Skip to content

Type 2 diabetes — biomarkers

TL;DR — HbA1c is the dominant diagnostic and monitoring biomarker but is an indirect glycation measure affected by erythrocyte lifespan, haemoglobin biology, kidney disease and treatment context. Glucose, C-peptide, autoantibodies, UACR and eGFR answer different questions; none is a complete “T2D biomarker.” Data-driven clusters, polygenic scores and multi-omics improve mechanistic description but have not yet demonstrated better treatment outcomes than ordinary clinical variables (Dennis 2019, PMID 31047901; Tobias 2023, PMID 37794253). The most useful current biomarkers are decision-linked: identify diagnosis, insulin deficiency, organ risk or response.

Diagnostic glycaemia

Test Diabetes threshold Strength Major limitation
HbA1c ≥6.5% No fasting; chronic exposure Red-cell/haemoglobin confounding
Fasting plasma glucose ≥126 mg/dL Standardised snapshot Day-to-day variability
2-hour OGTT ≥200 mg/dL Detects post-load disease Burdensome and less reproducible
Random glucose ≥200 mg/dL with symptoms Acute clinical utility Requires compatible presentation

Diagnostic criteria and confirmation rules are summarised in ADA Standards (ADA 2026, PMID 41358893).

HbA1c discordance

Cause Direction can be Response
Iron deficiency Falsely high in some settings Evaluate blood count/iron context
Haemolysis/blood loss Falsely low Use glucose-based assessment
Haemoglobin variant Assay-specific Check method and variant interference
CKD/erythropoiesis treatment Variable Integrate CGM/SMBG and clinical context
Recent rapid glycaemic change Lags current state Use contemporaneous glucose
Transfusion Uninterpretable transiently Delay or use glucose

HbA1c does not show hypoglycaemia or variability. A value of 7% can represent stable glucose or alternating severe highs and lows.

C-peptide and classification

C-peptide reflects endogenous insulin secretion and is most useful when classification or insulin deficiency is uncertain. Interpretation requires concurrent glucose, renal function, timing after food/stimulation and exogenous insulin context.

Question Biomarker combination
T1D/LADA vs T2D Islet autoantibodies + C-peptide
Severe insulin deficiency Stimulated or contextual C-peptide
Monogenic possibility Phenotype, family history, antibodies, C-peptide, genetics
Remission potential Duration, treatment, weight loss and secretion measures; no validated single test

Organ-risk biomarkers

Marker Domain Decision link
UACR Kidney/endothelial CKD risk and therapy intensity
eGFR Kidney Staging, dosing, SGLT2 eligibility
Troponin/BNP Cardiac in selected contexts HF/CV risk, not routine universal T2D diagnosis
Retinal imaging Microvascular Detects treatable eye disease
Monofilament/vibration Nerve/foot Ulcer-risk stratification
FIB-4 Liver fibrosis triage Determines need for second-line assessment

Subtype biomarkers

Ahlqvist clusters combine clinical and model-derived variables and predict different complication patterns (Ahlqvist 2018, PMID 29503172). However, categories are unstable across time and treatment, and simple continuous features performed at least as well for treatment response in trial datasets (Dennis 2019, PMID 31047901).

Genetics and multi-omics

Large genetic studies identify pathway-specific susceptibility across ancestries (Suzuki 2024, PMID 38374256). Multi-omic panels can improve incident-risk prediction in cohorts, but calibration, assay stability, cost and clinical-action thresholds limit use (Carrasco-Zanini 2024, PMID 37889320).

Technology Promise Required proof
Polygenic risk score Earlier lifetime-risk stratification Incremental utility and ancestry portability
Proteomics/metabolomics Near-term disease and complication signals External validation and decision impact
Epigenetics Exposure-linked risk Temporal stability and causal meaning
CGM-derived phenotypes Dynamic response patterns Outcome-improving treatment algorithm
Imaging ectopic fat Mechanistic remission prediction Scalable thresholds and prospective validation

Remission prediction

Weight loss and shorter disease duration currently outperform speculative molecular panels as practical predictors. DiRECT’s dose-response—86% remission with ≥15 kg loss—sets a high benchmark for added predictive value (Lean 2018, PMID 29221645).

Biomarker qualification pathway

Stage Required demonstration
Analytical validity Accurate, precise and stable measurement
Clinical validity Reproducible association with state/outcome
Incremental validity Adds beyond ordinary clinical variables
Clinical utility Changes a decision and improves outcome
Transportability Works across ancestry, sex, age and setting
Implementation Affordable, interpretable and actionable

Many candidate papers stop at association or internal discrimination. A high area under the curve in one cohort does not show that using the marker improves health.

