To explore the incidence of type 1 diabetes in children in relation to exposure to rotavirus infections.
A nationwide register-based ecological study on the 1995–2015 birth cohorts in Finland compared those born before and after the national implementation of the rotavirus vaccine in 2009.
When the prevaccine 2001–2005 birth cohorts were compared with the postvaccine birth cohorts, the number of children exposed to rotavirus infection by the age of 5 years decreased from 2,522 per 100,000 children (2.5%) to 171 per 100,000 children (0.2%), while the incidence of type 1 diabetes in those aged <5 years decreased from 71.5 to 54.4 per 100,000 person-years (incidence rate ratio 0.79, 95% CI 0.71–0.86).
At the population level, a decrease in exposure to rotavirus infections was associated with a decrease in the incidence of type 1 diabetes in young children.
Introduction
Rotavirus infections may be a trigger of islet autoimmunity leading to type 1 diabetes (1–4), and some studies have reported an association between national implementation of the vaccine and a decrease in the type 1 diabetes incidence in children younger than the age of 5 years (5–7). In Finland, rotavirus vaccines became commercially available in 2006 and were introduced into the Finnish vaccination program in July 2009, and the incidence of type 1 diabetes decreased after 2010 among children younger than the age of 5 compared with years 2003–2006 (8). We conducted an ecological study on the pre- and postvaccine birth cohorts in Finland to study changes in the diabetes incidence in relation to exposure to rotavirus infections.
Research Design and Methods
Study Design and Setting
In this ecological birth cohort study, we followed the 1995–2015 birth cohorts of 1,085,137 children from birth up to the age of 5 years for exposure to rotavirus infection (laboratory-confirmed infections in the National Infectious Diseases Register [NIDR]) and maximally up to the age of 14 years for the occurrence of type 1 diabetes (Diabetes in Finland [FinDM] database). The population at risk was determined using the public population database of Statistics Finland. The number of rotavirus infections and cases of type 1 diabetes (excluding those diagnosed younger than the age of 6 months as likely representing monogenic disease) were tabulated by birth year, 1-year age-group, and sex. In addition, four birth cohort groups were formed according to the availability and coverage of the rotavirus vaccine: the prevaccine (1995–2000 and 2001–2005), partly vaccinated (2006–2009), and the postvaccine (2010–2015) birth cohorts.
Statistical Analysis
Please see the statistical analysis section in the Supplemental Material.
Results
Of 18,154 children exposed to rotavirus infection younger than the age of 5 years, 14,910 (82%) belonged to the prevaccine birth cohorts (1995–2005). The number of exposed peaked in the 2001–2005 birth cohort group and decreased in children born thereafter, reaching a nadir in the postvaccine 2010–2015 birth cohort group (Supplementary Fig.1). The number of children exposed by the age of 5 years per 100,000 children was 2,522 (2.5%) in the 2001–2005 birth cohort group and 171 (0.2%) in the 2010–2012 birth cohorts, with an absolute reduction of 2,351 per 100,000 children (95% CI 2,291–2,411) (Table 1).
Children exposed to rotavirus infections (i.e., laboratory-confirmed infections) younger than the age of 5 years of the population at risk by birth cohort group and sex (follow-up between 1995 and 2016)
