Document resource
WHAT IS ALREADY KNOWN ON THIS TOPIC B and T cells have important roles in type 1 diabetes (T1D) pathogenesis through the loss of immune tolerance, antigen presentation, cytokine production, and destruction of pancreatic beta-cells, but the roles of other immune cells and combinations therein remain unclear.WHAT THIS STUDY ADDS We found immune cell ratio differences between T1D cases and controls before, during, and across seroconversion (SV) time points. Our findings highlight the complexity of immunodynamics around islet autoimmunity SV and the potential role for immune cell ratios in T1D risk stratification, early intervention or monitoring, and mechanistic insight.HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY In the foreseeable future, functional immune cell ratios derived from DNA methylation could be used to anticipate immunodynamics, enable disease stratification, and improve our understanding of T1D pathogenesis.Introduction Type 1 diabetes (T1D) remains one of the most common chronic autoimmune diseases affecting children globally. 1 Characterized by immune-mediated destruction of insulin-producing beta cells in the pancreas, it ultimately leads to dysglycemia, ketoacidosis, and a lifelong dependency on exogenous insulin therapy.1 2 Current global estimates point to increasing incidence rates, especially among children and adolescents, with estimated average annual increases of approximately 3%–4% over the last three decades, highlighting a significant public health burden.3–5 Prior to developing T1D, islet autoimmunity (IA) occurs, defined by the presence of autoantibodies to insulin (IAA), glutamic acid decarboxylase (GADA), insulinoma-associated protein 2 (IA-2A), or zinc transporter 8 (ZnT8A), which target pancreatic islet cells. IA signifies early faltering in immune tolerance and marks the beginning of a variable progression toward T1D, although some people do not progress to the disease.6 7 Although IA is a well-established precursor to T1D, gaps remain in our understanding of the underlying immune processes.Immune cell dynamics that occur around the time of IA seroconversion (SV) and during progression from IA to T1D have not been fully elucidated. Advances in immunophenotyping have augmented the utility of peripheral blood-based immune cell ratios, such as neutrophil-to-lymphocyte (NLR), CD4T/CD8T ratio, and memory/naive (mem/nv) lymphocyte ratios, as potential biomarkers of systemic inflammation, immune activation, and immune maturation status.8 9 Specifically, the NLR indicates systemic inflammation in innate and adaptive immunity (where higher values suggest more inflammation); the CD4T/CD8T ratio reflects immune senescence and activation (with lower values suggesting immune dysregulation through reduced helper T cells or expanded cytotoxic T cells); and the B and T cell mem/nv ratios reveal the functional state of the adaptive immune system, where higher ratios may suggest prior antigen experience and immune readiness, and lower ratios could indicate a predominance of naive cells, which may signal immaturity or impaired activation.8–11 From these ratios, we can glean insight into the balance between immune cell subsets rather than focusing on absolute or relative abundance alone, which can help to characterize underlying immune dynamics, such as shifts between innate and adaptive immune responses or naive and antigen-experienced states. Differential ratios of immune cell populations, therefore, can reflect underlying disease processes and may predict the onset or worsening of disease. Ratios offer modeling options that address the inherent compositional nature of peripheral blood-based cell proportion data (ie, the sum-to-one constraint, whereby a change in one cell proportion affects the others), capturing biologically meaningful contrasts while avoiding potential spurious associations driven by interdependence among components.12 13 However, few studies have used longitudinal data to examine changes in immune cell ratios before and after the loss of immune tolerance associated with the onset of IA in pediatric populations.This study aims to bridge these knowledge gaps by evaluating immune cell type heterogeneity and dynamics during critical time windows before and after IA seroconversion. Leveraging two well-established prospective cohorts—the