PRM64 Bayesian MTC Models to Combine Evidence From Different Trial Designs

Abstract

OBJECTIVES: Bayesian mixed treatment comparison models (MTCs) provide a powerful methodology to obtain estimates of relative efficacy between alternative treatments when head to head evidence is not available or not sufficient. Most evaluations only consider evidence from randomized controlled trials (RCTs), while information from other trial designs is ignored. In this work we propose 3 methods to extend MTC models to systematically include evidence from different trial designs using an application in Rheumatoid Arthritis (RA). METHODS: A systematic literature review identified 13 RCTs and 3 observational trials assessing the treatment effects of five anti-TNF agents currently licensed in Europe. Naive Pooling does not differentiate between designs, one simply pools across all studies. It is not possible to down-weight designs of lesser quality or to adjust for bias. Alternatively observational data can be analysed separately and the results used to inform the prior distribution for the RCT model. This allows for bias adjustments and controlling the influence on the overall effect. In addition to that, a 3-level hierarchical model allows the direct comparison of estimates on study type level to overall level. The method accounts for between trial design heterogeneity; overall estimates become more conservative when study type estimates differ. RESULTS: Including evidence from observational trials to estimate the relative efficacy between anti-TNF agents in RA has strengthened our belief in the effect estimates. Overall, the observational trial data found less difference between the agents than was suggested by RCT evidence only. CONCLUSIONS: Observational data is available for many disease areas providing additional information on treatment effectiveness. We think it is important for an informed decision making process to include all available evidence. The proposed techniques provide a framework for systematically including evidence from different trial designs in a MTC model.