Eirini Panoilia (1), Núria Buil-Bruna (1), Monica Simeoni (1)
(1) Clinical Pharmacology Modelling and Simulation, GSK, UK
Objectives: Atopic dermatitis (AD) is a long-term inflammatory skin disorder mainly characterized by eczematous lesions and pruritus [1, 2]. The assessment of disease severity and efficacy is based on various scoring tools including physician’s assessment such as the Eczema Area and Severity Index (EASI) and patient-reported outcomes e.g. pruritus Numerical Rating Scale (pNRS) [1]. The course of the disease is heterogeneous, and the treatment response varies, potentially affecting the magnitude of the placebo response in AD studies [1, 3]. Moreover, the limited availability of longitudinal placebo data coupled with the small sample size in early phase studies makes the optimal treatment decision in late phase studies more challenging. A model based meta-analysis (MBMA) was performed utilising publicly available efficacy data with the objective of characterising the time course of placebo response, as assessed by EASI and pNRS scores, in patients with moderate to severe AD to support design characteristics and provide higher confidence in dose-response analyses in late phase studies.
Methods: Certara’s Atopic Dermatitis Clinical Outcomes Database was used in the current MBMA [4]. Only an adult (age ≥18 yrs), moderate to severe AD population (baseline EASI score ≥16, IGA≥3, BSA involvement ≥10%) receiving placebo as randomised drug was considered. Placebo arms of studies having as background therapy topical corticosteroids, topical calcineurin inhibitor and antihistamines were excluded. The analysis was conducted using R (version 3.2.5) on a validated GSK modelling platform. A two-stage approach was employed whereby the standard deviation (sd) of the average placebo response was modelled, to allow the inclusion in the MBMA of the timepoints of placebo arms missing this variable, by utilising the simulated values. The second step was the model of the placebo response. Logistic transformation was applied to absolute EASI and pNRS score values [5] and the propagation error function associated with the transformation was derived. A global estimate for all transformed sd values was obtained using a regression-based approach. Exponential placebo response models were parameterised in terms of intercept, θeo, maximal placebo response at steady state, θeom, and a rate of change of placebo response over time, exp(θkt). Model development included the investigation of random effects and plausible covariates on relevant parameters. Overall model goodness of fit was assessed graphically and by statistical tests. An exploratory graphical analysis to assess the effect of study phase and primary drug on placebo response for EASI and pNRS scores was also conducted.
Results: 38 studies (3179 participants) and 28 studies (3060 participants) with EASI and pNRS score data, respectively, were included in the analysis. A three-parameter placebo response sd model was found to best describe the transformed sd values for EASI and pNRS scores whereas an exponential response model could adequately characterise the placebo response over time for transformed EASI and pNRS scores. The transformed sd of EASI and pNRS scores was found dependent on time. The maximal placebo response was -0.3086 and -0.53 for transformed EASI and pNRS scores, respectively. The predicted half-life of placebo response was 1.94 wks for the transformed EASI score and 3.37 wks for the transformed pNRS score. The effect of clinically relevant covariates (i.e., study phase, study arm, baseline) was tested in the models, however, no significant covariate effects could be identified. Based on an exploratory graphical analysis, Phase (Ph) 2 studies seem to have a wider range in baseline EASI and pNRS values compared to Ph3 studies. The sd of the pNRS score seems to increase over time in Ph3 studies whereas in Ph2 studies, there is a tendency to decrease up to wk3 and start stabilising beyond that timepoint. Of note, the Ph2 studies had a higher sd of pNRS score at baseline compared to the Ph3 studies.
Conclusions: An exponential model could adequately describe the time course of the placebo response for transformed EASI and pNRS scores. No significant covariate effects could be identified. Based on an exploratory graphical analysis, Ph2 studies seem to have a wider range in baseline EASI and pNRS score values compared to Ph3 studies. The proposed placebo models could be used as priors to support design characteristics and analyses in late phase AD studies.
References:
[1] Bieber T. Atopic dermatitis: an expanding therapeutic pipeline for a complex disease. Nat Rev Drug Discov 2022; 21:21-40. https://doi.org/10.1038/s41573-021-00266-6.
[2] Penton H, Jayade S, Selveindran S, Heisen M, Piketty C, Ulianov L et al. Assessing response in atopic dermatitis: A systematic review of the psychometric performance of measures used in HTAs and clinical trials. Dermatol Ther (Heidelb) 2023; 13:2549-2571. https://doi.org/10.1007/s13555-023-01038-3.
[3] Lee HH, Patel KR, Rastogi S, Singam V, Vakharia PP, Chopra R et al. Placebo responses in randomized controlled trials for systemic therapy in atopic dermatitis: A systematic review and meta-analysis. J Am Acad Dermatol 2020; 82(1):62-71. doi: 10.1016/j.jaad.2019.05.102.
[4] Certara. Atopic Dermatitis Clinical Outcomes Database version 20221215. https://codex.certara.com/codex/atopic-dermatitis/.
[5] Jauslin P, Largajolli A, Schindler E, Saito T, Loprete L, Wagner N et al. Modeling bounded scales for evaluation of treatment response to subcutaneous nemolizumab in moderate to severe atopic dermatitis. PAGE 28 2019; Abstr 9088. www.page-meeting.org/?abstract=9088.
Reference: PAGE 32 (2024) Abstr 11122 [www.page-meeting.org/?abstract=11122]
Poster: Drug/Disease Modelling - Other Topics