How would patients behave if they were continually informed of their blood glucose levels? A simulation study using a "virtual" patient.

Background: The emergence of real-time glucose sensors for people with diabetes may replace discontinuous monitoring (self-monitored blood glucose [BG]) in the future. In this study, we use a computer-generated "virtual" patient to predict changes in behavior that may result from an increased awareness of BG levels.

Methods: The employed strategy required educated patients with type 1 diabetes to simulate a virtual patient using the DIABLOG Scientific simulator, interactive computer program. Thirty patients with a mean age of 34 years and duration of diabetes of 18 years (15 with continuous subcutaneous insulin infusion, 15 with intensive conventional therapy) simulated several daily glucose profiles with conventional self-monitoring of BG, using the time-lapse function of the program. Thereafter they had access to the actual glucose value either in a watch-type display or in a graphical display. Behavioral changes were monitored and stored.

Results: Mean BG value improved from 154 to 139 mg/dL (P < 0.05). The analysis of the process and behavioral changes revealed that patients recognized an impending hypoglycemia with the sensor in 94% of cases (59% without sensor) and reacted adequately in 98% of cases in order to avoid hypoglycemia. The frequency of hypoglycemia could be reduced from 1.7 per week to 0.5 per week. Unnecessary interventions (mostly by administration of carbohydrates) doubled to 3.2 per week using the continuous measurement system. Impending hyperglycemia could only be prevented in 25% of cases without a sensor, and with sensor this ratio could only be marginally improved to 29%. Supplementary insulin administration resulted in hypoglycemia only in a few cases.

Conclusions: With the continuous measurement of their actual BG, subjects could reduce the frequency of hypoglycemia by 50% but increased the number of unnecessary interventions (excess carbohydrate intake or prematurely without a real hypoglycemia threat). Hyperglycemia prevention is a more difficult task. Simulation with an in silico disease model is a realistic alternative to studies in patients. A simulation program such as DIABLOG could be valuable for education in order to more rapidly and reliably recognize impending hypo- and hyperglycemia episodes.