Association of COVID-19 and Development of Type 1 Diabetes: A Danish Nationwide Register Study

In the fall of 2020, the U.S. Centers for Disease Control and Prevention reported an increased risk of diabetes within 30 days of a severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection in patients aged <18 years based on data from health insurance claims (1). However, the type of diabetes was not specified, and the incidence was evaluated using health insurance claims. Additionally, there were reports of increased rates of type 1 diabetes (T1D) diagnoses and diabetic ketoacidosis admissions among European pediatric populations during the coronavirus disease 2019 (COVID-19) pandemic (2–4). A study based on registries from 13 countries investigated time trends of diabetic ketoacidosis and found an increase in prevalence of diabetic ketoacidosis during the COVID-19 pandemic (5). Given the known association between other viral infections, especially other coronaviruses, and β-cell islet damage, it is possible that the observed increase in incidence and prevalence was causal (6). A large cohort study in Scotland used the nationwide Scottish Care Information–Diabetes registry to investigate a possible association between T1D and SARS-CoV-2 infection. This study found a positive association between T1D and SARS-CoV-2 infection within 30 days but no association after 30 days. However, the association was suspected to be due to a possible detection bias, as there were increased numbers of both positive and negative SARS-CoV-2 tests around the time of T1D presentation (7). Given the importance of replicating these findings in countries with widely available and validated nationwide health registers and the potential public health implications of a possible association between T1D and SARS-CoV-2 infection, we aimed to replicate the study. First, we investigated whether the incidence of T1D in individuals aged <30 years has increased during the COVID-19 pandemic in Denmark compared with previous years. Second, we investigated whether a positive SARS-CoV-2 test result was associated with an increased risk of developing T1D by creating a cohort of people testing positive for SARS-CoV-2 matched by age, sex, and vaccination status to three people not yet infected with SARS-CoV-2.

In Denmark, every citizen has equal access to health care, including primary and hospital care, and complete follow-up is possible through the use of identification numbers that allow for the cross linking of health and administrative databases at the individual level. For this study, data were collected from several nationwide registries: the Danish National Patient Registry, which contains information on all hospital admissions and outpatient contacts since 1977; the Danish National Prescription Registry, which holds information on all medicine prescriptions dispensed from a pharmacy since 1995; the Danish Cause of Death Registry, which contains information on date, cause, and place of death since 1970; the Danish Population Registry, which contains information on date of birth and sex; and the Danish Microbiology Database, which provides information on all SARS-CoV-2 PCR test results and vaccination status for the entire Danish population. Laboratory testing for SARS-CoV-2 was initially limited to hospitalized patients with severe respiratory illness during the early months of the COVID-19 pandemic in Denmark. Later, during the summer months of 2020, SARS-CoV-2 PCR tests became widely available to all individuals, regardless of symptom status and without the need for a referral from a medical professional, and patients were regularly tested for SARS-CoV-2 prior to any hospital contact.

The Danish government implemented a strict lockdown policy starting on 13 March 2020, which included the closure of schools, kindergartens, restaurants, and shopping malls, as well as a restriction on social gatherings to a maximum of 10 people (8). These measures were partially lifted in mid-April with the reopening of certain businesses and were further relaxed in mid-May and early June, allowing for larger social gatherings and the reopening of additional businesses (8). However, some of these measures were reintroduced in the autumn of 2020 and relaxed during the summer of 2021. By November 2021, the mask mandate was reinforced, but by February 2022, most restrictions were lifted, and COVID-19 was no longer regarded as a critical threat (8). Nationwide vaccination programs were initiated on 27 December 2020 and suspended in April 2022 (8).

All permanent residents in Denmark aged <30 years or individuals born after 1 January 2015 were included in the period 1 January 2015 to 31 December 2021. People with any prior diagnosis of diabetes, prior use of immunomodulating therapies, cancer, or diagnosis of autoimmune disease (celiac disease, psoriasis, Morbus Crohn disease, colitis ulcerosa, hypothyroidism, hyperthyroidism, or multiple sclerosis) were not included. To ensure the use of information on past exposure and covariates, people who immigrated to Denmark after 1 January 2010 were not included.

A positive SARS-CoV-2 PCR test result was used to define all cases in the cohort. Only first-time positive PCR tests were included.

Concomitant medical therapy was defined as the presence of at least one redeemed prescription 6 months prior to the inclusion date. Comorbidities were defined as the presence of an ICD-8 or ICD-10 diagnosis in hospital discharge records, including ambulatory contacts up to 10 years prior to inclusion date. The complete list on ICD-8, ICD-10, and Anatomical Therapeutic Chemical codes used to describe comorbidities are listed in Supplementary Table 1.