Monitoring-frequency logic

Marker Repeat when
HbA1c Therapy/exposure changes or stable interval review
UACR Confirm abnormality and track CKD risk
eGFR At risk-dependent interval and after relevant changes
C-peptide Classification/insulin-deficiency question, not routine serial target
Autoantibodies Classification uncertainty; repeated routine testing rarely useful
Retinal image Risk- and finding-dependent interval
FIB-4 Age/risk-dependent liver-fibrosis triage

Derived measures and their assumptions

Measure Inputs Fragility
HOMA2-B/HOMA2-IR Fasting glucose + insulin/C-peptide Invalid in unstable glycaemia/exogenous insulin contexts
eGFR Creatinine/cystatin C + demographics Muscle mass and assay effects
FIB-4 Age, AST, ALT, platelets Age inflation and non-liver influences
Time in range CGM readings Device wear and target definition
Glycaemic variability CGM/SMBG series Sampling density

Diagnostic misclassification control

Red flag Biomarker response
Ketosis or rapid weight loss Ketones, C-peptide, autoantibodies
Young onset with strong vertical family history Monogenic pathway
Very rapid oral-treatment failure Reassess insulin reserve/classification
Apparent HbA1c-glucose mismatch CBC, haemoglobin/renal context, alternate glycaemia
Lean phenotype with systemic disease Consider pancreatic/endocrine/medication causes

Biobank-to-clinic equity risk

Genetic and omic panels trained predominantly in European-ancestry cohorts can lose calibration elsewhere. A marker that improves average prediction but systematically underestimates risk in an underrepresented group can widen inequity. External validation must therefore report calibration and decision consequences, not only rank discrimination.

Precision cardiovascular prognostic models illustrate the gap between model performance and decision utility, while remission consensus shows why biomarker endpoints require agreed clinical states (Ahmad 2024, PMID 38253823; Riddle 2021, PMID 34462270).

Cross-domain evidence crosswalk

These adjacent studies constrain interpretation of this page and make explicit where its conclusions depend on prevention, organ-outcome, remission, burden or implementation evidence.

Verified evidence anchor Connection
(Sun 2022, PMID 34879977) Sun H, et al. IDF Diabetes Atlas: Global, regional and country-level diabetes prevalence estimates for 2021 and projections for 2045. Diabetes Res Clin Pract. 2022;183:109119
(GBD 2023, PMID 37356446) GBD 2021 Diabetes Collaborators. Global, regional, and national burden of diabetes from 1990 to 2021, with projections of prevalence to 2050. Lancet. 2023;402:203-234
(Taylor 2024, PMID 39038473) Taylor R. Understanding the cause of type 2 diabetes. Lancet Diabetes Endocrinol. 2024;12:664-673
(Taylor 2021, PMID 33289165) Taylor R. Type 2 diabetes and remission: practical management guided by pathophysiology. J Intern Med. 2021;289:754-770
(Taylor 2016, PMID 30058916) Taylor R. Calorie restriction and reversal of type 2 diabetes. Expert Rev Endocrinol Metab. 2016
(Knowler 2002, PMID 11832527) Knowler WC, et al. Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. N Engl J Med. 2002
(Diabetes 2015, PMID 26377054) Diabetes Prevention Program Research Group. Long-term effects of lifestyle intervention or metformin on diabetes development and microvascular complications over 15-year follow-up: the DPP Outcomes Study. Lancet Diabetes Endocrinol. 2015
(Frías 2021, PMID 34170647) Frías JP, et al. Tirzepatide versus Semaglutide. N Engl J Med. 2021
(GRADE 2022, PMID 36129997) GRADE Study Research Group. Glycemia Reduction in Type 2 Diabetes. N Engl J Med. 2022
(Adler 2024, PMID 38772405) Adler AI, et al. UKPDS 91: 24-year post-trial monitoring. Lancet. 2024;404:145-155
(TODAY 2021, PMID 34320286) TODAY Study Group. Long-Term Complications in Youth-Onset Type 2 Diabetes. N Engl J Med. 2021
(Marso 2016, PMID 27633186) Marso SP, et al. Semaglutide Cardiovascular Outcomes. N Engl J Med. 2016
(Perkovic 2019, PMID 30990260) Perkovic V, et al. Canagliflozin Renal Outcomes. N Engl J Med. 2019
(Perkovic 2024, PMID 38785209) Perkovic V, et al. Semaglutide on CKD. N Engl J Med. 2024
(Speight 2024, PMID 38128969) Speight J, et al. Bringing an end to diabetes stigma and discrimination: an international consensus statement on evidence and recommendations. Lancet Diabetes Endocrinol. 2024

Diagnostic thresholds trade sensitivity for specificity

Using OGTT as reference, a 37-study meta-analysis estimated HbA1c ≥6.5% sensitivity 50% (95% CI 42%–59%) and specificity 97.3% (95.3%–98.4%) in previously undiagnosed adults (Kaur 2020, PMID 33216783). A 73-study network meta-analysis produced similar estimates: HbA1c sensitivity/specificity 0.51/0.96, fasting glucose ≥126 mg/dL 0.49/0.98, and either-positive testing 0.64/0.95 (Duong 2023, PMID 36760402). Thus, concordant high results are specific, but any one test misses people identified by another.

