Association of Diabetic Ketoacidosis at Onset, Diabetes Technology Uptake, and Clinical Outcomes After 1 and 2 Years of Follow-up: A Collaborative Analysis of Pediatric Registries Involving 9,269 Children With Type 1 Diabetes From Nine Countries

Diabetic ketoacidosis (DKA) at presentation of type 1 diabetes is a serious acute and potentially life-threatening complication. In theory, most of DKA at diagnosis of type 1 diabetes could be avoided if the delay in diagnosis and appropriate management were eliminated and insulin replacement therapy started immediately, possibly through screening for type 1 diabetes (1). Unfortunately, the incidence of DKA at diagnosis increased in many health care systems around the world the decade prior to the coronavirus disease 2019 pandemic and rose even further during the pandemic (2). Timely diagnosis of type 1 diabetes is associated with long-term benefits, including lower HbA_1c and lower insulin requirements (3). In contrast, delayed treatment can lead to DKA-associated acute and chronic consequences, including diminished residual β-cell function (4), risk of recurring DKA (5), harmful neurocognitive outcomes (6), and persistently higher HbA_1c at follow-up (7,8).

The introduction of new technologies, such as automated insulin delivery (AID) systems, has transformed the management of type 1 diabetes, with AID systems now being recommended as the standard of care when accessible (9). Although the onset of type 1 diabetes with DKA has known consequences during follow-up, whether early initiation of advanced technologies can mitigate these outcomes remains less clear. Current studies addressing this issue have been limited by small sample sizes and have reported conflicting results (10,11).

Our longitudinal, multinational, population-based study including youth with type 1 diabetes investigated whether the presence of DKA at diagnosis is associated with lasting differences in glycemic outcomes, insulin requirements, BMI, and the incidence of acute complications, focusing on the role of different treatment modalities on these outcomes.

This international, multicenter population-based study on DKA at type 1 diabetes diagnosis used data from nine countries: Austria, Czechia, Germany, Italy, Luxembourg, New Zealand, Slovenia, Switzerland, and U.S. (Colorado). The study population included children and adolescents aged 0.5 to 15.9 years with type 1 diabetes diagnosed between 1 January 2019 and 31 December 2020 with a follow-up period until 31 December 2022 (12). We excluded individuals with known use of glucose-modifying therapies (other than insulin) and individuals with chronic multiple conditions (other than treated autoimmune thyroid disease and celiac disease). To exclude cases of neonatal diabetes, children <6 months were not included in the analysis. The coverage of the above data sources is listed in the Supplementary Material.

We analyzed demographic data (sex, age, weight, height, BMI SD score [SDS]) and DKA status at type 1 diabetes diagnosis and additionally HbA_1c, insulin requirement, BMI SDS (13), occurrence of severe hypoglycemia (SH) and subsequent DKA during insulin therapy, as well as treatment modality (continuous glucose monitoring [CGM], insulin pump, and/or AID system at follow-up). DKA was defined according to the International Society for Pediatric and Adolescent Diabetes (ISPAD) criteria as pH <7.3 and/or bicarbonate <15 mmol/L, and severe DKA as pH <7.1 and/or bicarbonate <5 mmol/L (14); cases not meeting criteria for severe DKA were categorized as nonsevere. The World Health Organization reference values for BMI were used to calculate BMI z scores using the LMS (Lambda, Mu, and Sigma) method (15,16).

The current study and audit received ethical approvals in each country, either from the authors’ institutional review boards or from relevant national bodies. Anonymized person-level data were analyzed at the Institute of Epidemiology and Medical Biometry, University of Ulm, Ulm, Germany. This report follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines for cohort studies.

Descriptive analyses were provided for the entire group, by age-group, and by country. To adjust for demographic differences between registries, linear and logistic regression analyses were implemented with age at diagnosis (<6, 6 to <10, 10 to <14, and 14 to <16 years), sex, and diabetes therapy after 1 or 2 years, respectively (multiple daily injection [MDI] or continuous subcutaneous insulin infusion [CSII]). To elucidate the potential association between diabetes technology use and the outcomes, additional models with adjustment for technology use were implemented. Models were based on complete case analysis with least-square means to depict adjusted outcome. We used hierarchic regression models (linear, logistic, or log-binomial according to the dependent variable) with a random intercept for each participating country. Partial likelihood was used for estimation, reporting least-square means for participants with and without DKA at onset, based on observed marginal distribution. Denominator degrees of freedom were estimated based on the between within method, using SAS proc glimmix on a Windows server mainframe. All analyses were performed with the SAS 9.4 software build TS1M8. A two-sided P value of <0.05 was defined as significant, applying the Tukey-Kramer method for correction of P values applied for multiple comparisons.

