Beta cell function in participants with single or multiple islet autoantibodies at baseline in the TEDDY Family Prevention Study: TEFA

Abstract Aim The aim of the present study was to assess beta cell function based on an oral glucose tolerance test (OGTT) in participants with single islet autoantibody or an intravenous glucose tolerance test (IvGTT) in participants with multiple islet autoantibodies. Materials and methods Healthy participants in Sweden and Finland, between 2 and 49.99 years of age previously identified as positive for a single (n = 30) autoantibody to either insulin, glutamic acid decarboxylase, islet antigen‐2, zinc transporter 8 or islet cell antibodies or multiple autoantibodies (n = 46), were included. Participants positive for a single autoantibody underwent a 6‐point OGTT while participants positive for multiple autoantibodies underwent an IvGTT. Glucose, insulin and C‐peptide were measured from OGTT and IvGTT samples. Results All participants positive for a single autoantibody had a normal glucose tolerance test with 120 minutes glucose below 7.70 mmol/L and HbA1c values within the normal range (<42 mmol/mol). Insulin responses to the glucose challenge on OGTT ranged between 13.0 and 143 mIU/L after 120 minutes with C‐peptide values between 0.74 and 4.60 nmol/L. In Swedish participants, the first‐phase insulin response (FPIR) on IvGTT was lower in those positive for three or more autoantibodies (n = 13; median 83.0 mIU/L; range 20.0‐343) compared to those with two autoantibodies (n = 15; median 146 mIU/L; range 19.0‐545; P = .0330). Conclusion Participants positive for a single autoantibody appeared to have a normal beta cell function. Participants positive for three or more autoantibodies had a lower FPIR as compared to participants with two autoantibodies, supporting the view that their beta cell function had deteriorated.


| INTRODUC TI ON
adults. 3 Metabolic derangement can also be assessed by measurement of HbA1c, which reflects the glucose levels over the last two to three months. HbA1c can be a useful biomarker, if it is regularly measured, to monitor the progression if the appearance of the first autoantibody to the clinical onset of type 1 diabetes is lengthy. 4 The finding of a single autoantibody when participants at risk are screened usually indicates a 10%-15% risk to progress to clinical diabetes over the next five to ten years. Multiple (two or more) autoantibodies indicate that more than 70% may progress to clinical onset within ten years. 5,6 The primary objective of the TEDDY Family Study (TEFA) is to evaluate the effect of a gluten free diet (GFD) over 18 months on beta cell function and glucose metabolism in a prospective randomized trial.
In the present baseline study, the aim was to assess beta cell function based on an oral glucose tolerance test (OGTT) in participants with single islet autoantibody or an intravenous glucose tolerance test (IvGTT) in participants with multiple islet autoantibodies.
Another aim was to compare participants with two different autoantibodies to those with three or more.

| Study population
Participants eligible for the TEFA study were selected from a population of autoantibody-positive first-degree relatives of participants in The Environmental Determinants of Diabetes (TEDDY) study or those identified as autoantibody positive in one of the following studies: The Diabetes Prediction in Skåne (DiPiS) study, the Type 1 Diabetes Prediction and Prevention (DIPP) study in Finland or in TrialNet screening. The major inclusion criteria for the TEFA randomized trial were as follows: (a) participants were between 2 and 49.99 years of age, (b) participants were persistently positive (positive in at least two consecutive visits) for one islet autoantibody (GADA, IA-2A, IAA or ZnT8R/W/QA) and had impaired glucose metabolism or were persistently positive for at least two autoantibodies. The exclusion criteria were ongoing immunosuppressant therapy (topical or inhaled steroids were accepted), diabetes, treatment with any anti-diabetic medications, clinically relevant abnormal haematology results at screening, participation in other clinical drug trials within the previous three months, history of hypercalcemia, presence of associated serious disease or diabetes-protective DQB1*06:02 allele.
Prior to inclusion into the randomized trial, potential participants were tested with either an OGTT in the case of positivity for a single autoantibody or an IvGTT if positive for multiple autoantibodies. Data from these two visits are presented in the current study.
Participants were eligible to the randomized trial if they had a single autoantibody combined with impaired glucose tolerance (IGT) based on the outcome from OGTT. Participants with multiple autoantibodies (two or more) qualified for inclusion irrespective of the glycemic status. All study participants signed an informed consent. The study was approved by the Regional Ethical Committee in Lund, Sweden and by the Ethics Committee of the Hospital District of Southwest Finland in Turku, Finland.

| Assessment of beta cell function
In Sweden, a total of 3738 family members and in Finland 1526 family members of participants enrolled in the TEDDY study were screened for autoantibodies.
Participants positive for a single autoantibody underwent an OGTT (time-points −10, 0, 30, 60, 90 and 120 minutes) at their first study visit to assess the beta cell function before enrolment into the randomized trial. Participants positive for multiple autoantibodies (2 or more) underwent an IvGTT (time-points −10, 0, 1, 3, 5, 7, 10, 30, 50, 70 and 90 minutes) at their first study visit. HbA1c was analysed in all participants. As the performance of the two tests for disposal of glucose is very different, since OGTT reflects the physiological uptake of glucose via the gut into the blood stream, while IvGTT reflects the disposal of glucose when injected intravenously, we only made a direct comparison from both tests for the fasting values.
K-values were not calculated in Finnish participants due to the short IvGTT (−10, 0, 3, 5 and 10 minutes) performed in these participants.

