Diabetes Classification: The Evidence Behind the Pathway

A 34-year-old schoolteacher comes to the outpatient clinic with her file. She was told she had type 2 diabetes four years ago, at 30. She is slim, with a body mass index of 23. She takes metformin and glimepiride, and her HbA_1c has crept up to 9.8 per cent in spite of them. She also takes levothyroxine for hypothyroidism. Her mother has diabetes. The resident beside you has written “add a third agent” on the sheet. This is an illustrative case, not a real patient.

The Diabetes Classification Pathway will help you with her. It asks the questions that narrow the type, tells you which test to send and which result means what, and prints the drug consequence of each answer. What it no longer carries is some of the why. The table of the names people use, and the Indian studies behind several of its steps, sat in its Evidence tab and have moved here, so that the tool stays quick at the bedside. By the end of this article you will know where those numbers come from, and how far each one can be trusted.

1. The names patients bring

Start with the words, because patients and residents arrive with them. The American Diabetes Association recognises four headings: type 1, type 2, specific types due to other causes, and gestational diabetes. The World Health Organization’s 2019 revision adds a fifth grouping, hybrid forms. Everything else is a label, and some labels mean nothing at all. Here is how each one stood when the tool carried this table, on 4 November 2026.

“Sir, so is type 3 diabetes real?” As a form of diabetes mellitus, no. The phrase was proposed by Steen, de la Monte and colleagues for a hypothesis about insulin signalling in Alzheimer’s disease (Steen 2005). Type 4 is a name from a mouse study (Bapat 2015). Type 5 is new and real, but its formal criteria are still being written (Wadivkar 2025). When a patient says “type 1.5”, hear LADA, and when someone says “type 2.5”, ask whether they mean double diabetes, and use that name.

2. Why a young, thin Indian can still have type 2 diabetes

“She was 30 and slim. Isn’t that type 1?” Not necessarily, and the reason is the Indian phenotype. The national picture first. ICMR-INDIAB, 113,043 people across mainland India reported in 2023: 11.4 per cent with diabetes and 15.3 per cent with prediabetes, roughly 101 million and 136 million people (Anjana 2023).

It arrives about a decade early. Prevalence peaks at 52 to 56 years in South Asians against 64 to 68 in white Europeans, with a sharp rise between 25 and 34. So a diagnosis at 30 argues against type 2 far more weakly here than the Western textbooks imply.

And the body mass index misleads. In a nationwide registry of 37,927 people with type 2 diabetes only 16 per cent had a body mass index below 25, yet 77 per cent of men and 71 per cent of women met a body-fat-percentage definition of obesity. That is the thin-fat phenotype: a normal-looking index with too much fat for the frame. How much Indian type 2 diabetes is truly lean is unsettled: 16 per cent in that urban registry, 52 per cent in a rural tertiary series, no reliable national figure. The figures above are studies, not denominators, and the tool does not print a frequency of its own.

The common mistake runs both ways. It is easy to call every young, thin patient type 1, and just as easy to call every adult type 2. Neither the age nor the index settles it. A test does.

3. Whom to test for LADA

Latent autoimmune diabetes in adults, LADA, is autoimmune diabetes that comes on slowly enough to be mistaken for type 2. The Fourlanos screening tool gives one point each for onset before 50, acute osmotic symptoms, a body mass index below 25, a personal autoimmune disease and a family autoimmune disease. Two or more, send the antibody. In its derivation the tool had a sensitivity of 90 per cent and a specificity of 71 per cent (Fourlanos 2006).

“Ninety per cent sounds good. So two points means she has LADA?” No. That is the part people get wrong. The positive predictive value is only about 21 per cent, so most patients who score two or more do not have LADA. The value of the tool is the other end: a score of 0 or 1 gives a negative predictive value of 99 per cent, and you can stop. It is a rule-out tool, and the antibody does the ruling in.

Our schoolteacher scores three: onset before 50, a body mass index below 25, and her own thyroid disease. Her GAD65 antibody comes back positive. She has LADA. That changes her drug: the consensus panel does not recommend sulfonylureas in LADA, and the tool prints that beside the result. For the antibody itself, see the article on reading an autoantibody report linked at the end.

4. MODY, and why Indian MODY may be different

“Her mother has diabetes too. Should we send a MODY panel instead?” The tool uses the Exeter criteria for that: diagnosis before 25 in at least one family member, two or three generations in a parent-to-child pattern, absent islet autoantibodies and preserved C-peptide. She fails two of them already: one affected parent and a positive antibody. So no.

