MATHEMATICAL MODELING AND IN-SILICO EVALUATION SYSTEM FOR THE DIAGNOSIS AND TREATMENT

Abstract

Improvement of the diagnosis and treatment of diabetes is concomitant with medical, economic, and social issues. To resolve these issues, mathematical modeling and in-silico evaluation approaches can be useful. They could aid in achieving a better understanding of the inherent characteristics of the glucose regulatory system and a diverse evaluation of glucose control algorithms. Accordingly, in this dissertation, the feasibility of mathematical modeling to obtain the inherent characteristics of the glucose regulatory system and assess the clinical indices is first demonstrated. In addition, the validity of using in-silico evaluation to assess various glucose control algorithms prior to clinical trials is demonstrated. The feasibility of mathematical modeling is assessed by three mathematical modeling studies for a better understanding of intrinsic glucose regulatory characteristics based on clinical data on Korean individuals. Firstly, a mathematical model that considered incretin hormones was identified based on the clinical data during the oral glucose tolerance test (OGTT) (n = 8). The model allows prediction of intrinsic hormone responses during an isoglycemic intravenous glucose infusion (IIGI) study required to calculate the incretin effect. Secondly, a computational method using the individualized model was developed to calculate glucose infusion rates more accurately during an IIGI study. The clinical trial was performed to evaluate the developed method by comparing it to the ad-hoc method (n = 18). The computational method exhibited higher correlation (0.95 ± 0.03 vs. 0.86 ± 0.10, P = 0.019) and lower error (root mean square error, 10.33 ± 1.99 mg/dL vs. 16.84 ± 4.43 mg/dL

P = 0.002) between the glucose levels from the OGTT and the IIGI study than the ad-hoc method. Lastly, a simpler mathematical model was applied to assess the ability of glucose tolerance in Korean individuals with normal glucose tolerance (NGT) (n = 8) and type 2 diabetes (T2D) (n = 14). Simulation results confirmed that the dynamic β-cell responsivity, static β-cell responsivity, total β-cell responsivity, and insulin sensitivity in the T2D group were lower than those in the NGT group. In addition, validity of using in-silico evaluations for assessing various glucose control algorithms prior to clinical trials is demonstrated by two studies focusing on characteristics of glucose control in hospitalized patients with critical and non-critical illnesses. In both studies, virtual patient models were implemented, and efficacy and safety of certain clinical glucose control algorithms were evaluated in computational environment. Firstly, the simulation results for glucose control in critically ill patients confirmed that all the implemented glucose control algorithms were effective in reducing the incidence of hyperglycemia but some were prone to potential incidences of hypoglycemia. Lastly, the simulation results for glucose control in non-critically ill hospitalized patients with T2D confirmed that basal bolus insulin therapy (BBIT) was more effective than sliding-scale insulin therapy (SSIT). This corresponds with what was reported in a previous clinical study. The performed in-silico trials indicated that BBIT, which includes daily adjustments of the total insulin dose, showed better glucose control than BBIT, which adjusts only the basal insulin dose. The performed in-silico trials also indicated that patients with a severe reduction in renal function were more vulnerable to rapid decreases in blood glucose levels, resulting in hypoglycemia. Thus, a gradual increase in the total daily dose, starting with a reduced dosage, is required in such patients. In conclusion, it was confirmed that mathematical modeling was useful to better understand the inherent characteristics of the glucose regulatory system and assess clinical indices, especially for Koreans. It was also demonstrated that in-silico evaluation was effective in assessing the performance of glucose control algorithms prior to clinical studies. Therefore, it is concluded that mathematical modeling and in-silico evaluation approaches are beneficial to improve the diagnosis and treatment of diabetes without increasing risks or costs.