The ADDRESS-2 cohort: database management & data analysis evaluating the characteristics and predictors of early remission in people with incident type 1 diabetes

Abstract

Type 1 diabetes (T1D) is an autoimmune disease characterised by loss of pancreatic β-cell function and is heterogeneous in its presentation and progression. Some recently diagnosed individuals may experience transient β-cell recovery (partial remission), characterised by low insulin requirements and good glycaemic control. ADDRESS-2 recruits individuals aged ≥5 years, within 6 months of a clinician-assigned diagnosis of T1D. Recruitment is facilitated by NIHR Clinical Research Network research nurses across 156 sites in the U.K. Participant’s demographic, medical and family history, clinical and biochemistry data within the first year of diagnosis are recorded on an online database. An optional blood sample is obtained for the measurement of islet autoantibodies, and storage of serum and DNA. This thesis first explores ADDRESS-2 data validation issues and the data preparation methodologies adopted to maintain data integrity, particularly in relation to insulin medication and clinical and biochemistry data structuring. Subsequent chapters then explore the research potential of ADDRESS-2 with analysis of (1) cohort characteristics in adults and children; (2) predictors of glycaemic control, measured by random and fasting plasma glucose (RPG and FPG) and HbA1c; and (3) predictors of partial remission. Adults were more likely than children to present in diabetic ketoacidosis (DKA), with weight loss and fatigue, a longer symptom duration, a higher HbA1c, and were more likely to be autoantibody negative. As measures of glycaemic control, RPG and FPG had multiple independent predictors at diagnosis, but no significant predictors subsequently. In contrast, HbA1c exhibited a broad range of predictors at and after diagnosis. Five definitions of partial remission were investigated. For all definitions, increasing age, male gender and the absence of DKA were independent predictors. Depending on the definition, 51-80% of individuals in partial remission had a c-peptide concentration > 300 pmol/l. In conclusion, after careful attention to data validation and structuring, this large, multicenter database could be highly informative regarding real-world T1D characteristics.