We have analysed randomized trials comparing unfractionated heparin and low molecular weight heparin in which symptomatic recurrent venous thromboembolism and major bleeding were objectively confirmed. Studies were classified as level 1 if they were double blind or if there was blinded assessment of outcome measures and level 2 if they did not provide assurance of blinded outcome assessment. In studies classified as level 1, the relative risk (RR) of recurrent venous thromboembolism during the first 15 days and over the entire period of anticoagulant therapy was 0.24 (95% confidence intervals [CI] 0.06-0.08, p = 0.02) and 0.39 (95% CI 0.3-0.4, p = 0.006), respectively, in favour of LMWH treatment. The RR for major bleeding was 0.42 (95% CI 0.2-0.9, p = 0.01), in favour of LMWH. In studies classified as level 2, no significant differences in the rates of recurrent venous thromboembolism and major bleeding were observed. Pooling level 1 and level 2 studies, the RR for overall mortality and mortality in cancer patients was 0.51 (95% CI 0.2-0.9, p = 0.01) and 0.33 (95% CI 0.1-1.8, p = 0.01) respectively, in favour of LMWH.