Forty RCTs met the inclusion criteria (62,480 participants, range 20 to 14,578): 21 conducted in medical patients and 19 in patients with acute stroke. Seventeen trials reported adequate allocation concealment, 22 were reported as double blind (six also blinded outcome assessors), six blinded outcome assessors only, 19 used ITT analysis in at least some outcomes and 28 adequately described withdrawals and drop-outs.
Medical patients: Compared to no heparin, heparin reduced the incidence of pulmonary embolism (OR 0.69, 95% CI 0.52 to 0.90; 10 RCTs) and increased the incidence of bleeding (OR 1.34, 95% CI 1.08 to 1.66; eight RCTs), but had no significant impact on mortality (10 RCTs), symptomatic DVT (five RCTs), fatal pulmonary embolism (five RCTs) and major bleeds (nine RCTs). There was no significant difference between unfractionated heparin and low molecular weight heparin (LMWH) in terms of mortality (nine RCTs), pulmonary embolism (nine RCTs), bleeding (eight RCTs), symptomatic DVT (three RCTs), fatal pulmonary embolism (three RCTs) and major bleeds (nine RCTs).
Acute stroke: Compared to no heparin, heparin increased the incidence of major bleeding (OR 1.66, 95% CI 1.20 to 2.28; eight RCTs), but had no significant impact on mortality (eight RCTs), symptomatic DVT (one RCT), pulmonary embolism (five RCTs), fatal pulmonary embolism (two RCTs) and all bleeding (six RCTs). There was no significant difference between unfractionated heparin and LMWH in terms of mortality (five RCTs), pulmonary embolism (four RCTs), bleeding (five RCTs), symptomatic DVT (two RCTs), fatal pulmonary embolism (four RCTs) and major bleeds (five RCTs). There was no significant impact of mechanical prophylaxis on mortality (three RCTs), symptomatic DVT (one RCT), pulmonary embolism (two RCTs). Mechanical prophylaxis increased skin damage (RR 4.02, 95% CI 2.34 to 6.91; one RCT).
Results were presented for medical and stroke patients combined and for sensitivity analyses. Results were tabulated for comparisons evaluated in single RCTs. Data from two RCTs that compared different durations of LMWH were reported. The authors reported evidence of publication bias in the analyses of heparin versus no heparin.