Our single-center retrospective cohort study was conducted at the Bókay Unit of the Pediatric Center, Semmelweis University, Budapest, Hungary. This center is responsible for the care of approximately 25% of all T1D children and adolescents in the country. Before exclusion, we obtained all patients’ data with diabetes mellitus (DM) who received insulin treatment and had at least one visit between 1^st of January, 2013 and 31^st of December, 2022 at the inpatient and/or outpatient Endocrinology and Diabetes Unit of our Center. Clinical data and laboratory results were obtained from Semmelweis University’s e-MedSolution software. Medical records were retrieved based on the BNO classification system, the Hungarian adaptation of the ICD-10 (The International Statistical Classification of Diseases and Related Health Problems, 10^th Revision) classification released by the World Health Organization (https://icd.who.int/browse10/2019/en, accessed: 09.10.2024).

Data was completed and verified by four medical doctors and two medical students in agreement with our pre-specified research protocol. All arising questions were discussed in detail within the data extraction team.

Our study was conducted in accordance with the World Medical Association’s Declaration of Helsinki. Due to the retrospective nature of our study, ethical approval was not required.

Inclusion criteria: all patients between 0 to 21 years of age (hereafter referred to collectively as “children”), who had been treated with T1D at our center between 2013 and 2022, and had at least one simultaneous ATPO and ATG measurements during our study period. For the purposes of this study, we have chosen 21 years as the upper age limit, consistently with our clinical practice, where care is provided to individuals up to 21 years of age. Data on TA were obtained from both newly diagnosed and follow-up cases.

Exclusion criteria: patients over 21 years of age, children with diabetes forms other than T1D, and those who never underwent TA screening throughout the duration of our study.

We collected the following parameters: clinical data of children (sex, date of birth, T1D diagnosis date), visit dates, thyroid autoantibody (ATPO, ATG) and TFTs (thyroid-stimulating hormone/TSH/, free thyroxine/fT4/) results from all years when the patient was screened as well as thyroid ultrasound findings for antibody-positive patients. Mean ages and time since T1D diagnosis were calculated according to the mid-point of each calendar year (1^st of July). Laboratory analysis were carried out at the Immunology Laboratory of Semmelweis University with assay methodology and normal ranges shown in Table 1 . If a child had more than one laboratory examination of a parameter in a calendar year, we used the last data for our analysis (or the last one before they started taking thyroid medicines in case of TSH and fT4). TSH levels above 6 mU/L or below 0.1 mU/L, and/or results that were consistently abnormal in a patient, were considered clinically relevant. TA was defined by at least one positive ATPO and/or ATG laboratory result obtained during our study period. Assay methodology and ranges of normal values of autoantibodies had changed multiple times over the examined period, therefore we calculated both absolute and relative values. Thyroid ultrasounds were deemed positive for thyroiditis if they showed enlargement and/or inhomogeneity and/or hyperemia in the thyroid gland. AIT was diagnosed if at least one autoantibody was positive and thyroiditis was confirmed by ultrasound examination.

Screening frequency of children was primarily based on ISPAD’s guidelines (18), therefore patients without thyroid comorbidity did not go through autoantibody measurements each year. However, if a patient had presented a positive thyroid autoantibody result, they had been examined in all years afterwards. Due to the different screening frequencies between positive and negative children, prevalence data would be unrealistically high if not adjusted. Thus, in years when no laboratory examinations were carried out, we used the following protocol:

If the antibody titers were negative in both the preceding and subsequent years’, we considered the intermediate year to be negative as well. When both were positive, we recognized it as positive. If they differed, the intermediate year was administered as missing data.

We defined the pre-pandemic period as 2013 to 2020 and the pandemic period as 2021 and 2022. In the year of 2020, we did not expect a significant impact of the pandemic on our children, as a large Hungarian national representative study indicated that there were very few pediatric COVID-19 cases during the first pandemic wave (19).

Statistical analysis was carried out by a professional biostatistician. Each year, we calculated the prevalence using data from all children under care at that time. Consequently, the data for most children are included in our database for multiple years, corresponding to each year they were under care at our center. To calculate and compare prevalence data, and perform descriptive statistics, MedCalc Statistical Software version 22.023 was used (MedCalc Software bv, Ostend, Belgium; https://www.medcalc.org; 2020). Chi square comparison were conducted using StataCorp. (Stata Statistical Software: Release 18.5). The visualization was created using Microsoft Excel (Microsoft Corporation, 2016).

Between January 1, 2013 and December 31, 2022, a total of 1,667 patients with insulin-dependent DM were treated at our Clinic. We excluded 14 adults over 21 years old, 65 children who had another condition rather than T1D and 227 patients due to missing data and/or who were treated primarily at other centers. All in all, we included 1,361 T1D children’s data in our final analysis. Mean (± SD) follow-up period was 4.7 (± 2.8) years. The girl-to-boy ratio was 637:724 (46.8% vs 53.2%). Detailed annual data are provided in Table 2 .

The overall prevalence of thyroid autoimmunity among our T1D children was found to be 22.8% ([20.3;25.5], 310 children). Girls were significantly more affected. (TA group: 106 boys, 204 girls, non-TA group: 618 boys, 433 girls; p<0.001, RR: 1.60 [1.43;1.78]).

