Skip navigation

Anticoagulants or antiplatelets for AF

Clinical bottom line

Three meta-analyses have examined substantially the same trials. They have extracted subtly different numbers of events (depending on how stroke is defined, perhaps). There is no doubt that anticoagulants are beneficial. The differences between anticoagulant or antiplatelet therapy are less well accentuated, but on balance anticoagulants win.

Bandolier has abstracted the first of these reviews. The other two have used substantially the same trials, and abstracted substantially the same data in terms of strokes and major bleeds. Two things make the comparisons in each meta-analysis subtly different:

  1. There are primary and secondary prevention studies. Some reviews choose to combine them, others choose to separate them.
  2. The meta-analyses cannot agree about the duration of follow up. So the same trial in two reviews can have differing reported follow up rates.
  3. The numbers of events used for calculations differs between different reviews. Without going back and doing a fourth review, it is impossible to be dogmatic about who is right.

The one we found most helpful was the original meta-analysis by Hart and colleagues. It read very clearly, and gave good and understandable reasons why it had chosen the outcomes it had. The Segal review tries to do much in a little space, and looks as if it is part of an ACHPR work. Though studious, and involving a US Cochrane Centre, look out for mistakes in the tables (inversion of treatments, and unaccountable differences between the number of groups in trials). The Taylor review concentrates on perceived heterogeneity between trials, on the basis of the annual risk of stroke with placebo. This is actually rather interesting, and not totally accounted for by using different follow up times from the Segal review.

We found nothing in the Segal or Taylor reviews that required any additional abstracting. We were disappointed that while both Segal and Taylor acknowledged the earlier review, neither told us why they thought another necessary, nor what the differences were between the new and existing review, not the reasons for those differences. On balance, we're happy to stick with Hart.


RG Hart, O Benavente, R McBride, LA Pearce. Antithrombotic therapy to prevent stroke in patients with atrial fibrillation: a meta-analysis. Annals of Internal Medicine 1999 131: 492-501.

JB Segal et al. Prevention of thromboembolism in atrial fibrillation. A meta-analysis of trials of anticoagulants and antiplatelet drugs. J Gen Intern Med 2000 15: 56-67.

FC Taylor et al. Systematic review of long term anticoagulation or antiplatelet treatment in patients with non-rheumatic atrial fibrillation. BMJ 2001 322: 321-326.


This is tricky territory, and these trials are difficult. There is much to consider, and the confines of a paper published in a journal with word limits is not always the place for a wider perspective easily to be drawn. One does not wish to be overly critical, but these three meta-analyses in the same area are fascinating. They differ, in style, in clarity, in setting the problems and the solutions. They would make a wonderful master class in critical appraisal, but one would need the original randomised trials to hand in order to work out who (if any) was right.