Abstract
OBJECTIVE
We investigated islet autoimmunity (IA) incidence trends among Colorado children born 1993–2010 in Diabetes Autoimmunity Study in the Young (DAISY) and The Environmental Determinants of Diabetes in the Young (TEDDY) and whether HLA genotype or early-life environmental exposures contributed to observed increases.
RESEARCH DESIGN AND METHODS
We analyzed the risk of IA among 2,734 Colorado children representing three birth cohorts (BCs): BC1 (1993–1998), BC2 (1999-August 2004), and BC3 (September 2004–2010). Cox models evaluated IA by BC, with time-varying coefficients, BC-HLA interaction, and stratification by type 1 diabetes family history. Mediation analyses examined age at gluten introduction, weight-for-age Z-score, and maternal smoking.
RESULTS
IA increased across BC1–BC3 (7.0%, 10.7%, and 12.7%). At age 1 year, moderate-risk HLA participants in BC2 and BC3 had, respectively, 2.57 and 6.19 times the IA risk of BC1. By age 5 years, effects attenuated. No selected environmental exposures mediated the increased incidence.
CONCLUSIONS
IA increased among Colorado children born 1993–2010, with the strongest effects among moderate-risk HLA genotypes in later BCs.
Graphical Abstract
Introduction
Type 1 diabetes incidence has increased in children over recent decades (1). We investigated whether islet autoimmunity (IA) incidence was increasing among children born between 1993 and 2010 using the Diabetes Autoimmunity Study in the Young (DAISY) and The Environmental Determinants of Diabetes in the Young (TEDDY) Colorado study. We also examined whether increasing penetrance of moderate- or high-risk HLA genotypes was driving the rising IA incidence, as seen in type 1 diabetes (2), and whether early-life environmental exposures previously associated with type 1 diabetes contributed to the increase.
Research Design and Methods
Study Populations
We included longitudinal participant data from DAISY and TEDDY Colorado, totaling 2,734 participants. Study designs and populations are described elsewhere (3,4). Briefly, participants were recruited through newborn screening in Denver metropolitan area hospitals based on HLA-DR risk alleles or through clinical settings as unaffected relatives of individuals with type 1 diabetes. DAISY recruited from one hospital, while TEDDY recruited from several. Participants were categorized into three birth cohorts (BC1, BC2, and BC3) based on their dates of birth: BC1, 1993–1998 (n = 892); BC2, 1999–August 2004 (n = 504); and BC3, September 2004–2010 (n = 1,338). DAISY included participants from BC1 and BC2, while TEDDY represented BC3. Participants were followed since birth. We classified participants with DR3/4 HLA genotypes as high risk, while DR3/3, DR4/4, and DR4/X were grouped as moderate risk. Participants with DR3/X or DRX/X HLA genotypes were excluded for cohort comparability due to the low sample size in TEDDY (n = 4).
Outcome Ascertainment
Islet autoantibodies to insulin, IA2, GAD, or ZnT8 were measured consistently in DAISY and TEDDY by the same laboratory at the Barbara Davis Center for Diabetes in Denver, Colorado, using radio-binding assays. IA was defined as the persistent presence of at least one islet autoantibody at two or more consecutive study visits, or once if the child was diagnosed with type 1 diabetes at the next visit (5–7). For consistency, IA for TEDDY participants was redefined using the DAISY definition above.
Environmental Exposures
Three environmental exposures were selected a priori based on previous associations with IA or type 1 diabetes, including age at gluten introduction, weight-for-age Z-score, and maternal smoking during pregnancy (8–12). Age at gluten introduction was categorized as 0–4, 5–6 (reference), or ≥7 months. World Health Organization reference ranges were used for the weight-for-age Z-score indicator (≥85th percentile) and were calculated from weight measurements collected during the first visit that occurred between 14 and 25 months of age for each participant. Maternal smoking during pregnancy was captured as an indicator (yes vs. no).
