Decreased Incidence of Type 1 Diabetes in Young Finnish Children

Type 1 diabetes is an immune-mediated disease in which genetic predisposition, combined with environmental factors, lead to autoimmune destruction of the insulin-producing β-cells in the pancreatic islets. A preclinical period of variable length, marked by the appearance of diabetes-associated autoantibodies, precedes clinical symptoms and subsequent diagnosis. Type 1 diabetes is one of the most common chronic illnesses of childhood, particularly in Finland, where the world’s highest disease incidence has been repeatedly recorded (1–4). Alarmingly, the worldwide incidence of type 1 diabetes has been on the rise, particularly in young children (1,2,5–7). This has previously been the case also in Finland. The incidence rate (IR) in Finnish children younger than 15 years of age doubled between 1980 and 2005, with the greatest annual increase of 4.7% in children younger than 5 years of age, and the incidence was predicted to double again by 2020 (8). However, there was a reduction in the rate of increase in some high-incidence countries between 2004 and 2013 (5,9,10), and a plateau in IR was observed between 2006 and 2011 in Finland (11). We set out to evaluate recent trends in the incidence of type 1 diabetes in Finnish children.

In this study, children younger than 15 years of age, who had been diagnosed with type 1 diabetes between 2003 and 2018, were identified from the Finnish Pediatric Diabetes Register (12). Data for this register have been collected since 2002, and annually >90% of Finnish children with newly diagnosed type 1 diabetes participate in the register (13). We excluded children younger than 6 months of age from the analyses because their diabetes is likely to represent monogenic disease (14). The sex distribution of this register and of all Finnish cases is comparable (8), implying no potential sex bias in the current analyses. The study protocol for the Finnish Pediatric Diabetes Register was approved by the Ethics Committee of the Hospital District of Helsinki and Uusimaa (Helsinki, Finland). To determine the appropriate population for the analyses, annual age- and sex-specific population data were obtained from the Finnish Population Register Center.

We calculated overall, sex-specific, and age-specific IRs of type 1 diabetes per 100,000 person-years (PY). The three age categories comprised children aged 0.50–4.99 years, 5.00–9.99 years, and 10.00–14.99 years. To better account for random fluctuation in the IRs, we aggregated data into four 4-year periods (2003–2006, 2007–2010, 2011–2014, and 2015–2018), for which we calculated the overall, age-specific, and sex-specific IRs. The exact 95% CIs were assessed by assuming Poisson distributed rates. To better quantify possible changes over time, we used the 2003–2006 period as a reference and assessed IR ratios (IRRs) and their 95% CIs by fitting multiplicative Poisson regression models to the numbers of children diagnosed with type 1 diabetes, when using the ln of PYs as an offset term. In addition, we visually inspected the distribution of age at diagnosis by the above-mentioned 4-year periods and tested for possible differences between the periods. Owing to the nonnormality of these data, we calculated medians and interquartile ranges (IQRs) of age at diagnosis and used the Kruskal-Wallis test. For all the analyses, a two-tailed P value of <0.05 was considered significant. We also performed sensitivity analysis to examine the consistency of the results with respect to age division. We used a division into two age categories, 0.50–6.99 and 7.00–14.99 years. This division is in line with the concept of different age-related endotypes of type 1 diabetes (15).

Data may be available upon request from the corresponding author (M.K.), but will be subject to ethical and legal considerations.

Based on the Finnish Pediatric Diabetes Register, we determined that 7,871 children younger than age 15 were diagnosed with type 1 diabetes in Finland between 2003 and 2018. Among these children, 4,417 (56.1%) were boys and 3,454 (43.9%) were girls. The overall IR/100,000 PY during the study period was 54.9 (95% CI 53.7–56.1) and was 60.3 (95% CI 58.5–62.1) in boys and 49.2 (95% CI 47.6–50.9) in girls. The annual IRs of type 1 diabetes fluctuated considerably both within and between the three age-groups (Fig. 1). In the whole study cohort, the highest IR/100,000 PY (62.4 [95% CI 57.3–67.7]) was seen in 2006, and the nadir (48.2 [95% CI 43.7–52.9]) occurred in 2013 (Supplementary Fig. 1). For the division into three age-groups, the age-specific IRs/100,000 PY during the study period were 46.0 (95% CI 44.1–48.0) in the youngest age-group (boys 47.7 [95% CI 44.9–50.5], girls 44.2 [95% CI 41.5–47.0]), 62.1 (95% CI 59.9–64.4) in the middle age-group (boys 66.7 [95% CI 63.5–70.0], girls 57.4 [95% CI 54.2–60.4]), and 56.4 (95% CI 54.3–58.5) in the oldest age-group (boys 66.1 [95% CI 62.9–69.3], girls 46.2 [95% CI 43.5–49.0]). The overall, age-, and sex-specific IRs/100,000 PY and IRRs for the 4-year periods using the division into three age-groups are summarized in Table 1 and visually presented in Supplementary Fig. 2.

Data on the three age-groups, presented in Table 1, show that the overall IR/100,000 PY decreased from 57.9 in 2003–2006 to 52.2 in 2015–2018, resulting in an IRR of 0.90 (95% CI 0.85–0.96, P = 0.001). The most apparent decline was seen in the youngest children (IRR 0.77; P < 0.001) and to a lesser degree in the middle age-group (IRR 0,86; P = 0.002), whereas no change was observed in the oldest age-group. In the youngest age-group, the decline was significant in both sexes, though more pronounced in boys. In the middle age-group, a decrease was observed only in girls. In sensitivity analysis with the division into two age-groups, we observed similar results for the youngest age-group (see Supplementary Table 1). There was a significant decrease in the IR among both boys and girls younger than the age of 7 years and no significant change in the IR in those aged 7–14 years.