From 2013 to 2022, the prevalence of TA increased from 15.9% to 20.6% (p=0.041). We observed this increase during the pre-pandemic years (2013 to 2019) but not in the COVID-19 era. Whilst analyzing the yearly prevalence differences, we found no significant difference between any consecutive years. Annual prevalence rates with confidence intervals (CI) are shown in Figure 1 .

Of the children with TA who were screened with ultrasound at least once (n=260), 208 (80.0%) showed signs of thyroiditis on ultrasound, representing 67.1% of all children with TA. This resulted in an overall AIT prevalence of 15.3% in our study group. Among the 208 children with ultrasound-proven thyroiditis, one was later diagnosed with papillary thyroid carcinoma, as was one other child who did not show signs of thyroiditis. Although we performed an increasing number of ultrasounds throughout the years, the positivity rate remained stable. Yearly ATPO, ATG and ultrasound positivity rates can be seen in Table 3 .

Given the multiple changes in assay methodology of ATPO and ATG titers during our examined period, we calculated both absolute and relative values in case of both antibodies. Detailed results of autoantibody titers are provided in Supplementary Tables 1 , 2 .

We also analyzed the rate of thyroid hormone changes and data related to thyroid medication over this 10-year period in our children with TA. From the TFTs’ analysis, we excluded children who did not have at least one TSH result during the study period (n=8). For children who were prescribed thyroid medication, we included only data preceding the initiation of treatment. Therefore, children who started their medication before 2013 (the beginning of our study period) were also excluded (n=20). Among the included 282 children, 44 (15.6%) exhibited clinically relevant elevation in TSH levels, while eight (2.8%) showed TSH reduction during the study period. Increased fT4 levels were observed in four cases, while decreased fT4 levels were noted in two children. The annual results are presented in Table 4 . Detailed description of TSH and fT4 values can be seen in Supplementary Table 3 .

Among the 310 children with TA, 74 (23.9%) were treated for hypo- and/or hyperthyroidism. Of these, 63 were initially prescribed levothyroxine to manage subclinical or clinical hypothyroidism, while two required levothyroxine following a total thyroidectomy due to thyroid carcinoma. Additionally, nine children were originally treated with antithyroid (thyrostatic) medications (thiamazole or propylthiouracil). During our study period, three children required treatment with both types of medication. Two of them began with thiamazole and later transitioned to levothyroxine due to hypothyroidism, while one child started on levothyroxine and subsequently began taking thiamazole.

Altogether, out of the 302 children with TA who had at least one TSH result during the study period, 86 (28.5%) had clinically relevant TSH abnormalities and/or were prescribed thyroid medication.

We examined the connection between ultrasound positivity and thyroid function abnormalities among the children with at least one ultrasound and one TSH result (n=258). From this analysis, we excluded the two aforementioned children with papillary thyroid carcinoma, as they had started hormone replacement therapy following total thyroidectomy. The contingency table summarizing these data is presented as Table 5 . Children with ultrasound-proven thyroiditis had a significantly elevated risk of having TSH abnormalities and/or being prescribed thyroid medication (p<0.001, RR: 6.24 [2.05;18.98]).

The main strength of our study is that it covers ten years of annual data on more than 1,300 children allowing us to assess the prevalence of TA longitudinally. The decade-long study period enables for a comprehensive assessment of trends in TA over time. By analyzing TA prevalence on a year-by-year basis, the study provides a more nuanced understanding of the trends compared to pooled data analysis. It has the ability to reflect on the increasing trend during a long-term period, while with the comparison of pre-pandemic and pandemic groups, one might assume that the difference between the pooled prevalences could be attributed to the impact of the pandemic. The inclusion of both pre-pandemic and pandemic years adds an important dimension to the understanding of the trends before and during the COVID-19 pandemic. Additionally, given that our center is responsible for the care of about 25% of all T1D children in Hungary and is one of the largest in Central Europe, we believe that our findings are representative of this region.

As this was a retrospective study, we encountered instances of incomplete or missing data in certain cases. Possible reasons for these were missed annual check-ups by some patients, variations in protocols among specialists, overlooked screenings or insufficient blood sample volumes. Furthermore, the documentation regarding medication may have been inconsistent. Moreover, the laboratory test method used for the measurement of ATPO and ATG changed multiple times during our study period, thus we needed to present relative values of ATPO and ATG besides absolute ones. Unfortunately, as to our experiences, neither of them is ideal for precise comparison, nevertheless we indicated both results. Finally, it should be noted, that autoimmune diseases may develop months or years after an infection, so our study (ending in 2022) might not capture the long-term effects of the pandemic.

In our retrospective cohort study, we found an increase in the prevalence of thyroid autoimmunity among children with type 1 diabetes from 2013 to 2022. Unexpectedly, this growth was remarkable during the studied pre-pandemic period and it just stopped during the COVID-19 pandemic years. Therefore, our results among T1D children do not support the assumed role of SARS-CoV-2 in the development of thyroid autoimmunity.