| Sex | Birth cohort group | Exposed (n) | Population (n) | Exposed (%) | Exposed n /100,000 person-years (95% CI) |
|---|---|---|---|---|---|
| All | 1995–2000 | 7,649 | 386,415 | 2.0 | 1,979 (1,936–2,024) |
| 2001–2005 | 7,261 | 287,851 | 2.5 | 2,522 (2,466–2,580) | |
| 2006–2009 | 2,935 | 230,390 | 1.3 | 1,274 (1,229–1,321) | |
| 2010–2012 | 309 | 180,481 | 0.2 | 171 (153–191) | |
| Males | 1995–2000 | 4,181 | 196,961 | 2.1 | 2,123 (2,060–2,187) |
| 2001–2005 | 4,094 | 147,021 | 2.8 | 2,785 (2,702–2,870) | |
| 2006–2009 | 1,635 | 117,723 | 1.4 | 1,389 (1,324–1,457) | |
| 2010–2012 | 178 | 92,309 | 0.2 | 193 (167–223) | |
| Females | 1995–2000 | 3,468 | 189,454 | 1.8 | 1,831 (1,771–1,892) |
| 2001–2005 | 3,167 | 140,830 | 2.3 | 2,249 (2,173–2,328) | |
| 2006–2009 | 1,300 | 112,667 | 1.2 | 1,154 (1,093–1,218) | |
| 2010–2012 | 131 | 88,172 | 0.2 | 149 (125–176) |
| Sex | Birth cohort group | Exposed (n) | Population (n) | Exposed (%) | Exposed n /100,000 person-years (95% CI) |
|---|---|---|---|---|---|
| All | 1995–2000 | 7,649 | 386,415 | 2.0 | 1,979 (1,936–2,024) |
| 2001–2005 | 7,261 | 287,851 | 2.5 | 2,522 (2,466–2,580) | |
| 2006–2009 | 2,935 | 230,390 | 1.3 | 1,274 (1,229–1,321) | |
| 2010–2012 | 309 | 180,481 | 0.2 | 171 (153–191) | |
| Males | 1995–2000 | 4,181 | 196,961 | 2.1 | 2,123 (2,060–2,187) |
| 2001–2005 | 4,094 | 147,021 | 2.8 | 2,785 (2,702–2,870) | |
| 2006–2009 | 1,635 | 117,723 | 1.4 | 1,389 (1,324–1,457) | |
| 2010–2012 | 178 | 92,309 | 0.2 | 193 (167–223) | |
| Females | 1995–2000 | 3,468 | 189,454 | 1.8 | 1,831 (1,771–1,892) |
| 2001–2005 | 3,167 | 140,830 | 2.3 | 2,249 (2,173–2,328) | |
| 2006–2009 | 1,300 | 112,667 | 1.2 | 1,154 (1,093–1,218) | |
| 2010–2012 | 131 | 88,172 | 0.2 | 149 (125–176) |
Birth cohorts 2013–2015 were excluded from the analyses due to incompletely observed age span.
A total of 8,674 individuals were diagnosed with type 1 diabetes before the age of 15 years from a total follow-up of ∼13 million person-years. In the youngest age-group (both boys and girls), the incidence followed the same trend as the exposure to rotavirus infections (i.e., peaked in the 2001–2005 birth cohort group, and decreased thereafter) (Fig. 1 and Supplementary Table 1).
Age-specific incidence rates per 100,000 person-years (PY) with 95% CIs of type 1 diabetes in Finnish children and adolescents in four birth cohort groups between 1995 and 2015.
Age-specific incidence rates per 100,000 person-years (PY) with 95% CIs of type 1 diabetes in Finnish children and adolescents in four birth cohort groups between 1995 and 2015.
Trends in the diabetes incidence coincided only partly when calculated using observed and both observed and imputed counts, due to the incompletely observed age range (Supplementary Fig. 2). We focused on the relative differences that were statistically significant and consistent in both the main and the sensitivity analyses. Compared with the 2001–2005 birth cohort group, the relative differences in the type 1 diabetes incidence (Table 2) were statistically significant and consistent in both the main and the sensitivity analyses for the following birth cohort and age-groups: for the 1995–2000 birth cohort group, the overall reduction was 5% and in the youngest age-group 21%; for the partially vaccinated 2006–2009 birth cohort group, the incidence was 11% higher among those aged 5–9 years; for the postvaccine 2010–2015 birth cohort group, a reduction of 21% was seen in the youngest age-group, with a corresponding absolute reduction of 17.1 cases per 100,000 person-years (95% CI 10.9–23.3).
