Diabetic ketoacidosis (DKA) is a life-threatening complication of type 1 diabetes mellitus (T1DM) that results from absolute insulin deficiency and is marked by acidosis, ketosis, and hyperglycemia (1). Therefore, prevention of DKA is one goal in T1DM care, but recent data indicate increased incidence (2).
For adult patients, only limited data are available on rates and risk factors for development of DKA, and this complication remains epidemiologically poorly characterized. The Diabetes Prospective Follow-up Registry (DPV) has followed patients with diabetes from 1995. Data for this study were collected from 2000 to 2016. Inclusion criteria were diagnosis of T1DM, age at diabetes onset ≥6 months, patient age at follow-up ≥18 years, and diabetes duration ≥1 year to exclude DKA at manifestation. DKA was defined as serum pH <7.3, and rates of DKA were analyzed based on a Poisson regression model accounting for overdisperson stratified by sex, age, diabetes duration, HbA1c, treatment regimen, size of diabetes center, and migration background, i.e., whether the patient or one or both parents were born outside the countries of the registry. A diabetes center that treated ≥50 adult patients with T1DM in the year 2016 was considered large.
In total, 46,966 patients were included in this study (average age 38.5 years [median 21.2], 47.6% female). The median HbA1c was 7.7% (61 mmol/mol), median diabetes duration was 13.6 years, and 58.3% of the patients were treated in large diabetes centers.
On average, 2.5 DKA-related hospital admissions per 100 patient-years (PY) were observed (95% CI 2.1–3.0). The rate was highest in patients aged 18–30 years (4.03/100 PY) and gradually declined with increasing age (Fig. 1A). No significant differences between males (2.46/100 PY) and females (2.59/100 PY) were found (Fig. 1B). Individuals with a migration background had significantly higher hospitalization rates than nonmigrants (3.40/100 PY vs. 2.47/100 PY; P < 0.05) (Fig. 1C). Diabetes duration ≤2 years or >20 years was associated with lower DKA rates (2.03 and 1.62/100 PY, respectively) in comparison with diabetes duration between 5 and 10 years (Fig. 1D). Patients with HbA1c levels <7% (53 mmol/mol) had significantly fewer DKA admissions than patients with HbA1c ≥9% (75 mmol/mol) (0.88/100 PY vs. 6.04/100 PY; P < 0.001) (Fig. 1E).
Regarding therapy, use of an insulin pump (continuous subcutaneous insulin infusion [CSII]) was not associated with higher DKA rates (Fig. 1F), while patients aged 31–50 years on CSII showed lower rates than patients using multiple daily injections (2.21 vs. 3.12/100 PY; adjusted P < 0.05) (Fig. 1G). Treatment in a large center was associated with lower DKA-related hospital admissions (Fig. 1H).
The overall DKA rate was lower than in a previous DPV analysis of patients that were ≤20 years old (3). While glycemic control in children and adolescents with T1DM can be more challenging, parents of children with T1DM may be less reluctant to visit the emergency room than the adult patient. In both adults and children, poor metabolic control was the strongest predictor of hospital admission due to DKA. This might be explained by suboptimal diabetes treatment and/or inadequate patient adherence possibly because of poor patient education and/or other factors. In contrast to the findings in younger patients (3), the DKA rate was not higher in females, possibly demonstrating that omitting insulin injections in order to lose weight might be less prevalent in adult patients. Migrants showed increased risk for DKA, possibly due to lower household income, a separate DKA risk factor (4). Furthermore, diabetes education may be insufficient because of cultural and/or language barriers.
CSII was not associated with higher DKA rates, which stands in contrast to older studies (5). This may be attributable to technological advances in current CSII and catheter technology as well as improved patient education. In this study, patients were rarely equipped with glucose sensors; therefore, the increased use of continuous glucose monitoring since 2016 might further reduce DKA rates, especially in patients with CSII. Conventional insulin therapy is not considered a standard treatment of T1DM. Since DKA rates are low, it is a safe option for patients that need the simplest possible therapy, e.g., bedridden or mentally restricted patients.
One strength of this study is its population-based multicenter database, permitting analysis of demographic and treatment variables. Limitations are its observational design, which precludes detection of causal associations, and possible reporting biases due to the structure of the registry.
In conclusion, the results of this study identify patients with T1DM at risk for DKA (high HbA1c, diabetes duration 5–10 years, migrants, age 30 years and younger) in real-life diabetes care. These at-risk individuals may need specific attention since structured diabetes education has been demonstrated to specifically reduce and prevent this acute complication.
Acknowledgments. The authors thank Andreas Hungele and Ramona Ranz for support and development of the DPV documentation software and Katharina Fink and Esther Bollow for DPV data management (each a clinical data manager at the University of Ulm).
Funding. This study was supported by the German Diabetes Society (Deutsche Diabetes Gesellschaft) and the Federal Ministry of Education and Research (Berlin, Germany), integrated into the German Center for Diabetes Research (DZD) (FKZ 82DZD14A02). This project received funding from the Innovative Medicines Initiative 2 Joint Undertaking under grant agreement no. 115797 (INNODIA project).
Duality of Interest. Boehringer Ingelheim Pharma GmbH & Co. KG (contacts: Dr. Jan Marquard, Dr. Stefan Kaspers) provided support for this study. No other potential conflicts of interest relevant to this article were reported.
Author Contributions. H.K. and J.S. wrote the manuscript and analyzed the data. S.L., J.R., and R.W.H. researched and analyzed the data and reviewed and edited the manuscript. W.Ka., D.B., J.K.M., S.Z., S.M.S., M.H., and W.Ke. contributed to discussion and reviewed and edited the manuscript. S.L. 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.
H.K. and J.S. contributed equally to this work.