Statistical Analysis
We used R 4.4.0 statistical software (R Core Team, 2024) to conduct all analyses (13). To investigate trends in IA incidence, we included BC as the primary explanatory variable, with BC1 as the reference. Cox proportional hazards (PH) models evaluated the time to IA incidence within 16 years by BC, adjusting for HLA genotype (moderate- vs. high-risk alleles) and stratifying by dichotomous type 1 diabetes family history in first-degree relatives. Since BC violated the PH assumption, we included a time-varying coefficient using the logarithm of age at follow-up. To characterize potential changes in HLA penetrance, we included an interaction term for HLA risk group (14). Finally, to examine potential mediating effects of the three environmental exposures, we used a combination of Cox PH and logistic regression models to conduct mediation analyses as implemented in the mediator R package (15). It estimated the proportion mediated with 95% CIs. We included all participants who had complete data within each environmental exposure: age at gluten introduction (n = 2,695), weight-for-age Z-score (n = 2,259), and maternal smoking (n = 2,651).
Sensitivity Analysis
Weight-for-age Z-score visit-related missingness did not differ across birth cohorts (P = 0.2595). We also examined stratified results by type 1 diabetes family history.
Data and Resource Availability
Investigators may request access to DAISY and TEDDY data through their respective National Institute of Diabetes and Digestive and Kidney Diseases repositories: https://repository.niddk.nih.gov/study/206 and https://repository.niddk.nih.gov/study/24.
Results
Participant characteristics for the DAISY and TEDDY cohorts are outlined in Table 1 and Supplementary Table 1. Overall, there were slightly fewer female participants (48.6%), and most were non-Hispanic White (63.4%). BC2 had more first-degree relatives (27.4%) compared with BC1 (16.0%) and BC3 (11.3%).
Table 1.
Participant characteristics for DAISY and TEDDY by BC
| DAISY | TEDDY (Colorado) | Overall | ||
|---|---|---|---|---|
| BC1: 1993–1998 | BC2: 1999–Aug. 2004 | BC3: Sep. 2004–2010 | 1993–2010 | |
| (n = 892) | (n = 504) | (n = 1,338) | (N = 2,734) | |
| Female sex | 413 (46.3) | 248 (49.2) | 669 (50.0) | 1,330 (48.6) |
| Non-Hispanic White | 598 (67.0) | 332 (65.9) | 802 (59.9) | 1,732 (63.4) |
| First-degree relative with T1D | 143 (16.0) | 138 (27.4) | 151 (11.3) | 432 (15.8) |
| HLA high-risk group | 290 (32.5) | 149 (29.6) | 475 (35.5) | 914 (33.4) |
| Developed IA | 62 (7.0) | 54 (10.7) | 170 (12.7) | 286 (10.5) |
| Follow-up time, mean (SD) (years) | 8.9 (6.1) | 9.5 (6.1) | 9.5 (6.0) | 9.3 (6.0) |
| Age at gluten introduction (months) | ||||
| 0–4 | 78 (8.7) | 24 (4.8) | 138 (10.3) | 240 (8.8) |
| 5–6 | 262 (29.4) | 212 (42.1) | 617 (46.1) | 1091 (39.9) |
| ≥7 | 515 (57.7) | 266 (52.8) | 583 (43.6) | 1364 (49.9) |
| Maternal smoking during pregnancy | 115 (12.9) | 37 (7.3) | 150 (11.2) | 302 (11.0) |
| Weight-for-age Z-score (>85th perc.) | 93 (10.4) | 50 (9.9) | 100 (7.5) | 243 (8.9) |
Data are presented as n (%) unless indicated otherwise as mean (SD). perc., percentile; T1D, type 1 diabetes. The high-risk HLA group consists of participants with DR3/4 genotypes, and the moderate-risk HLA group includes DR3/3, DR4/4, and DR4/X. Mean follow-up ages in months for the weight-for-age Z-score analysis in BC1, BC2, and BC3 were 16.0, 16.1, and 15.5, respectively.
