Nassim Djebli (1), Clémence Rauch (1), Sandrine Turpault (2) and David Fabre (1)
(1) SANOFI R&D, Montpellier, France ; (2) SANOFI R&D, Bridgewater, United States.
Objectives: To explore the relationship between 12 safety and 6 efficacy variables and steady-state mean Teriflunomide trough concentrations (MCONC) in RMS patients with the inclusion of significant covariates, after 7 and 14 mg once daily (OD) Teriflunomide administration.
Methods: Data were collected in 2 clinical studies (Phases II and III) for the evaluation of safety (up to 1265 patients) and in 1 Phase III study (up to 1036 patients) for efficacy. The PK/PD models were developed and validated with R following: (a) Exploratory plots: PD variables were plotted vs. MCONC to assess possible relationship between MCONC and PD; (b) Statistical model (inter-individual variability) both additive and proportional, were tested; (c) MCONC effect model: linear, exponential, Emax (and Sigmoidal Emax models when necessary) were tested – logistic regression (binomial R function) or quasi-Poisson distribution models were used to evaluate categorical variables; (d) Covariate Screening: a minimum increase of the ΔOFV of 15.1 (p<0.0001) was necessary to retain a covariate and (e) Qualification of the PK/PD models was performed via graphical and statistical evaluations
Results: Regarding safety variables, final PK/PD models were mainly Emax models with an inter-individual variability and covariate inclusion on baseline for ALT (Alanine aminotransferase), Neutrophils, Lymphocytes, White Blood Cells, Diastolic Blood Pressure, Phosphate and Uric Acid. A linear model was selected for Amylase and a logistic regression model for Hair thinning. No significant relationship between MCONC and Lipase, Systolic Blood Pressure and CLCR was found.
Regarding efficacy variables: (a) A survival analysis model (Kaplan-Meier method) by MCONC category was performed and the Cox regression model showed a significant (p=0.033) decreased risk for disability progression as MCONC increased; (b) Number of Gadolinium-enhanced T1 lesions and number of unique active lesions per number of scans: quasi-Poisson models were selected with a low predictive performance; (c) Number of patients free of active lesions: a logistic regression model was selected; (d) Annual Relapse Rate: a trend was observed regarding relationship with MCONC; (e) Burden of disease: no significant relationship was observed.
Conclusions: This analysis allowed the development and qualification of PK/PD models (safety and efficacy) for OD Teriflunomide administered to RMS patients at doses of 7 and 14 mg.
Reference: PAGE 21 () Abstr 2562 [www.page-meeting.org/?abstract=2562]
Poster: Covariate/Variability Model Building