Longitudinal model-based meta-analysis in rheumatoid arthritis: an application towards model based drug development
I. Demin(1), B. Hamrén(1), O. Luttringer(1), G. Pillai(1), T. Jung(2)
(1)Novartis Pharma AG, Department of Modeling and Simulation, Basel, Switzerland; (2)Novartis Pharma AG, Translational Medicine, Basel, Switzerland
Objectives: The aim was two-fold, first to quantify the longitudinal behavior of the key clinical measure of signs and symptomes (ACR20) in rheumatoid arthritis (RA), over time and across drug treatment and patient population and second, to apply this knowledge in the decision making process for an internal Novartis compound, canakinumab.
Methods: Summary level data was extracted from 39 phase II and III studies including data for all currently approved biologics (nine), in total 105 treatment arms and about 17,000 patients. The longitudinal meta-analysis model describes the full time course of ACR20 of all nine, currently approved biologics, standard of care (methotrexate) as well as true placebo across different patient populations.
Results: This provides insight into the clinical efficacy of the different treatment options and allows a quantitative assessment of the efficacy observed in clinical studies of existing biological treatments in RA. This knowledge was used retrospectively to assess a go/no go decision for canakinumab by incorporating the phase IIb results into the meta-analysis framework. The integrated analysis supported the interpretation that canakinumab had an effect on ACR20, however it also showed that the probability to be as good as the current most effective treatments in terms of the magnitude of effect was low, thereby supporting the decision not to progress canakinumab in RA.
Conclusions: This integrated approach, can be extended to any other compound targeting RA, supporting internal, and external decision making at all clinical development stages.
