Skip to content

Type 1 diabetes — biomarkers

TL;DR — No single biomarker measures the entire T1D process. Autoantibodies identify islet autoimmunity and multiplicity defines stage 1/2; glucose marks metabolic progression; stimulated C-peptide measures residual β-cell function; genetic scores estimate inherited susceptibility (Insel 2015, PMID 26404926; Palmer 2004, PMID 14693724; Sharp 2019, PMID 30655379). Children with multiple autoantibodies had 69.7% 10-year progression in pooled birth cohorts, but individual timing remains uncertain (Ziegler 2013, PMID 23780460). Composite longitudinal scores outperform isolated measurements for timing, while direct β-cell-death assays remain research tools (Weiss 2022, PMID 36028774; Usmani-Brown 2014, PMID 25004096).

Biomarker map

Question Biomarker class Example Principal limitation
Who is susceptible? Germline genetics HLA, T1D genetic risk score Ancestry calibration; risk is not disease
Has autoimmunity begun? Autoantibodies IAA, GADA, IA-2A, ZnT8A Indirect; assay and persistence matter
How close is stage 3? Metabolic trajectory OGTT glucose, HbA1c drift, PLS Sampling burden and variable tempo
How much function remains? Secretory function Mixed-meal C-peptide AUC Influenced by glucose, age, renal function
Is cell death active? Cell-free DNA Unmethylated INS DNA Low abundance and limited validation
Is therapy engaging target? Immune pharmacodynamics Lymphocyte subsets, exhaustion markers Engagement need not predict clinical benefit

Autoantibodies

Two or more persistent antibodies plus normoglycemia define stage 1; dysglycemia defines stage 2 (Insel 2015, PMID 26404926). In pooled prospective childhood cohorts, multiple positivity predicted 69.7% progression at 10 years and 84% at 15 years after seroconversion, versus 14.5% 10-year risk with one antibody (Ziegler 2013, PMID 23780460).

Pattern Inference Do not infer
One unconfirmed positive Possible risk marker Established stage 1
Persistent single antibody Heterogeneous elevated risk Inevitable progression
Multiple confirmed antibodies Early-stage T1D if metabolic criteria fit Exact diagnosis date
Antibody-negative clinical diabetes Requires classification work-up Automatically type 2 diabetes

Confirmation on a second sample is recommended because low-prevalence screening magnifies the consequences of false positives (Phillip 2024, PMID 38912694).

C-peptide

C-peptide is secreted equimolarly with endogenous insulin and is not present in injected insulin. The ADA workshop established stimulated C-peptide as the principal β-cell-function outcome for preservation trials (Palmer 2004, PMID 14693724).

T1D decline is biphasic in longitudinal data: a faster initial exponential fall followed by a more stable phase in many participants (Shields 2018, PMID 29880650). Age and duration strongly affect interpretation; a detectable value does not exclude T1D, especially in adult onset (Evans-Molina 2025, PMID 40230204).

C-peptide context Strength Confounder
Mixed-meal AUC Standard trial outcome Labor and standardization
Fasting Simple Lower sensitivity; concurrent glucose
Random with glucose Practical classification aid Meal timing and renal clearance
Longitudinal change Measures trajectory Assay/platform drift

Metabolic prediction

OGTT glucose abnormalities define stage 2, but repeated testing is burdensome. The progression-likelihood score integrates longitudinal glucose and C-peptide and separated presymptomatic substages in screened children (Weiss 2022, PMID 36028774). Its transport to adult onset, different ancestries, and routine laboratories remains unresolved.

Genetics

HLA contributes the largest inherited effect; DR3/DR4-DQ8 confers high risk, while DRB115:01-DQA101:02-DQB1*06:02 can remain protective even among autoantibody-positive relatives (Noble 2012, PMID 22315720; Pugliese 2016, PMID 26822082). GWAS identified more than 40 loci by 2009 and subsequent fine mapping expanded functional hypotheses (Barrett 2009, PMID 19430480; Tomlinson 2014, PMID 25008175).

The improved T1D genetic risk score supports newborn enrichment and adult classification, but ancestry portability is an empirical requirement, not an assumption (Sharp 2019, PMID 30655379; Hu 2025, PMID 40569436).

Direct β-cell injury markers

Differential methylation of insulin DNA can distinguish β-cell-derived circulating DNA. Droplet digital PCR detected β-cell death signals in research settings (Usmani-Brown 2014, PMID 25004096). Low concentration, rapid clearance, tissue specificity, preanalytics, and uncertain relationship to recoverable function prevent routine diagnostic use.

Biomarkers in treatment selection

Teplizumab increased KLRG1+TIGIT+CD8+ cells while delaying stage 3, but exploratory immune correlates require validation before becoming treatment-selection tests (Herold 2019, PMID 31180194). Abatacept changed T-follicular-helper, naïve CD4, and regulatory T-cell frequencies without meeting the stage-1 primary progression endpoint (Russell 2023, PMID 36920087). Pharmacodynamic change is therefore not a surrogate for clinical success.

