PII-008 - C-PEPTIDE DISEASE PROGRESSION MODEL IN INDIVIDUALS WITH NEW-ONSET TYPE 1 DIABETES

K. Collins, A. Lozano, F. Walker, B. Greeno, S. Ma, J. Hedrick; Critical Path Institute.

Background: HbA1c is a validated endpoint for clinical trials in type 1 diabetes (T1D), however studies have shown that preservation of c-peptide is also associated with improved clinical outcomes, including reductions in hypoglycemia, neuropathy and retinopathy. Our objective was to understand the predictors of c-peptide decline in new-onset T1D patients, defined as less than 100 days from diagnosis, to inform the design of future clinical trials focused on c-peptide as an endpoint.

Methods: Subject-level data from 1980 individuals with new-onset T1D were pooled from 20 randomized clinical trials. Data were supplied from the NIDDK Central Repository (TN02, TN05, TN08, TN09, TN14, TN19), GlaxoSmithKline (DEFEND1, DEFEND2, GSKALB), ExTOD, NIAID (EXTEND), Diamyd, Novo Nordisk, MacroGenics (Protégé), Immune Tolerance Network (RETAIN, START, T1DAL), Janssen Research and Development, LLC (T1GER), Yale University (AbATE) and UCSF. A disease progression model, using log-transformed c-peptide AUC (nmol/L) measured by a 2-hour mixed meal tolerance test, was trained and validated using an 80/20 data split. Stepwise covariate model building was used to test the following covariates: baseline (c-peptide, age, BMI, disease duration), sex, race, ethnicity, and HLA genotypes. In addition, a drug effect model was applied after grouping active treatment arms by mechanisms of action: T-cell inhibition/deactivation, T-cell depletion, antigen-specific, cytokine-mediated activity and beta-cell function. Model performance was guided by standard goodness-of-fit measures and visual predictive checks.

Results: C-peptide decline was best described by the sigmoidal Emax function (E0: 0.64, Emax: -3.87, EC50: 3.5 years, Hill coefficient: 1.68) and lognormal residual and inter-individual variability. Significant predictors included baseline BMI on E0, baseline age on EC50, and baseline c-peptide on Emax (ΔOFV: -532). Visual predictive checks on the training and validation dataset showed good performance.

Conclusion: Understanding the predictors of c-peptide decline can help inform the design of future clinical trials. Next steps included utilizing this model to develop an interactive, clinical trial simulation tool that allows users to specify ranges of predictive covariates, sample size and anticipated drug effect to determine clinical trial power.