Prediction of type 1 diabetes is based on the detection of multiple islet autoantibodies in subjects who are at increased genetic risk. Prediction of the timing of diagnosis is challenging, however. We assessed the utility of HbA1c levels in predicting the clinical disease in genetically predisposed children with multiple autoantibodies. Cord blood samples from 168,055 newborn infants were screened for class II HLA genotypes in Finland, and 14,876 children with increased genetic risk for type 1 diabetes were invited to participate in regular follow-ups, including screening for diabetes-associated autoantibodies. When two or more autoantibodies were detected, HbA1c levels were analyzed at each visit. During follow-up, multiple (two or more) autoantibodies developed in 466 children; type 1 diabetes was diagnosed in 201 of these children (43%, progressors), while 265 children remained disease free (nonprogressors) by December 2011. A 10% increase in HbA1c levels in samples obtained 3–12 months apart predicted the diagnosis of clinical disease (hazard ratio [HR] 5.7 [95% CI 4.1–7.9]) after a median time of 1.1 years (interquartile range [IQR] 0.6–3.1 years) from the observed rise of HbA1c. If the HbA1c level was ≥5.9% (41 mmol/mol) in two consecutive samples, the median time to diagnosis was 0.9 years (IQR 0.3–1.5, HR 11.9 [95% CI 8.8–16.0]). In conclusion, HbA1c is a useful biochemical marker when predicting the time to diagnosis of type 1 diabetes in children with multiple autoantibodies.

The development of type 1 diabetes is characterized by immune-mediated destruction of the pancreatic β-cells, which eventually stop producing insulin. The incidence of type 1 diabetes has been growing worldwide, with the highest incidence rate observed in Finland, where the rate has more than doubled from 1980 to 2005 in children under the age of 15 years (13).

More than 50 genes conferring susceptibility to human type 1 diabetes have been identified (4). The major risk genes are located in the class II HLA region on chromosome 6p21 (5). Environmental risk factors are believed to interact with susceptibility genes and thereby contribute to the disease process (6,7). Currently, type 1 diabetes is predicted by analyzing islet autoantibodies of various specificities in an individual with increased genetic disease susceptibility. The presence of multiple (more than two) autoantibodies provides a cumulative disease risk of 50–60% over the next 5 years in children with risk-conferring HLA class II genotypes (8), and during the 15-year follow-up the risk of the development of diabetes is very high (84%) in such children (9).

Although oral glucose tolerance tests (OGTTs) have been shown to provide relatively good accuracy in the prediction of type 1 diabetes (10,11), more practical markers of glucose metabolism are needed. Glycated hemoglobin A1c (HbA1c) levels could be superior to OGTT results as a predictive measure, since HbA1c is widely used in clinical practice as an indicator of metabolic control in patients with diabetes, and is generally not affected by short-term variations in food intake and physical activity (12,13). The World Health Organization and American Diabetes Association have adapted the recommendation of the International Expert Committee that an HbA1c value ≥6.5% (48 mmol/mol) is diagnostic for type 2 diabetes (14), but no threshold value for the diagnosis of type 1 diabetes has been determined (15). As a laboratory test, HbA1c is cost effective and easy to obtain, and it could be useful also in the prediction of type 1 diabetes (16). Although positivity for multiple islet autoantibodies has been shown to be relatively sensitive in the prediction of type 1 diabetes, there is a lack of a reliable method to predict the onset of type 1 diabetes. In this study, the goal was to monitor HbA1c levels from the start of positivity for multiple (more than two) autoantibodies and to assess its potential as a predictive marker for type 1 diabetes.

Study Design

The Type 1 Diabetes Prediction and Prevention (DIPP) study is a Finnish population-based study in which children with HLA-conferred susceptibility to type 1 diabetes are observed prospectively from birth (17,18). Recruitment of newborn infants for the DIPP study started in November 1994 and still continues in three university hospitals (Oulu, Tampere, and Turku) in Finland. Screening for genetic risk is performed from cord blood. Families with an infant carrying HLA genotypes associated with type 1 diabetes are invited to participate in prospective follow-up at 3- to 12-month intervals until the age of 15 years or until clinical disease is diagnosed. Islet autoantibodies are analyzed at each visit; monitoring of HbA1c is initiated in children who test positive for at least two different autoantibodies in the same sample, and positivity is confirmed in the subsequent sample. The diagnosis of type 1 diabetes is based on typical symptoms and random high plasma glucose levels, or in asymptomatic patients' diagnostic plasma glucose values in two separate OGTTs, as the World Health Organization recommends (19). The study population for the current analysis comprises 466 DIPP children who tested positive for multiple autoantibodies in a minimum of two consecutive samples and had at least one HbA1c sample taken between November 1994 and December 2011. The study has been approved by the ethics committees of the participating university hospital districts. All families participating in the study have provided written informed consent.

