Pharmacokinetics, Pharmacodynamics, and Proposed Dosing of the Oral JAK1 and JAK2 Inhibitor Baricitinib in Pediatric and Young Adult CANDLE and SAVI Patients

Abstract

Population pharmacokinetic (popPK) modeling was used to characterize the PK profile of the oral Janus kinase (JAK)1/JAK2 inhibitor, baricitinib, in 18 patients with Mendelian interferonopathies who are enrolled in a compassionate use program. Patients received doses between 0.1 to 17 mg per day. Covariates of weight and renal function significantly influenced volume-of-distribution and clearance, respectively. The half-life of baricitinib in patients less than 40 kg was substantially shorter than in adult populations, requiring the need for dosing up to 4 times daily. On therapeutic doses, the mean area-under-the-concentration-vs.-time curve was 2,388 nM*hr, which is 1.83-fold higher than mean baricitinib exposures in adult patients with rheumatoid arthritis receiving doses of 4 mg once-daily. Dose-dependent decreases in interferon (IFN) biomarkers confirmed an in vivo effect of baricitinib on type-1 IFN signaling. PopPK and pharmacodynamic data support a proposal for a weight- and estimated glomerular filtration rate-based dosing regimen in guiding baricitinib dosing in patients with rare interferonopathies.

Trial registration: ClinicalTrials.gov NCT01724580 NCT02974595.

© 2017 American Society for Clinical Pharmacology and Therapeutics.

Figures

Figure 1. Covariate plots for Relationship between (A) body weight and volume of distribution, V/F and (B) glomerular filtration rate (GFR) on clearance, CL/F

Pt. C1 is an outlier for the V/F (L) estimates (red circle) but not for the clearance (red arrow) (see Supplementary Methods). Scatter plot of covariates versus estimated PK parameters are shown. The solid colored circles represent the individual estimated PK parameters and covariates. Data from the same individual are connected by a solid colored line. The solid black lines represent the relationships between the typical PK parameters and the covariates. The equations for the relationships are: (for A) V/F = 67.1 × e0.0192·(WTV−27.2) and (for B) CL/F=9.74×(eGFR146.82)0.444.

Figure 2. IC50 values by Cell Type for (A) IFNα-stimulated STAT1 Phosphorylation (pSTAT1), (B) IL-6 stimulated STAT3 Phosphorylation (pSTAT3), and (C) STAT-1 phosphorylation at baricitinib trough levels

(A), (B). Scatter plots of actual data versus model curves for pSTAT by cell type are shown with the median fluorescence intensity (MFI) ratio (stimulated divided by un-stimulated) minus 1 versus the peripheral blood drug level in nM. The solid line and light blue band are the best-fit curve and 95% predictive interval, respectively. The table shows the IC50 (nM) calculated based on this modeling (estimate) with standard error, coefficient of variance, and 95% confidence intervals. * p-value < 0.01, ** p-value < 0.001, ns: not significant.

Figure 2. IC50 values by Cell Type for (A) IFNα-stimulated STAT1 Phosphorylation (pSTAT1), (B) IL-6 stimulated STAT3 Phosphorylation (pSTAT3), and (C) STAT-1 phosphorylation at baricitinib trough levels

(A), (B). Scatter plots of actual data versus model curves for pSTAT by cell type are shown with the median fluorescence intensity (MFI) ratio (stimulated divided by un-stimulated) minus 1 versus the peripheral blood drug level in nM. The solid line and light blue band are the best-fit curve and 95% predictive interval, respectively. The table shows the IC50 (nM) calculated based on this modeling (estimate) with standard error, coefficient of variance, and 95% confidence intervals. * p-value < 0.01, ** p-value < 0.001, ns: not significant.

Figure 2. IC50 values by Cell Type for (A) IFNα-stimulated STAT1 Phosphorylation (pSTAT1), (B) IL-6 stimulated STAT3 Phosphorylation (pSTAT3), and (C) STAT-1 phosphorylation at baricitinib trough levels

(A), (B). Scatter plots of actual data versus model curves for pSTAT by cell type are shown with the median fluorescence intensity (MFI) ratio (stimulated divided by un-stimulated) minus 1 versus the peripheral blood drug level in nM. The solid line and light blue band are the best-fit curve and 95% predictive interval, respectively. The table shows the IC50 (nM) calculated based on this modeling (estimate) with standard error, coefficient of variance, and 95% confidence intervals. * p-value < 0.01, ** p-value < 0.001, ns: not significant.

Figure 3. Correlation of AUC with (A) 25-gene IFN score and (B) IP-10

Drug exposure (AUC24,SS) is significantly negatively correlated with 25-gene IFN score and serum IP-10 levels by F-test for AUC24,SS effect. Each individual is denoted with a different colored line and dots. The heavy red line represents the overall line for the group and the equation is listed below. (A) Correlation of AUC with 25-Gene IFN Score includes 327 observations in 18 patients (all diagnoses). The dashed line represents cutoff of normal 25-gene IFN score based on healthy controls. Equation: IFN STD 25 = 431.98 – 0.07220 (AUC24,SS), slope= −0.096 p-value for slope = 0.049. (B) Correlation of AUC with serum IP-10 levels includes189 observations in 18 patients (all diagnoses). Equation: IP-10 = 9060.45 – 2.1432 (AUC24,SS), slope= −0.197, p-value for slope = 0.0002.