The outcome was incident T1D defined as a first T1D diagnosis and/or a hospital admission for ketoacidosis. T1D diagnosis was defined as first use of insulin at an age <30 years and/or any first hospital contact with a discharge diagnosis of T1D. The study population was followed from inclusion until the outcome of interest, the age of 30, end of study (31 December 2021), emigration, type 2 diabetes (T2D) development, date of first cancer, date of initiating immunomodulating therapy, or development of autoimmune disease (celiac disease, psoriasis, Morbus Crohn disease, colitis ulcerosa, hypothyroidism, hyperthyroidism, or multiple sclerosis), whichever came first. We followed patients only to 30 years of age because we wanted to reduce the potential misclassification of T1D and T2D that might arise in older populations.

Patient characteristics were summarized at year of T1D diagnosis. Continuous variables are presented as medians with interquartile ranges (IQRs) and categorical variables as counts with percentages. Differences among the covariates were tested using the χ² test for categorical variables, Mann-Whitney U test for nonnormally distributed continuous variables, and t test for normally distributed continuous variables. Median follow-up time was estimated using the reverse Kaplan-Meier method. Crude incidence rates for the four quarters (January–March, April–June, July–September, and October–December) of the years 2015–2021 were calculated per 100,000 person-years along with crude incidence rate ratios (IRRs), and the exact Poisson method was used to calculate 95% CIs. A multivariable time-dependent Poisson regression model was fitted to compare the adjusted IRR of T1D in the four quarters of the years 2021 and 2020, with 2015–2019 as the reference. The model was adjusted for age-group (<15 years or >15 years), calendar year, and sex. To estimate potential differences in the IRR of T1D among subgroups of age and sex, we stratified the cohort according to prespecified subgroups of sex (male, female) and age-groups (<15 years, >15 years). All estimates were reported with 95% CIs. To investigate the temporal relationship between exposure (SARS-CoV-2 infection) and outcome (T1D), we performed a matched cohort study according to the directed acyclic graph shown in Supplementary Fig. 1. After excluding ineligible individuals based on the exclusion criteria outlined above, we used exposure density sampling to match all individuals testing positive for SARS-CoV-2 on the date of the positive PCR test result with three control individuals not yet having tested positive for SARS-CoV-2. We also matched on sex, birth year, and vaccination status prior to match date. If one or more of the three control individuals subsequently tested positive for SARS-CoV-2, we censored the entire risk set at the earliest date of a positive SARS-CoV-2 test result. The cohort was followed from date of match until date of T1D, first date of positive SARS-CoV-2 test result among the control individuals, development of T2D, age 30 years, emigration, development of autoimmune disease, cancer, end of study (31 December 2021), or all-cause death, whichever came first. Median follow-up time was estimated using the reverse Kaplan-Meier method. To estimate the 30-, 90-, and 180-day risk of T1D following a positive SARS-CoV-2 test result, the Aalen-Johansen estimator was applied, with all-cause death as a competing risk. A cause-specific Cox regression model was applied to estimate the hazard ratio (HR) of T1D among individuals testing positive for SARS-CoV-2 compared with those with no positive SARS-CoV-2 test result. To estimate the HR within the first 30 days and after the 30 days, we constructed two landmark cohorts. The first cohort comprised all individuals alive at the time of inclusion, and this was followed until day 30, date of T1D diagnosis, death, or end of study, whichever came first. The second cohort comprised all individuals who were event free and alive on day 31, and they were again followed to the date of T1D, death, or end of study. To account for possible detection bias and reverse causality, we repeated the first cohort and excluded individuals with T1D who were diagnosed with SARS-CoV-2 on the same day of T1D diagnosis. All analyses were performed using R version 3.6.1 statistical software (R Foundation for Statistical Computing) (9).

Register-based research is granted permission by Danish law to be performed without the requirement of ethical approval or informed consent. This project was approved by the Knowledge Centre on Data Protection Compliance–The Capitol Region of Denmark (approval no. P-2010-191).

The data sets generated during and/or analyzed in the current study are not available because of the data protection policies issued by Statistics Denmark.

