The most robust evidence for the efficacy of ASA comes from the APTC meta-analysis (APTC, 1994) and the PEP trial (Rodgers et al, 2000). All other evidence is from small or non-randomized trials and does not change the conclusions derived from these two publications.
The Antiplatelet Trialists’ Collaboration meta-analysis
The APTC meta-analysis of antiplatelet therapy for VTE prevention included randomized trials in which DVT was systematically sought by either contrast venography or fibrinogen leg scanning, and reported data on over 9000 patients at high-risk of VTE (APTC, 1994). In all the included studies, PE was suspected clinically and confirmed by objective diagnostic testing (e.g., ventilation perfusion scanning) or at necropsy.
The results of the APTC meta-analysis indicated that any antiplatelet therapy (ASA, ASA plus dipyridamole, hydroxychloroquine, or ticlopidine) compared to control was associated with a 26% reduction in the risk of DVT (24·8 vs. 33·6%, P < 0·0001) and a 63% reduction in PE (1 vs. 2·7%; P < 0·0001). Among 1619 patients in 16 trials, ASA reduced DVT by 20% (24·1–30%) and PE by 69% (1–3%) when compared with control. Antiplatelet therapy compared to control significantly increased the need for transfusion (from 0·4 to 0·7%, P = 0·04) as well as wound haematomas, reoperation and infections related to bleeding (7·8 vs. 5·6%; P = 0·003).
The APTC meta-analysis has important strengths. It included all unconfounded randomized trials, both published and unpublished, and thus captured the totality of the evidence available prior to 1990. A consistent benefit of antiplatelet therapy was evident across subsets of patients undergoing different surgical procedures and high-risk medical patients. Surgery included general, elective orthopaedic procedures and surgery for fractures. The trials involving high-risk medical patients included patients with stroke, recurrent venous thrombosis, congestive heart failure and myocardial infarction.
Two methodological issues that have been the topic of much debate (Cohen et al, 1994) are relevant to the interpretation of the results of the APTC meta-analysis. The first pertains to the lack of blinding in many of the trials included in the APTC meta-analysis. Randomized trials in which patients, care providers, and outcome adjudicators are blinded to treatment assignment, yield higher quality evidence than open-label trials. In a trial where treatment assignment is open-label, blinded adjudication of outcomes can reduce bias, particularly if the outcome is measured in all trial participants (e.g., mandatory venography). On the other hand, in VTE prevention trials, if the decision to perform an objective diagnostic test for PE or DVT is driven by the occurrence of subjective symptoms and signs, there is a strong potential for bias. DVT was sought by systematic fibrinogen leg scanning or venography in 53 of the 62 studies included in the APTC meta-analysis and 17 of these studies were open label. All 62 studies reported PE, 19 of which were open label (APTC, 1994). The benefits of antiplatelet therapy were reported to be almost identical irrespective of whether patients and health care providers were blinded to treatment allocation (Odds reduction of 38% for the placebo-controlled studies and 39% in the whole meta-analysis) (APTC, 1994; Collins et al, 1994). Thus, despite the shortcoming that almost a third (31%) of trials were open label, the results are likely to be valid.
The second methodological issue in the APTC analysis pertains to the method used for diagnosing DVT. The traditional diagnostic gold standard for evaluating the efficacy of thromboprophylaxis is ascending contrast venography adjudicated by observers blinded to treatment allocation. Other less accurate outcome measures have the potential to systematically bias treatment estimates (Rodgers & MacMahon, 1995). Most early studies assessed the efficacy of thromboprophylaxis using I125-fibrinogen scanning, which has much lower sensitivity and specificity than venography for deep vein thrombosis (Cruickshank et al, 1989). The APTC investigators showed that the effectiveness of antiplatelet therapy on reducing DVT outcomes was evident irrespective of whether venography or fibrinogen scanning was used to assess outcome (odds reduction of 49% for the venographic trials compared to 39% for the overall meta-analysis) (APTC, 1994). Furthermore, the estimates of benefit of antiplatelet therapy obtained from the APTC meta-analysis might be conservative because of the tendency of fibrinogen scans to underestimate the magnitude of the treatment effect when active therapy is compared with no active treatment (Rodgers & MacMahon, 1995). Indeed, Rodgers and MacMahon (1995), using data from the meta-analysis showed that ASA produced smaller relative risk reductions (RRR) for the prevention of DVT compared with placebo when the outcomes were assessed using fibrinogen leg scanning compared with venography (RRR 33 vs. 56% respectively).
The Pulmonary Embolism Prevention trial
The only large trial specifically designed to test the efficacy and safety of ASA for VTE prevention is the PEP study (Rodgers et al, 2000). The PEP trial was a double blind randomized trial that included more than 13 000 patients undergoing surgery for hip fracture and about 4000 patients undergoing elective joint replacement surgery. Patients were randomized to receive either 160 mg per d of ASA or placebo, beginning preoperatively and continued for 35 d. About 18% of the randomized patients received treatment with unfractionated heparin (UFH) and about 26% received low molecular weight heparins (LMWH). The primary outcome was the occurrence of symptomatic VTE at 35 d. In contrast to earlier studies of VTE prophylaxis, ascertainment of outcomes in the PEP trial did not involve mandatory screening for DVT; only clinically suspected DVTs that were objectively confirmed by venography or duplex ultrasound were counted as outcomes. Likewise, all episodes of pulmonary embolism were suspected clinically and confirmed by objective testing using ventilation perfusion scanning, pulmonary angiography, or at necropsy. Outcomes were adjudicated by a committee that was masked to treatment assignment.
The results of the PEP study demonstrated that ASA produced proportional reductions in DVT of 28% (95% confidence interval (CI) 4–45%; P = 0·02) and PE of 39% (95% CI 16–54; P < 0·005) (Collins et al, 2000). Among the subgroup of patients undergoing surgery for hip fracture, ASA reduced the risk of DVT by 29% (95% CI 3–48%; P = 0·03) and PE by 43% (95% CI 18–60; P = 0·002). This translates into a reduction of 9 VTE events for every 1000 patients treated. The benefits of ASA were partly offset by an increase of six postoperative bleeds that required transfusion for every 1000 patients treated with ASA (2·9 vs. 2·4%; P = 0·04), and an excess of haematemesis and melena, which did not require transfusion (2·7 vs. 1·8%; P = 0·0005). A consistent reduction in VTE was found in patients undergoing elective arthroplasty (HR 0·83, 95% CI 0·47–1·47; P for heterogeneity 0·4) although it was not statistically significant. The benefits of ASA were evident irrespective of whether patients received concomitant LMWH therapy. The authors of the PEP study have been criticized appropriately for de-emphasizing their prespecified primary endpoints of any vascular death, and major non-fatal vascular events (PE, myocardial infarction or stroke) (MacMahon et al, 1994; Cohen & Quinlan, 2000). The PEP investigators apparently made the decision to focus on secondary outcomes (DVT, PE and any VTE) in preference to their pre-specified primary endpoint after the results were known to them. This oversight weakens the conclusions as these are derived from what are essentially post-hoc analyses (Collins et al, 2000).
Nevertheless, criticism notwithstanding, data from the PEP trial, supported by the results of the APTC meta-analysis, provide convincing evidence that ASA, compared to placebo, significantly reduces DVT and PE in high-risk patients.