A sensitivity analysis omitting studies of low quality
was planned to assess the effect of the methodological
quality of the primary studies on the overall outcome.
Two subgroup analyses were done after the fact.
One analysis assessed the risk of bleeding for patients
maintained at higher international normalized ratios
(INRs) and included studies in which the mean INR
of the study group was > 3. A larger treatment effect
was hypothesized for these studies. Another analysis was
done for studies that used antifibrinolytic agents; a
smaller treatment effect was hypothesized for these
studies.