IV-22 Paul Matthias Diderichsen

Characterising the onset of response using public domain data and MBMA: application cases in Atopic Dermatitis and Psoriatic Arthritis

Maria Luisa Sardu, Alienor Berges, Carly Barnett, Tony Ning, Jaap Mandema, Li Qin, Paul Matthias Diderichsen, Monica Simeoni

Certara USA, Inc., Princeton NJ; Clinical pharmacology modelling and simulation, GSK UK; Biostatistics, GSK UK; Clinical Development, GSK USA

Objectives: Immune-mediated inflammatory diseases (IMIDs) comprise a group of conditions sharing common immune pathways and are often treated with same drugs and drug classes [1]. Public domain data is instrumental in addressing strategic clinical questions in drug-development. Model-Based Meta Analysis (MBMA) provides a quantitative framework to leverage public and proprietary information within and across different diseases. MBMA has been applied in immunology to support dose-bridging across different diseases, e.g., between rheumatoid arthritis (RA) and axial spondyloarthritis (AxSpA) [2] and between inflammatory bowel diseases (IBDs) and other immune-related inflammatory diseases [3]. Characterising the onset of response across drug classes and understanding when and if the plateau is reached is a clinical question that has a direct impact on study design and primary timepoint definition for clinical endpoints. An MBMA characterizing the onset of the response in AxSpA [4] was recently conducted. In this work, a preliminary analysis of the onset of mean EASI score in atopic dermatitis (AD) and onset of ACR20 score response in psoriatic arthritis (PSA) is presented. A Full MBMA analysis based on time-course ACR20 score data in PSA was also performed to support a study futility analysis planning for an asset currently in clinical development.

Methods: 

Certara’s clinical trial outcome databases for AD and PSA [5] were used as data sources in the analysis (supported by GlaxoSmithKline). Two preliminary analyses focusing on mean EASI score and on ACR20 score data up to week 16 were performed. Placebo-corrected scores data were normalised to the response at week 16 and analysed using nonlinear regression for binary endpoints, implemented in the generalised nonlinear least squares (gnls) routine in R (3.5.3.). An exponential time-course of the onset of the responses was assumed.

For ACR20, a full MBMA assuming a drug-class-independent unstructured model for the description of onset of effect was performed. Further, placebo response was described by an unstructured model estimating placebo response by study and visit. As the focus of this analysis was on the dynamics of the response, constant effect across treatment regimens (doses, dosing intervals) was assumed for most of the drugs beside the ones showing clear dose-response relationships. The effects of covariates such as the use of methotrexate as background therapy and the prior use of biologics were evaluated. Based on the modelling framework, the fractional treatment effect relative to week 16 was simulated.

Results: 

Preliminary analyses: mean EASI score information at week 16 was reported in 23 studies of drugs in the following drug classes: anti-IL-13, anti-IL-13 in combination with corticosteroid, anti-il-31, anti-IL-4/13, Janus kinase (JAK) inhibitors and JAK inhibitors in combination with corticosteroid drug classes. ACR20 responses at week 16 were reported in 31 studies of anti-CD28, anti-IL-12/23, anti-IL-17, anti-IL-23, anti-IL-6 and JAK inhibitor drugs. Both preliminary analyses showed that models based on exponential onset and power function loss of response could adequately describe the observed data. For both analyses, most of the drug classes were found to reach the plateau of the response at around 8 weeks.

Full MBMA on ACR20: A model describing treatment effect by drug could be estimated. Dose-response relationship was identified only for secukinumab (estimated ED50 = 67.25 mg). Onset of response was estimated within time bins across classes. Use of background methotrexate was found to be a significant covariate suggesting a negative correlation between the percent of patients on methotrexate treatment and ACR20 response. Simulated treatment effect relative to week 16 indicated that approximately 90% of total week 16 response was achieved around week 8.

Conclusions: Similar MBMA methodology was successfully applied across different immune diseases. For both preliminary analyses on mean EASI in AD and ACR20 in PSA, onset of response was found to reach the plateau at approximately 8 weeks. These results were confirmed also in the full MBMA analysis where an unstructured onset model was assumed.

References:

  1. McInnes IB, Gravallese EM. Immune-mediated inflammatory disease therapeutics: past, present and future. Nat Rev Immunol 2021; 21(10): 680-686.
  2. Simeoni M, Mandema J, Zamuner S and Gupta A, A Model Based Meta-Analysis for Bridging Treatment Doses of Rheumatoid Arthritis with Axial Spondyloarthritis, 29th Virtual PAGE Meeting 2021
  3. Buil Bruna N, Qin l, Sardu ML, Berges A, Diderichsen PM and Simeoni M, Can mAb dose selection in IBD be supported by public domain data from other immune‐mediated inflammatory diseases?, 30th PAGE Meeting 2022
  4. Gupta A, Collignon O, Fernandes S, Simeoni M, Mandema J, Characterisation of time course of response using model based meta-analysis of public domain data to support trial design in axial spondyloarthritis, 30th PAGE Meeting 2022
  5. Certara. Clinical Outcomes Database Explorer (CODEX) portal. In, https://codex.certara.com 2022.

Reference: PAGE 30 (2022) Abstr 10162 [www.page-meeting.org/?abstract=10162]

Poster: Methodology - New Modelling Approaches

PDF poster / presentation (click to open)