Use case Biomarker logic Failure mode
Population screening Lower threshold can increase sensitivity More false positives and confirmatory testing
Clinical diagnosis Standard thresholds preserve specificity and prognostic continuity Discordant HbA1c/FPG/OGTT is common
Rapid symptomatic deterioration Glucose and ketones reflect current state HbA1c may lag and misclassify acute change
Haemoglobin/erythrocyte disorder Glucose-based tests bypass red-cell bias Single fasting/OGTT results have biological variability

C-peptide: useful classification, unstable standardisation

C-peptide's strongest role is in insulin-treated people when classification changes management. Persistence of substantial secretion three to five years after diagnosis supports T2D/monogenic disease, whereas absent C-peptide indicates absolute insulin requirement regardless of phenotype (Jones 2013, PMID 23413806). Proposed thresholds—<0.2 nmol/L suggesting T1D and >0.6 nmol/L suggesting T2D—cannot be treated as universal cut-points because assays remain materially non-standardised (Schleicher 2025, PMID 40781799).

Interpretation must state fasting/random/stimulated context, concurrent glucose, kidney function and exogenous insulin use. A biomarker with excellent analytic discrimination can still have low clinical utility if it does not change treatment or if a mixed phenotype lies between cut-points.

Genetics and proteomics: incremental prediction is modest

Platform Quantified evidence Clinical-utility boundary
Polygenic score in an Indigenous cohort Adult clinical AUC 0.728→0.735; youth 0.805→0.812; birth cohort 0.614→0.685 Statistical improvement does not establish better prevention outcomes (Wedekind 2023, PMID 36862161)
Multi-ancestry T2D GWAS 1.4 million people; 568 associations, including 318 newly reported loci; complication associations Locus discovery and drug-target interaction are not prescribing tests (Vujkovic 2020, PMID 32541925)
Trans-ancestry glycaemic-trait GWAS Up to 281,416 people, 30% non-European; 242 loci, 99 novel; credible sets narrowed median 37.5% Better mapping reduces but does not eliminate ancestry inequity (Chen 2021, PMID 34059833)
Plasma-protein ASCVD score C-index +0.014 (95% CI 0.002–0.028) beyond clinical factors in primary prevention Modest discrimination gain; not T2D-specific and no strategy trial (Helgason 2023, PMID 37606673)

Genomic reviews reach opposing but compatible conclusions: polygenic scores predict incident T2D, yet BMI and ordinary clinical variables can identify risk groups more efficiently, and risk disclosure alone has not improved health (Meigs 2019, PMID 31332628). Scores may be more useful for mechanistic subphenotypes or early-life risk, but that proposition needs prospective, ancestry-diverse intervention trials.

A live PubMed E-utilities search through 2026-08-30 found biomarker-association, validation and prediction studies, but no randomised biomarker-use-versus-usual-assessment strategy trial demonstrating better patient outcomes. This is a dated implementation-evidence gap, not evidence that biomarker discovery has stopped.

Biomarker controversies

  • Diagnosis versus prediction: HbA1c thresholds define disease; proteomic/genetic scores estimate future risk. Their calibration questions differ.
  • Association versus utility: a hazard ratio or AUC increment is insufficient; use of the test must improve a decision or outcome.
  • Universal cut-points versus context: C-peptide, FIB-4 and HbA1c all change meaning with assay, age, kidney function or physiology.
  • Equity versus portability: multi-ancestry discovery improves resolution, but validation, calibration and access must be local.
  • Subtype labels versus continuous traits: forcing continuous biology into clusters may simplify communication while losing predictive information.

Open questions

  • Can a biomarker-guided therapy strategy improve hard outcomes over clinical risk-based treatment?
  • Which marker predicts durable remission independently of achieved weight loss? (Lean 2024, PMID 38423026)
  • How should HbA1c targets be adjusted when red-cell biology creates persistent discordance?
  • Can multi-omic risk tools transfer across ancestries without widening inequity?