The median age at diagnosis was 8.02 years (IQR 4.67–11.24), 8.28 (IQR 4.90–11.59) in boys and 7.71 (IQR 4.44–10.83) in girls. The median age at diagnosis increased from 7.88 years in 2003–2006 to 8.33 years in 2015–2018. We observed significant differences in the distribution of age at diagnosis between the 4-year periods analyzed in the whole study cohort and in boys but not in girls (Supplementary Fig. 3). Based on visual inspection of the distribution of age at diagnosis, we noted the most apparent change appeared in boys younger than 5 years of age, among whom a noticeable drop occurred in the frequencies over time (Fig. 2).

In this observational register-based study on Finnish children younger than 15 years of age, the overall IR of type 1 diabetes decreased during the 16-year study period between 2003 and 2018. The observed change was mainly due to the decrease in disease incidence among the youngest children. The decrease was consistent both in those aged <5 years and <7 years, and in both sexes. No consistent changes in disease incidence were seen among older children. An increase in median age at diagnosis was also observed, which, when stratified by sex, was statistically significant in boys only.

Our current findings were unexpected in light of previous reports from Finland and other countries showing that the incidence of type 1 diabetes is rising (1,2,5,6,8), or at most, reaching a plateau (9–11,16,17). In addition, the previously reported increase in incidence has often been especially prominent in children younger than 5 years of age, thus leading to a younger age at diagnosis (1,2,6,8).

Recently, the endotype concept has been introduced in type 1 diabetes research to describe disease heterogeneity based on underlying biological mechanisms (18). The Exeter group has generated data suggesting that children diagnosed younger than the age of 7 years differ from those diagnosed at an older age with respect to immune profile (19) and proinsulin processing in their pancreatic islets (15). Irrespective of the division of the cohort by age at diagnosis into three (<5, 5–9, and 10–14 years) or two age-groups (<7 and 7–14 years), we observed that the IR of type 1 diabetes decreased over the observation period in the youngest age-group. We also found a decreased IR in girls aged 5–9 years but not in boys, whereas no changes were seen in either sex aged ≥7 years. Apparently, the changes observed in the current study are related to age, but our findings cannot be attributed to the endotype concept based on the age division alone. However, age divisions other than the traditional one (<5, 5–9, and 10–14 years) should be considered in epidemiological studies, especially in hypothesis-generating research.

Changes in the IR of type 1 diabetes have mainly been attributed to changes in exposure to immunomodulatory environmental factors affecting progression toward islet autoimmunity. Many environmental factors have been implicated, but none have been conclusively proven to have a causal effect. Identification of environmental factors affecting the incidence of type 1 diabetes is of key importance, considering its possible implications for disease prevention. Even though the nature of our study allows only speculation on factors having temporal association with the observed decrease in incidence, because the register does not provide data on exposure to environmental factors, the current observation may provide clues for the quest toward determining causality.

Our current findings imply that environmental factors affecting individuals early on in life, and having undergone changes in exposure level during the study period, may be associated with the decrease in incidence observed among young children. Because such environmental factors are comparatively restricted, it may be possible to investigate causality for the major changes having affected this part of the population. Such factors include increased intake of vitamin D after the policy-based fortification of milk products with vitamin D since 2002 and a doubling of the fortification dose in 2010 (20–22), decreased Calmette-Guérin vaccine coverage (from >98% to 6%) after changes in the National Immunization Program in 2006 regarding the vaccine’s target population (23,24), decreased exposure to rotavirus infections after commercial availability of the rotavirus vaccine since 2006 and introduction of the vaccine to the National Immunization Program in 2009 (25–30), and increased use of peroral probiotics by infants and young children during the study period (31–33).

The current nationwide study was performed in the country with the highest disease incidence globally. A limitation of our study is that our data source, the Finnish Pediatric Diabetes Register, does not cover all cases of type 1 diabetes in Finnish children, and we do not have continuous adherence and surveillance data for the study period. The register does, however, cover >90% of cases (13) and is based on uniform diagnostic criteria allowing for reliable distinction of type 1 diabetes from other forms of diabetes, thus controlling for ascertainment bias. Differential misclassification leading to bias is also unlikely, because the diagnostic criteria and register protocol both remained unchanged during the study period. Another limitation, which is common to studies on the incidence of type 1 diabetes (34), is that the register does not provide information on those who have been diagnosed at age ≥15. Likewise, given that the decrease in the incidence of type 1 diabetes appeared recently, the current study provides no information on further, possibly ongoing, changes. Therefore, it remains uncertain whether the total IR is truly decreasing in Finland or whether the disease manifestation has merely shifted to an older age. Follow-up studies on the incidence trends of type 1 diabetes in high-incidence countries are required to assess whether the tide of incidence has truly turned or whether the observed decrease merely represents a temporary fluctuation or regression to the mean.

Acknowledgments. The authors thank all of the participants, hospitals, and personnel of the Finnish Pediatric Diabetes Register.

Funding. The study was supported by the Academy of Finland (decision number 292538), the Sigrid Jusélius Foundation, Finska Läkaresällskapet, and the Liv and Hälsa Fund.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. A.P. analyzed the data and wrote and edited the manuscript. A.B. performed part of the statistical analyses and reviewed and edited the manuscript. H.S. reviewed and edited the manuscript. M.K. planned the study, contributed to the discussion, and reviewed and edited the manuscript. All authors approved the final manuscript. M.K. 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.