We collected and analyzed data from 9,269 children and adolescents diagnosed with type 1 diabetes across nine countries between 1 January 2019 and 31 December 2020 (Supplementary Material). The mean (SD) age of the study population at diagnosis was 9.0 (4.0) years, 5,081 (54.8%) were male. Overall, BMI z score for age and sex (BMI SDS) was −0.02 (1.4). There were 3,172 individuals (34.2%) who presented with DKA at the diagnosis of type 1 diabetes, of which 1,189 (12.8% of whole study population) presented with severe and 1,983 (21.3%) with nonsevere DKA. The baseline characteristics of the study population stratified by DKA status are reported in Table 1.

After 1 year of follow-up, overall mean (SD) unadjusted HbA_1c was 7.34% (1.2) or 56.7 mmol/mol (12.7), and 40.3% of study participants achieved the recommended glycemic target HbA_1c <7%. After 2 years, mean HbA_1c was 7.51% (1.2) or 58.6 mmol/mol (13.4), and 35.1% of youth with type 1 diabetes achieved HbA_1c <7%. Overall, we noted 193 subsequent DKA and 261 SH events during the first year of follow-up and 244 DKA and 251 SH events during the second year of follow-up.

In the regression analysis adjusted for age at type 1 diabetes diagnosis, sex, treatment modality, and country, severe DKA at diagnosis was associated with higher adjusted HbA_1c compared with nonsevere DKA over the whole observation period after 1 year (7.41% [SD 0.13] vs. 7.23% [0.13], P = 0.001 after 1 year) and after 2 years (7.58% [0.15] vs. 7.38% [0.15], P < 0.001). Similar results were seen when the severe DKA group was compared with the no DKA group after 1 year (7.41% [0.13] vs. 7.14% [0.13], P < 0.001) and after 2 years (7.58% [0.15] vs. 7.32% [0.15], P < 0.001) (Table 2 and Fig. 1).

Severe DKA at diagnosis was associated with higher insulin requirement compared with the no DKA group after 1 year (0.72 IU/kg/day vs. 0.62 IU/kg/day, P < 0.001) as well as after 2 years of diabetes (0.76 IU/kg/day vs. 0.71 IU/kg/day, P = 0.003), but not compared with the nonsevere DKA group (Table 2 and Supplementary Material). After 1 year of type 1 diabetes, BMI SDS was 0.86 in the severe DKA group compared with 0.61 in the nonsevere DKA group (P < 0.001) and 0.52 in the no DKA group (P < 0.001). A similarly significant trend was observed for the second year of observation (Table 2). The proportion of individuals experiencing SH or subsequent DKA events was low. There were no differences in SH events between the groups, while the group with severe DKA at diagnosis experienced more subsequent DKA events compared with the no DKA group both after 1 (P = 0.005) and 2 years (P = 0.030) (Table 2 and Supplementary Material).

Technology use within the first year demonstrated 89.8% of study participants started to use CGM, 41.9% used an insulin pump, and 10.7% were using AID. After 2 years, 94.2% were CGM users, 52.1% were insulin pump users, and 19.6% were AID users. The adoption of CGM was similarly high in all DKA groups (P = 0.916 for the first year and P = 0.283 for the second year). Similarly, there were no differences in insulin pump use between the DKA groups (P = 0.638 and P = 0.231, respectively). However, significant differences were present in AID uptake stratified by DKA at onset groups at 1 year: 11.2% in the severe DKA group versus 13.2% in the nonsevere group versus 9.7% in the no DKA group (P < 0.001). After 2 years, these differences were no longer significant (19.5% in the severe DKA group vs. 21.8% in the nonsevere group vs. 18.9% in the no DKA group, P = 0.171).

