Lucia Siovitz (1), Sumit Bhatnagar (2), Mohamed-Eslam F. Mohamed (2), Sven Mensing (1), Sven Stodtmann (1)
(1) Clinical Pharmacology, Deutschland GmbH Co. KG, Ludwigshafen am Rhein, Germany (2) Clinical Pharmacology, AbbVie Inc, North Chicago, IL, USA
Introduction:
Upadacitinib (UPA) is an orally administered Janus Kinase inhibitor which has been approved or is in development for the treatment of several autoimmune conditions. Interim analyses in Phase 2 trials inform Phase 3 decisions, but data may be limited for later visits. To improve predictions, all available information should be utilized. In this analysis of a Phase 2 study of UPA in subjects with non-segmental vitiligo (NSV), a chronic autoimmune skin disorder 1 without any approved systemic therapies, the continuous and binary endpoints are correlated, with binary endpoints being thresholds of their respective continuous endpoints. Therefore, a longitudinal time model of the continuous endpoints, specifically the change from baseline in Facial/Total Vitiligo Area Scoring Index (F/T-VASI), can provide well-informed predictions for all visits, doses, and endpoints, including the binary endpoints. However, to achieve accuracy, the models must fit the median trend as well as the variability in the endpoint across the full time-course to predict clinically relevant binary endpoints.
Methods:
Data for the exposure-response analysis were collected from 181 subjects, with a total of 1,274 F-VASI and 1,280 T-VASI evaluations through 52 weeks of treatment. The model was developed after the last subject completed their Week 24 visit and used the average UPA concentration as a predictor for the change from baseline in F/T-VASI. Model convergence, robustness, GOF, VPC plots and NPCs were used to assess the adequacy of the model. In addition to “standard” VPCs stratified by treatment groups and exposure quartiles, that assess the full time-course, a cut at Week 24, representing the primary endpoint, was made to evaluate the predicted change of T/F-VASI across all exposures against observations. Furthermore, the models’ ability to predict the binary endpoints was assessed by using the simulations generated for the VPCs and applying the thresholds at each replicate, followed by summarizing the relative frequencies of patients reaching the level of improvement.
Results:
The placebo effect on change from baseline in F/T-VASI was modelled as a mixture of two subgroups: a non-improver subgroup where no change occurred and a subgroup with a linear improvement over time in F-VASI [point estimate (95% CI) of parameters]: 0.00106 (0.00057,0.00197)1/week and T-VASI: 0.000354 (0.000124, 0.00101) 1/week. UPA treatment was modeled by an Emax-shaped drug effect which accumulated over time in addition to the placebo effect (F-VASI: EC50 0.177 (0.-1.30, 1.66) ng/mL, EMAX 0.0528 (0.0325, 0.0732) 1/week; and T-VASI: EC50 0.625 (0.447, 0.802) ng/mL, and EMAX 0.0190 (0.0147, 0.0234) 1/week). This indicated that as the drug concentration increases, the effect on change from baseline in F/T-VASI also increases until a maximum effect is reached. Delay compartments were used to account for the gradual onset of drug effect before affecting the F/T-VASI compartment. No covariate showed a significant influence when tested on the placebo effect, or parameters of the Emax-shaped treatment effect. The models were validated with all data after the study was completed, confirming its reliability and accuracy. The models were used to predict the change from baseline in F/T-VASI at Week 24 and 52 following a daily dosing of 6,11 or 22 mg UPA. The difference between predicted and observed response rates of ≥ 50%/75% improvement in T/F-VASI were ≤ 10%.
Conclusions:
Two models were developed to explain the change in NSV scoring indices over time, based on average UPA concentration. These models effectively described the changes from baseline and the response rates of the connected binary endpoints. Validation with later trial data confirmed their reliability and accuracy. This analysis effectively underscores the indispensability of Pharmacometrics in decision-making during drug development. By utilizing the most informative data, specifically longitudinal data of a continuous endpoint, the model was able to simultaneously describe multiple continuous and binary endpoints. This approach allowed for predictions of outcomes at later timepoints, including placebo-corrected response rates beyond the observed placebo-controlled period. It also enables more accurate predictions for potential Phase 3 designs and provides predictions for untested doses and timepoints.
References:
[1] Oiso N, Suzuki T, Fukai K, et al. Nonsegmental vitiligo and autoimmune mechanism. Dermatol ResPract. 2011;2011:518090.
[2] Thierry Passeron, Khaled Ezzedine, Iltefat Hamzavi, Nanja van Geel, Bethanee J Schlosser, Xiaofei Hu, Xiaohong Huang, David Rosmarin, John E Harris, Heidi S Camp, Amit G Pandya, 501 – Efficacy and safety of upadacitinib in a phase 2 randomized, double-blind, dose-ranging study of adults with extensive non-segmental vitiligo, British Journal of Dermatology, Volume 190, Issue Supplement_2, February 2024, Pages ii65–ii66, https://doi.org/10.1093/bjd/ljad498.066
Disclosures:
All authors are employees of AbbVie and may hold AbbVie stock. AbbVie funded this study and participated in the study design, research, analysis, data collection, interpretation of data, reviewing, and approval of the publication. All authors had access to relevant data and participated in the drafting, review, and approval of this publication. No honoraria or payments were made for authorship.
Acknoledgements:
AbbVie and authors thank all the trial investigators and the patients who participated in this clinical trial.
Reference: PAGE 32 (2024) Abstr 10900 [www.page-meeting.org/?abstract=10900]
Poster: Drug/Disease Modelling - Other Topics