T1T2 Classification
Diagnosing type 1 and 2 diabetes
Diagnosing diabetes type is fundamental for treatment: treatment guidelines are very different. These differences are driven by the development of severe insulin deficiency in Type 1 diabetes which leads to absolute insulin requirement, high glucose variability (and therefore need for physiological insulin replacement), and limited benefit from most non-insulin therapies.
In many cases it can be difficult to distinguish between type 1 and type 2 diabetes. Features pointing to both type 1 and type 2 diabetes may be present (for example young age but high body weight), and biomarker tests at diagnosis are imperfect: islet autoantibody testing can have both false negative and false positive results. A beta version of a prediction model that combines clinical features and, where available, islet autoantibody test results (currently only applicable to patients aged 18-50 at diagnosis, of white European origin) is available here: https://www.diabetesgenes.org/t1dt2d-prediction-model/
Using clinical features to differentiate type 1 and 2 diabetes
While textbooks and guidelines list many features for differentiating type 1 and type 2 diabetes most of these have very little evidence, and may add little to discrimination. On systematic review in 2015, age at diagnosis and BMI were the only evidence based features at diagnosis that robustly discriminated type 1 and type 2 diabetes, with age the most discriminatory factor.
Results from a recent study of 1048 newly diagnosed participants aged 18-50 are shown below, ordered by most to least discriminative clinical feature, with diagnosis based on a combination of C-peptide and antibody testing. Importantly no single feature confirms type 1 diabetes (T1D) particularly in older people where type 1 diabetes is less likely (as illustrated by the 5% prevalence shown here).
| Feature | AUCROC^ | Sensitivity (T1D) | Specificity (T1D) | Positive predictive value with 33% T1D* | Positive predictive value with 5% T1D+ |
| BMI (<25 for sens/spec) | 0.85 | 48.7% | 92.4% | 76.1% | 25% |
| Age at diagnosis (<30 for sens/spec) | 0.78 | 49.6% | 89.7% | 70.5% | – |
| Weight loss | 0.76 | 80.1% | 74.7% | 61.1% | 14.3% |
| Diagnosis HbA1c
(>115mmol/mol for sens/spec) |
0.72 | 35.1% | 87.1% | 57.5% | 12.5% |
| Diagnosis Glucose (>20mmol/L) | 0.72 | 51.5% | 76.3% | 51.9% | 10.3% |
| Lack of FH diabetes | 0.69 | 58.7% | 79.6% | 58.8% | 13.2% |
| Osmotic symptoms | 0.61 | 94.7% | 27.6% | 39.4% | 6.4% |
| ketoacidosis | 0.61 | 30.3% | 91.4% | 63.7% | 15.6% |
| Other autoimmune disease | 0.52 | 10.2% | 94.2% | 46.6% | 8.5% |
^Area under receiver operating characteristic curve. A higher value represented higher discrimination, 1=perfect test, 0.5 = no discrimination.
* cohort prevalence Type 1 diabetes – higher than true prevalence for this age group (approximately 15%) due to an excess of secondary care recruitment
+ Approximate population prevalence Type 1 diabetes at age 50.
Early insulin requirement strongly predicts type 1 diabetes
People diagnosed with type 2 diabetes who progress to insulin within 3 years have a high probability of Type 1 diabetes with subsequent absolute insulin deficiency. Therefore all participants in this group should undergo biomarker testing for type 1 diabetes.
Pitfalls for diagnosing type 1 diabetes in older people
Diagnosing type 1 diabetes in the elderly is difficult, as even those with ‘classic’ clinical features of type 1 diabetes (such as ketoacidosis) may be more likely to have type 2 diabetes. This is because type 2 diabetes is extremely common in this age group, so the ‘prior prevalence’ of type 1 diabetes is low. Therefore a clinical diagnosis of type 1 diabetes should be confirmed by islet autoantibody or C-peptide testing (see below). Due to low prior prevalence a negative islet antibody test result in an older adult makes Type 1 diabetes unlikely. Islet antibodies should not be tested in older adults who do not have features to suggest type 1 diabetes, as in this situation a large proportion of positive islet antibody results will be false positive (>50% with many assays).
Use of islet antibody and C-peptide tests to differentiate type 1 and 2 diabetes
Further information on islet antibody and C-peptide tests are given here
C-peptide: https://www.exeterlaboratory.com/test/c-peptide-plasma/
Antibody testing: https://www.exeterlaboratory.com/test/gad-antibodies/
Close to diagnosis (within 3 years) islet autoantibody tests (GADA, IA2 and/or Znt8) are the most appropriate biomarker test for diagnosing diabetes subtypes. C-peptide may be relatively preserved at diagnosis of type 1 diabetes, particularly in the presence of obesity (insulin resistance), so while a low result may confirm type 1 diabetes a high result close to diagnosis is usually unhelpful. In long duration diabetes C-peptide is the most appropriate diagnostic test, as it is loss of endogenous insulin secretion that defines the glycaemic treatment requirements of Type 1 and 2 diabetes.