An evaluation of associations between treatment modalities and glycemic outcomes achieved following DKA at onset showed the differences between groups persisted among MDI + CGM and CSII + CGM subgroups (Fig. 2and Supplementary Material). No differences between the DKA groups were observed in the group using MDI + blood glucose monitoring (BGM), yet this group had the highest adjusted HbA_1c levels after 1 year (7.59% [SD 0.33] in severe DKA vs. 7.47% [0.32] in nonsevere DKA and 7.44% [0.31] in the no DKA group, P = 0.622) and after 2 years of follow-up (8.30% [0.37] vs. 7.91% [0.35] vs. 7.93% [0.31], respectively; P = 0.350). In a subgroup of youth using AID systems, there were no differences between the severe DKA, nonsevere DKA, no DKA groups after 2 years (7.31% (0.14) vs. 7.15% (0.13) vs. 7.11% (0.12), respectively; P = 0.072), while after 1 year of follow-up the differences were marginal (7.31% (0.13) vs. 7.13% (0.11) vs. 7.03% (0.10), respectively; P = 0.033).

The presence of DKA, especially severe DKA, at the diagnosis of type 1 diabetes is associated with higher HbA_1c levels, greater insulin requirements, and higher BMI SDS in the first 2 years following diagnosis. Our study is in line with several studies that previously described persistent differences in glycemic outcomes linked to DKA status at onset (7,8,17). Importantly, since the glycemic outcomes have a direct effect on the long-term risks of cardiovascular (macrovascular) and renal (microvascular) complications, it is paramount to achieve tight glycemic control regardless of the DKA status (18,19). In the current study, there were clear trends toward lower HbA_1c at each time point with both CGM and AID use, technologies that facilitate more intensive diabetes management than MDI and BGM approaches.

We observed high prevalence of DKA at type 1 diabetes diagnosis. In an international multicenter study that included several registries from the current analysis, Birkebaek et al. (2) reported a long-term prepandemic increase in the prevalence of diabetic ketoacidosis at the diagnosis of pediatric type 1 diabetes with a marked additional escalation during the coronavirus disease 2019 pandemic in 2020 and 2021, when the prevalence of DKA at diabetes diagnosis reached almost 40%.

We observed that predicted differences in long term HbA_1c based on DKA at diagnoses are mitigated by AID. A recent single-center study evaluating the initiation of AID systems within 1 month after the diagnosis of type 1 diabetes demonstrated that AID mitigated the negative glycemic effects of DKA at diagnosis despite the relatively higher insulin requirement among those who had experienced DKA in 51 children with type 1 diabetes (10). Importantly, a pilot study demonstrated reductions in hyperglycemia and glucose variation, accompanied by gains in standardized IQ scores and multiple metrics of brain development and function with early AID initiation (20). This strongly indicates a tendency toward “normalization” in the AID group compared with the control group. In the current study, AID users had lower HbA_1c levels compared with the overall population and moderated differences between DKA groups after 2 years of follow-up. Despite the relatively low proportion of the youth with diabetes using the AID systems in our current study, we observed a positive trend in AID uptake, with a doubling in numbers of children using AID from ∼10% after year 1 to ∼20% after year 2 and even more children using it in both DKA subgroups compared with the no DKA group. These results and other recently published data further emphasize the role of the AID systems, and early and equitable access to AID could further improve outcomes in this population (21).

With the rapid advent and broader availability of advanced diabetes technologies, timely introduction of these technologies to all interested users could be increasingly important to equalize the access to diabetes care (22). We observed very high prevalence of CGM use, with almost 90% of participants, regardless of DKA status at diagnosis, using CGM 1 year after the type 1 diabetes diagnosis. Less than 10% used CGM and less than 5% used CSII at the same time points. Although these trends show high technology uptake among these relatively wealthy countries, there may remain further room for improvement by increasing technology uptake within the first year or months following a diabetes diagnosis. Indeed, very early CGM initiation improved HbA_1c when initiated within 1 month of diagnosis in a structured clinical program (22).

We observed an almost 10% higher insulin requirement after 1 year in the group who experienced severe DKA at diagnosis compared with the group without DKA, consistent with prior studies. DKA coinciding with the diagnosis of type 1 diabetes is associated with a higher depletion of β-cell mass, possibly increasing exogenous insulin needs during follow-up (23). Recently, data from the Innovative Approach Toward Understanding and Arresting Type 1 Diabetes (INNODIA) consortium reported a possible association between DKA status at onset and progressive decline of C-peptide (4). Additionally, Hammersen et al. (7) observed an association between early clinical diagnosis of type 1 diabetes and a higher rate of partial remission after 3 years of follow-up. Preserving residual endogenous insulin secretion is associated with less glycemic variability and improved glycemic outcomes (24–26). Together, these data suggest that diagnosis of type 1 diabetes before the onset of DKA preserves β-cell function and that the early CGM and AID initiation may further slow β-cell loss through more intensive management.