Cumulative IA incidence by age 16 years increased from the earliest to the latest birth cohort: 7.0% (BC1), 10.7% (BC2), and 12.7% (BC3) (Table 1). We found significant interaction between the effects of cohort and HLA risk group (P = 0.0124 [BC2], P = 0.0101 [BC3]); thus, we provided stratified graphs and hazard ratio estimates (Fig. 1, Table 2, and Supplementary Fig. 1). At 1 year of age, moderate-risk HLA participants in BC2 and BC3 had 2.57 (95% CI 1.17, 5.61) and 6.19 (95% CI 3.21, 11.92) times the IA risk of BC1, respectively; high-risk HLA participants in BC3 had 2.81 (95% CI 1.53, 5.16) times the IA risk. At 5 years of age, moderate-risk HLA participants in BC2 and BC3 had 2.20 (95% CI 1.29, 3.77) and 2.55 (95% CI 1.57, 4.12) times the IA risk of BC1, respectively; the high-risk HLA group had no differences in IA hazard between birth cohorts. At 10 years of age, moderate-risk HLA participants in BC2 had 2.06 (95% CI 1.11, 3.84) times the IA risk of BC1, with no difference for BC3; the high-risk HLA group had no difference in IA hazard between birth cohorts (Supplementary Table 2).
Figure 1.
IA incidence by BC in DAISY and TEDDY Colorado over 16 years of follow-up. Plot A represents the cumulative incidence of islet autoimmunity among the BCs from DAISY (BC1 and BC2) and TEDDY (BC3), while plots B and C separate out the cumulative incidences by moderate and high HLA risk, respectively.
Table 2.
Time-varying IA risk by BC and HLA group
| IA hazard ratio at 1 year | IA hazard ratio at 5 years | |||
|---|---|---|---|---|
| (95% CI) | P value | (95% CI) | P value | |
| BC2 ∗ HLA | 0.0124 | — | ||
| BC3 ∗ HLA | 0.0101 | — | ||
| Moderate-risk HLA | ||||
| BC1 (reference) | 1.00 | — | 1.00 | — |
| BC2 | 2.57 (1.17, 5.61) | 0.0183 | 2.20 (1.29, 3.77) | 0.0041 |
| BC3 | 6.19 (3.21, 11.92) | 5.10 × 10−8 | 2.55 (1.57, 4.12) | 1.40 × 10−4 |
| High-risk HLA | ||||
| BC1 (reference) | 1.00 | — | 1.00 | — |
| BC2 | 0.97 (0.43, 2.18) | 0.9367 | 0.83 (0.48, 1.44) | 0.5061 |
| BC3 | 2.81 (1.53, 5.16) | 0.0008 | 1.16 (0.78, 1.71) | 0.4652 |
These hazard ratios reflect Cox PH models with BC as the primary explanatory variable and IA incidence as the outcome at 1 and 5 years of age. The model was stratified by type 1 diabetes family history and included an interaction between birth cohort and HLA risk group. The interaction P values comparing BC2 with BC1 and BC3 with BC1 were statistically significant at 1 year of age.
Associations between the environmental exposures (age at gluten introduction, weight-for-age Z-score, and maternal smoking during pregnancy) and IA are detailed in Table 3. We found a marginally significant interaction between HLA risk group and age at gluten introduction (P = 0.0531); thus, we stratified the effects by HLA risk group. Among participants with moderate-risk HLA genotypes, gluten introduction at 0–4 months of age was associated with an 81% lower hazard of IA compared with 5–6 months of age (P = 0.0202), while children introduced to gluten at ≥7 months of age were not significantly different from those introduced between 5 and 6 months of age. Children with high-risk HLA alleles did not have significant differences between earlier or later ages of gluten introduction. Finally, we did not find significant differences between weight-for-age Z-score or maternal smoking during pregnancy and risk of IA. We confirmed through mediation analysis that the selected environmental exposures did not mediate the association between birth cohort and IA hazard (Supplementary Table 3). Our results were robust to sensitivity analyses.
Table 3.
