Efficacy of automatic bolus calculator with automatic speech recognition in patients with type 1 diabetes: A randomized cross-over trial

Background: Patients using an insulin pump as part of their diabetes treatment need to calculate insulin bolus doses to compensate for a meal. Some patients do not modify their meal boluses according to changes in the amount and composition of food products in a meal. The lack of correct meal boluses leads to unstable, and therefore harmful, blood glucose levels. The aim of the present study was to test a system supporting bolus determination based on a voice description of a meal. Methods: The bolus calculator developed (VoiceDiab) consists of a smartphone application and three remote servers for automatic speech recognition, text analysis, and insulin dosage calculation. Forty-four people with type 1 diabetes (T1D) treated with continuous subcutaneous insulin infusion finished the randomized cross-over study. Patients were randomly allocated to the group in which the VoiceDiab system supported bolus calculation or to an unsupported group, in which patients or their caregivers calculated boluses. After a 14-day washout period, patients from the supported group were switched to the unsupported group, whereas those in the unsupported group were switched to the supported group. Results: There was a significant difference between the supported and unsupported groups in the percentage of patients with 2-h postprandial glycemia within the 70-180 mg/dL range (58.6% vs 46.6%, respectively; P = 0.031). Conclusions: The VoiceDiab system improves postprandial glucose control without increasing the time in hyperglycemia or hypoglycemia. Therefore, it may be useful in the treatment of patients with diabetes on intensive insulin therapy.