Classifying and characterising type 1 diabetes in adults

Abstract

Around half of all type 1 diabetes cases occur in adults but the phenotype of the disease in this age group is poorly understood. A major difficulty studying type 1 diabetes in adults is differentiating cases from type 2 diabetes.

The aim of this thesis was to both determine strategies for improving type 1 diabetes classification in adults and evaluate the impact of onset age on clinical phenotype in robustly defined adult-onset type 1 diabetes.

Chapter 1 explores the reported characteristics of adult-onset type 1 diabetes and how this varies with case definition. I discuss difficulties classifying adult-onset type 1 diabetes and how this might impact the observed phenotype.

Chapter 2 evaluated the commonly reported reduction in positive islet autoantibodies with increasing onset age in type 1 diabetes. We show a significant reduction in genetic predisposition to type 1 diabetes in autoantibody negative adults concluding that most of these cases probably have non-autoimmune diabetes.

Chapter 3 examines in adults diagnosed with type 1 diabetes the clinical benefit of measuring and reporting autoantibody results to patients and their clinicians given in Chapter 2 a high proportion without autoantibodies likley have non-autoimmune diabetes. This showed some autoantibody negative cases stopping insulin without detrimental impact on glycaemic control.

Chapter 4 explores the conflicted evidence for the loss of HLA (DR15-DQ6) associated protection against type 1 diabetes with increasing onset age. We show that in adults when type 1 diabetes is defined robustly, protection remains.

Chapter 5 further builds on the importance of robustly classifying type 1 diabetes in adults. We show that presentation and progression of type 1 diabetes in adults is equivalently severe irrespective of onset age.

Chapter 6 evaluated commonly used approaches for defining diabetes type in datasets where biomarkers for diabetes classification are unavailable. This identified the optimum approach to be BMI and Age of diagnosis used continually within a prediction model. We produced an online tool allowing researchers to select the optimum approach for their research question.

An overview of the major finding of each chapter, their implications and potential future research are discussed in Chapter 7.