Immunological Screening

DIPP participants who seroconverted to positivity for any of the diabetes-associated autoantibodies (islet cell antibodies, insulin autoantibodies, GAD antibodies, and autoantibodies specific for the insulinoma-2–associated antigen) were scheduled for follow-up visits at 3-month intervals. The age at seroconversion was defined as the age at which at least one of the diabetes-associated autoantibodies was detected for the first time. The age at multiple autoantibody positivity was defined as the time point when at least two autoantibodies were detected in the same sample.

Autoantibodies were analyzed as described previously (8). In the Diabetes Autoantibody Standardization (DASP) workshop in 2005, the following sensitivities and specificities were reported: insulin autoantibodies 58% and 96%; GAD antibodies 82% and 96%; and insulinoma-2–associated antigen antibodies 72% and 100%.

Genetic Screening

HLA-conferred susceptibility to type 1 diabetes was analyzed centrally using cord blood, as described previously (20). According to various HLA DRB1–DQA1–DQB1 haplotype combinations, genotypes conferring high, moderate, and low risk for the disease were determined. High risk was defined as heterozygosity for the two risk-associated haplotypes DRB1*04:01/2/4/5/8-DQA1*03-DQB1*03:02/4 and [DRB1*03]-DQA1*05-DQB1*02. Moderate risk was defined as homozygosity for any of the two risk haplotypes or DRB1*04:01/2/4/5/8-DQA1*03-DQB1*03:02/4 combined with a neutral haplotype, or the [DRB1*03]-DQA1*05-DQB1*02/[DRB1*09]-DQA1*03-DQB1*03 genotype. Low risk was conferred by other genotypes as previously defined (21).

HbA1c Assays and Measurements

All HbA1c values were measured from venous blood samples in the clinical laboratories of the three university hospitals. Methods for HbA1c analyses varied slightly among the centers.

In the Oulu University Hospital, an immunoassay-based method was applied throughout the study. Until 6 May 2004, Cobas Integra 700 (Hoffman-La Roche, Basel, Switzerland) was used with the reagent HbA1c 2054302 (Hoffmann-La Roche). From 7 May 2004 until 20 April 2009, Advia 2400 (Siemens, Munich, Germany) was used with reagent B01–4797–01 (Bayer Healthcare, Leverkusen, Germany). From 20 April 2009 onward, Advia 1800 (Siemens) has been used with the reagent HbA1c 06854744 (Siemens).

In the Tampere University Hospital, fast-protein liquid chromatography was used to measure HbA1c levels until 9 June 1999 (Mono S for HbA1c reagent 17–1040–01; Pharmacia, Uppsala, Sweden). From 10 June 1999 until 14 November 2007, an immunoassay-based method was used (Cobas Integra with the reagent Hemoglobin A1C; Hoffmann-La Roche). From 15 November 2007 onward, Cobas Integra has been used with the reagent Tina-quant Hemoglobin A1C Gen 2.

In the Turku University Hospital, the fast-protein liquid chromatography method was used to measure HbA1c levels until 5 August 1996 (Pharmacia). Thereafter, high-performance liquid chromatography has been used. From 6 August 1996 until 13 July 2001, the Variant Classic instrumentation (Bio-Rad, Hercules, CA) was used with the reagents in VARIANT HbA1c Reorder Pack #270–0003 (Bio-Rad). From 14 July 2001 until 20 November 2007, the VARIANT II instrumentation (Bio-Rad) was used with the reagents in VARIANT II HbA1c Reorder Pack #270–2101 (Bio-Rad). From 21 November 2007 until 13 October 2008, the VARIANT II instrumentation (Bio-Rad) was still used but with the reagents in VARIANT II HbA1c Reorder Pack #270–2101 NU (Bio-Rad). From 14 October 2008 until 2 March 2010, the VARIANT II Turbo instrumentation (Bio-Rad) was used with the reagents in VARIANT II Turbo HbA1c Reorder Pack #270–2415 (Bio-Rad), which were changed to the reagents in VARIANT II Turbo HbA1c Kit 2.0 #270–2455EX (Bio-Rad) starting from 3 March 2010.