The overall (0.5–14.9 years) and age-specific incidence rate ratios of type 1 diabetes in Finnish children and adolescents calculated for birth cohort groups 1995–2000, 2006–2009, and 2010–2015 compared with birth cohort group 2001–2005 (reference)
| Birth cohort group, analyses performed | Age-group | |||||||
|---|---|---|---|---|---|---|---|---|
| 0.5–14.9 years | 0.5–4.9 years | 5.0–9.9 years | 10.0–14.9 years | |||||
| Crude IRR (95% CI) | Adjusted* IRR (95% CI) | Crude IRR (95% CI) | Adjusted* IRR (95% CI) | Crude IRR (95% CI) | Adjusted* IRR (95% CI) | Crude IRR (95% CI) | Adjusted* IRR (95% CI) | |
| 1995–2000, main analyses | 0.94 (0.90–0.99) | 0.95 (0.90–1.00***) | 0.79 (0.72–0.86) | 0.79 (0.72–0.86) | 1.03 (0.95–1.11) | 1.03 (0.95–1.11) | 1.00 (0.92–1.09) | 1.03 (0.94–1.12) |
| 1995–2000, sensitivity analyses | 0.94 (0.89–0.98) | 0.94 (0.89–0.98) | 0.79 (0.72–0.86) | 0.79 (0.72–0.86) | 1.03 (0.95–1.11) | 1.03 (0.95–1.11) | 0.97 (0.90–1.05) | 0.97 (0.90–1.05) |
| 2001–2005** | Reference | Reference | Reference | Reference | Reference | Reference | Reference | Reference |
| 2006–2009, main analyses | 1.04 (0.97–1.10) | 1.04 (0.98–1.11) | 0.95 (0.86–1.05) | 0.95 (0.86–1.05) | 1.11 (1.01–1.22) | 1.11 (1.01–1.22) | 0.79 (0.56–1.13) | 0.74 (0.52–1.07) |
| 2006–2009, sensitivity analyses | 1.06 (1.00***–1.12) | 1.06 (1.01–1.12) | 0.95 (0.86–1.05) | 0.95 (0.86–1.05) | 1.11 (1.01–1.21) | 1.11 (1.01–1.21) | 1.11 (1.01–1.21) | 1.10 (1.01–1.21) |
| 2010–2015, main analyses | 0.83 (0.77–0.90) | 0.89 (0.82–0.97) | 0.76 (0.69–0.84) | 0.79 (0.71–0.87) | 1.09 (0.91–1.31) | 1.08 (0.89–1.30) | Not available | Not available |
| 2010–2015, sensitivity analyses | 1.02 (0.98–1.07) | 1.03 (0.98–1.08) | 0.81 (0.74–0.88) | 0.81 (0.74–0.88) | 1.09 (1.01–1.18) | 1.09 (1.00***–1.18) | 1.16 (1.07–1.26) | 1.16 (1.07–1.26) |
| Birth cohort group, analyses performed | Age-group | |||||||
|---|---|---|---|---|---|---|---|---|
| 0.5–14.9 years | 0.5–4.9 years | 5.0–9.9 years | 10.0–14.9 years | |||||
| Crude IRR (95% CI) | Adjusted* IRR (95% CI) | Crude IRR (95% CI) | Adjusted* IRR (95% CI) | Crude IRR (95% CI) | Adjusted* IRR (95% CI) | Crude IRR (95% CI) | Adjusted* IRR (95% CI) | |
| 1995–2000, main analyses | 0.94 (0.90–0.99) | 0.95 (0.90–1.00***) | 0.79 (0.72–0.86) | 0.79 (0.72–0.86) | 1.03 (0.95–1.11) | 1.03 (0.95–1.11) | 1.00 (0.92–1.09) | 1.03 (0.94–1.12) |
| 1995–2000, sensitivity analyses | 0.94 (0.89–0.98) | 0.94 (0.89–0.98) | 0.79 (0.72–0.86) | 0.79 (0.72–0.86) | 1.03 (0.95–1.11) | 1.03 (0.95–1.11) | 0.97 (0.90–1.05) | 0.97 (0.90–1.05) |
| 2001–2005** | Reference | Reference | Reference | Reference | Reference | Reference | Reference | Reference |
| 2006–2009, main analyses | 1.04 (0.97–1.10) | 1.04 (0.98–1.11) | 0.95 (0.86–1.05) | 0.95 (0.86–1.05) | 1.11 (1.01–1.22) | 1.11 (1.01–1.22) | 0.79 (0.56–1.13) | 0.74 (0.52–1.07) |
| 2006–2009, sensitivity analyses | 1.06 (1.00***–1.12) | 1.06 (1.01–1.12) | 0.95 (0.86–1.05) | 0.95 (0.86–1.05) | 1.11 (1.01–1.21) | 1.11 (1.01–1.21) | 1.11 (1.01–1.21) | 1.10 (1.01–1.21) |
| 2010–2015, main analyses | 0.83 (0.77–0.90) | 0.89 (0.82–0.97) | 0.76 (0.69–0.84) | 0.79 (0.71–0.87) | 1.09 (0.91–1.31) | 1.08 (0.89–1.30) | Not available | Not available |
| 2010–2015, sensitivity analyses | 1.02 (0.98–1.07) | 1.03 (0.98–1.08) | 0.81 (0.74–0.88) | 0.81 (0.74–0.88) | 1.09 (1.01–1.18) | 1.09 (1.00***–1.18) | 1.16 (1.07–1.26) | 1.16 (1.07–1.26) |
In the main analyses, calculations were based on cases of type 1 diabetes diagnosed during 1995–2016 (incomplete age span for birth cohorts 2003–2016). Main analyses were calculated using observed data, and sensitivity analyses were calculated using imputed data. Statistically significant results are marked in bold. IRR, incidence rate ratio.