But when you do send a MODY test in India, know what came back in the one Indian series. Indian MODY may not be the same disease genetically. Targeted sequencing of 56 clinically diagnosed patients at CMC Vellore found a pathogenic variant in only 19 per cent, against about 30 per cent or more in Western laboratories, and the commonest gene was NEUROD1, not GCK or HNF1A, with HNF4A next (Chapla 2015). Eleven mutation-positive patients is a signal, not a frequency. A negative panel here does not prove the clinical picture wrong, and a panel that tests only the three common Western genes may miss the Indian ones.

5. Two Indian forms with small numbers behind them

Fibrocalculous pancreatic diabetes. A lean young adult, years of abdominal pain, pancreatic calculi on a plain film, then brittle diabetes that does not go into ketosis. Southern India has the highest known prevalence; a Chennai survey put it at 0.36 per cent of self-reported diabetes and 0.019 per cent of the population (Mohan 2008). Rare, then, but worth naming, because enzyme replacement changes the patient’s life and the cancer risk is high.

Ketosis-prone diabetes. A patient who presents in ketoacidosis and later comes off insulin. The first Indian series, CMC Vellore: 11 adults with unprovoked ketoacidosis, antibody-negative with preserved beta cell function, against 23 with classical antibody-positive type 1. All 11 came off insulin within three to four months (Gupta 2017). A case series, not a denominator. It is why the tool asks for an antibody and a C-peptide at three months before calling the insulin lifelong.

6. DIPSI, and what it misses

India’s national programme tests pregnant women with DIPSI: 75 g of glucose whatever the last meal, one venous sample at two hours, 140 mg/dL or above. It exists because bringing a woman back fasting for a three-sample test is often not possible. It captures no fasting value and is the less sensitive of the two tests. How much less? One analysis found it missed about 16 per cent of the women IADPSG identified, a Sri Lankan study reported a sensitivity of 40.6 per cent, and a 2023 meta-analysis of eight comparative Indian studies gave a pooled sensitivity of 0.44, 95% CI 0.29 to 0.60, with a specificity of 0.97.

“So should we stop using DIPSI?” No. A positive DIPSI is very likely right, because the specificity is high. A negative one is less reassuring. Add a fasting sample where the woman can return fasting; do not abandon a test that is national guidance and that many women can complete. The tool says exactly that.

Back to the clinic

Our schoolteacher did not get a third tablet. She had LADA. The glimepiride was stopped, and her treatment moved towards insulin as her C-peptide and her glucose required. The age and the index had pointed to type 2; the history pointed elsewhere, and the antibody settled it. Each of those steps is a line in the tool.

Try this

A 22-year-old man is admitted in diabetic ketoacidosis. He is overweight, with acanthosis on his neck, and his father and grandfather both have diabetes. On the second day the resident sends a C-peptide to “prove type 1”. It comes back low. What do you do? Illustrative case.

Answer. Do not believe that C-peptide yet. A sample taken within two weeks of a hyperglycaemic emergency reads low, because high glucose itself suppresses insulin secretion. Send the antibodies, treat the ketoacidosis, and repeat the C-peptide on follow-up. If he is antibody-negative with preserved C-peptide at three months, he has ketosis-prone type 2 diabetes, and he may come off insulin, as all 11 patients in the Vellore series did.

Take-home points

• In India a diagnosis at 30 and a normal body mass index do not exclude type 2 diabetes; nor do they prove type 1.
• The Fourlanos tool rules LADA out at 0 or 1; at 2 or more it only tells you to send GAD65.
• LADA and HNF1A or HNF4A MODY need opposite drugs, so a wrong label inverts the treatment.
• Indian MODY may involve other genes than the Western three; a negative panel is not the last word.
• DIPSI is specific but misses many cases; add a fasting sample whenever the woman can return fasting.

Read with this: Reading an Autoantibody Report: The Evidence Behind the Numbers, for the antibodies themselves, and Pregnancy Dating and the Antenatal Schedule: The Evidence Behind the Tool, for the screening schedule in pregnancy.

Declaration: the author built and maintains the tool. It is free to use. The figures in this article are those the tool carried in its Evidence tab on 4 November 2026. Where a figure is attributed, the attribution is the tool’s own reference whose title matches; the figures given without a name (the age peaks, the registry of 37,927, the rural series, the 16 per cent analysis and the Sri Lankan study) were printed by the tool without a source, and are reproduced as it printed them. The tool is in administrator preview, pending clinical review.