Diabetes Autoimmunity Study in the Young (DAISY) and The Environmental Determinants of Diabetes in the Young (TEDDY)—we estimated immune cell ratios derived from DNA methylation (DNAm) in peripheral blood to better understand the immunological shifts occurring in children genetically susceptible to T1D during a critical stage of disease pathogenesis.Research design and methods DAISY study population The prospective DAISY cohort enrolled and followed 2547 children based in Denver, Colorado, beginning in 1993. They were recruited either as part of population-based newborn screening at St Joseph’s Hospital for specific human leukocyte antigen (HLA) genotypes known to have increased genetic risk of T1D or from unaffected first-degree relatives of current patients and then followed for the development of IA (up to 20 years of age) and T1D (until diagnosis or study withdrawal). Additional study information is detailed elsewhere. 14 Clinic visits with blood sample collection occurred at 9, 15, and 24 months of age, then annually thereafter until the development of IA. Participants who developed IA came in for visits every 3–6 months until T1D diagnosis, based on American Diabetes Association (ADA) clinical guidelines. Most participants identified as non-Hispanic White. Informed consent was obtained from the parents or primary caretakers of all participating children, and assent was obtained from children 7 years of age and older.A nested case-control subset of DAISY study participants was selected to evaluate DNAm and other omics measurements in the context of IA and T1D. T1D cases were frequency matched on age at IA seroconversion, race-ethnicity, and sample availability across five time points—birth, infancy (age 9–16 months), pre-IA, onset of IA, and prior to diagnosis of T1D. This analysis included T1D cases and controls (IA negative) with good-quality DNAm data available pre-IA, at IA onset, or both (n=151). DAISY defined IA as the persistent presence of one or more autoantibodies (IAA, GADA, IA-2A, or ZnT8A) detected across two consecutive visits, or as a single positive autoantibody result followed by a T1D diagnosis at the next visit.15–17TEDDY study population TEDDY represents a multinational prospective cohort study of children followed from birth with an increased genetic risk of T1D, based on their HLA genotype, and includes 8676 participants who were followed through the development of IA until T1D diagnosis. Additional study design and follow-up details are described elsewhere. 18 19 Recruitment occurred across six clinical research centers: Colorado, Georgia/Florida, and Washington, in the USA, Germany, Finland, and Sweden, starting in 2004. Visits were conducted every 3 months until age 4, then every 6 months until age 15. Similar to DAISY, an accelerated schedule was used after the development of IA, which included visits every 3 months until age 15 years or diagnosis of T1D based on ADA clinical criteria. Study participants are primarily composed of persons identifying as non-Hispanic White. An external review committee formed by the National Institutes of Health monitored the study. Informed consent was obtained from the parents or primary caretakers of all participating children, and assent was obtained from children 7 years of age and older.From the TEDDY cohort, a T1D risk set based on a nested case-control cohort with DNAm data was selected. T1D cases were matched 1:1 with controls based on age, sex, T1D family history, and clinical center. We included T1D cases with known ages for developing IA and who had DNAm data available pre-IA, at the onset of IA, or both, along with samples from matched controls who were IA negative (n=166). TEDDY defined IA as testing positive for the same autoantibody (IAA, GADA, or IA-2A) across two consecutive visits.20DNA methylation DNAm measures were obtained from whole blood samples on the Infinium HumanMethylation450 (450K) and MethylationEPIC (EPIC) BeadChip arrays (Illumina, San Diego, California, USA), which interrogate over 450 000 and 850 000 methylation sites, respectively, across the genome at single-nucleotide resolution. DAISY participants had samples from both 450K and EPIC arrays, while TEDDY participants only included samples from EPIC arrays. Samples of low quality or with discordant values for sex were removed. Technical replicate samples were included during preprocessing for quality control, as detailed in previous work in DAISY 21 22 and TEDDY (unpublished), and were removed prior to statistical analysis. Additional information regarding sample preprocessing is included in online supplemental appendix 1.SP110.1136/bmjdrc-2026-006015.supp1Supplementary dataCell deconvolution We performed cell deconvolution using the estimateCellCounts2 function within FlowSorted.Blood.EPIC package (V.2.8.0) and the cell references from FlowSorted.BloodExtended.EPIC package (V.1.1.2) in R, using the recommended parameters for CpG selection and normalization. 