Longitudinal Model-Based Meta-Analysis (MBMA) for rheumatoid arthritis with Monolix
Geraldine Ayral (1), Jonathan Chauvin (1)
(1) Lixoft, Antony, France
Objectives: Model-based meta-analysis (MBMA) uses published aggregate data from many studies to develop a study-level model and support the decision process. Because in an MBMA approach one considers studies instead of individuals, the formulation of the problem as a mixed effect model differs slightly from the typical PK/PD formulation. Here we present how MBMA models can be implemented, analyzed and used for decision support in Monolix and Simulx. As an example, we focus on longitudinal data of the clinical efficacy of drugs for rheumatoid arthritis (RA), following [1]. The goal is to evaluate the efficacy of Canakinumab in comparison to two drugs already on the market (Adalimumab and Abatacept).
Methods: We first collected literature data, incorporating new studies compared to the work presented in [1]. We focus on the ACR20 as endpoint, i.e the percentage of patients achieving 20% improvement. We then formulate a longitudinal mixed effect model with: (i) an Emax structural model, (ii) between-study variability (BSV), (iii) between treatment arm variability (BTAV), and (iv) a residual error. The variances of the BTAV and residual error terms is weighted by the number of individuals per arm, which requires a careful implementation, that we demonstrate using the Monolix software. The model is then used to simulate the true effect of the three drugs, taking into account the uncertainty of the parameter estimates.
Results: The proposed model satisfactorily describe the longitudinal ACR20 data for the three drugs. After parameter estimation, the model is used to predict the chances of Canakinumab to be a more efficacious drug than Adalimumab and Abatacept. To compare the true efficacy (over a infinitely large population), we perform a large number of simulations for the 3 treatments, drawing the Emax population value from its uncertainty distribution. These simulations can easily be done using Simulx, giving the correlation matrix estimated by Monolix as argument. The results show that there are only 6% chances that Canakinumab is better than Abatacept and 16% chance that it is better than Adalimumab.
Conclusions: We have shown that chances are low that Canakinumab performs better than the two drugs already on the market. As a consequence of study [1], the development of this drug has been stopped, thus saving the high costs of the phase III clinical trials. The MonolixSuite offers a powerful environment for longitudinal MBMA analysis.
References:
[1] Demin, I., Hamrén, B., Luttringer, O., Pillai, G., & Jung, T. (2012). Longitudinal model-based meta-analysis in rheumatoid arthritis: an application toward model-based drug development. Clinical Pharmacology and Therapeutics, 92(3), 352–9.