| Analysis of beta cell autoantibodies
GADA, IA-2A and IAA were analysed using in-house radio binding assays in two separate laboratories. 7 Swedish samples were analysed

HbA1c levels
Plasma glucose, serum insulin and serum C-peptide were all analysed using standard assays at local Departments of Clinical Chemistry at University Hospitals in Malmö, Oulu and Turku.
In Malmö, an enzymatic assay with absorbance measurement was used for measurement of p-glucose (Roche Diagnostics, Basel, Switzerland). A one-step immunometric sandwich assay, electrochemilumiscence immunoassay (ECLIA; Roche Diagnostics, Basel, Switzerland), was used for measurement of s-insulin and s-C-peptide. In Malmö, HbA1c was analysed using a spectrophotometric assay (Capillary 3 Tera; Sebia, Paris, France).
In Oulu, plasma glucose was measured by an enzymatic hexokinase assay (ADVIA Chemistry XPT, Munich, Germany). Serum insulin and C-peptide were analysed by two-site sandwich immunoassay

| Statistical methods
IBM SPSS statistics 25, was used for the statistical calculations.
Student's t test was used to compare variables with normal distributions. Variables with non-normal distributions or low number of participants were compared using Mann-Whitney U test or Kruskal-Wallis test, where appropriate. P-values below 0.05 were considered as significant. Correlations were estimated using Pearson's correlation coefficient for normally distributed variables and Spearman's rank correlation coefficient for those at non-normal distributions.
Areas under the curve (AUCs) for glucose, insulin and C-peptide based on 5 time-points in the OGTT (0, 30, 60, 90 and 120 minutes) were calculated using the auc function in the R-package (MESS version 0.5.2 in R; www.r-proje ct.org) with linear interpolation using the trapezoidal rule. We only report AUCs based on complete data (5 measurements, 29 for glucose, 23 for insulin and 24 for C-peptide).
The analysis was stratified by sex, as well as by age, below and above 18 years, in order to examine the possible impact of these variables.
Fasting levels of glucose, insulin and C-peptide were compared between participants with a single autoantibody and participants with multiple autoantibodies. Otherwise, we were not able to compare glucose, insulin and C-peptide levels between participants positive for a single autoantibody and participants positive for multiple autoantibodies since those with single autoantibody underwent an OGTT at their first visit and those with multiple autoantibodies underwent an IvGTT.  Table S1 and S2).

| The demographics of the study participants
The participants were between 2-49.99 years, but there were no participants between 18 and 30 years, and the median age of those under 18 years was 10 years.
The mean age, in both Sweden and in Finland, was older in participants who underwent an OGTT (mean 37.7 ± 8.72 years) and (mean 28.2 ± 16.0 years) as compared to those who underwent an IvGTT (mean 17.3 ± 11.8 years; P < .001) and (mean 8.87 ± 7.99 years; P < .001), respectively. Furthermore, participants from Finland who underwent an IvGTT were younger (mean 8.87 ± 7.99 years) than those from Sweden (mean 17.3 ± 11.8 years; P = .0140). There was no difference in age or BMI between men and women with multiple autoantibodies in Sweden nor in Finland (Table 1).
In Swedish participants positive for a single autoantibody, HbA1c values were higher in men (34.7 ± 2.71 mmol/mol) than in women (31.8 ± 2.25 mmol/mol; P = .0180). There were too few participants in Finland to perform this comparison. There was no difference in HbA1c, vitamin D levels or glucose between men and women in Sweden or in Finland or when participant below and above 18 years were compared. (Supplementary online Table S3). There was no difference in AUC for glucose from OGTT in Swedish and Finnish participants (P = .518; Table 2). AUC for insulin from OGTT was higher in the Finnish participants (n = 6; 11 676 ± 8356) than in the Swedish participants (n = 17; 5333 ± 3194; P = .0360; Table 2). AUC for C-peptide was similar in Swedish and Finnish participants (P = .836; Table 2). There were no differences in AUC of glucose, insulin or C-peptide with respect to gender. No differences were identified in AUC of glucose, insulin or C-peptide in participants below or above 18 years of age in Sweden (Supplementary online   Table S5). There were no differences in glucose values at any timepoint in participants positive for GADA alone (n = 10) or IAA alone (n = 11) in Sweden (Supplementary online Table S6). This analysis was not performed separately for participants from Finland due to too few participants in each group. There was only one participant from Finland that was positive for IA-2A alone and none in Sweden.

| OGTT in single autoantibody participants
No differences in glucose metabolism were detected among those with a single autoantibody, either GADA (n = 17) or IAA (n = 13), with respect to autoantibody levels. The tested parameters were HbA1c, −10 minutes glucose, −10 minutes C-peptide and 120 minutes insulin, 120 minutes glucose, 120 minutes C-peptide and 120 minutes insulin (Supplementary online Table S7).