23 24 Expanded details on the procedure are in online supplemental appendix 2. The latter included an expanded reference for 12 immune cell proportions: neutrophils (Neu), eosinophils (Eos), basophils (Baso), monocytes (Mono), naive and memory B cells (Bnv and Bmem), naive and memory CD4+ and CD8+ T cells (CD4Tnv, CD4Tmem, CD8Tnv, and CD8Tmem), natural killer (NK), and T regulatory (Treg) cells. If any individuals had zero values in their cell proportions, we imputed them via substitution by taking the minimum value within each cell type and then dividing it by two.25 We used previously published immune cell references to gauge the accuracy of estimated cell proportions.26 We also examined how immune cell proportions changed across time within TEDDY samples that passed quality control and study exclusions.Statistical approach We selected one sample per subject at each available time point of interest: before IA (pre-SV) and at IA SV, when participants converted from seronegative to seropositive with the detection of islet autoantibodies, along with matched controls. The median years between pre-SV and SV visits for DAISY and TEDDY were 1.06 and 0.31, respectively. Figure 1 shows the analytical sample size by time point, case-control status, and study. The pre-SV timepoints in DAISY and TEDDY included 84 and 122 samples, respectively. The SV timepoints in DAISY and TEDDY included 140 and 160 samples, respectively.Figure 1UpSet plots with sample breakdowns by study cohort. (A) Captures the different sample set sizes and intersections from DAISY participants, while (B) depicts samples from TEDDY participants. In both cohorts, more SV samples were available compared with pre-SV samples, based on the quality and sample selection criteria for the present study (Research Design and Methods). DAISY, Diabetes Autoimmunity Study in the Young; SV, seroconversion; TEDDY, The Environmental Determinants of Diabetes in the Young; T1D, type 1 diabetes.We tested whether immune cell ratios differed between T1D cases and matched controls at pre-SV, SV, or changed differently from pre-SV to SV using linear modeling of immune cell ratios as the outcome and an interaction between T1D status and sample time point as the primary explanatory variables of interest. For the outcomes, we used log-transformed immune cell ratios noted in previous work,23 which included the following: NLR, monocyte/lymphocyte, B-memory/B-naive (B-mem/nv), CD4T-memory/CD4T-naive (CD4T-mem/nv), CD8T-memory/CD8T-naive (CD8T-mem/nv), CD4T/CD8T, and B-CD4T-CD8T-memory/B-CD4T-CD8T-naive (B-CD4T-CD8T-mem/nv). Lymphocytes consisted of CD4+ and CD8+ T cells, B cells, NK cells, and T regulatory cells.In DAISY, additional covariates for age, DR3/4 HLA status, and platform were included to either adjust for confounding or increase precision. For TEDDY, we included DR3/4 HLA status as a fixed effect and the case-control index variable (comprised of age, sex, T1D family history, and clinical center) as a random intercept to account for the matching strata. T1D family history represented a dichotomous variable (yes vs no). For reporting, we exponentiated each log immune cell ratio, yielding the immune cell ratio between T1D cases and controls (geometric mean), and presented them as percentage differences. The lm base function in R and the lmer function from the lme4 package (V.1.1-35.5) were applied to construct the linear and linear mixed-effects models, respectively. We used an alpha level of 0.05 to determine statistical significance across all analyses.We estimated the effects of interest separately by study. Then, given the similarities in study design (eligibility, follow-up, IA, and T1D estimation), we combined them with inverse