Nine RCTs were included in the review (n=19,218 participants). All trials were double-blinded and most had adequate randomisation and allocation concealment
Rivaroxaban was superior to enoxaparin for the prevention of venous thromboembolism (RR 0.56, 95% CI 0.43 to 0.73; I2=54%; six RCTs; n=11,009 participants), but with a trend for increased haemorrhage (RR 1.26, 95% CI 0.94 to 1.69; I2=28%; six RCTs; n=11,009 participants).
Dabigatran was not superior to enoxaparin for prevention of venous thromboembolism (RR 1.12, 95% CI 0.97 to 1.29; I2=41%; three RCTs; n=8,209 participants). The risk of haemorrhage with dabigatran and enoxaparin were the same.
Rivaroxaban was superior to dabigatran in preventing venous thromboembolism (RR 0.50, 95% CI 0.37 to 0.68), but with a slight trend towards increased haemorrhage (RR 1.14, 95% CI 0.80 to 1.64).
The benefit:harm ratio (derived from the change in event rates from the use of rivaroxaban instead of dabigatran) was reported to vary from three to 15 venous thromboembolism events saved by rivaroxaban for every additional haemorrhage caused.
A dose of dabigatran 150mg daily was significantly less effective than enoxaparin (RR 1.21, 95% CI 1.05 to1.39), whereas a dabigatran dose of 220mg daily was similar to enoxaparin in preventing total venous thromboembolism events.
Sensitivity analysis using fixed-effect instead of random-effects models did not significantly change the results of the main analysis. Observed effects remained consistent in intention-to-treat analysis or the evaluable-for-primary-outcome and safety populations as the denominator.