Diabetic ketoacidosis in children newly diagnosed with type 1 diabetes mellitus: Role of demographic, clinical, and biochemical features along with genetic and immunological markers as risk factors. A 20‐year experience in a tertiary Belgian center

Diabetic ketoacidosis (DKA) is the leading cause of morbidity and mortality in children with type 1 diabetes (T1D). Little is known about the association between genetic and immunological markers and the risk for DKA at onset of T1D. The aim of this study was to create a model foreseeing the onset of DKA in newly diagnosed patients.

Diabetic ketoacidosis is still the leading cause of morbidity and mortality in children with T1D. Mortality is predominantly related to cerebral edema. DKA at T1D diagnosis results in morphologic and functional brain changes. 1,2 There is a wide geographic variation in the frequency of DKA at T1D onset 3,4 ; rates vary from 14.7% (Denmark) to 79.8% (Saudi Arabia) 5 and inversely correlate with the regional incidence of the disease. 3,4 A decreasing rate of DKA at diagnosis has been observed over time. 6 Increasing medical information and awareness might have resulted in changes in the clinical presentation of T1D in developed countries. 3,7,8 Previous studies have shown different risk factors that might increase the likelihood of presenting with DKA. [8][9][10] Risk factors for DKA in newly diagnosed cases include: younger age [<2 years (yr) old], ethnic minority, lower Body Mass Index (BMI), delayed diagnosis as well as lower socioeconomic status. Protective factors are: having a first-degree relative (FDR) with T1D at the time of diagnosis, higher parental education, and higher background incidence of T1D. 8,11 Specific human leukocyte antigen (HLA) class II gene alleles at the HLA-DRB1, DQA1, and DQB1 loci are the major genetic determinants for predisposition and resistance to T1D. In Caucasian populations, the highest risk for developing the disease has been associated with the DR3-DQA1*0501-DQB1*0201/DR4-DQA1*0301-DQB1*0302 genotype that increases 20-fold the risk for T1D. [12][13][14] In Belgium, the Belgian Diabetes Registry (BDR) has shown that this high-risk genotype confers a 21.5-fold increased relative risk of developing diabetes before the age of 40. 15 In 1986, Ludvigsson et al. noted that DR3 patients had more often a mild disease with less ketonuria at diagnosis in North America and Europe. 16 Currently, little is known about the association between various HLA class II genotypes and the risk of DKA at T1D onset. 17,18 Islet autoantibodies (AAb) are also useful preclinical markers for risk of developing T1D, including: islet cell antibody (ICA), insulin autoantibody (IAA), glutamic acid decarboxylase antibody (GADA),

| Clinical and biological assessments
A retrospective review of patients' files was analyzed in order to collect data at diagnosis.
Age at onset of T1D was expressed in years.
Ethnicity was self-reported and determined on the basis of the

| HLA-genotyping and associated risk categories
Blood samples were collected within the first days of diagnosis and sent to the BDR for identification of HLA-genotyping. HLA-DQ genotype was determined using polymerase chain reaction and oligotyping with allele specific probes as described previously. 16 In 22 patients, no sample was available for analysis; however, they had one or more positive diabetes-associated AAb reflecting β-cell autoimmunity. HLAassociated risk categories for developing T1D were defined according

| Statistical analysis
Univariate, bivariate, and multivariate analysis were performed on our data in order to investigate risk factors related to DKA and its severity at diagnosis.
Quantitative parameters were described by their means and standard deviations or medians and interquartile ranges (IQR), qualitative parameters as frequencies and percentages.
In bivariate analysis, continuous factors were tested using Student's t tests or ANOVA for equality of means and Mann-Whitney U tests for equality of medians. Categorical factors were performed with χ 2 or Fisher's exact tests of independence.
A multivariable logistic regression was performed in order to investigate the effect of multiple risk factors on children presenting with DKA at T1D onset and the results were reported as odds ratio (OR) and 95% confidence interval (CI).
Statistical analysis were performed at a 5% level of significance.
Results were considered statistically significant when P value was <0.05. Statistical analysis were performed using SAS 9.2. Software. The mean age at diagnosis was 7.8 years (range: 0.2-17.5); 15% were aged 0 to 3 years.
The median duration of the symptoms was 2 weeks (range: 1-3).

| Comparison of characteristics between groups with and without DKA
Overall 42% of patients presented DKA at diagnosis, 51% of whom were boys, and 49% were girls, giving a male-to-female ratio of 1:1.
Ethnic minorities presented with DKA significantly more frequently than ethnic majorities (P = .004) ( Table 1). Moreover there was no difference in the occurrence of DKA between EC and MC (P = .335, data not shown).
The prevalence of DKA was significantly higher in the 0 to 3 age group compared to children over 3 years (P = .001) ( Table 1).
T A B L E 1 Comparison of characteristics between groups with and without diabetic ketoacidsis (DKA) at diagnosis; frequencies with percentages for qualitative variables and means with standard deviations (SD) or medians with interquartile ranges (IQR) for quantitative variables  23,24 ) and all the above variables were compared among the groups.
Severe form of DKA was observed in 29% of children presenting with DKA at diagnosis.
When considering ethnic minorities and majorities, there were no differences in frequencies of degrees of DKA.
The three groups differed significantly for mean age at diagnosis; the lower the age, the more severe the DKA (P = .009). Above all, children below 3 years presented with severe DKA statistically more frequently than older children (P = .026).
The GCS scores never declined to less than 14 and no other clinical signs of cerebral edema were reported.
No deaths linked to DKA occurred in patients with DKA of any degree at diagnosis. There were no significant differences in the prevalence of any multiple AAb positivity at diagnosis neither concerning the occurrence of DKA nor its severity (data not shown).

