Patients with deep venous thrombosis are at a short-term risk of symptomatic or even life-threatening pulmonary embolism, and a long-term risk of post-thrombotic syndrome, characterised by lower-limb pain, varicose veins, oedema, and sometimes skin ulcers. What is the best choice of initial antithrombotic therapy following deep venous thrombosis or pulmonary embolism, in terms of mortality and short-term and long-term complications? How do the harm-benefit balances of the different options compare? To answer these questions, we reviewed the available literature using the standard Prescrire methodology. Unfractionated heparin has documented efficacy in reducing mortality and recurrent thromboembolic events in patients with pulmonary embolism or symptomatic proximal (above-knee) deep venous thrombosis. The authors of a systematic review selected 23 trials of low-molecular-weight heparin (LMWH) versus adjusted-dose unfractionated heparin in a total of 9587 patients. Deaths, recurrences and major bleeds were less frequent with LMWH than with unfractionated heparin. The results of other meta-analyses are similar, but all are undermined by a probable publication bias and methodological flaws. Compared to unfractionated heparin, LMWHs have the advantage of fixed-dose administration, once or twice daily, by subcutaneous injection. All available LMWHs seem to have similar efficacy. Those with the longest experience of use are enoxaparin, dalteparin and nadroparin. The harm-benefit balances of fondaparinux and rivaroxaban do not appear more favourable than that of an LMWH followed by an adjusted-dose vitamin K antagonist. A meta-analysis included 12 trials comparing thrombolysis with anticoagulation alone in 700 patients with deep venous thrombosis. Adding a thrombolytic drug did not reduce mortality or the incidence of pulmonary embolism, whereas it increased the incidence of bleeding. A meta-analysis of 13 trials failed to show that adding a thrombolytic drug to initial anticoagulant therapy reduced mortality or recurrences after pulmonary embolism. In the 5 trials that included patients with massive pulmonary embolism, thrombolytic therapy appeared to reduce mortality by about one-half (6% versus 13%). This difference is noteworthy, even if it did not reach the usual threshold of statistical significance. The results of the 6 trials involving patients with deep venous thrombosis, and those of 2 trials and 8 cohort studies in patients with pulmonary embolism at low risk of complications, suggest that outpatient management is acceptable in some cases. Clinical practice guidelines largely agree on the use of LMWH or fondaparinux as initial therapy for most patients with deep venous thrombosis or pulmonary embolism. Unfractionated heparin is generally recommended for patients with renal failure. Thrombolysis is recommended for massive pulmonary embolism and, in some guidelines, for iliofemoral venous thrombosis. In practice, initial treatment of deep venous thrombosis and pulmonary embolism should be based on LMWH in patients without renal failure. Thrombolytic agents may be useful in case of massive pulmonary embolism, but more evaluation is needed. Bleeding and heparin thrombocytopenia are the main adverse effects of these treatments.