Adjusted for age and sex when fitted to the whole data set and for sex in the analyses stratified by age.
Observed data as a reference for main analyses, imputed data as a reference for sensitivity analyses.
1 is not included in the CI, although it appears so due to rounding of decimals.
In both main and sensitivity analyses, a reduction of 1 percentage point in the proportion of children with laboratory-confirmed rotavirus infections was associated with a 5% decrease in type 1 diabetes incidence in children aged 0.5–14.9 years (Table 3). Results were consistent for both main and sensitivity analyses in the youngest age-group only, where a reduction of 1 percentage point in the proportion of children with laboratory-confirmed rotavirus infections was associated with an 8% decrease in the incidence of type 1 diabetes (Table 3).
Crude and adjusted incidence rate ratios with 95% CIs, demonstrating changes in the incidence rate of type 1 diabetes per 1 percentage point decrease in the exposure to laboratory-confirmed rotavirus infections
| Analyses performed | Age-group | |||||||
|---|---|---|---|---|---|---|---|---|
| 0.5–14.9 years | 0.5–4.9 years | 5.0–9.9 years | 10.0–14.9 years | |||||
| Crude IRR (95% CI) | Adjusted* IRR (95% CI) | Crude IRR (95% CI) | Adjusted** IRR (95% CI) | Crude IRR (95% CI) | Adjusted** IRR (95% CI) | Crude IRR (95% CI) | Adjusted** IRR (95% CI) | |
| Main analyses | 0.88 (0.85–0.90) | 0.95 (0.91–0.98) | 0.83 (0.80–0.87) | 0.92 (0.89–0.96) | 1.03 (0.97–1.09) | 1.05 (0.99–1.12) | 0.65 (0.56–0.76) | 0.91 (0.74–1.11) |
| Sensitivity analyses | 1.00 (0.98–1.02) | 0.95 (0.92–0.98) | 0.90 (0.87–0.94) | 0.94 (0.90–0.97) | 1.03 (1.00–1.06) | 1.04 (1.01–1.07) | 1.06 (1.03–1.09) | 1.08 (1.04–1.11) |
| Analyses performed | Age-group | |||||||
|---|---|---|---|---|---|---|---|---|
| 0.5–14.9 years | 0.5–4.9 years | 5.0–9.9 years | 10.0–14.9 years | |||||
| Crude IRR (95% CI) | Adjusted* IRR (95% CI) | Crude IRR (95% CI) | Adjusted** IRR (95% CI) | Crude IRR (95% CI) | Adjusted** IRR (95% CI) | Crude IRR (95% CI) | Adjusted** IRR (95% CI) | |
| Main analyses | 0.88 (0.85–0.90) | 0.95 (0.91–0.98) | 0.83 (0.80–0.87) | 0.92 (0.89–0.96) | 1.03 (0.97–1.09) | 1.05 (0.99–1.12) | 0.65 (0.56–0.76) | 0.91 (0.74–1.11) |
| Sensitivity analyses | 1.00 (0.98–1.02) | 0.95 (0.92–0.98) | 0.90 (0.87–0.94) | 0.94 (0.90–0.97) | 1.03 (1.00–1.06) | 1.04 (1.01–1.07) | 1.06 (1.03–1.09) | 1.08 (1.04–1.11) |
Birth cohorts 1995–2015, stratified into three age-groups. Main analyses were calculated using observed data, and sensitivity analyses were calculated using imputed data. Statistically significant results are marked in bold. IRR, incidence rate ratio.
Adjusted for age, sex, and year of diagnosis.
Adjusted for age and sex.