|
No statistically significant differences were found in the prevalence of HLA risk genotypes neither in the occurrence of DKA nor in its degrees of severity (data not shown).
Neutral genotypes had significantly higher prevalence in the ethnic minority group whether the ethnic majority group carried more frequently susceptible or protective genotypes (P = .001) (Figure 1).

| Multivariable logistic regression model
This model aimed to define the predictive markers for DKA at T1D onset ( Figure 2 Neu et al. 27 and Szypowska and Skórka 10 suggested that children under 2 years of age remain the most prone to DKA. The short duration of symptoms and high frequency and severity of DKA in this age group have been attributed to a more rapid decline in β-cell function, further aggravated by higher incidence of intercurrent infections precipitating acidosis, higher risk of dehydration, and less developed mechanism of metabolic compensation. 28 The prevalence of DKA in boys and girls was the same in our study, as already shown in a systematic review by Usher-Smith et al. 11 Ketoacidosis is often considered to be the result of a late diagnosis. 29 However, in our study, the duration of symptoms was similar (2 weeks) in children with and without DKA at presentation and at any degree of acidosis.
As expected, children presenting with DKA at diagnosis had higher glycaemia and HbA1c and lower random C-peptide level, related to poorer residual endogenous insulin secretion, as previously reported. 30,31 Although well-established data revealed that a precocious diagnosis preserves the residual endogenous insulin secretion during a longer period, 31 it has been shown that early intensive therapy did not provide benefits in preserving β-cell function. 32 Messaaoui et al. showed that younger age and lower C-peptide level at T1D onset are correlated with an increased glomerular filtration rate which could be responsible for later nephropathy. 33 Like other authors, 8,18 we also observed that having a FDR with T1D is a preventing factor against DKA at diagnosis. Moreover, the prevalence of IA-2A is higher in children than in adults. 37 It has been suggested that islet AAb screenings may lead to F I G U R E 1 HLA risk genotypes (protective, neutral, and susceptible) by ethnicity an earlier diagnosis of T1D and, thereby, prevent DKA at onset and improve the clinical course after diagnosis. 19,21 The nature and prevalence of HLA-DQ and HLA-DR haplotypes and genotypes susceptibility for T1D may differ between populations throughout the world. [38][39][40] In On the contrary, in an Italian study, subjects with increased HLAassociated risk genotypes were more likely to present DKA at T1D onset, compared to subjects with neutral or decreased HLAassociated risk genotypes. 18 It is noteworthy that, in our study, children with neutral genotypes had a 1.5-fold increased risk of DKA at diagnosis than children with susceptible or protective genotypes, an observation not previously reported. Interestingly, neutral genotypes were more frequent in ethnic minorities and, paradoxically, these parameters were both independent predictors of DKA at T1D onset in the logistic regression analysis.
Standardized criteria for definition of DKA were proposed by ISPAD in 2014 and 2018. 23,24 Irrespective of different definitions of DKA, the general observation is that the prevalence of DKA is decreasing when the incidence of T1D is increasing. 41 International studies observed a continuous increase of the incidence of T1D in children. 42 In Belgium, before age 40, the incidence of type 1 diabetes averages 10 new cases per 100 000 persons per year, rates increasing by 3.7% annually in children <15 years. The incidence rate of T1D doubled over a 15-year period in Belgian children <15 years and more than doubled in those <5 years. 43,44 In our population we noted that the frequency of DKA at disease onset did not vary over time although the incidence of type 1 diabetes more than doubled during the 20-year period of observation (data not shown).
Earlier diagnosis through genetic and immunological screenings of high-risk children could decrease DKA incidence at diabetes onset. 45,46 The Environmental Determinants of Diabetes in the Young (TEDDY) study recently showed a decreased prevalence of DKA in youth <5 years of age participating in a longitudinal follow-up. 47 In Belgium, the BDR systematically performs genetic and annual immunological assessments in FDR of T1D patients <40 years.
Based on the analysis of risk factors for DKA at T1D onset discussed above, our multivariable model could help to identify high-risk children for DKA at T1D diagnosis.
Therefore, in order to prevent the incidence of DKA in newly diagnosed children, a medical follow-up, as well as parental learning to recognize early symptoms of T1D, should be carried out on mainly the following five target groups: children with a FDR affected by T1D, younger children (<3 years), those belonging to ethnic minorities, children with IA-2A positivity at immunological screenings and the ones carrying neutral genotypes at the genetic assessment. In our opinion, the last two markers are original.
The ROC curve in our study showed that this model would be able to correctly predict the probability of DKA at diagnosis in almost 70% of cases (Figure 3) Although it involved a large and multi-ethnic population of children, this study had limitations. This was a single-center retrospective study; moreover, the number of patients non EC or MC was relatively small. Therefore, caution should be used in generalizing these findings to other populations.

| CONCLUSIONS
As DKA at diagnosis of T1D is still a cause of mortality in children with T1D, the identification of risk factors for DKA is of the utmost importance.
We confirmed that children aged <3 years, those without a FDR affected by T1D and ethnic minorities are at higher risk of developing DKA.
To our knowledge, this was the first reported implication of IA-2A positivity and neutral genotypes predisposing to DKA at diagnosis regardless of its severity. The proposed multivariable model could help to predict the probability of DKA in approximately 70% of newly diagnosed cases.
Prevention programs should consider the application of special measures to avoid DKA in these high-risk groups.