There was no difference in K-values between men and women in
Sweden, neither in children below and above 10 years in Sweden or when participants below and above 18 years were compared (Supplementary online Table S3). K-values were not calculated in Finnish participants due to the short IvGTT performed in these participants. There was no difference in glucose levels at any time-point in participants below and above 18 years of age (Supplementary online Table S9).

TA B L E 1 Clinical characteristics of participants enrolled in the TEFA study in Sweden and in Finland
There was no difference in glucose values at −10 minutes or FPIR between men and women in Sweden or in Finland, neither in FPIR in children below and above 10 years in Sweden or in Finland or when participant below and above 18 years were compared.
FPIR was lower in participants with three or more autoantibodies (n = 13) compared to those with only two autoantibodies (n = 15) in Sweden (P = .0330) but not in Finland (n = 5). K-values were similar between those with two autoantibodies as compared to those with three or more (Table 3).
At time-point −10 minutes in the Swedish participants, there were significant correlations between BMI and fasting insulin levels (P = .0170) and BMI and fasting C-peptide levels (P = .0160), but not between BMI and fasting glucose levels. In the Finnish participants, there were significant correlations between BMI and fasting glucose levels (P = .0410) and BMI and fasting C-peptide levels (P = .0180), but no significant correlation between BMI and fasting insulin levels (  Abbreviation: M10 = 10 minutes prior to ingestion of glucose.

| Comparison of fasting levels of glucose, insulin and C-peptide between participants with a single autoantibody or with multiple autoantibodies
There was no difference in fasting glucose levels between participants with two or three and more autoantibodies, neither in  (Table 5).

| D ISCUSS I ON
The most important finding of this study was that all participants with only a single autoantibody had normal beta cell function based on glucose, C-peptide and insulin levels in the 6-point OGTT test.
Therefore, none of the participants with a single autoantibody was eligible to enter the randomized trial with either gluten free or regular diet. This finding is in line with a recent study reporting that single islet autoantibody at clinical diagnosis of type 1 diabetes is associated with older age in the patients as well as insulin resistance and higher BMI. 8 It was important to test single autoantibody-positive participants as 15% (483/3165) newly diagnosed 1-to 18-year-olds had only one islet autoantibody at the time of clinical diagnosis. 9 Therefore, it could not be excluded that single autoantibody-positive participants with impaired glucose tolerance would have been eligible for the TEFA study.
Interestingly, HbA1c levels were higher but still within the reference range in men with a single autoantibody as compared with women in Swedish participants, but no difference in age was observed. In some studies, HbA1c has been proven to be a useful marker to predict the progression to type 1 diabetes. 4,10 An increase in HbA1c levels has previously been reported as a specific but not sensitive indicator of progression to type 1 diabetes in children. 10 In a study of children with multiple autoantibodies, it was shown that a 10% increase in HbA1c was highly predictive for type 1 diabetes within one year. 4  Possibly, the difference in the age groups between Sweden and Finland, and the low number of participants in Finland, might have had an impact on these outcomes.
Another important finding was that participants who were positive for three or more autoantibodies had significantly lower FPIR as compared to participants positive for only two autoantibodies. This finding indicates that the higher the number of autoantibodies, the more severe is the disturbance in insulin production. A higher number of autoantibodies is an indication of an approaching manifest type 1 diabetes, a finding supported in other studies. 15,16 Declined FPIR was found several years before the onset of clinical type 1 diabetes in the DIPP children with multiple autoantibodies but not in those with a single autoantibody. [17][18][19] The strength of the present study is that all participants (except one) were persistently islet autoantibody positive prior to enrolment into this study and that all participants followed a coordinated protocol. A major limitation on the present study is that participants, positive for a single autoantibody underwent an OGTT, while participants, positive for multiple autoantibodies underwent an IvGTT.
Thus, the outcome of these glucose challengers is not comparable.
Another limitation of the present study is that measurements of glucose, insulin and C-peptide were analysed in three different clinical laboratories with different reference values. Differences in assay performances therefore precluded direct comparisons between the three different laboratories. An additional limitation is that autoantibody measurement, both prior to and during the study, were performed in two different laboratories (one for each country). Another major limitation is the low number of participants raising concerns for statistical power; nevertheless, it requires a large screening population of healthy participants to identify autoantibody-positive participants. In Sweden, a total of 3738 family members were screened, and in Finland, a total of 1526 family members were screened, to a participant enrolled in the TEDDY study.
In conclusion, participants with only a single autoantibody appeared to have a normal beta cell function. Participants positive for three or more autoantibodies have a lower FPIR as compared to participants positive for two autoantibodies, indicating a more severe beta cell lesion.