INTRODUCTION: The three main novel anticoagulants (NOACs) currently licensed in Europe, apixaban, dabigatran and rivaroxaban, have all been directly compared against warfarin in randomised controlled trials. However, none of the three drugs have been directly compared against each other. Thus, there has been an increase in the number of meta-analyses and indirect comparisons published comparing the relative efficacy and safety of these novel anticoagulants against each other via warfarin as a common comparator.

OBJECTIVES: Systematically review all meta-analyses and indirect comparisons evaluating the NOACs against warfarin for the prevention of stroke in patients with AF and critically appraise the statistical methods used to do so.

METHODS: Systematic searches of EMBASE, MedLine, EBM Reviews, EconLIT as well as manual searches of ClincalTrials.gov, the Cochrane Library, CADTH, NICE, NHSEED and HTA were conducted. Data was abstracted from any citation applying statistical methods to compare the efficacy and safety of NOACs for the prevention of AF-related stroke. Information regarding the statistical approach; model assumptions; data presentation; interpretation of the evidence; and discussions of internal and external validity was used to quality rate each study.

RESULTS: Bucher’s method of adjusted indirect comparison was most widely used. There were generally three main model assumptions required: the similarity, homogeneity and consistency assumptions, each being investigated with varying scrutiny in the studies reviewed. According to the quality assessment, the indirect comparison conducted by Wells and colleagues (2012) is of the highest relative quality.

CONCLUSIONS: The limited number of RCTs available comparing the NOACs to standard therapy, creates considerable uncertainty surrounding the comparative efficacy and safety of these anticoagulants. In order to establish which individual NOAC is most likely to benefit a given patient population, indirect comparisons and meta-analyses are increasingly used. However, the quality of indirect comparison studies are variable and results should be interpreted with care.