Overall and HLA-stratified hazard ratios for the association between three environmental exposures and IA
| Sample size | Hazard ratio for IA | ||
|---|---|---|---|
| Environmental exposure (mediator) | n (% total) | (95% CI) | P value |
| Age at gluten introduction (months) | 2,695 (98.6) | ||
| Overall | |||
| Month 0–4 vs. 5–6 | 0.54 (0.28, 1.03) | 0.0623 | |
| Month ≥7 vs. 5–6 | 1.00 (0.79, 1.27) | 0.9819 | |
| Age at gluten introduction ∗ HLA (interaction) | 0.0531 | ||
| Moderate HLA allele | |||
| Month 0–4 vs. 5–6 | 0.19 (0.05, 0.77) | 0.0202 | |
| Month ≥7 vs. 5–6 | 0.82 (0.59, 1.14) | 0.2367 | |
| High HLA risk allele | |||
| Month 0–4 vs. 5–6 | 0.99 (0.46, 2.13) | 0.9880 | |
| Month ≥7 vs. 5–6 | 1.23 (0.87, 1.75) | 0.2449 | |
| Weight-for-age Z-score (World Health Organization) | 2,259 (82.6) | ||
| Overall | |||
| >85th percentile | 0.77 (0.57, 1.05) | 0.0968 | |
| Weight-for-age Z-score ∗ HLA (interaction) | 0.2049 | ||
| Moderate HLA allele | |||
| >85th percentile | 0.65 (0.42, 1.01) | 0.0553 | |
| High HLA risk allele | |||
| >85th percentile | 0.97 (0.63, 1.48) | 0.8800 | |
| Maternal smoking during pregnancy | 2,651 (97.0) | ||
| Overall | |||
| Yes vs. no | 0.94 (0.61, 1.42) | 0.7533 | |
| Maternal smoking during pregnancy ∗ HLA (interaction) | 0.8046 | ||
| Moderate HLA allele | |||
| Yes vs. no | 0.97 (0.54, 1.74) | 0.9079 | |
| High HLA risk allele | |||
| Yes vs. no | 0.87 (0.48, 1.58) | 0.6446 |
The overall model represents the crude association between the environmental exposure and the hazard ratio of IA between the reference and comparison groups. The moderate and high HLA risk models reflect the stratified associations between the environmental exposure and the hazard ratio of IA between the reference and comparison groups.
Conclusions
We report an increasing incidence of IA across birth cohorts among children enrolled in DAISY and TEDDY Colorado between 1993 and 2010. Children with moderate-risk HLA genotypes exhibited stronger secular trends in IA incidence, suggesting the rising IA burden may be accounted for by increasing involvement of a larger, moderate-risk segment of the population, as reported previously in type 1 diabetes (2,16). Thus, the secular increase in type 1 diabetes incidence likely has its roots in increased penetration of IA triggers (17). If so, prevention and early identification efforts may need to extend beyond the highest-risk HLA strata.
The selected environmental exposures previously implicated in IA or type 1 diabetes did not explain the observed increase in IA incidence. Earlier gluten introduction (0–4 months) was associated with a reduced hazard of IA among children with moderate-risk HLA genotypes, consistent with prior findings (9), but this effect did not mediate the BC trend. Previous work noted that weight Z-scores were associated with IA at 12 months, but not at later ages (18), which might explain why we found no association at 14–25 months. Energy intake and BMI have been shown to influence IA development (19), but BC trends were not examined, and we used weight measurements. Finally, although maternal smoking has been inversely associated with IA and type 1 diabetes (20), its decrease did not explain increasing IA incidence. These results underscore the complexity of IA etiology and suggest that other environmental or behavioral factors may be contributing to the rising incidence.
Our study possessed several strengths. Having two well-characterized prospective cohorts enabled evaluation of temporal trends in IA incidence. Integration of genetic data with time-to-event IA outcomes provided a unique opportunity to evaluate whether an increasing penetrance of moderate-risk HLA genotypes has contributed to increasing IA incidence over time.