The mean HbA1c level measured in the Oulu University Hospital was 5.57% (SD 0.78); in Tampere University Hospital, it was 5.56% (SD 0.46); and in Turku University Hospital, it was 5.57% (SD 0.51). The covariate adjustment was used in the linear mixed model (LMM) analysis to take into account the possible confounding effect of variable assay levels in the three hospitals.

Statistical Analyses

We established a priori three time-dependent decision rules for HbA1c. First, children who had undergone at least two HbA1c measurements within 3–12 months were classified into two groups according to whether they experienced an increase of at least 10% in their HbA1c values or not (decision rule 1). Second, children who had undergone at least two HbA1c measurements within 3–12 months and a third HbA1c measurement during the next 6 months were classified into two groups according to whether they had a 10% increase in HbA1c values within 3–12 months and any additional increase during the next 6 months or not (decision rule 2). Third, children were divided into two groups according to whether they had two consecutive HbA1c values ≥5.9% (41 mmol/mol) or not (decision rule 3). The use of an increase of 10% as a decision rule was based on the data from the three laboratories involved. The interassay coefficient of variation was 3.3% in Oulu University Hospital, 2.0% in Tampere University Hospital, and 2.7% in Turku University Hospital, and the reported uncertainty of measurement values were 10%, 5.8%, and 5.9%, respectively. The calculation of the uncertainty was based on current recommendations (22). The highest reported percentage of uncertainty was selected for the analysis. The cutoff HbA1c value of ≥5.9% (41 mmol/mol) has been suggested in previous studies of the prediction of diabetes (2325). Two sequential measurements were required to minimize the biological and analytical variations. A time window of 3–12 months was used in the analyses according to the DIPP protocol in which follow-up visits occur every 3 months.

Cox regression with these time-dependent covariates was used to evaluate the association between HbA1c and the risk of diabetes. The entry time to the analysis was the date when at least two islet autoantibodies were positive. All subjects were considered to be unexposed until the time-dependent covariate fulfilled the decision rule, and thereafter the status of the subject was exposed. Consequently, if the subject had only one (decision rules 1 and 3) or two (decision rule 2) HbA1c measurements, he/she belonged to the unexposed group. Samples taken on the day of diagnosis were excluded from the analyses considering the predictive use of HbA1c (decision rules), but not from the LMM analysis. The total follow-up time was partitioned into intervals according to the cut points of the decision rule. Univariate Cox regression analysis was performed to identify the risk factors of diabetes, and survival curve estimates were calculated to assess the timing and probability of diabetes after fulfilling a decision rule. Sensitivity and specificity for the three decision rules were also calculated. Multivariate Cox regression with the backward stepwise model was used to identify a set of predictors that are most effective in disease prediction. In the multivariate analysis, the explanatory variable candidates were a 10% increase in HbA1c values within 3–12 months, a 10% increase in HbA1c values within 3–12 months and any additional increase during the next 6 months, two consecutive HbA1c values ≥5.9% (41 mmol/mol), age at the detection of multiple autoantibodies, type 1 diabetes in a first-degree relative (FDR), HLA risk, and sex.

LMM analysis with a random intercept and first-order autoregressive covariance structure for repeated measurements was used to analyze HbA1c levels over time between the progressors (children with multiple [two or more] autoantibodies who progressed to overt type 1 diabetes) and nonprogressors (children with multiple autoantibodies in whom clinical disease did not develop during the follow-up). The random intercepts and repeated measurements were nested within subjects, and subjects were nested within the hospital. The group-by-time interaction was included in the model to test differences between group means at each time point. Sex, age at sampling, age at seroconversion, age at multiple autoantibody positivity, type 1 diabetes in an FDR, and HLA risk were included in the LMM analysis as fixed variables. Differences in proportions were tested using the standardized normal deviate test.

All analyses were performed using IBM SPSS version 20.0.0 for Windows, Stata/IC version 11.2 for Windows, and StatsDirect statistical software version 2.7.9. The figures were drawn using OriginPro version 8.6.0 and Stata/IC version 11.2.

Between November 1994 and December 2011, a total of 168,055 newborn infants were screened for HLA-conferred susceptibility to type 1 diabetes. Altogether, 14,876 children with increased genetic risk were enrolled for regular follow-up into the DIPP study. During follow-up, multiple positive autoantibodies developed in 466 children, and 201 of these children (43%) progressed to type 1 diabetes (progressors), whereas 265 children remained disease free until the end of December 2011 (nonprogressors). In the progressor group, at least two positive autoantibodies were detected in the first positive sample of 136 children (68%), whereas in the nonprogressor group, at least two positive antibodies were detected in the first positive sample of 110 children (42%) (P < 0.001). The baseline characteristics of the children are presented in Table 1.