variance-weighted, fixed-effects meta-analysis using the meta package in R.27 We evaluated heterogeneity based on Cochran’s Q and I2. To evaluate the robustness of our findings, we performed two sensitivity analyses—one that excluded samples with imputed cell proportion values and another that assessed differences by three-level T1D family history (mom, dad/sibling, and none). A combination of a priori knowledge, Akaike information criterion (AIC), and consistency of effect size and direction was examined to determine the most appropriate models.Data and resource availability A subset of the DNAm data analyzed during the current study is available through NCBI GEO: GSE142512. Investigators may request access to additional data through the DAISY NIDDK repository: https://repository.niddk.nih.gov/study/206. DNAm data for TEDDY may be requested through an ancillary study (https://teddy.epi.usf.edu/research/). Additional data can be found in the TEDDY NIDDK repository: https://repository.niddk.nih.gov/study/24. All analyses were performed using R Statistical Software V.4.4.0.28 Code used in the statistical analysis is available on our lab GitHub: https://github.com/rkjcollab/paper_immune_cell_ratio_BMJODRC.Results Online supplemental table 1 provides a snapshot of several immune cell subsets (B, CD4T, CD8T, and NK) for TEDDY samples collected between 1 and 2 years of age, which includes the average age before and during SV. Compared with published immune cell references,26 estimates of immune cell proportions were lower in TEDDY participants, although the 10th and 90th percentiles shared some overlap. All immune cell subsets exhibited a similar shift toward the lower end of the reference ranges. Trends in immune cell proportions over age are shown in online supplemental figure 1. As expected, neutrophils comprise the largest proportion, and basophils comprise the smallest proportion across all ages. All memory cell populations increase with age, as expected, as children experience new immune challenges during early life.Participant characteristics for the DAISY and TEDDY cohorts are summarized in table 1. In DAISY, a total of 151 children were included in the analysis, where 76 developed T1D. Values for the frequency-matched variables (age at SV, race/ethnicity, and sample availability) remained similar between groups. T1D cases had a higher proportion of first-degree relatives with T1D as well as DR3/4 HLA genotypes compared with controls. In TEDDY, a total of 166 children were included for analysis, with 83 developing T1D. Matching on age, sex, T1D family history, and clinical center was confirmed between groups. T1D cases had a higher proportion of DR3/4 HLA genotypes compared with controls. The average age at which DAISY cases developed T1D was older than TEDDY at 9.48 years compared with 2.86 years, respectively.Table 1Participant characteristics for DAISY and TEDDYDAISYTEDDYT1D control N=75T1D case N=76Total N=151T1D control N=83T1D case N=83Total N=166N (%) or mean (SD)N (%) or mean (SD)Sex (female)30 (40.0)37 (48.7)67 (44.4)36 (43.4)36 (43.4)72 (43.4)Non-Hispanic White (yes)68 (90.7)68 (89.5)136 (90.1)57 (68.7)55 (66.3)112 (67.5)T1D family history (yes)41 (54.7)54 (71.1)95 (62.9)27 (32.5)27 (32.5)54 (32.5)Age at islet autoimmunity onset (years)–3.78 (3.12)––1.34 (0.72)–Age at T1D onset (years)–9.48 (4.79)––2.86 (1.37)–Human leukocyte antigen-DR3/4 (yes)14 (18.7)34 (44.7)48 (31.8)32 (38.6)48 (57.8)80 (48.2)Platform (450K)37 (49.3)38 (50.0)75 (49.7)0 (0)0 (0)0 (0)Clinical center Georgia/Florida5 (6.0)5 (6.0)10 (6.0) Washington5 (6.0)5 (6.0)10 (6.0) Colorado75 (100)76 (100)151 (100)12 (14.5)12 (14.5)24 (14.5) Sweden25 (30.1)25 (30.1)50 (30.1) Germany10 (12.0)10 (12.0)20 (12.0) Finland26 (31.3)26 (31.3)52 (31.3)DAISY, Diabetes Autoimmunity Study in the Young; T1D, type 1 diabetes; TEDDY, The Environmental Determinants of Diabetes in the Young.We identified multiple immune ratios associated with T1D (figure 2). In the pre-SV period, the NLR was 15% higher (ratio: 1.15, 95% CI 1.00 to 1.33; p=0.0442), while the CD4T/CD8T was 9% lower (ratio: 0.91, 95% CI 0.83 to 1.00; p=0.0448) in T1D cases compared with controls after adjusting for covariates. Examining the individual immune cell components, these changes appear driven by decreases in CD4T and increases in CD8T and an increase in neutrophil and