Although lower BMI is associated with higher rate of DKA at diagnosis (17), less is known regarding BMI at follow-up. We observed that children who experienced severe DKA tend to have persistently higher BMI SDS at follow-up compared with both nonsevere DKA and no DKA groups. Socioeconomic disadvantage provides a greater risk for obesity (26) and a higher DKA rate at onset (27), but how DKA and risk for obesity may be connected is unclear. Insulin resistance associated with obesity may accelerate the risk of DKA, but the diagnosis of obesity may be hidden by the dehydration and catabolism seen in DKA. Therefore, comprehensive strategies are needed to identify and address these interconnected concerns early in the course of disease. Early diagnosis, improved education, and equitable health care access may improve outcomes for children with diabetes and mitigate the long-term impacts of these disparities.

The severe DKA group experienced significantly more subsequent DKA episodes than the other groups, suggesting socioeconomic factors might possibly persist and create an environment more prone to DKA repetition (28,29). This should be addressed, as any single DKA episode has an impact on further development, including neurocognitive functions (30) and AID could be the means to reduce its occurrence.

Our study has certain limitations. The design does not allow for any causal inferences. Additionally, the association between race, ethnicity, and especially socioeconomic status on the prevalence of DKA and outcomes during the follow-up period could not be analyzed because of the heterogeneity of collected data across the registries and varying definitions for racial and ethnic background across registries. This heterogeneity also applies to the HbA_1c targets that might differ across the participating countries/centers. Setting and communication of healthier HbA_1c targets and benchmarking activities at a local diabetes center level are known to impact outcomes and might have affected the study outcomes (31). Also, some registries cover only a minority of children with type 1 diabetes of their country and are therefore not representative of the whole country. Furthermore, our data did not capture comorbidities or concomitant medications at follow-up that could have had a minor impact on our outcome data.

Our study has several strengths, with the first being the large sample size and the multinational design. Secondly, the length of the longitudinal observation period supports the durability of our findings. The adoption of insulin pumps and CGM was high and similar between DKA status groups, indicating equity in providing this technology for youth with type 1 diabetes regardless of DKA status.

In conclusion, DKA at type 1 diabetes diagnosis is associated with persistent disadvantageous differences in HbA_1c, BMI, and insulin dose 1 and 2 years after onset. Early adoption of CGM and AID systems might improve long-term glycemic management and moderate the predictive effect of DKA on long-term higher HbA_1c.

Acknowledgments. The authors acknowledge all of the international registries for making data available for this international project. The authors thank Andreas Hungele and Ramona Ranz (University of Ulm, Ulm, Germany) for their support with data management and Julia Grimsmann (University of Ulm, Ulm, Germany) for validating the statistical analysis. A detailed list of the registries contributing data to this analysis can be found in the Supplementary Material.

Funding. The Ulm University received funding from German Center for Diabetes Research (grant number 82DZD14E03) and German Robert Koch Institute. U.S. data was supported by the University of Colorado Diabetes Research Center Clinical Resources Core National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases grant P30-DK116073. The Slovenian Childhood Diabetes Registry is supported by Slovenian Research Agency grants J3-4521, J3-4528, and P3-0343.

Duality of Interest. G.T.A. has been an advisory board member for Mannkind. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. K.D., V.N., and M.d.B. interpreted the analyses, visualized the results, and wrote the initial draft of the manuscript. R.W.H. designed the study, performed the data analysis, contributed to writing, and critically revised the article. K.D., V.N., G.G., V.C., G.T.A., M.F., C.B., C.d.B., R.W.H., and M.d.B. collected data, contributed intellectually to this study, and critically reviewed the scientific content of the manuscript. All authors had full access to all the data in the study, had final responsibility for the decision to submit for publication. and approved the final manuscript as submitted. R.W.H. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis

Prior Presentation. Parts of this work were presented in abstract format and as a poster presentation at the International Society for Pediatric and Adolescent Diabetes (ISPAD) 50th Annual Conference, 2024, in Lisbon, Portugal, 16–19 October 2024.

Handling Editors. The journal editors responsible for overseeing the review of the manuscript were Elizabeth Selvin and Thomas P.A. Danne.