Simulating The Effect of Sugar Substitutes on Blood Glucose and Lipid Levels in Patients with Type 2 Diabetes

Abstract

This simulation study aims to evaluate the effect of sugar substitutes (stevia, erythritol, and aspartame) on blood glucose levels and lipid profiles in patients with type 2 diabetes. A 12-week clinical simulation was conducted with 150 patients divided into six groups. The methodology involved generating time-dependent data for fasting glucose (FBS), hemoglobin A1c, cholesterol, and triglycerides, which were then analyzed using a bidirectional LSTM model to predict temporal changes with an accuracy of up to 99.5%.
Simulation results showed that stevia and erythritol reduced FBS from 175–178 mg/dL to 130–143 mg/dL and triglycerides from 208–212 mg/dL to 182–190 mg/dL in the compliant groups (≥70% dietary adherence), while the non-compliant groups maintained elevated values (>160 mg/dL FBS, >210 mg/dL TG) with fluctuations. The LSTM model outperformed conventional models (R² ≥0.995, RMSE <2 mg/dL FBS), supporting its use in designing personalized dietary plans.