decrease in lymphocyte proportions in cases compared with controls for the CD4T/CD8T and NLR, respectively (online supplemental figures 2–5). The remaining immune cell ratio associations were null for the pre-SV timepoint. In the SV period, the B-CD4T-CD8T-mem/nv ratio was estimated to be 26% lower (ratio: 0.74, 95% CI 0.57 to 0.98; p=0.0325) among T1D cases compared with controls, after covariate adjustments. Based on the individual components, this difference seems to follow from lower memory and higher naive lymphocytes among cases versus controls (online supplemental figures 2–5). The other immune cell ratio associations during SV did not yield statistically significant effects.Figure 2Meta-analysis: differences in mean immune cell ratios between T1D cases and controls. The x-axis represents the immune cell ratio between T1D cases and controls, and the y-axis signifies the different seroconversion time points. Interpretation: an immune cell ratio <1 indicates a lower immune cell ratio among T1D cases compared with controls, and vice versa for an immune cell ratio >1. In the pre-SV to SV change, an immune cell ratio <1 indicates a reduced trajectory among T1D cases compared with controls, and vice versa for an immune cell ratio >1. B-CD4T-CD8T-mem/nv, B-CD4T-CD8T-memory/B-CD4T-CD8T-naive ratio; B-mem/nv, B-memory/B-naive ratio; CD4T/CD8T, CD4T/CD8T ratio; CD4T-mem/nv, CD4T-memory/ CD4T-naive ratio; CD8T-mem/nv, CD8T-memory/CD8T-naive ratio; DAISY, Diabetes Autoimmunity Study in the Young; MLR, monocyte/lymphocyte ratio; NLR, neutrophil/lymphocyte ratio; SV, seroconversion; TEDDY, The Environmental Determinants of Diabetes in the Young; T1D, type 1 diabetes.Results for the pre-SV to SV change are detailed in figures 2 and 3. For the pre-SV to SV change analysis, we found that the B-mem/nv, B-CD4T-CD8T-mem/nv, and NLR ratio trajectories were reduced by 35% (ratio: 0.65, 95% CI 0.43 to 0.96; p=0.0324), 38% (ratio: 0.62, 95% CI 0.41 to 0.95; p=0.0273), and 21% (ratio: 0.79, 95% CI 0.66 to 0.95; p=0.0119), respectively, among T1D cases compared with controls, after adjusting for covariates (figure 2). These shifts appear related to smaller increases in B and T memory cells and smaller decreases in B and T naive cells, along with smaller increases in neutrophils and smaller decreases in lymphocytes, when comparing cases to controls (online supplemental figures 6 and 7). T1D cases started with higher immune cell ratios for the B-mem/nv, B-CD4T-CD8T-mem/nv, and NLR, but controls had more pronounced increases at seroconversion (figure 3). No other significant associations were found for the additional immune cell ratios during the pre-SV to SV change.Figure 3Meta-analysis: immune cell ratio changes between T1D cases and controls around seroconversion. The x-axis signifies the different seroconversion time points, and the y-axis represents the estimated immune cell ratio values, which are stratified by T1D cases (red) and controls (blue). Interpretation: immune cell ratio changes between T1D cases and controls across the pre-SV and SV time points. B-CD4T-CD8T-mem/nv, B-CD4T-CD8T-memory/B-CD4T-CD8T-naive ratio; B-mem/nv, B-memory/B-naive ratio; CD4T/CD8T, CD4T/CD8T ratio; CD4T-mem/nv, CD4T-memory/CD4T-naive ratio; CD8T-mem/nv, CD8T-memory/CD8T-naive ratio; MLR, monocyte/lymphocyte ratio; NLR, neutrophil/lymphocyte ratio; SV, seroconversion; T1D, type 1 diabetes.The effects of the DAISY and TEDDY studies were largely consistent (figure 2), and we did not observe significant heterogeneity across studies. Results were robust to sensitivity analyses excluding samples with imputed values, and we did not observe immune cell ratio differences by T1D family history, though, we were likely underpowered for the three-way interaction.Conclusions Among children who developed T1D, we identified early and dynamic changes in DNAm-derived immune cell ratios compared with controls around the time of seroconversion to IA. T1D cases displayed elevated NLR and reduced CD4T/CD8T cell ratios prior to SV, followed by lower B-CD4T-CD8T-mem/nv ratios during SV and attenuated increases in the NLR, B-mem/nv, and B-CD4T-CD8T-mem/nv ratios over time. Early NLR and CD4T/CD8T alterations were observed at median age values of 3.53 (DAISY) and 0.27 (TEDDY) years prior to the first detection of persistent autoantibodies, highlighting