We also had some limitations. Our analysis only included participants from Colorado, which may limit generalizability. TEDDY used stricter genetic risk recruitment criteria than DAISY for participants with first-degree relatives with type 1 diabetes, potentially reducing comparability of moderate-risk HLA groups between studies. The weight-for-age Z-score had some visit-related missingness, although this did not differ across BCs. Continuous birth year was not available in both cohorts; thus, we used categorical BCs, reducing statistical power but easing interpretability. The timing of follow-up visits was different between DAISY and TEDDY, but a sensitivity analysis suggested only mild attenuations in the effects (Supplementary Appendix 1). Lastly, genetic risk scores were not available in both cohorts to assess overall genetic risk versus HLA risk.
Future research should assess whether the observed increase in IA incidence persists, explore regional and international variability, and investigate environmental exposures or gene-environment interactions that may underlie the increasing incidence of IA and subsequent rise in type 1 diabetes.
This article contains supplementary material online at https://doi.org/10.2337/figshare.32715546.
Article Information
This manuscript is the result of funding from the NIH and is subject to the NIH Public Access Policy. Through acceptance of this federal funding, NIH has been given a right to make this manuscript publicly available in PubMed Central upon the Official Date of Publication, as defined by NIH. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. K.R.H. contributed to formal analysis. K.R.H. contributed to software. K.R.H. contributed to creating figures and tables to communicate research findings. K.R.H. wrote the original draft of the manuscript. K.R.H., F.D., K.W., A.K.S., T.B., RKJ, P.M.C., B.F., M.R., and J.M.N. reviewed and edited the manuscript. K.R.H., F.D., K.W., and T.B. curated the data. K.R.H., M.R., and J.M.N. contributed to methodology. K.R.H., M.R., and J.M.N. contributed to ensuring the accuracy, precision, reliability, quality, and the overall integrity of the research. M.R. contributed to the investigation. M.R. contributed to providing resources. M.R. and J.M.N. conceptualized the study. M.R. and J.M.N. acquired funding. M.R. and J.M.N. administered the project. M.R. and J.M.N. contributed to supervision. J.M.N. 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.
Prior Presentation. A previous version of this work was presented as a poster at the Immunology of Diabetes Society’s 20th Immunology of Diabetes Congress, Bruges, Belgium, 4–8 November, 2024.
Handling Editors. The journal editors responsible for overseeing the review of the manuscript were John B. Buse and Michael J. Haller.
Funding Statement
This work was funded by the National Institutes of Health (NIH) through the National Institute of Diabetes and Digestive and Kidney Diseases grant awards R01-DK32493, R01-DK104351, and P30-DK116073, and National Institute of Allergy and Infectious Diseases grant award R21-AI142483. The TEDDY Study is funded by U01 DK63829, U01 DK63861, U01 DK63821, U01 DK63865, U01 DK63863, U01 DK63836, U01 DK63790, UC4 DK63829, UC4 DK63861, UC4 DK63821, UC4 DK63865, UC4 DK63863, UC4 DK63836, UC4 DK95300, UC4 DK100238, UC4 DK106955, UC4 DK112243, UC4 DK117483, U01 DK124166, and U01 DK128847, and Contract No. HHSN267200700014C from the National Institute of Diabetes and Digestive and Kidney Diseases, National Institute of Allergy and Infectious Diseases, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institute of Environmental Health Sciences, Centers for Disease Control and Prevention, and Breakthrough T1D (formerly JDRF). This work is supported in part by the NIH National Center for Advancing Translational Sciences Clinical and Translational Science Awards to the University of Florida (UL1 TR000064) and the University of Colorado (UL1 TR002535).
