Inclusion and exclusion criteria
The inclusion crit er ia for this meta-analysis were as follows: original prospective, randomized research articles on human subjects diagnosed with nvaf; articles presented in English; subjects selected from the general population, regardless of age. Although studies with subjects who had taken any anticoagulants or antiplateletdrugswithin six months before entry into the study were excluded, studies with subjects who had taken other Canadian medications (including acetylsalicylic acid [ASA]) during the study period were included. Moreover, subjects were not restricted by warfarin dose or International Normalized Ratio (inr), and subjects were assessed based on any of the outcomes of interest including stroke, transient ischemic attack (tia), systemic embolism, death or adverse effects (eg, hemorrhage). Studies did not have to be placebo controlled or blinded. Excluded were case reports, editorials, reviews and studies in which the data were not extractable.
To avoid selection bias, two judges were blinded (noninvestigators of the meta-analysis) and asked to determine which studies should be included in the meta-analysis. The Method section of all articles was photocopied. The investigators removed all identifying statements and titles (names of investigators, studies, journals, geographical locations, dates) in each Method sect ion. These revised Method sect ions were then photocopied and numbered using random numbers generated by a calculator. Each judge was given the same set of Method sections and a list of criteria for acceptability and asked to evaluate each study according the inclusion criteria based on the information presented. If there was disagreement between the two judges on the eligibility of a study for the meta-analysis, another judge (noninvestigator) was asked make a final decision. Accepted studies were then classified into prospective or retrospective trials, blinded or not blinded and placebo/ ASA-controlled.