|
Abstract:
|
Patients, providers, and payers need to make evidence-based decision in terms of treatment choice and expected outcomes. Network meta-analysis (NMA) synthesizes both direct and indirect evidence, which plays an important role in quantifying relative efficacy and safety especially when head-to-head comparison is lacking. Although many guidance documents on designing, conducting, and interpreting NMA have been developed by professional societies like ISPOR and key HTA bodies as UK NICE, they mainly rely on aggregated study level data from randomized clinical trials. As increasing number of observational studies become available, different methodologies have been proposed to synthesize both randomized and non-randomized studies. These include design-adjusted synthesis, use of non-randomized evidence as prior information, and three-level hierarchical models. We will report the underlying assumptions, pros and cons, performance of different methods and recommend practical considerations in undertaking such research endeavors.
|