References

  1. ADA Professional Practice Committee. Diagnosis and Classification of Diabetes: Standards of Care-2026. Diabetes Care. 2026. PMID 41358893
  2. Ahlqvist E, et al. Novel subgroups of adult-onset diabetes. Lancet Diabetes Endocrinol. 2018;6:361-369. PMID 29503172
  3. Dennis JM, et al. Disease progression and treatment response in T2D subgroups. Lancet Diabetes Endocrinol. 2019;7:442-451. PMID 31047901
  4. Tobias DK, et al. Precision diabetes medicine consensus. Nat Med. 2023. PMID 37794253
  5. Suzuki K, et al. Genetic drivers of heterogeneity in T2D pathophysiology. Nature. 2024;627:347-357. PMID 38374256
  6. Carrasco-Zanini J, et al. Multi-omic prediction of incident T2D. Diabetologia. 2024;67:102-112. PMID 37889320
  7. Lean MEJ, et al. Primary care-led weight management for remission. Lancet. 2018. PMID 29221645
  8. Lean MEJ, et al. Five-year follow-up of DiRECT. Lancet Diabetes Endocrinol. 2024. PMID 38423026
  9. Riddle MC, et al. Definition and Interpretation of Remission. Diabetes Care. 2021. PMID 34462270
  10. Ahmad A, et al. Precision prognostics for cardiovascular disease in T2D. Commun Med. 2024;4:11. PMID 38253823
  11. Sun H, et al. IDF Diabetes Atlas: Global, regional and country-level diabetes prevalence estimates for 2021 and projections for 2045. Diabetes Res Clin Pract. 2022;183:109119. PMID 34879977
  12. GBD 2021 Diabetes Collaborators. Global, regional, and national burden of diabetes from 1990 to 2021, with projections of prevalence to 2050. Lancet. 2023;402:203-234. PMID 37356446
  13. Taylor R. Understanding the cause of type 2 diabetes. Lancet Diabetes Endocrinol. 2024;12:664-673. PMID 39038473
  14. Taylor R. Type 2 diabetes and remission: practical management guided by pathophysiology. J Intern Med. 2021;289:754-770. PMID 33289165
  15. Taylor R. Calorie restriction and reversal of type 2 diabetes. Expert Rev Endocrinol Metab. 2016. PMID 30058916
  16. Knowler WC, et al. Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. N Engl J Med. 2002. PMID 11832527
  17. Diabetes Prevention Program Research Group. Long-term effects of lifestyle intervention or metformin on diabetes development and microvascular complications over 15-year follow-up: the DPP Outcomes Study. Lancet Diabetes Endocrinol. 2015. PMID 26377054
  18. Frías JP, et al. Tirzepatide versus Semaglutide. N Engl J Med. 2021. PMID 34170647
  19. GRADE Study Research Group. Glycemia Reduction in Type 2 Diabetes. N Engl J Med. 2022. PMID 36129997
  20. Adler AI, et al. UKPDS 91: 24-year post-trial monitoring. Lancet. 2024;404:145-155. PMID 38772405
  21. TODAY Study Group. Long-Term Complications in Youth-Onset Type 2 Diabetes. N Engl J Med. 2021. PMID 34320286
  22. Marso SP, et al. Semaglutide Cardiovascular Outcomes. N Engl J Med. 2016. PMID 27633186
  23. Perkovic V, et al. Canagliflozin Renal Outcomes. N Engl J Med. 2019. PMID 30990260
  24. Perkovic V, et al. Semaglutide on CKD. N Engl J Med. 2024. PMID 38785209
  25. Speight J, et al. Bringing an end to diabetes stigma and discrimination: an international consensus statement on evidence and recommendations. Lancet Diabetes Endocrinol. 2024. PMID 38128969
  26. Kaur G, et al. Diagnostic accuracy of tests for type 2 diabetes and prediabetes. PLoS One. 2020;15:e0242415. PMID 33216783
  27. Duong KNC, et al. Comparison of diagnostic accuracy for diabetes diagnosis. Front Med. 2023;10:1016381. PMID 36760402
  28. Jones AG, Hattersley AT. Clinical utility of C-peptide measurement. Diabet Med. 2013;30:803-817. PMID 23413806
  29. Schleicher E, et al. Call for Standardization of C-Peptide Measurement. Clin Chem Lab Med. 2025. PMID 40781799
  30. Meigs JB. Genetic Epidemiology of Type 2 Diabetes: Opportunities for Health Translation. Curr Diab Rep. 2019;19:62. PMID 31332628
  31. Wedekind LE, et al. Utility of a T2D polygenic score in an Indigenous study population. Diabetologia. 2023;66:1401-1416. PMID 36862161
  32. Vujkovic M, et al. Discovery of 318 new risk loci for T2D and vascular outcomes. Nat Genet. 2020;52:680-691. PMID 32541925
  33. Chen J, et al. Trans-ancestral genomic architecture of glycaemic traits. Nat Genet. 2021;53:840-860. PMID 34059833
  34. Helgason H, et al. Evaluation of Large-Scale Proteomics for Prediction of Cardiovascular Events. JAMA. 2023;330:725-735. PMID 37606673