their potential as early biomarkers for risk prediction and stratification. The attenuated changes in mem/nv ratios during the appearance of autoimmunity may indicate immune dysfunction or immaturity in children who go on to develop T1D.We found that T1D cases started with higher NLR pre-SV, but they did not increase as much pre-SV to SV compared with controls. The NLR is considered a biomarker of systemic inflammation or activation of the innate and adaptive immune response, yet its role has not been examined in children at risk for T1D. Reduced peripheral neutrophil counts have been previously identified among IA cases,29 including in a subset of TEDDY children,30 and may reflect neutrophilic infiltration in the pancreas.29 The elevated NLR at pre-SV in T1D cases may be capturing an innate immune response to an environmental stimulus or otherwise detecting neutrophil activation along with a decline in peripheral lymphocytes, both of which may precede or initiate autoimmunity.31 Thus, the NLR could have potential utility as an early marker of T1D risk prior to IA. The reduced NLR change from pre-SV to SV in T1D cases may suggest several physiological developments after the initial inflammatory response. Specifically, the decrease in circulating neutrophils could reflect pancreatic infiltration and contribution to insulitis, while the increase in lymphocytes may indicate B, T, and NK cell recruitment as part of autoimmunity.T1D cases had lower CD4T/CD8T ratios at pre-SV compared with controls, which appear to be driven by concurrent decreases in CD4T and increases in CD8T. The ratio between CD4+ and CD8+ T cells marks immune senescence and activation, as well as the balance between helper and cytotoxic T cell functions within the adaptive immune system. It has seen predominant use in studies of HIV susceptibility and ongoing viral replication, where lower CD4T/CD8T ratios reflected reduced activity of CD4+ T helper cells and augmented activity of CD8+ cytotoxic T cells.32 Reductions in the CD4T/CD8T ratio among participants who developed T1D have been observed previously, as early as 1989.10 Prior studies of individual cell subsets are similarly consistent. Adults with new-onset T1D have decreased CD4+ and increased CD8+ T cell proportions, suggesting a similar relationship between helper and cytotoxic T cells.9 Altered CD4T and CD8T cell proportions also precede the appearance of autoimmunity in high-risk children, though the specific changes may differ based on the first-appearing autoantibody.33 Interestingly, HLA class II genetic variation may contribute to CD4T/CD8T ratio levels in T1D.34 These findings support the clinical utility of the CD4T/CD8T ratio as a biomarker, not just as a measure of effective immune modulation in clinical trials35 but also as an early indicator of disease risk. This early shift in helper versus cytotoxic T cells could suggest functional impairment of the CD4+ cells, including regulatory T cells, and subsequent activation of CD8+ cells into targeting pancreatic beta cells.We also found that T1D cases had lower B-CD4T-CD8T-mem/nv ratios at SV and attenuated changes for B-mem/nv and B-CD4T-CD8T-mem/nv ratios compared with controls. The composite B and T cell memory/naive ratios serve as relative measures of balance between the cells of the adaptive immune system. While little information exists on how the ratios change with time during the development of T1D, increases in individual naive B cell and T cell subsets have been observed in new-onset T1D previously9 36 37 and may be useful predictors of C-peptide decline.37 Prior studies highlight the role of B cells in T1D pathogenesis, including therapeutic B cell depletion with rituximab, which preserved beta-cell function for 1 year,38 and variations in B cell proportions that may involve a loss of anergic B cells and reduced immune tolerance, contributing to autoreactivity.39 Age-related differences in B cell infiltration have also been observed, with younger patients exhibiting higher proportions of B cells in pancreatic tissue and higher levels among those who lost insulin secretion more quickly,37 40 41 which may be relevant given the younger age of our study population. Our findings may signify an expansion or persistence of naive B and T cells and a reduction in memory B and T cells during IA initiation, resulting