References

1. American Diabetes Association Professional Practice Committee. Diagnosis and classification of diabetes: standards of care in diabetes 2026. Diabetes Care. 2026;49(suppl 1):S27-S49.
2. World Health Organization. Classification of diabetes mellitus. Geneva: WHO; 2019.
3. Buzzetti R, Tuomi T, Mauricio D, et al. Management of latent autoimmune diabetes in adults: a consensus statement from an international expert panel. Diabetes. 2020;69(10):2037-2047.
4. Fourlanos S, Perry C, Stein MS, Stankovich J, Harrison LC, Colman PG. A clinical screening tool identifies autoimmune diabetes in adults. Diabetes Care. 2006;29(5):970-975.
5. Ellard S, Bellanne-Chantelot C, Hattersley AT. Best practice guidelines for the molecular genetic diagnosis of maturity-onset diabetes of the young. Diabetologia. 2008;51(4):546-553.
6. Shields BM, McDonald TJ, Ellard S, Campbell MJ, Hyde C, Hattersley AT. The development and validation of a clinical prediction model to determine the probability of MODY in patients with young-onset diabetes. Diabetologia. 2012;55(5):1265-1272.
7. Chapla A, Mruthyunjaya MD, Asha HS, et al. Maturity onset diabetes of the young in India: a distinctive mutation pattern identified through targeted next-generation sequencing. Clin Endocrinol (Oxf). 2015;82(4):533-542.
8. Holt RIG, DeVries JH, Hess-Fischl A, et al. The management of type 1 diabetes in adults: a consensus report by the American Diabetes Association and the European Association for the Study of Diabetes. Diabetologia. 2021;64(12):2609-2652.
9. Hope SV, Wienand-Barnett S, Shepherd M, et al. Practical classification guidelines for diabetes in patients treated with insulin: a cross-sectional study of the accuracy of diabetes diagnosis. Br J Gen Pract. 2016;66(646):e315-e322.
10. Balasubramanyam A, Garza G, Rodriguez L, et al. Accuracy and predictive value of classification schemes for ketosis-prone diabetes. Diabetes Care. 2006;29(12):2575-2579.
11. Gupta RD, Ramachandran R, Gangadhara P, et al. Clinical characteristics, beta-cell dysfunction and treatment outcomes in patients with A-β+ ketosis-prone diabetes: the first identified cohort amongst Asian Indians. J Diabetes Complications. 2017;31(9):1401-1407.
12. Unnikrishnan AG, Bhake RC, Kumaran S, Kalra S. Fibrocalculous pancreatic diabetes. In: Endotext. South Dartmouth: MDText.com; updated 13 September 2024.
13. Mohan V, Farooq S, Deepa M. Prevalence of fibrocalculous pancreatic diabetes in Chennai in south India. JOP. 2008;9(4):489-492.
14. Ewald N, Bretzel RG. Diabetes mellitus secondary to pancreatic diseases (type 3c): are we neglecting an important disease? Eur J Intern Med. 2013;24(3):203-206.
15. Wadivkar P, Jebasingh F, Thomas N, et al. Classifying a distinct form of diabetes in lean individuals with a history of undernutrition: an international consensus statement. Lancet Glob Health. 2025;13(10):e1771-e1776.
16. International Diabetes Federation. New IDF type 5 diabetes working group. Brussels: IDF; 15 April 2025.
17. Sharif A, Chakkera H, de Vries APJ, et al. International consensus on post-transplantation diabetes mellitus. Nephrol Dial Transplant. 2024;39(3):531-549.
18. Suh S, Park MK. Glucocorticoid-induced diabetes mellitus: an important but overlooked problem. Endocrinol Metab (Seoul). 2017;32(2):180-189.
19. Scheen AJ, Feve B. When therapeutic drugs lead to diabetes. Diabetologia. 2022;65(5):751-762.
20. Anjana RM, Unnikrishnan R, Deepa M, et al. Metabolic non-communicable disease health report of India: the ICMR-INDIAB national cross-sectional study. Lancet Diabetes Endocrinol. 2023;11(7):474-489.
21. Research Society for the Study of Diabetes in India. RSSDI clinical practice recommendations for the management of type 2 diabetes mellitus 2022.
22. Ministry of Health and Family Welfare, Government of India. Diagnosis and management of gestational diabetes mellitus: national guidelines. New Delhi: National Health Mission; 2018.
23. Meta-analysis comparing Diabetes in Pregnancy Study Group India (DIPSI) versus the International Association of the Diabetes and Pregnancy Study Groups (IADPSG) criteria for the diagnosis of gestational diabetes among the Indian population. Taiwan J Obstet Gynecol. 2023;62(4):498-505.
24. World Health Organization. Diagnostic criteria and classification of hyperglycaemia first detected in pregnancy. Geneva: WHO; 2013.
25. Steen E, Terry BM, Rivera EJ, et al. Impaired insulin and insulin-like growth factor expression and signaling mechanisms in Alzheimer’s disease: is this type 3 diabetes? J Alzheimers Dis. 2005;7(1):63-80.
26. Bapat SP, Myoung Suh J, Fang S, et al. Depletion of fat-resident Treg cells prevents age-associated insulin resistance. Nature. 2015;528(7580):137-141.