Altogether, 4,270 HbA1c samples were analyzed during the prediabetic period, 1,613 from the progressors and 2,657 from the nonprogressors. An average of 8.0 measurements per child (95% CI 7.0–9.0, range 1–37) were obtained in the group of progressors and 10.0 (95% CI 8.9–11.1, range 1–44) in the group of nonprogressors.

When retrospectively comparing HbA1c levels between the progressors and nonprogressors, we observed that HbA1c levels started to be consistently higher in the progressors 2.0 years before the diagnosis (Fig. 1). More specifically, during the period 1.8–2.0 years before diagnosis, the adjusted mean HbA1c level was 5.5% (37 mmol/mol [95% CI 5.4–5.7 (36–39)]) among the progressors compared with 5.4% (36 mmol/mol [95% CI 5.2–5.4 (33–36)]) in the nonprogressors (P = 0.025). During the following year, a small but significant difference was consistently observed between the two groups. Thereafter, the adjusted mean HbA1c level in the progressors started to increase more steeply, as follows: 5.9% (41 mmol/mol) during the period 0.4–0.6 years before diagnosis, 6.1% (43 mmol/mol) during the period 0.2–0.4 years before diagnosis, and 6.8% (51 mmol/mol) during the period 0.01–0.2 years before diagnosis. At diagnosis, the adjusted mean HbA1c level in the progressors was 7.6% (60 mmol/mol [95% CI 7.5–7.7 (58–61)]), compared with 5.5% (37 mmol/mol [95% CI 5.4–5.6 (36–38)]) among the nonprogressors in their last measurement (P < 0.001). The range of crude HbA1c measurements at diagnosis was 4.9–12.9% (30–117 mmol/mol) in the progressors and 4.4–6.5% (25–48 mmol/mol) in the nonprogressors at the end of follow-up.

A 10% increase in the HbA1c values taken 3–12 months apart increased the disease risk almost sixfold (hazard ratio [HR] 5.7 [95% CI 4.1–7.9]) (Table 2), and clinical disease developed in half of these children within 1.1 years (interquartile range [IQR] 0.6–3.1 years) (Fig. 2B) according to the univariate Cox regression analysis. When including any additional rise in HbA1c levels in a third measurement during the subsequent 6 months, the HR was 5.1 (95% CI 3.6–7.2), and overt disease developed in 50% of the children during the next 0.7 years (IQR 0.7–1.5 years) (Table 2 and Fig. 2C). If a child was found to have two consecutive HbA1c values ≥5.9% (41 mmol/mol), the risk of diabetes was almost 12-fold compared with the remaining children (HR 11.9 [95% CI 8.8–16.0]), and the median time to type 1 diabetes was 0.9 years (IQR 0.3–1.5) (Table 2 and Fig. 2D). The most effective set of predictive variables in multivariate Cox regression analysis is shown in Table 3.

Sensitivities and specificities were calculated for the suggested decision rules. A 10% increase in HbA1c levels over 3–12 months provided a sensitivity of 57% (95% CI 50–64) and a specificity of 66% (95% CI 60–72). When including an additional rise during the next 6 months, the sensitivity of the test was 22% (95% CI 17–29) and the specificity was 91% (95% CI 87–94). Two consecutive values of HbA1c ≥5.9% (41 mmol/mol) presented a sensitivity of 42% (95% CI 35–49) and a specificity of 86% (95% CI 82–90).

When multiple positive autoantibodies were detected, an increase of 10% in HbA1c values appeared after a mean time of 2.5 years (SD 2.0 years). For an additional rise, the time was 2.8 years (SD 2.1 years); and with two HbA1c values ≥5.9% (41 mmol/mol), the mean time was 3.4 years (SD 2.4 years).

An HbA1c level of ≥6.5% (48 mmol/mol) was detected in 61% of the progressors (122 of 201 progressors) and in only 2% of the nonprogressors (5 of 265 nonprogressors). However, 83% of the HbA1c values ≥6.5% (48 mmol/mol) (101 of 122 values) in the group of progressors were observed at the time of diagnosis.

