Petra Jauslin (1), Pooja Kulkarni (1), Russ Wada (1), Suresh Vatakuti (1), Azher Hussain (2), Larissa Wenning (2), Thomas Kerbusch (1)
(1) Certara Inc., Princeton, NJ, USA, (2) Merck & Co Inc., Kenilworth, NJ, USA
Objectives: Tildrakizumab is an anti-IL-23p19 monoclonal antibody (mAb) in development for the treatment of chronic plaque psoriasis. The objectives of this analysis were to characterize the population pharmacokinetics of tildrakizumab and to identify intrinsic and extrinsic factors influencing its exposure in healthy volunteers and subjects with psoriasis across all clinical development phases.
Methods: Subcutaneous (SC) administration arms of six Phase 1, 2b and 3 trials were included in the analysis data set, containing 2098 individuals and 13967 observation records in total. Model development was performed in NONMEM 7.3 [1] / PsN 4.2.0 [2]. A formal covariate analysis was conducted; covariates were evaluated in a stepwise procedure with forward addition (α = 0.01) followed by backward elimination (α = 0.001). Covariates of interest included body weight, gender, age, race, ethnicity, Japanese origin, patient status, serum albumin, creatinine clearance, prior treatment with a biological agent and formulation type. The model was validated by a prediction-corrected visual predictive check [3] and a bootstrap analysis [4]. The impact of covariates and need for dose-adjustment was assessed by conducting univariate and multivariate covariate simulations.
Results: The base model was a 1-compartment model (parameterized in terms of clearance (CL) and volume of distribution (V)) with first order absorption and elimination, and inter-individual variability on CL, V and absorption rate constant (KA). The residual error structure was combined proportional and additive. Relative bioavailability was significantly higher in healthy subjects than in psoriatic subjects. Including this covariate in the base model was necessary to obtain an acceptable fit. The difference between healthy subjects and psoriatic subjects was partially explained by body weight. Hence, body weight was included as structural covariate on apparent total clearance (CL/F) and apparent volume of distribution (V/F) in the base model as well.
Though most covariates (body weight, gender, age, race, ethnicity, patient status, serum albumin, creatinine clearance and formulation) turned out to have a statistically significant effect on one or several model parameters, these covariate effects – with the exception body weight and patient status (healthy versus psoriasis patient) – were found to be small to modest. Inclusion of additional covariates on top of body weight and patient status (both part of the base model) had little effect on inter-individual variability (IIV CL decreased by 3%, other IIVs unchanged) and no effect on residual error estimates.
The final population PK model indicated that psoriatic subjects were characterized by a geometric mean (%CV) clearance of tildrakizumab of 0.32 L/day (38%), volume of distribution of 10.8 L (24%), absorption and elimination half-life (t½) of 1.5 days (18%) and 23.4 days (23%), respectively, and an absorption lag time of 0.05 days (1.2 hours).
Univariate and multivariate simulations showed that the effects of all identified covariates on tildrakizumab steady-state AUC and Cmax were within the established clinical comparability bounds that would be expected to result in no important change in tildrakizumab efficacy or safety.
Conclusions: All covariate effects except body weight and patient status were found to be small to modest. Patient status is not relevant in clinical practice, as the drug will not be administered to healthy subjects. Simulations showed that the body weight effect was still contained within clinical comparability bounds. Based on PK data only, there is no need for dosage adjustment for the evaluated intrinsic and extrinsic factors. Nonetheless, body weight was influential and was subsequently evaluated in a PKPD analysis.
References:
[1] Beal SL, Sheiner LB, Boeckmann AJ, Bauer RJ. NONMEM Users Guides. Icon Development Solutions, Ellicott City, Maryland, USA. 1989-2011.
[2] Lindbom L, Pihlgren P, Jonsson EN. PsN-Toolkit–a collection of computer intensive statistical methods for non-linear mixed effect modeling using NONMEM. Comput Methods Programs Biomed. 2005;79(3):241-57.
[3] Karlsson MO and Holford N. A Tutorial on Visual Predictive Checks. 2008. PAGE 17 Abstr 1434 [www.page-meeting.org/?abstract=1434].
[4] Bootstrap user guide. 2011. Available from: http://psn.sourceforge.net/pdfdocs/bootstrap_userguide.pdf
Reference: PAGE 27 (2018) Abstr 8457 [www.page-meeting.org/?abstract=8457]
Poster: Drug/Disease Modelling - Other Topics