Virtual population analysis of a Quantitative Systems Pharmacology (QSP) model of atopic dermatitis (AtD) to evaluate AtD treatments
Joey Leung (1), Loveleena Bansal (2), Jaimit Parikh (2), Núria Buil Bruna (3), Valeriu Damian (2)
(1) Systems Modeling and Translational Biology (SMTB), GSK, Stevenage, UK, (2) SMTB, GSK, Collegeville, PA, USA, (3) Clinical Pharmacology Modelling and Simulation (CPMS), GSK, Stevenage, UK
Introduction: Atopic dermatitis (AtD) is a chronic inflammatory skin condition characterized by eczematous lesions and intense itch [1]. The exact cause of AtD is unclear but involves a combination of factors including genetics, skin barrier abnormalities, and overactive immune responses to antigen exposure [1]. A QSP model of AtD has been developed to link its heterogeneous disease mechanisms to clinical outcomes and evaluate different treatments for this condition. A virtual patient population for the model has been generated and calibrated using baseline disease biomarkers and clinical trial data.
Objectives:
- Predict the efficacy of treatments of interest in comparison to standard of care
- Identify patient sub-groups for which the treatment has superior efficacy
- Discover biomarkers predictive of treatment response and for patient stratification
- Evaluate combinations of targets/treatments for additive or synergistic effects
Methods:
A recently developed QSP model of AtD in GSK was further updated to evaluate Dupilumab and support GSK programs. In the model, infiltration of allergens and microbes activates keratinocytes and immune cells, which produce cytokines that initiate the itch-scratch cycle and leads to skin barrier disruption. The barrier dysfunction in turn allows more infiltration of allergens/microbes and perpetuates the disease. The model has been updated with a blood compartment for the prediction of serum biomarkers. In addition, updates to the model structure have been made to enable targeted modulation of Th2 cytokines (IL4, IL5 and IL13) and to include additional immune cell types important in AtD such as Th1 cells. Further updates to the model allow for prediction of the AtD clinical score Eczema Area and Severity Index (EASI) [2]. These model improvements allow for the evaluation of existing therapies including Dupilumab (antagonist of IL4 and IL13 receptors approved in 2017) as well as anti-IL13 treatments such as Tralokinumab (approved in 2021) and Lebrikizumab (in phase 3 trial). A virtual patient population has been generated for the QSP model by sampling over parameters identified by global sensitivity analysis as important in determining clinical outcomes. The virtual population was filtered using baseline patient biomarker data and calibrated with a Bayesian algorithm [3] to match the biomarker statistics and EASI score response to Dupilumab [4].
Results: The virtual population prediction is consistent with clinical data on reduction of EASI and itch by Dupilumab [4,5] and IL13 inhibition [6,7]. The model predicts that IL4 inhibition could lead to elevation of IFNg via activation of Th1 cells in a small subpopulation of patients, which may be associated with adverse events such as conjunctivitis [8]. Responder analysis on the model predictions suggest that by targeting both IL4 and IL13, Dupilumab achieves a broader response in the patient population. Finally, reduction in IL31 level is associated with adequate response to Dupilumab and could serve as a biomarker for Dupilumab response [9].
Conclusion: Using Dupilumab as an example, these findings demonstrate the utility of the QSP model in evaluating the mechanisms of treatment response and identifying important biomarkers that can predict treatment success. Using the virtual population approach, the QSP model can also help formulate strategies for patient stratification to optimize treatment response. By integrating in vitro, in vivo, and clinical data, the QSP model can support model-informed drug development (MIDD) to inform critical decisions such as treatment line position, dosing strategy, and clinical trial design.
References:
[1] Weidinger and Novak, Lancet 387, 1109 (2016).
[2] Hanifin et al., Exp. Dermatol. 10, 11 (2001).
[3] Butler et al., SIAM J. Sci. Comput. 40, A984 (2018).
[4] Hamilton et al., J. Investig. Dermatol. 134, S104 (2014).
[5] Guttman-Yassky et al., J. Allergy Clin. Immunol. 143, 155 (2019).
[6] Silverberg et al., Br. J. Dermatol. 184, 450 (2021).
[7] Simpson et al., J. Am. Acad. Dermatol. 78, P863 (2018).
[8] Bakker et al., Allergy 76, 3814 (2021).
[9] Wu et al., Br. J. Dermatol., in press, ljad032 (2023).