The prediction of type 1 diabetes has so far been based mainly on the presence of islet autoantibodies in subjects who are at increased genetic risk (8,26,27). Young age at seroconversion also has a clear impact on the risk of progression to clinical disease (17). In addition to autoantibodies, consecutive OGTT results have been of interest in the prediction of type 1 diabetes, giving at best almost 90% accuracy during a 2-year follow-up among the relatives of patients affected by type 1 diabetes (10,11,2831). When evaluating the potential of HbA1c level as an additional marker in the prediction of type 1 diabetes, we looked for new and practical ways to predict the timing of the diagnosis. The following three new criteria were established: a 10% rise in HbA1c values taken 3–12 months apart, an additional rise during the subsequent 6 months, and two consecutive values of ≥5.9% (41 mmol/mol) that could be used to predict the clinical disease in a child with multiple autoantibodies. These results are important since families with a child positive for multiple islet autoantibodies are concerned about the time remaining until the development of clinical disease. They are well aware of the high disease risk and need expert counseling. Stable HbA1c values in consecutive measurements suggest that the child is not going to present with overt diabetes in the near future. In contrast, an increase in HbA1c level gives a warning of incipient disease, which may help in early diagnosis and thereby reduce the risk of acute complications, such as diabetic ketoacidosis (32).

Age at multiple (two or more) autoantibody positivity, the time from seroconversion to multiple autoantibodies, the presence of affected FDRs, and a high-risk HLA class II genotype have previously been identified as risk factors for type 1 diabetes (8,16,17,33). In our analyses, age at multiple autoantibody positivity was a significant predictive factor in both univariate and multivariate analyses. Children with an affected FDR had more than a twofold disease risk, demonstrating that genetic factors also have an important effect on the progression from β-cell autoimmunity to clinical disease. The HLA class II genotype appeared to play a role in disease progression, although with a lower HR than the HbA1c variables.

Our prediabetic cohort of 201 Finnish children with multiple autoantibodies represents the largest series of young children who have participated in a long-term intensive follow-up from birth to diagnosis of type 1 diabetes. Only one study (16) on the evolution of HbA1c levels during prediabetes has been published earlier. Although that study included a relatively small number of 28 islet autoantibody–positive children who progressed to clinical diabetes, the results were similar to our study's, showing that HbA1c levels start to increase within the normal reference range and that the rise is steepest during the last 6 months before diagnosis. The strength of our study is the fact that with a considerably larger sample size and higher number of HbA1c measurements per child we were able to get more accurate and reliable estimates when analyzing the predictive characteristics of HbA1c. Our study population came from three clinical centers that used different methods to analyze HbA1c, which might have produced slightly different absolute values. This was taken into account using covariate adjustment in the analyses. Furthermore, we used a relative increase in HbA1c values rather than absolute values in our predictive model to make the results of our study more generalizable. Currently, there are no exact threshold values for a significant change in HbA1c levels. Our decision rules were based on the available literature and variances reported by the university hospital laboratories involved, and they are described in more detail in the 2research design and methods section.

The suggested decision rules of a 10% rise in HbA1c level during 3–12 months and two consecutive HbA1c values ≥5.9% (41 mmol/mol) both provide high HRs with relatively short median time to diagnosis of type 1 diabetes (1.1 and 0.9 years, respectively). However, a 10% rise provides superior sensitivity (57% vs. 42%), whereas two consecutive results of ≥5.9% (41 mmol/mol) has a higher specificity (86% vs. 66%). Since children with multiple (two or more) autoantibodies have a high risk of the development of type 1 diabetes over a variable period of time, it is probably more practical from the clinical point of view to use two consecutive HbA1c results ≥5.9% in the prediction of the timing of diagnosis. It is still noteworthy that a relative increase of 10% in HbA1c level is more independent from methodological differences and reference values, therefore remaining important alongside the absolute values.

Our retrospective analysis showed that HbA1c values start to increase ~2 years before the diagnosis, reflecting the gradual deterioration in endogenous insulin secretion and increasing fluctuation in plasma glucose levels. There were some differences in the demographic characteristics between the nonprogressors and the progressors, however. The nonprogressors were older both at initial seroconversion and when multiple autoantibodies developed. The mean age of the nonprogressors was 10.9 years at the end of follow-up, whereas the age of the progressors at diagnosis was 6.3 years. The nonprogressors had affected family members less often. On the other hand, our data show that the mean HbA1c level remained very stable in the children in the nonprogressor group during the follow-up, varying from 5.3% (34 mmol/mol) to 5.6% (38 mmol/mol), and therefore the differences in age and follow-up time are probably not interfering with this analysis.