Conclusions
Our ecological study on the 1995–2015 birth cohorts in Finland found an association between changes in exposure to rotavirus infections and changes in the type 1 diabetes incidence in children younger than the age of 5. Exposure to rotavirus infections in children younger than the age of 5 years first increased to 2,552 per 100,000 children in the prevaccine 2001–2005 birth cohort group and then decreased to 171 per 100,000 children in the postvaccine 2010–2015 birth cohort group. The type 1 diabetes incidence in children aged <5 years changed in parallel and was 21% lower in the 2010–2015 birth cohorts than in the 2001–2005 cohorts. The lower incidence in the 1995–2000 birth cohort group is in line with previous observations that the increase in incidence rate was faster than ever before in the time period 2000–2005 and that the increase was fastest among those <5 years (9). At the population level, in the 1995–2015 birth cohorts, a reduction of 1 percentage point in the proportion of exposed children associated with a decrease of 8% in the incidence of type 1 diabetes in children younger than the age of 5 years.
Several studies, including a recent meta-analysis, have reached conclusions similar to the current ones; that is, that there may be an association between the introduction of a nationwide rotavirus vaccination program (and subsequent decrease in rotavirus infections) and a decrease in the type 1 diabetes incidence rate, specifically in children younger than the age of 5 years (5–7,10,11). The aforementioned meta-analysis included a number of individual-level cohort studies (12–16) that provided no support for such a protective association. Nonetheless, the conclusion was that vaccinated children younger than the age of 5 had a decreased risk of type 1 diabetes (relative risk 0.84, 95% CI 0.75–0.95) (7).
The main strengths and limitations of our study relate to its design. To our knowledge, the current study is unique in its approach to this conundrum, examining the association between type 1 diabetes and the rotavirus vaccine from the point of view of changes in the magnitude of exposure to rotavirus infections. As the study is conducted at the population level in the country with the highest incidence of type 1 diabetes globally and using nationwide register data, it provides a large sample size, and the ∼13 million person-years and 8,674 cases of type 1 diabetes accumulated in our study provide statistical power sufficient even for subgroup analyses.
However, population-level studies inevitably suffer from a weaker level of evidence than individual-level studies. This study is restricted to exploring the association at the population level and cannot claim causality at the individual level, as this would represent ecological fallacy. We are unable to account for potential confounding factors, including other environmental factors changing over time. Moreover, as a population-level study, our study does not include individual-level data, such as whether laboratory-confirmed rotavirus infection and type 1 diabetes occurred within the same individual.
Comparisons of present and historical cohorts have several caveats, including potential differences in case detection or medical practices in the periods or possibility of noncomparable baseline transmission. However, as far as we are aware, no such changes in the diagnostic criteria of rotavirus infections or type 1 diabetes have been implemented during the 20-year study period. One aspect to keep in mind is that the NIDR data represent rotavirus infections with confirmed microbial etiology (i.e., the tip of the iceberg). In the current study, we assumed that the number of cases of rotavirus infections recorded in the NIDR, although incomplete, reflect the magnitude and changes in the underlying overall exposure at the population level and that the 93% reduction we observed between the pre- and postvaccine periods is in line with the changes reported in the study by Leino et al. (17), where cases were detected by ICD-10 diagnostic codes.
The incompletely observed age range is an unfortunate limitation, which can be remedied only once more time has passed since widespread rotavirus vaccination and more follow-up data are available. As an attempt to account for the uncertainty of the results due to the incompletely observed age range, we performed sensitivity analysis using imputed data and considered the associations robust only if they were consistent in both the observed and imputed data.
The current study adds to the evidence supporting the role of the rotavirus as a trigger of type 1 diabetes, and hence, the protective role of the rotavirus vaccination in young children, indicating a need for further individual-level studies with sufficient statistical power and follow-up time.
This article contains supplementary material online at https://doi.org/10.2337/figshare.24256219.
Article Information
Funding. This research was supported by the Sigrid Jusélius Foundation and the Helsinki University Hospital (State Research Funding).
Duality of Interest. M.K. is a member of the board of Vactech Ltd., which develops vaccines against picornaviruses. No other potential conflicts of interest relevant to this article were reported.
Author Contributions. A.P. wrote the first draft of the manuscript. A.P., A.B., R.S., M.A., H.S., and M.K. contributed to the study concept and design. A.P., A.B., R.S., M.A., H.S., and M.K. contributed to the discussion and reviewed and edited the manuscript. A.B. performed the statistical analysis. 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.
