CoDa-thesis at UdG (GIrona, Spain)

10/07/2019

CONDITION ASSESSMENT OF PATIENTS WITH TYPE 1 DIABETES USING COMPOSITIONAl DATA ANALYSIS

 

On the 9th of July 2019, LYVIA REGINA BIAGI SILVA BERTACHI, a member of our association defended her thesis, supervised by Dr Josep Vehí Casellas and Dr Josep Antoni Martín-Fernández. The jury members were Dr José Luís Díaz (UPV), Dr Glòria Mateu (UdG) and Dr Cecilio Angulo (UPC).

 

Type 1 Diabetes Mellitus (T1DM) is a disease related to the autoimmune process of pancreatic Beta-cell destruction that leads to absolute insulin deficiency. Although T1DM occurs most frequently in children and adolescents, it can develop at any age. People with T1DM need exogenous insulin to maintain glucose at proper levels and avoid hyperglycemia. However, if insulin is over delivered, it may cause hypoglycemic events. If not properly treated T1DM may lead to several complications over time, including blindness, kidney failure, cardiovascular complications, and even death. Insulin infusion in T1DM patients can be performed with multiple daily injections (MDI); however, treatment with continuous subcutaneous insulin infusion (CSII) therapy provides improvements in glycemic control. Current T1DM management technology allows the integration of continuous glucose monitoring (CGM) and CSII. One example of such technology is the artificial pancreas (AP), which is a closed-loop (CL) system for the automatic control of glucose. The AP integrates a CGM device, an insulin pump and a control algorithm. However, even with the advances in diabetes technology, achieving optimal glycemic control is still a major hurdle due to the large intra-patient variability, and CGM plays an essential role for individuals with T1DM, allowing them to follow blood glucose levels in real time. These devices are recognized for improvements in glucose control and for providing relevant information not only regarding patient’s glucose profile but also lifestyle. Moreover, CGM technology has been changing how glycemic control is assessed, and there is an open discussion on considering the values obtained by CGM into different ranges of glucose instead of considering only the glycated hemoglobin test.

 

This thesis is devoted to describe the condition assessment of patients with T1DM through the analysis of glucose data obtained from CGM. Firstly, this work focuses on understanding and dissecting the measures obtained from CGM sensors. For that, a model of the error of a CGM sensor has been obtained and the accuracy of the CGM has been assessed during challenging conditions. Secondly, a novel approach for the categorization of daily glucose profiles based on the analysis of compositional data (CoDa) is proposed. This methodology considers the analysis of time spent in different glucose ranges by individuals with T1DM. CoDa analysis is related to the analysis of vectors of positive components that describe the contribution of parts to some whole. Last, a probabilistic model of transition between different categories of periods of glucose data obtained with CoDa analysis is presented. The aforementioned approaches were evaluated considering T1DM data sets obtained from real patients using CGM devices. The obtained results are promising and could contribute to the advances in the development of technologies and also to assist both physicians and T1DM patients in the management of T1DM.