Our results provide important data on the natural evolution of HbA1c levels during the prediabetic period in young children, which can be used in future prevention trials. In secondary prevention trials aimed at slowing down or reversing the progression of β-cell dysfunction in subjects with islet autoantibodies, there is a clear need for markers that can be used to monitor the disease process. HbA1c level is an indirect measure of glucose control and cannot be used as a direct estimate of β-cell function. It is noteworthy that the time interval between the HbA1c measurements used to calculate the presence or absence of a 10% increase varied between 3 and 12 months, and may represent a potential confounder, given that a 10% increase occurring during a 3-month interval might have a different weight in terms of prediction compared with a similar increase in samples derived from a 12-month interval. As HbA1c level represents a variation in plasma glucose levels during the lifetime of erythrocytes (120 days), with the weight on the preceding 6–8 weeks, it seems unnecessary to measure HbA1c levels more often than every 3 months. Since the three laboratories involved reported rather high uncertainty of measurements varying from 5.8% to 10% and the methodology changed over the study period, one has to consider the possibility that some changes in HbA1c levels are due to variations in methodology. However, this possible source of error affects the progressor and nonprogressor groups equally. The HbA1c values suggested on the basis of the current analyses for the prediction of clinical type 1 diabetes are from a single study, and validation of the results is needed in additional populations.

In conclusion, we have shown that HbA1c level is a useful biochemical marker for the estimation of the time to the diagnosis of type 1 diabetes in genetically susceptible children with multiple islet autoantibodies.

O.H. and S.A. contributed equally to this work.

Acknowledgments. The authors thank the dedicated personnel of the Type 1 Diabetes Prediction and Prevention study in Oulu, Turku, and Tampere, Finland, and the study children and their families for their essential contributions.

Funding. This work was supported by the Juvenile Diabetes Research Foundation Internationalhttp://dx.doi.org/10.13039/100000901 (grants 4-1998-274, 4-1999-731, and 4-2001-435); the European Union (grant BMH4-CT98-3314); the Novo Nordisk Foundation; the Academy of Finland (Centre of Excellence in Molecular Systems Immunology and Physiology Research 2012–2017, Decision No. 250114); the TEKES National Technology Agency of Finland; Special Research Funds for University Hospitals in Finland; the Finnish Office for Health Technology Assessment; the Diabetes Research Foundation, Finland; the Sigrid Juselius Foundation; the Emil Aaltonen Foundation; the Jalmari and Rauha Ahokas Foundation; the Signe and Ane Gyllenberg Foundation; the Research Foundation of Orion Corporation; the Foundation for Pediatric Research; and the Päivikki and Sakari Sohlberg Foundation.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

The sponsors had no role in creating this report at any stage.

Author Contributions. O.H. helped to acquire, analyze, and interpret the data; to draft the manuscript; to critically revise the manuscript for important intellectual content; and to provide administrative, technical, or material support. S.A. helped to acquire, analyze, and interpret the data; to draft the manuscript; and to critically revise the manuscript for important intellectual content. T.P. helped to analyze and interpret the data; to draft the manuscript; to critically revise the manuscript for important intellectual content; and to provide administrative, technical, or material support; and performed the statistical analysis. M.-R.H. helped to acquire the data and to critically revise the manuscript for important intellectual content. N.H. helped to acquire the data, to draft the manuscript, and to critically revise the manuscript for important intellectual content. J.L. helped to acquire, analyze, and interpret the data; to critically revise the manuscript for important intellectual content; and to provide administrative, technical, or material support. J.I. helped to conceive and design the study; to acquire the data; to critically revise the manuscript for important intellectual content; and to provide administrative, technical, or material support. O.S. helped to conceive and design the study and to acquire the data. M.K. helped to conceive and design the study; to acquire, analyze, and interpret the data; to critically revise the manuscript for important intellectual content; to provide administrative, technical, or material support; and to supervise the study. R.V. helped to conceive and the design the study; to acquire, analyze, and interpret the data; to draft and critically revise the manuscript for important intellectual content; to provide administrative, technical, or material support; and to supervise the study. O.H. and R.V. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