Minimum validation pathway

  1. Analytic validity: precision, calibration, stability, interference.
  2. Clinical validity: prospective discrimination and calibration.
  3. Added value beyond age, antibodies, and glucose.
  4. External validation across ancestry, age, geography, and assay platform.
  5. Clinical utility: a changed decision that improves net outcomes.
  6. Monitoring for inequitable false-positive and false-negative rates.

C-peptide is an outcome, not a complete clinical surrogate

In 944 DCCT/EDIC participants with mean T1D duration 35 years, 117 (12.4%) retained meal-stimulated C-peptide. Severe hypoglycemia history occurred in 27% with peak C-peptide >0.2 nmol/L and 48% with 0.03–0.2, versus 74% with 0.003–0.03 and 70% with none; advanced microvascular complication rates were similar (Gubitosi-Klug 2021, PMID 33529168). Residual secretion therefore relates strongly to hypoglycemia but is not validated as a surrogate for every complication.

C-peptide proposition Supporting result Limitation
Mixed-meal AUC is sensitive Standard trial endpoint; captures stimulated reserve Two- to four-hour visit burden
Fasting C-peptide can triage severe deficiency 96% sensitivity and 100% specificity in 57 Chinese youth Small, single-population study
Detectability persists for decades 12.4% responders after mean 35 years in DCCT/EDIC Selected survivor cohort
More secretion reduces severe hypoglycemia Marked gradient above 0.03 nmol/L Observational association
Stimulation adds yield Dutch cohort detected residual secretion in 10% more than fasting Clinical significance varies by level

The Dutch Biomarkers of Heterogeneity cohort enrolled 611 people with ≥5 years’ duration and a second group of 160 with ≥35 years. Stimulated testing detected secretion in an additional 10%, and fasting secretion was associated with lower impaired-awareness risk (Aanstoot 2024, PMID 38904129).

Longitudinal metabolic markers outperform snapshots

Among 749 antibody-positive children with similarly abnormal baseline DPTRS, progressors and nonprogressors diverged within six months in glucose AUC, C-peptide/glucose ratio, and Index60. Nonprogressors’ longitudinal patterns resembled antibody-negative relatives despite marked baseline impairment (Sims 2022, PMID 36326757). A high-risk snapshot can thus regress or stabilize.

In 93 multiple-antibody-positive TrialNet relatives, 29/48 CGM metrics differed by later progression, but individual CGM AUCs were only 0.50–0.69; four of seven OGTT measures were about 0.80 and Index60/DPTRS outperformed adjusted CGM models (Ylescupidez 2023, PMID 37572381). Dynamic CGM methods remain exploratory: entropy and Poincaré area discriminated low versus high immunological risk with AUC 0.72 (95% CI 0.58–0.86) and 0.66 (0.47–0.86), respectively (Montaser 2024, PMID 38820084).

Autoantibody features beyond count

Feature Evidence Use boundary
Titer Five-year risk 19% for lowest-quartile GADA to 60% for highest-quartile IA-2A Assay-specific thresholds
Affinity Persistent high-affinity single IAA can behave unlike low-affinity IAA Specialized methods
Reversion More common in later and low-affinity single positivity Repeat sampling required
Epitope specificity Can distinguish risk within GADA and IAA Limited routine standardization
Sequence/age IAA-first peaks in infancy; GADA-first is later Childhood cohort evidence

Across 24,662 children, type-specific titer thresholds stratified five-year risk from 6% to 75% (Ng 2022, PMID 34758977). In two German birth cohorts, 44% of single-IAA and 24% of single-GADA children spread to multiple antibodies; persistent high-affinity single IAA or GADA positive by both radiobinding and ELISA carried >50% ten-year progression (Giannopoulou 2015, PMID 26138334). These data argue against treating all single positives alike (So 2021, PMID 33881515).

Direct β-cell-death assays remain analytically fragile

Unmethylated INS DNA exploits β-cell-specific methylation, but assay versions differ in primers, bisulfite recovery, reference denominator, and low-copy handling. A multiplex preproinsulin assay tracked the expected death sequence in three mouse models (Fisher 2013, PMID 23825129), while reviews conclude that no circulating stress/death marker is fully validated for presymptomatic clinical use (Mirmira 2016, PMID 27541297). Biological specificity does not remove technical noise.

Predictive versus response biomarkers

A predictive biomarker identifies differential treatment effect; a prognostic marker predicts outcome regardless of treatment. Seven-year AbATE follow-up found durable immunologic changes and less C-peptide loss among previously defined teplizumab responders, but no lower HbA1c or insulin use (Perdigoto 2019, PMID 30569273). Post-randomization responder definitions can generate hypotheses but cannot establish a baseline selection test.

Reanalysis of stage-1 abatacept suggests baseline insulin secretion may modify effect: high secretors gained 15.8 progression-free months (95% CI 4.85–26.68) and had HR 0.46 (0.25–0.84), with interaction HR 2.92 (1.23–6.96) (Galderisi 2026, PMID 41237315). Because the parent trial missed its primary endpoint, this subgroup requires prospective confirmation before treatment selection.