A total of 2,381,348 individuals were included between 1 January 2015 and 31 December 2021 (Fig. 1). The median age of the population at time of inclusion was 12.8 years (IQR 3.5–21.1), and the population consisted of 51.5% males. The median follow-up time was 7.0 years (IQR 4.2–7.0), and the entire population contributed 13,365,574 person-years of observation. Characteristics of the patients with T1D according to year of diagnosis are shown in Table 1. Patients diagnosed with T1D in 2020 and 2021 were similar with regard to age of diagnosis, diagnosis, and mode of diagnosis. During follow-up, 3,579 individuals developed T1D, of whom 571 were diagnosed first by a ketoacidosis hospitalization, yielding an overall rate of T1D and ketoacidosis admission of 26.8 (95% CI 25.9; 27.7) and 4.3 (4.0; 4.7) per 100,000 person-years, respectively. The crude rates of T1D according to annual quarter in age-groups 0–15 years and 16–30 years are shown in Supplementary Fig. 2. The adjusted IRRs of T1D in each quarter of years 2020 and 2021, with 2015–2019 as the reference, are shown in Fig. 2. When stratifying the cohort by age and sex, we found that the increase in IRR observed for the full cohort in April–June 2021 was mainly driven by an increase in incidence among females aged 0–15 years and males aged 16–30 years (Supplementary Fig. 3).

A total of 338,670 individuals with a positive SARS-CoV-2 test were matched with 1,004,688 control individuals. Baseline demographic characteristics are shown in Supplementary Table 2. The two groups were similar in terms of age, sex, and vaccination status prior to the match date. During a median observation period of 42 days (IQR 11–225), 130 individuals developed T1D (30 in the SARS-CoV-2–positive group, 100 in the SARS-CoV-2–negative group) of whom 25 (3 in the SARS-CoV-2–positive group, 22 in the SARS-CoV-2–negative group) were diagnosed by admission to the hospital with ketoacidosis. The risk of developing T1D 30, 90, and 180 days after testing positive for SARS-CoV-2 was 0.0042% (95% CI 0.0017; 0.0068), 0.0081% (0.0041; 0.0121), and 0.013% (0.0072; 0.018), respectively. The 30-, 90-, and 180-day risk of developing T1D in the SARS-CoV-2–negative group was 0.0022% (0.0012; 0.0033), 0.0058% (0.0038; 0.0079), and 0.012% (0.0084; 0.015), respectively (Fig. 3). The HR of incident T1D following a positive SARS-CoV-2 test result was 0.90 (95% CI 0.60; 1.35) compared with not yet having a positive SARS-CoV-2 test result. When looking into the landmark cohorts of patients who were followed from 0–30 days and from day 31, we found HRs of 1.94 (0.91; 4.14) and 0.69 (0.42; 1.13), respectively. Among the individuals with T1D who tested positive for SARS-CoV-2, a total of 11 (36.7%) cases of T1D were diagnosed in the first 30 days compared with 17 (17.0%) among the individuals with T1D who tested negative for SARS-CoV-2. When excluded individuals diagnosed with T1D the same day as testing positive for SARS-CoV-2–were included in the study, the increased risk observed diminished (HR 1.06 [0.42; 2.69]) (Supplementary Table 3).