1.
DIAMOND Project Group
.
Incidence and trends of childhood type 1 diabetes worldwide 1990-1999
.
Diabet Med
2006
;
23
:
857
866
[PubMed]
2.
Patterson
CC
,
Gyürüs
E
,
Rosenbauer
J
, et al
.
Trends in childhood type 1 diabetes incidence in Europe during 1989-2008: evidence of non-uniformity over time in rates of increase
.
Diabetologia
2012
;
55
:
2142
2147
[PubMed]
3.
Harjutsalo
V
,
Sjöberg
L
,
Tuomilehto
J
.
Time trends in the incidence of type 1 diabetes in Finnish children: a cohort study
.
Lancet
2008
;
371
:
1777
1782
[PubMed]
4.
Cooper
JD
,
Howson
JM
,
Smyth
D
, et al.;
Type 1 Diabetes Genetics Consortium
.
Confirmation of novel type 1 diabetes risk loci in families
.
Diabetologia
2012
;
55
:
996
1000
[PubMed]
5.
Ilonen
J
,
Sjöroos
M
,
Knip
M
, et al
.
Estimation of genetic risk for type 1 diabetes
.
Am J Med Genet
2002
;
115
:
30
36
[PubMed]
6.
Knip
M
,
Veijola
R
,
Virtanen
SM
,
Hyöty
H
,
Vaarala
O
,
Akerblom
HK
.
Environmental triggers and determinants of type 1 diabetes
.
Diabetes
2005
;
54
(
Suppl. 2
):
S125
S136
[PubMed]
7.
Gillespie
KM
,
Bain
SC
,
Barnett
AH
, et al
.
The rising incidence of childhood type 1 diabetes and reduced contribution of high-risk HLA haplotypes
.
Lancet
2004
;
364
:
1699
1700
[PubMed]
8.
Siljander
HT
,
Simell
S
,
Hekkala
A
, et al
.
Predictive characteristics of diabetes-associated autoantibodies among children with HLA-conferred disease susceptibility in the general population
.
Diabetes
2009
;
58
:
2835
2842
[PubMed]
9.
Ziegler
AG
,
Rewers
M
,
Simell
O
, et al
.
Seroconversion to multiple islet autoantibodies and risk of progression to diabetes in children
.
JAMA
2013
;
309
:
2473
2479
[PubMed]
10.
Sosenko
JM
,
Palmer
JP
,
Rafkin-Mervis
L
, et al.;
Diabetes Prevention Trial-Type 1 Study Group
.
Incident dysglycemia and progression to type 1 diabetes among participants in the Diabetes Prevention Trial-Type 1
.
Diabetes Care
2009
;
32
:
1603
1607
[PubMed]
11.
Sosenko
JM
,
Skyler
JS
,
Mahon
J
, et al.;
Type 1 Diabetes TrialNet Study Group
;
Diabetes Prevention Trial-Type 1 Study Group
.
The application of the diabetes prevention trial-type 1 risk score for identifying a preclinical state of type 1 diabetes
.
Diabetes Care
2012
;
35
:
1552
1555
[PubMed]
12.
Jeffcoate
SL
.
Diabetes control and complications: the role of glycated haemoglobin, 25 years on
.
Diabet Med
2004
;
21
:
657
665
[PubMed]
13.
Rohlfing
CL
,
Wiedmeyer
HM
,
Little
RR
,
England
JD
,
Tennill
A
,
Goldstein
DE
.
Defining the relationship between plasma glucose and HbA(1c): analysis of glucose profiles and HbA(1c) in the Diabetes Control and Complications Trial
.
Diabetes Care
2002
;
25
:
275
278
[PubMed]
14.
International Expert Committee
.
International Expert Committee report on the role of the A1C assay in the diagnosis of diabetes
.
Diabetes Care
2009
;
32
:
1327
1334
[PubMed]
15.
Vehik
K
,
Cuthbertson
D
,
Boulware
D
, et al.;
TEDDY, TRIGR, Diabetes Prevention Trial–Type 1, and Type 1 Diabetes TrialNet Natural History Study Groups
.
Performance of HbA1c as an early diagnostic indicator of type 1 diabetes in children and youth
.
Diabetes Care
2012
;
35
:
1821
1825
[PubMed]
16.
Stene
LC
,
Barriga
K
,
Hoffman
M
, et al
.
Normal but increasing hemoglobin A1c levels predict progression from islet autoimmunity to overt type 1 diabetes: Diabetes Autoimmunity Study in the Young (DAISY)
.
Pediatr Diabetes
2006
;
7
:
247
253
[PubMed]
17.
Parikka
V
,
Näntö-Salonen
K
,
Saarinen
M
, et al
.
Early seroconversion and rapidly increasing autoantibody concentrations predict prepubertal manifestation of type 1 diabetes in children at genetic risk
.
Diabetologia
2012
;
55
:
1926
1936
[PubMed]
18.
Näntö-Salonen
K
,
Kupila
A
,
Simell
S
, et al
.
