The most expensive error in the module is automatically writing "E11" for every adult with hyperglycaemia and a BMI of 29. ADA 2026 estimates that up to 40% of adults with new type 1 diabetes are initially mistaken for type 2. The MoH RK 2022 protocol gives a convenient but crude table: young age and acute onset with ketonuria versus obesity, hypertension and family history. It is crude because in a real clinic both columns are mixed together.
ADA 2026 Recommendation 2.5: classify people with hyperglycaemia into the appropriate diagnostic categories in order to personalise management (level E). The categories: (1) type 1 diabetes — autoimmune destruction of β-cells, including latent autoimmune diabetes in adults; (2) type 2 diabetes — non-autoimmune progressive loss of secretion, often against a background of resistance; (3) specific types (monogenic, diseases of the exocrine pancreas, drug-induced); (4) gestational.
Clinical anchors that shift the probability towards type 1 in an adult (ADA 2026, Fig. 2.1, based mainly on white European populations): age at diagnosis <35 years, BMI <25 kg/m², unintentional weight loss, ketoacidosis, plasma glucose >360 mg/dL (>20 mmol/L) at presentation. No single feature works alone. ADA offers the mnemonic AABBCC: Age, Autoimmunity, Body habitus, Background (family history of type 1 diabetes), Control (failure of non-insulin regimens), Comorbidities (oncology checkpoint inhibitors).
What to do about antibodies. Recommendation 2.10: standardised islet autoantibodies are indicated for classification in adults whose phenotypic features overlap with type 1 diabetes. The panel: GAD (first), then IA-2 and/or ZnT8; in those not yet treated with insulin, insulin autoantibodies may be useful. The MoH RK 2022 protocol includes the same specifications in its differential diagnosis table. A negative panel does not exclude type 1 diabetes: 5–10% of adults with type 1 diabetes have no antibodies.
C-peptide is not "a test for type on the day of a crisis". ADA: do not measure it within 2 weeks of a hyperglycaemic catastrophe; in those already on insulin, measure a random sample with simultaneous glucose. Reference points from Fig. 2.1: <200 pmol/L supports type 1 diabetes; >600 pmol/L favours type 2 or preserved secretion; 200–600 is a grey zone. These data are calibrated mainly on white European cohorts; protocol No. 158 does not formulate blind transfer of the 200 pmol/L threshold as "RK law".
A special trap is ketosis-prone diabetes: a person with a type 2 phenotype (often from particular racial/ethnic groups; ADA names Black and Hispanic/Latino adults) may present with DKA and then temporarily not need insulin. The initial need for insulin here is often transient. The opposite trap is latent autoimmune diabetes in adults (LADA): slow autoimmune loss in an adult, months or years without insulin. ADA includes LADA under type 1 diabetes so that insulin is not delayed until DKA.
Immune checkpoint inhibitors (ICIs, anti-PD-1/PD-L1) cause autoimmune diabetes in 0.6–1.4% of those receiving therapy; fewer than half have antibodies, and there is usually no remission as in classic type 1 diabetes. This is not "type 2 diabetes in a cancer patient".
The conclusion. Classification is a working hypothesis that is revised if "typical T2DM" unexpectedly progresses to ketoacidosis or cannot be maintained without insulin. An error in type is not an academic dispute: it means missed insulin or unnecessary insulin.