Liang Yang (1), Carolina Llanos-Paez (1), Shuying Yang (2), Claire Ambery (2), Alienor Berges (2), Maria Kjellsson (1), Mats O. Karlsson (1)
(1) Department of Pharmacy, Uppsala University, Uppsala, Sweden (2) Clinical Pharmacology Modelling and Simulation, GSK, London, UK
Introduction:
Model-based meta-analysis (MBMA) allows integration of aggregated-level data (AD) from different clinical trials in one pharmacometrics model to assess and compare efficacy/safety of different treatments and estimate interstudy variability (ISV) with covariate relations. However, AD data are limited in information about interindividual variability (IIV) and the covariate relations that can explain such variability. Combined modeling with both AD and individual patient-level data (IPD) may take advantage of the information in both data types and allow more informed clinical trial simulations for designing studies. Chronic obstructive pulmonary disease (COPD) is the seventh leading cause of poor health and the third leading cause of death according to the World Health Organization [1]. A legacy MBMA model [2] of AD from 298 studies evaluated mono-, dual-, triple-therapies of bronchodilators and anti-inflammatories, where the endpoint is forced expiratory volume in one second (FEV1). IPD of two clinical trials which were included in this MBMA dataset were available. The two clinical trials [3,4] explored therapies of fluticasone furoate/vilanterol, which could be combined to the model for more precise estimation of IIV and covariate effects.
Objectives:
This study aims to establish a combined model for the AD and IPD of COPD clinical trials, including to find a suitable method for estimating ISV and IIV for combined modelling of AD and IPD.
Methods:
Stochastic simulation and estimations (SSE) were performed to evaluate different methods to estimate ISV and IIV for the combined AD and IPD model. The previously established AD model (298 studies representing data from 250,543 patients) [2] was modified, where all the covariate relationships were revised as linear to avoid aggregation bias in the combined model. A separate model for IPD (2 studies, 2241 patients) was established, and covariates were added to the IPD model linearly using stepwise covariate model building by scmplus [5]. Then the significant covariates from both the legacy MBMA AD model and the IPD model by scmplus were added to the combined model. The AD data corresponding to the two IPD studies were excluded from the dataset to avoid duplication. Finally, the significance of the covariates of the combined model was evaluated by the backwards stepwise modelling. The SSE and scmplus and model estimation were performed via NONMEM7.5.1 and PsN5.3.1.
Results:
SSE results indicated a method as best for combined modelling of AD (298 studies) and IPD (2 studies) using $LEVEL: for the AD part, ISV values from AD model were fixed and separate ETAs for each arm were used to estimate IIV (scaled with the size of arm); and the IPD part shared the fixed ISV as AD model and estimated the IIV. The combined model structure included FEV1 baseline (1.03 L, relative standard error (RSE) 1.1%), linear disease progression over time (0.0358 L/year, RSE 6.1%), placebo effects (-0.00641 L, RSE 32%) and EMAX/constant dose-responses were described for 23 compounds. Applying the same model structure (i.e. same covariates as AD model and adding those identified in a separate IPD analysis), the identified covariates of the combined model included: age, female sex and disease severity showed negative relations with FEV1 baseline; the FEV1 baseline positively correlated with the maximal drug effects of anti-inflammation and bronchodilator drugs. In the combined model, the coefficients of age on FEV1 baseline and FEV1 baseline on maximal drug effects were estimated with smaller RSE compared to the AD model (RSE decreased from 21.5% to 3.6% for age coefficient on baseline, 37.1% to 16.4% for baseline on anti-inflammatories, 33.1% to 19.9% for baseline on bronchodilator). The baseline IIV RSE decreased from 7.4% to 1.4% and residual variability RSE decreased from 3.1% to 0.5%. The above results indicated the importance of IPD in estimation. All the covariates/IIV were added linearly in the combined model to avoid aggregation bias. However, linear relationship is possibly not optimal to describe the data, and nonlinear relationship will be investigated in the future.
Conclusions:
The combined modelling of aggregated and individual patient-level data was established for multiple COPD trials. New significant covariates and multiple-level variabilities were identified and covariate/variability parameter precisions were improved based on the combined modelling.
References:
[1] https://www.who.int/news-room/fact-sheets/detail/chronic-obstructive-pulmonary-disease-(copd)
[2] Llanos-Paez et al. J Pharmacokinet Pharmacodyn. (2023) 50(4):297-314.
[3] Kerwin et al. Respir Med. (2013) 107(4):560-9.
[4] Martinez et al. Respir Med. (2013) 107(4):550-9.
[5] Svensson et al. CPT Pharmacometrics Syst Pharmacol. (2022) 11(9):1210-1222.
Funding Information: GSK funded this research in the form of a Research payment to Uppsala University.
Conflict of Interest: LY, CL-P, MJ and MOK declare that they have no conflict of interest. SY, CA and AB are GSK employees and hold GSK shares.
Reference: PAGE 32 (2024) Abstr 10848 [www.page-meeting.org/?abstract=10848]
Poster: Drug/Disease Modelling - Other Topics