Data Cleaning, Preliminary Summary and Evaluation of Diagnostic Criteria of T-Cell Data in a Juvenile Onset Diabetes Cohort

Type 1 diabetes mellitus (T1DM) is an autoimmune disease manifested by an autoimmune attack on pancreatic beta-islet cells. T1DM can occur at any age. However, it is most often diagnosed in children, adolescents, or young adults. My thesis is derived from a large longitudinal study of Juvenile Onset Diabetes (JOD) at Children’s Hospital of Pittsburgh. The objectives are: 1) Data cleaning and preliminary summary of the cohort with respect to T-cell data. 2) Evaluating the T-cell data criteria used for the prediction of the diabetes. \ud An extensive data examination was made for accuracy and consistency. A preliminary summary of the stimulation index (SI) for the test analytes and the number of positive antigens was performed by demographic sub-groups, HLA-DQ serotype, and follow up time. Using the ROC analysis, an evaluation of diagnosis test performance based on two different criteria was performed.\ud The JOD dataset had few errors with an error rate under 0.5%. The accuracy and consistency of the data is good. New onsets and first degree relatives (FDRs) nonconverters had a relatively stable SI as well as positive antigen tests results. The SI level and positive test results are higher in new onsets when compared with FDRs. FDR-converters (those subsequently developing diabetes) prior to using insulin have SIs and number of positive antigens similar to FDR-nonconverters; and FDR-converters after starting insulin have results similar to new onsets. The recommended SI cutoff of 1.5 indicating positive response appears reasonable. However, the cutoff still may be optimized for better prediction. Evidence suggests that a lower cutoff within 1.25 to 1.5 may be better and the number of positive antigens could move from ≥4 to greater than 5 or 6.\ud Public health significance: Development of a better understanding of the pattern of T-cell response in diabetes and non-diabetic children, and those progressing to diabetes, may give us tools to predict the early onset of disease. It is this point in time where therapeutic intervention could be focused to help stem the development of T1DM or to dramatically reduce its severity