Open questions

  • Can a minimally invasive composite predict stage 3 within 6–12 months across ages? (Weiss 2022, PMID 36028774)
  • Which single-antibody states represent abortive autoimmunity? (Ziegler 2013, PMID 23780460)
  • Can cell-free β-cell DNA become quantitatively reproducible enough for response monitoring? (Usmani-Brown 2014, PMID 25004096)
  • How should genetic scores be recalibrated without reifying ancestry categories? (Sharp 2019, PMID 30655379; Hu 2025, PMID 40569436)

References

  1. Palmer JP, et al. C-peptide is the appropriate outcome measure for T1D trials. Diabetes. 2004;53:250-264. PMID 14693724
  2. Noble JA, Erlich HA. Genetics of type 1 diabetes. Cold Spring Harb Perspect Med. 2012;2:a007732. PMID 22315720
  3. Ziegler AG, et al. Seroconversion to multiple islet autoantibodies and progression. JAMA. 2013;309:2473-2479. PMID 23780460
  4. Usmani-Brown S, et al. Analysis of β-cell death by droplet digital PCR. Endocrinology. 2014;155:3694-3698. PMID 25004096
  5. Insel RA, et al. Staging presymptomatic T1D. Diabetes Care. 2015;38:1964-1974. PMID 26404926
  6. Pugliese A, et al. Protective HLA haplotype in autoantibody-positive relatives. Diabetes. 2016;65:1109-1119. PMID 26822082
  7. Shields BM, et al. C-Peptide Decline in T1D Has Two Phases. Diabetes Care. 2018;41:1486-1492. PMID 29880650
  8. Sharp SA, et al. Improved Type 1 Diabetes Genetic Risk Score. Diabetes Care. 2019;42:200-207. PMID 30655379
  9. Weiss A, et al. Progression likelihood score. Diabetologia. 2022;65:2121-2131. PMID 36028774
  10. Phillip M, et al. Monitoring IAb-Positive Pre-Stage 3 T1D. Diabetes Care. 2024;47:1276-1298. PMID 38912694
  11. Evans-Molina C, et al. T1D presenting in adults. Diabetes Obes Metab. 2025. PMID 40230204
  12. Barrett JC, et al. GWAS identifies over 40 T1D loci. Nat Genet. 2009;41:703-707. PMID 19430480
  13. Tomlinson MJ, et al. Fine mapping at chromosome 16p13.13. Diabetes. 2014;63:4360-4368. PMID 25008175
  14. Herold KC, et al. Teplizumab in Relatives at Risk. N Engl J Med. 2019;381:603-613. PMID 31180194
  15. Russell WE, et al. Abatacept for Delay of Stage-1 T1D Progression. Diabetes Care. 2023;46:1005-1013. PMID 36920087
  16. Hu J, et al. T1D genetic risk score in a Chinese population. Diabetologia. 2025;68:1969-1982. PMID 40569436
  17. Gubitosi-Klug RA, et al. Residual β cell function in long-term type 1 diabetes associates with reduced incidence of hypoglycemia. J Clin Invest. 2021;131. PMID 33529168
  18. Aanstoot HJ, et al. Cohort profile: Biomarkers of heterogeneity in type 1 diabetes. BMJ Open. 2024;14:e082453. PMID 38904129
  19. Sims EK, et al. Persistence of β-Cell Responsiveness for Over Two Years in Autoantibody-Positive Children. Diabetes Care. 2022;45:2982-2990. PMID 36326757
  20. Ylescupidez A, et al. OGTT Metrics Surpass CGM Data for T1D Prediction in Multiple-Autoantibody-Positive Individuals. J Clin Endocrinol Metab. 2024;109:57-67. PMID 37572381
  21. Montaser E, et al. Novel Detection and Progression Markers for Diabetes Based on CGM Data Dynamics. J Clin Endocrinol Metab. 2025;110:254-262. PMID 38820084
  22. Ng K, et al. Islet Autoantibody Type-Specific Titer Thresholds Improve Risk Stratification. Diabetes Care. 2022;45:160-168. PMID 34758977
  23. Giannopoulou EZ, et al. Islet autoantibody phenotypes and incidence in children at increased risk. Diabetologia. 2015;58:2317-2323. PMID 26138334
  24. So M, et al. Advances in Type 1 Diabetes Prediction Using Islet Autoantibodies: Beyond a Simple Count. Endocr Rev. 2021;42:584-604. PMID 33881515
  25. Fisher MM, et al. Detection of islet β-cell death in vivo by multiplex PCR of differentially methylated DNA. Endocrinology. 2013;154:3476-3481. PMID 23825129
  26. Mirmira RG, et al. Biomarkers of β-Cell Stress and Death in Type 1 Diabetes. Curr Diab Rep. 2016;16:95. PMID 27541297
  27. Perdigoto AL, et al. Treatment of type 1 diabetes with teplizumab: follow-up after 7 years. Diabetologia. 2019;62:655-664. PMID 30569273
  28. Galderisi A, et al. Baseline Insulin Secretion Determines Response to Abatacept in Stage 1 Type 1 Diabetes. Diabetes. 2026;75:229-240. PMID 41237315