in lower mem/nv ratios at SV. These patterns could reflect an impaired ability to activate memory subsets following antigen exposure. In this way, people who go on to develop T1D may possess an immature immunological phenotype that predisposes them to autoimmunity and subsequent T1D.A major strength of our study includes longitudinal DNAm data prior to the onset of IA and thereafter, which allowed this novel look at immune cell profiles just before and during seroconversion. Given the natural changes in immune cell composition with age, having well-matched controls was another major strength. Meta-analysis of DAISY and TEDDY improved our power to detect these changes and established generalizability to high-risk children in Europe and the USA. Results were consistent across these two prominent birth cohort studies. Use of the newest high-resolution cell deconvolution reference panel provides greater granularity over the common six-cell panel, including allowing us to distinguish between memory and naive cell populations, although through indirect measures of immune cell composition.Our study also possessed some limitations. Not having access to direct flow cytometry data, we utilized cell deconvolution of peripheral blood samples for immune cell estimates with a reference panel derived from healthy adult donors, which included both sexes and several ethnic and ancestral backgrounds. Despite the unavailability of flow cytometry data, we leveraged DNAm data to estimate these proportions with a highly accurate, reference-based immune profiling method previously used in cancer settings.42 Regardless of the difference in age for the reference, our immune cell proportions seemed to correspond to previously reported reference ranges for this age group, although at lower overall ranges, which may be related to the autoimmune process.26 Even with differences in age at T1D and IA onset across DAISY and TEDDY, immune profile changes were robust across studies, independent of age. The cross-sectional comparisons at pre-SV and SV may be more susceptible to unmeasured confounding or between-person heterogeneity (eg, baseline differences in immune composition related to HLA) and should be interpreted with this in mind. Still, these cross-sectional results were consistent across the two study groups. The pre-SV to SV change analyses are inherently more robust to such confounding and represent our most reliable findings. Future research should consider the creation and utilization of a peripheral blood reference panel for DNAm data specific to children and adolescents, since immunological development and subsequent changes occur with age. Even though we did not find differential proportions of NK cells or have the ability to distinguish other immune cells, such as macrophages and dendritic cells, previous work highlights their importance in the autoimmune process for T1D and warrants further investigation.33 43In this epidemiological analysis leveraging two well-characterized, prospective cohorts, we identified distinct alterations in immune cell ratios that precede and accompany IA in children who progress to T1D. The findings herein suggest potential roles for the NLR and CD4T/CD8T ratio in serving as early biomarkers for T1D risk stratification, even before the onset of IA. Having the ability to identify these high-risk individuals provides a unique opportunity for public health intervention. In addition, children who develop T1D appear to have lower B-CD4T-CD8T mem/nv ratios and decreased trajectories for the NLR, B mem/nv, and B-CD4T-CD8T mem/nv ratios and may serve as indicators of immune dysfunction or immaturity and T1D progression. Together, these findings underscore the importance of temporal dynamics in immune cell profiles as novel biomarkers of early disease progression and offer potential targets for improved risk stratification and timely intervention aimed at delaying or preventing T1D. Moreover, high-resolution cell deconvolution represents a cost-effective, scalable approach for observational studies with DNAm data to profile immunodynamics specific to disease in the absence of flow cytometry. Future work should verify these results through functional analyses to determine the biological relevance of the variations in the DNAm-derived immune cell proportions and immune cell ratios.