Supporting information
References
- 1. Norris JM, Johnson RK, Stene LC.. Type 1 diabetes-early life origins and changing epidemiology. Lancet Diabetes Endocrinol 2020;8:226–238 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Fourlanos S, Varney MD, Tait BD, et al. The rising incidence of type 1 diabetes is accounted for by cases with lower-risk human leukocyte antigen genotypes. Diabetes Care 2008;31:1546–1549 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Rewers M, Bugawan TL, Norris JM, et al. Newborn screening for HLA markers associated with IDDM: diabetes autoimmunity study in the young (DAISY). Diabetologia 1996;39:807–812 [DOI] [PubMed] [Google Scholar]
- 4. TEDDY Study Group . The Environmental Determinants of Diabetes in the Young (TEDDY) study: study design. Pediatr Diabetes 2007;8:286–298 [DOI] [PubMed] [Google Scholar]
- 5. Rewers M, Norris JM, Eisenbarth GS, et al. Beta-cell autoantibodies in infants and toddlers without IDDM relatives: diabetes autoimmunity study in the young (DAISY). J Autoimmun 1996;9:405–410 [DOI] [PubMed] [Google Scholar]
- 6. Barker JM, Barriga KJ, Yu L, et al.; Diabetes Autoimmunity Study in the Young . Prediction of autoantibody positivity and progression to type 1 diabetes: Diabetes Autoimmunity Study in the Young (DAISY). J Clin Endocrinol Metab 2004;89:3896–3902 [DOI] [PubMed] [Google Scholar]
- 7. TEDDY Study Group . The Environmental Determinants of Diabetes in the Young (TEDDY) Study. Ann N Y Acad Sci 2008;1150:1–13 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Norris JM, Barriga K, Klingensmith G, et al. Timing of initial cereal exposure in infancy and risk of islet autoimmunity. JAMA 2003;290:1713–1720 [DOI] [PubMed] [Google Scholar]
- 9. Uusitalo U, Lee H-S, Andrén Aronsson C, et al.; TEDDY Study Group . Early infant diet and islet autoimmunity in the TEDDY Study. Diabetes Care 2018;41:522–530 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Lamb MM, Yin X, Zerbe GO, et al. Height growth velocity, islet autoimmunity and type 1 diabetes development: the Diabetes Autoimmunity Study in the Young. Diabetologia 2009;52:2064–2071 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Liu X, Vehik K, Huang Y, et al.; TEDDY Study Group . Distinct growth phases in early life associated with the risk of type 1 diabetes: the TEDDY Study. Diabetes Care 2020;43:556–562 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Begum M, Pilkington RM, Chittleborough CR, et al. Effect of maternal smoking during pregnancy on childhood type 1 diabetes: a whole-of-population study. Diabetologia 2020;63:1162–1173 [DOI] [PubMed] [Google Scholar]
- 13. R Core Team . R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing: Vienna, Austria, 2024 [Google Scholar]
- 14. Valeri L, Vanderweele TJ.. Mediation analysis allowing for exposure-mediator interactions and causal interpretation: theoretical assumptions and implementation with SAS and SPSS macros. Psychol Methods 2013;18:137–150 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Creed J, Gerke T.. Mediator: causal mediation analysis. 2025, R package version 0.0.0.9000
- 16. Vehik K, Hamman RF, Lezotte D, et al. Trends in high-risk HLA susceptibility genes among Colorado youth with type 1 diabetes. Diabetes Care 2008;31:1392–1396 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Ziegler A-G, Pflueger M, Winkler C, et al. Accelerated progression from islet autoimmunity to diabetes is causing the escalating incidence of type 1 diabetes in young children. J Autoimmun 2011;37:3–7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Elding Larsson H, Vehik K, Haller MJ, et al.; TEDDY Study Group . Growth and risk for islet autoimmunity and progression to type 1 diabetes in early childhood: The Environmental Determinants of Diabetes in the Young Study. Diabetes 2016;65:1988–1995 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Aronsson CA, Tamura R, Vehik K, et al. Dietary intake and body mass index influence the risk of islet autoimmunity in genetically at-risk children: a mediation analysis using the TEDDY Cohort. Pediatr Diabetes 2023;2023:3945064. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Vanderlinden LA, Wong E, Johnson RK, et al. DNA methylation smoking scores and risk of islet autoimmunity and type 1 diabetes. Diabetes Care 2025;48:1628–1636 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.