Development and validation of a nonalcoholic fatty liver disease-based self-diagnosis tool for diabetes.

Background: Prediction of type 2 diabetes mellitus (DM) has been studied widely. However, a hospital visit was necessary to apply previous prediction models for the evaluation of DM. This study was conducted to develop and validate a hospital visit-free self-diagnosis tool for DM. Methods: Participants who underwent health screening between 2017-2018 (n=7,519; training cohort) and 2019-2020 (n=7,564; validation cohort) were extracted from the Korea National Health and Nutrition Examination Survey (KNHANES). DM was defined as doctor-diagnosed DM in a questionnaire. Logistic regression was used to determine independent predictors for DM, and a multivariable logistic regression-based nomogram was developed for the prediction of DM, which was validated in a cohort consisting of an independent population. The presence of nonalcoholic fatty liver disease (NAFLD) was operationally defined using the KNHANES-NAFLD score. Results: Age, sex, waist circumference, systolic blood pressure, total cholesterol, triglyceride, aspartate aminotransferase, blood urea nitrogen, urinary protein, urinary glucose, and NAFLD were identified as independent predictors for DM. After excluding laboratory variables that require laboratory tests, a simplified multivariable model was conducted based on hospital visit-free variables, including age, sex, waist circumference, systolic blood pressure, and NAFLD. The full and simplified prediction models for DM were presented as nomograms. In the independent validation cohort, the full and simplified DM prediction models were validated with an area under the curve values of 0.903 and 0.824 from the receiver operating characteristic curves, respectively. Conclusions: Involvement of NAFLD has allowed satisfactory prediction of DM without laboratory tests that require a hospital visit. The developed model may be promising in terms of early diagnosis of DM among individuals without hospital visits and may reduce the socioeconomic burden of DM in the real-world, which awaits future prospective trials to confirm.