Antiplatelet therapy for thromboprophylaxis: the need for careful consideration of the evidence from randomised trials

Publisher Website
Google Scholar
Abstract
Venous thromboses and pulmonary embolism remain an important cause of morbidity and mortality both in surgical patients and in immobilised medical patients.*RF 1-5* Various thromboprophylactic treatments have, therefore, been devised to prevent or limit thromboembolism. Our previous systematic overview (or meta-analysis) of randomised trials of perioperative subcutaneous heparin found that among surgical patients such treatment can roughly halve the risk not only of deep venous thrombosis but, more importantly, of pulmonary embolism6 (see fig 1). Subcutaneous heparin is now widely recommended for surgical or medical patients at high risk of venous occlusion.*RF 3-5* The recent Antiplatelet Trialists' Collaboration overview of the thromboprophylactic effects of antiplatelet therapy used prospectively determined criteria for trial inclusion and treatment comparisons that were similar to those of the previous heparin overview.*RF 6-8* The aim was to include all unconfounded properly randomised trials of antiplatelet versus no antiplatelet therapy (or of one antiplatelet regimen versus another) that could have been available for review by March 1990 in which deep venous thrombosis was systematically and unbiasedly monitored. (Parts I and III of the previous overview report give a fuller description of the methods used.1,7 The appropriateness of using “assumption free” statistical methods rather than the “random effects” model when combining trial results, as when combining results from different centres in a multicentre trial, has been discussed in detail previously.9,10) Such randomised trials were to be included whether or not the treatment comparison was “blinded” by placebo control. This was also the case in the heparin overview, where exclusion of informative “open” trials (in particular, the important open international multicentre trials coordinated by Professor V V Kakkar11) would have been equally inappropriate. Analyses confined to placebo controlled studies, which may be less subject to treatment dependent biases in the assessment …