Nine RCTs (3,999 patients) were included.
All studies used proper methods to generate the randomised treatment allocation and all trials appeared to use adequate concealment. In 7 of the 9 studies, neither the patient nor investigator was aware of the treatment allocation.
The follow-up was 100% in 6 RCTs and greater than 90% in the other 3 RCTs.
Out-of hospital symptomatic venous thromboembolism. Extended-duration prophylaxis significantly reduced the risk of symptomatic venous thromboembolism, compared with placebo or untreated control; the OR was 0.38 (95% CI: 0.24, 0.61) and the NNT was 50. There was no evidence of statistical heterogeneity (chi-squared 5.65, d.f.=8, p=0.69). There was a significant reduction in deep vein thrombosis (OR 0.41, 95% CI: 0.24, 0.68; NNT = 62), but a low event rate for pulmonary embolism with a non statistically-significant reduction (OR 0.43, 95% CI: 0.17, 1.06; NNT = 250).
There was a greater risk reduction in patients undergoing hip replacement (OR 0.33, 95% CI: 0.19, 0.56; NNT = 34), compared with knee replacement (the OR from 2 RCTs was 0.74, 95% CI: 0.26, 2.15; NNT = 250).
Out-of hospital symptomatic venous thromboembolism (1,901 patients).
Extended-duration prophylaxis significantly reduced deep vein thrombosis; the OR was 0.48 (95% CI: 0.36, 0.63) and the NNT was 10. There was no evidence of statistical heterogeneity (chi-squared 5.89, d.f.=6, p=0.43).
There was no evidence of heterogeneity between: UFH and LMWH (p=0.96); different durations of in-hospital prophylaxis; or studies that used mandatory bilateral venography at hospital discharge, compared with trials not undertaking this investigation.
The results were similar after excluding each individual study in turn and after excluding lower-quality studies. There were no important differences between the results from fixed- and random- effects models.
A funnel plot was consistent with no evidence for publication bias.
Adverse reactions.
Extended-duration prophylaxis did not increase major bleeding (0.1% versus 0.3%) or all-cause mortality (0.1% versus 0.3%). However, it was associated with an increased risk of minor bleeding (3.7% versus 2.5%); the OR was 1.56 (95% CI: 1.08, 2.26; NNH = 83).