Background: We designed this study to develop and validate a prevalence model for latent autoimmune diabetes in adults (LADA) among people initially diagnosed with type 2 diabetes mellitus (T2DM).

Material/Methods: The study recruited 930 patients aged 318 years who were diagnosed with T2DM within the past year. Demographic information, medical history, and clinical biochemistry records were collected. Logistic regression was used to develop a regression model to distinguish LADA from T2DM. Predictors of LADA were identified in a subgroup of patients (n=632) by univariate logistic regression analysis. From this we developed a prediction model using multivariate logistic regression analysis and tested its sensitivity and specificity among the remaining patients (n=298).

Results: Among 930 recruited patients, 880 had T2DM (96.4%) and 50 had LADA (5.4%). Compared to T2DM patients, LADA patients had fewer surviving b cells and reduced insulin production. We identified age, ketosis, history of tobacco smoking, 1-hour plasma glucose (1hPG-AUC), and 2-hour C-peptide (2hCP-AUC) as the main predictive factors for LADA (P<0.05). Based on this, we developed a multivariable logistic regression model: Y=-8.249-0.035(X1)+1.755(X2)+1.008(X3)+0.321(X4)-0.126(X5), where Y is diabetes status (0=T2DM, 1=LADA), X1 is age, X2 is ketosis (1=no, 2=yes), X3 is history of tobacco smoking (1=no, 2=yes), X4 is 1hPG-AUC, and X5 is 2hCP-AUC. The model has high sensitivity (78.57%) and selectivity (67.96%).

Conclusions: This model can be applied to people newly diagnosed with T2DM. When Y 30.0472, total autoantibody screening is recommended to assess LADA.