In this Danish nationwide study, the incidence of T1D among individuals aged <30 years was found to be increased in the April–June 2021 quarter compared with the same quarter in the period 2015–2019. However, the detected increase in T1D incidence during the COVID-19 pandemic could not be explained by SARS-CoV-2 infection itself, as there was no significant association found between having a positive SARS-CoV-2 test result and T1D after excluding cases diagnosed with a positive SARS-CoV-2 test on the day of T1D diagnosis. This study aimed to investigate the potential link between a positive SARS-CoV-2 test result and the development of incident T1D, as an overlooked causal relationship could have significant socioeconomic and public health consequences. The rising incidence of T1D during the COVID-19 pandemic has been reported in several European countries, but the findings are not entirely consistent. Large Finnish, Scottish, Romanian, and German studies have identified an increased incidence of T1D during the COVID-19 pandemic compared with prepandemic years (2,4,7,10). Conversely, studies from Saudi Arabia and Italy found no increase in T1D during the COVID-19 pandemic compared with earlier years (11,12). The difference in reporting could be a result from differences in study design, data resources available, and duration of observation time. Interestingly, studies using the same data resources as in the German study, but with a shorter observation time, showed no increase in T1D during the COVID-19 pandemic compared with previous years (3,10). Although the increase in incidence of T1D could indicate the presence of an association, and the known β-cell islet damage associated with other types of coronaviruses could lead one to expect a causal link, we found no association in our study. To ensure the temporal relationship, we constructed a matched cohort and tried to account for all possible measured confounders either by restriction or matching (age, sex, vaccination status). We matched the cohort on the date of the first positive SARS-CoV-2 test result, ensuring that information on exposure was present prior to information on outcome. The median observation time was 42 days, and the number of events was very low, yielding very-low 30-, 90-, and 180-day risks. Overall (including the HRs), the estimates had very wide CIs and, therefore, a high degree of uncertainty in the estimates. The HR for developing T1D was reduced (although not significantly) for individuals with a positive SARS-CoV-2 test result. A Scottish study with similar nationwide coverage and inclusion criteria found similar HRs for individuals with a positive SARS-CoV-2 test result >30 days earlier to being diagnosed with T1D. By stratifying our analyses according to the duration of time until the diagnosis of T1D, our results mirrored those of the Scottish study. Specifically, we identified an elevated risk of T1D (i.e., among those diagnosed within 30 days of T1D onset). Conversely, we identified a decreased risk of T1D among individuals diagnosed after 30 days. As depicted in Fig. 3, the percentage of cases in which SARS-CoV-2 infection was present does not begin at 0%, suggesting that T1D cases were being diagnosed even before the onset of the COVID-19 pandemic. Moreover, development of T1D is most often a latent, progressive process that takes >30 days. When we excluded these cases, the increased risk we observed during the first 30 days was no longer apparent. It is important to note that since the onset of the COVID-19 pandemic, it has been mandatory for all individuals seeking medical attention to be tested for SARS-CoV-2 infection prior to being admitted to the public health care system. As a result, individuals presenting with ketoacidosis or newly onset T1D would be tested for SARS-CoV-2 regardless of whether they were exhibiting symptoms and, thus, could lead to an increase in the incidence of SARS-CoV-2 infection among individuals diagnosed with T1D. Additionally, the majority of our cohort comprised individuals aged <15 years, and young children are known to be more likely to be asymptomatic carriers of SARS-CoV-2, which may have contributed to a detection bias. However, it is important to note that the presence of asymptomatic or randomly detected SARS-CoV-2 infection at the time of T1D diagnosis does not necessarily rule out β-cell islet damage as a contributing factor in the development of T1D. The estimated time from infection to new-onset T1D is ∼25 days, and if such an association exists, we would expect to observe it even when excluding individuals diagnosed with SARS-CoV-2 on the same day as their T1D diagnosis (13). The increase in incidence could also be explained by other factors related to COVID-19 containment measures that we were unable to include in our study. Such factors may include delayed contact with the health system or suppression of other viral infections during the lockdown periods and then a reactive surge after lockdown restrictions were lifted.

The strengths of this study are the nationwide completeness of data and the number of patients included. We used well-validated measures to assess exposure and outcome (14–17), but although we relied on diagnosis codes of T1D, hospital admission for ketoacidosis, and prescriptions of insulin, we lacked information on important biochemical markers of insulin production, which could have strengthened the definition of T1D. The testing capacity in Denmark was reduced in the beginning of the pandemic, and new-onset T1D cases occurring after SARS-CoV-2 infection might have been missed, making the study susceptible of exposure misclassification and ultimately biasing the results toward a null finding. More importantly, we lacked information on the presence of other viral infections, as there is a known association between certain viral infections and the development of T1D. We have presented an association, not causation, and residual confounding cannot be ruled out since we lacked important information on behavioral risk factors, such as adherence to containment measures, family history of diabetes, and other environmental triggers.

In conclusion, we observed an increase in T1D incidence during April–June 2021 compared with April–June 2015–2019. However, this observation could not be explained by a SARS-CoV-2 infection alone, as no statistically significant association was found between a positive SARS-CoV-2 test result and development of T1D.

Funding. This project was funded by Nordsjaellands Hospital, The Capitol Region of Denmark.

Duality of Interest. U.P.-B. has served on advisory boards for Sanofi, Novo Nordisk, and Vertex and has received lecture fees from Abbott, Sanofi, and Novo Nordisk. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. B.Z. analyzed the data and drafted the manuscript. B.Z., C.T.-P., and R.L.M.N. researched data, contributed to the discussion, and reviewed and edited the manuscript. B.Z. and R.L.M.N. designed the study. K.K.S., P.A.E., T.K.F., P.L.K., M.E.L., and U.P.-B. contributed to the discussion and reviewed and edited the manuscript. All authors approved the final version of the manuscript. B.Z. 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.