Nasal insulin to prevent type 1 diabetes in children with HLA genotypes and autoantibodies conferring increased risk of disease: a double-blind, randomised controlled trial
.
Lancet
2008
;
372
:
1746
1755
[PubMed]
19.
Alberti
KG
,
Zimmet
PZ
.
Definition, diagnosis and classification of diabetes mellitus and its complications. Part 1: diagnosis and classification of diabetes mellitus provisional report of a WHO consultation
.
Diabet Med
1998
;
15
:
539
553
[PubMed]
20.
Hermann
R
,
Knip
M
,
Veijola
R
, et al.;
FinnDiane Study Group
.
Temporal changes in the frequencies of HLA genotypes in patients with type 1 diabetes—indication of an increased environmental pressure
?
Diabetologia
2003
;
46
:
420
425
[PubMed]
21.
Hekkala
A
,
Ilonen
J
,
Knip
M
,
Veijola
R
.;
Finnish Paediatric Diabetes Register
.
Family history of diabetes and distribution of class II HLA genotypes in children with newly diagnosed type 1 diabetes: effect on diabetic ketoacidosis
.
Eur J Endocrinol
2011
;
165
:
813
817
[PubMed]
22.
Siekmann
L
.
Requirements for reference (calibration) laboratories in laboratory medicine
.
Clin Biochem Rev
2007
;
28
:
149
154
[PubMed]
23.
Braga
F
,
Dolci
A
,
Montagnana
M
, et al
.
Revaluation of biological variation of glycated hemoglobin (HbA(1c)) using an accurately designed protocol and an assay traceable to the IFCC reference system
.
Clin Chim Acta
2011
;
412
:
1412
1416
[PubMed]
24.
Choi
SH
,
Kim
TH
,
Lim
S
,
Park
KS
,
Jang
HC
,
Cho
NH
.
Hemoglobin A1c as a diagnostic tool for diabetes screening and new-onset diabetes prediction: a 6-year community-based prospective study
.
Diabetes Care
2011
;
34
:
944
949
[PubMed]
25.
Braga
F
,
Dolci
A
,
Mosca
A
,
Panteghini
M
.
Biological variability of glycated hemoglobin
.
Clin Chim Acta
2010
;
411
:
1606
1610
[PubMed]
26.
Aly
TA
,
Ide
A
,
Jahromi
MM
, et al
.
Extreme genetic risk for type 1A diabetes
.
Proc Natl Acad Sci U S A
2006
;
103
:
14074
14079
[PubMed]
27.
Siljander
HT
,
Veijola
R
,
Reunanen
A
,
Virtanen
SM
,
Akerblom
HK
,
Knip
M
.
Prediction of type 1 diabetes among siblings of affected children and in the general population
.
Diabetologia
2007
;
50
:
2272
2275
[PubMed]
28.
Sosenko
JM
,
Krischer
JP
,
Palmer
JP
, et al.;
Diabetes Prevention Trial-Type 1 Study Group
.
A risk score for type 1 diabetes derived from autoantibody-positive participants in the diabetes prevention trial-type 1
.
Diabetes Care
2008
;
31
:
528
533
[PubMed]
29.
Sosenko
JM
,
Skyler
JS
,
Krischer
JP
, et al.;
Diabetes Prevention Trial-Type 1 Study Group
.
Glucose excursions between states of glycemia with progression to type 1 diabetes in the diabetes prevention trial-type 1 (DPT-1)
.
Diabetes
2010
;
59
:
2386
2389
[PubMed]
30.
Sosenko
JM
,
Skyler
JS
,
Herold
KC
,
Palmer
JP
.;
Type 1 Diabetes TrialNet and Diabetes Prevention Trial–Type 1 Study Groups
.
The metabolic progression to type 1 diabetes as indicated by serial oral glucose tolerance testing in the Diabetes Prevention Trial-type 1
.
Diabetes
2012
;
61
:
1331
1337
[PubMed]
31.
Sosenko
JM
,
Skyler
JS
,
Palmer
JP
, et al.;
Type 1 Diabetes TrialNet Study Group
;
Diabetes Prevention Trial-Type 1 Study Group
.
The prediction of type 1 diabetes by multiple autoantibody levels and their incorporation into an autoantibody risk score in relatives of type 1 diabetic patients
.
Diabetes Care
2013
;
36
:
2615
2620
[PubMed]
32.
Barker
JM
,
Goehrig
SH
,
Barriga
K
, et al.;
DAISY study
.
Clinical characteristics of children diagnosed with type 1 diabetes through intensive screening and follow-up
.
Diabetes Care
2004
;
27
:
1399
1404
[PubMed]
33.
Ziegler
AG
,
Hummel
M
,
Schenker
M
,
Bonifacio
E
.
Autoantibody appearance and risk for development of childhood diabetes in offspring of parents with type 1 diabetes: the 2-year analysis of the German BABYDIAB Study
.
Diabetes
1999
;
48
:
460
468
[PubMed]