 16 of 410 (3.9%) subjects; 22 of 498 subjects (4.4%) had at least two positive markers (either ICA or IA-2A and GADA or IAA) for type 1 diabetes. Three autoantibodies were measured in 360 subjects, and 5 (1.4%) had three positive markers. Four autoantibodies were measured in 274 subjects, and 3 (1%) had four positive markers. No significant difference in the demographic characteristics between the tested and untested subjects was found.

OGTTs were normal in 70.5% of 748 subjects, impaired in 27% of subjects, and abnormal in 2.5% of subjects. Patients with abnormal OGTT were excluded from the follow-up study.

The FPIR was <1st centile, according to normal values for pubertal stages, in 105 of 409 (25.6%) subjects. No significant difference in the demographic characteristics between the tested and untested subjects was found.

Regarding HLA-DR typing, performed in 210 subjects, HLA-DR3/DR4, DR4/DR4, and DR3/DR3 genotypes were detected in 3.5%, HLA-DR3/other in 18.6%, HLA-DR4/other in 22.1%, and HLA-DR other/other in 48.8%. HLA-DR3/DR3 and HLA-DR4/other alleles were more frequent in hyperglycemic children than in control subjects (P = 0.012 and P = 0.005, respectively), whereas HLA-DR other/other genotypes were less frequent in hyperglycemic children than in control subjects (P = 0.000027).

Regarding DQA1 and DQB1 typing, four susceptible DQα-β heterodimers were present in 4.2% of subjects, two heterodimers in 25.5%, one heterodimer in 24.5%, and 0 heterodimers in 45.7%. Two or more heterodimers were more frequent in hyperglycemic children than in control subjects (P = 0.025).

Subjects were followed for 1 month to 7 years (median 42 months) after enrollment into the study. The 3-year cumulative risk of developing type 1 diabetes was 2.5% (95% CI 1.5–4.0) in the whole population and 26.4% (16.7–40.4) in subjects with positive antibodies. This is comparable with 0.9% (0.5–1.5) and 9.7% (5.1–17.9), respectively, in 1,641 first-degree relatives in our family study (17). During the follow-up period, 16 (2.1%) subjects with incidental hyperglycemia became insulin-dependent. The characteristics of these 16 subjects at enrollment and diabetes diagnosis are reported in Tables 1 and 2. Of the 16 subjects, 11 were males and 5 were females, and the age at first evaluation ranged from 4.1 to 13.9 years. ICAs and/or IAAs, GADAs, and IA-2As were present in all of these subjects, and 13 subjects had two or more autoantibodies. FPIR was <1st centile in 14 subjects. OGTTs were impaired in 13 subjects. Seven of eight subjects with HLA-DR typing had a DR3- or DR4-containing genotype, and 10 of 11 subjects with DQ typing had at least one diabetes-susceptible DQα-β heterodimer. Type 1 diabetes developed at 1 to 28 months (median 10) from initial screening. A total of 11 patients had diabetic symptoms, and 4 had ketonuria. Diagnosis was done either during routine screening or because parents were informed to check urinalysis 1) randomly and 2) in case of symptoms compatible with diabetes.

To assess the association of risk factors with recurrence of diabetes during follow-up, Cox regression was applied, and results are summarized in Table 3. The cumulative risk of type 1 diabetes is similar in males and females and similar in subjects under or over 10 years of age, but it is increased in subjects with ICAs, IAAs, GADAs, and IA-2As (Fig. 1). The cumulative risk of type 1 diabetes is increased in subjects with two antibodies (Fig. 2) and in subjects with an FPIR ≤1st centile (Fig. 3).

At the time of type 1 diabetes diagnosis, >10% (4) of children are in a coma and ∼1 of 200 children die in diabetic ketoacidosis (26). Morbidity and mortality would be significantly reduced by early diagnosis and institution of appropriate therapy before severe metabolic decompensation occurs (4).

First-degree relatives of type 1 diabetic patients carry a higher risk of acquiring diabetes than the general population (27); however, most children with type 1 diabetes do not have a family history of diabetes. Incidental hyperglycemia is a relatively common finding in the pediatric population (28). It has been reported that the risk of progression to type 1 diabetes is low when transient hyperglycemia occurs during a serious intercurrent illness (10,11). In contrast, the risk is high in children with transient hyperglycemia without a serious illness. We prospectively studied only children with incidental hyperglycemia without an intercurrent illness and found that these subjects had a risk of developing type 1 diabetes that was at least the same as that of first-degree relatives. These subjects had an increased prevalence of islet autoantibodies, and an elevated risk of type 1 diabetes was confined to those with this serological marker of prediabetes. Risk was also high in subjects with low FPIR. These findings are consistent with a previous study in children with transient hyperglycemia (10).

Therefore, our data suggest that children and adolescents found to have incidental hyperglycemia (or glycosuria) without intercurrent illness should be considered at risk of developing type 1 diabetes and should be screened for islet autoantibodies and metabolic markers.

In the absence of immunological and immunogenetic markers, incidental finding of hyperglycemia in children and adolescents appears unlikely to be associated with progression to insulin-dependent diabetes. Moreover, we would like to underline that in 30.5% of our subjects with incidental hyperglycemia without immunological, metabolic, or immunogenetic markers for type 1 diabetes and with a family history of type 2 diabetes, a clinical diagnosis of maturity-onset diabetes of the young (MODY) was made and confirmed by genetic analysis (29). This could explain the finding of an FPIR <1st centile in 25.6% of our hyperglycemic subjects, but clinical diabetes developed only in 2.5% of them.

In conclusion, our observations confirm that the finding of transient hyperglycemia without serious intercurrent illness is a risk factor for diabetes. We suggest that all pediatricians should look for evidence of markers for type 1 diabetes and MODY in children and adolescents with incidental hyperglycemia and closely follow subjects with these markers.

Co-authors are the members of the Pediatric Italian Study Group of Prediabetes: V. Cherubini, Ancona; L. Cavallo, F. Dammacco, Bari; G. Cerasoli, M. Pocecco, Cesena; S. Tumini, F. Chiarelli, Chieti; M. Mancuso, Catania; P. Banin, Ferrara; S. Toni, M. Martinucci, Firenze; C. Barella, M. Cotellessa, Genova; A. Monaci, Grosseto; F. Lombardo, Messina; F. Meschi, G. Chiumello, Milan; L. Iughetti, S. Bernasconi, Modena; D. Iafusco, F. Prisco, A. Franzese, Napoli; F. Cadario, Novara; F. Cardella, Palermo; C. Monciotti, Padova; R. Giacchero, M.A. Avanzini, Pavia; L. Crespin, L. Calisti, Pisa; G. Chiari, M. Vanelli, Parma; L. Guazzarotti, Recanati; N. Sulli, G. Multari, A. Crinò, G. Marietti, S. Piccinini, M.L. Manca Bitti, Rome; P. Sacchini, A. Maciani, Rimini; C. Sacchetti, F. Cerutti, Torino; G. Tonini, Trieste; A. Visentin, Treviso; G. Contreas, L. Pinelli, Verona, Italy.

This study has been supported by feasibility grants from IRCCS G. Gaslini Institute, Genoa; IRCCS Policlinico, San Matteo, Pavia; and IRCCS Burlo Garofalo Institute, Trieste, Italy.

We thank Elena Bazzigaluppi from IRCCS San Raffaele, Milan, for technical assistance in autoantibody measurements.

We also thank Anna Capurro for her help in revising the manuscript.