An overview of conducting systematic reviews with network meta-analysis

摘要: Systematic reviews with network meta-analysis (NMA) are published increasing frequency in the health care literature. Prior to 2008, very few systematic contained a NMA [1]; however, there has been marked increase, mid-2012 Lee recorded 201 networks [2]. The statistical method available since 2002 [3,4] and owes its origins much earlier work [5,6]. matured models for all types of underlying data summary effect measures [7-12] can be readily implemented both frequentist Bayesian frameworks pre-written programmes widely used softwares [8,13-15]. Recently, focus shifted making more accessible [16,17]; conduct received less attention [18]. In this special thematic series on meta-analysis, editors Reviews encouraging submissions methodological papers concerning reporting meta-analyses results (http://www.systematicreviewsjournal.com/about/update/SysRevCFP). As preface series, editorial provides an overview basic principles summarises some key challenges those conducting review. The need comparative effectiveness research Why increased popularity? To illustrate, consider relative six psychotherapies vs. treatment as usual moderate severe depression [19]. pairwise reviewer three synthesis options: (1) “lump” together form single comparator, (2) separate review, or (3) reviews. If question interest decision-maker is “which psychotherapy should I recommend depression?” syntheses do not satisfactorily translate into practice. A clinician does “average” patient but specific one, such cognitive behavioural therapy. use from options 2 3, must summarise across multiple analyses/reviews without formal assessment whether body evidence was coherent similar enough recommendation. Such approach makes estimates problematic interpret recommended [20]. NMA came prominence within decision-making context [21,22]. simultaneous comparison competing treatments model [23]. simplest form, it combination direct indirect effect, where refers C B obtained studies. This commonly depicted by equation θBCI=θACD-θABD θ denotes true estimate (e.g. log odds ratio, mean difference, etc.) superscript either Direct Indirect evidence. available, they pooled produce internally set each every other have compared head-to-head trials. It also possible calculate probability one being best outcome. Treatment then ranked worst