Clinical practice guidelines regarding the use of high-flux haemodialysis membranes vary widely. We aimed to analyse the current evidence reported for the benefits and harms of high-flux and low-flux haemodialysis. We searched Cochrane Renal Group's specialised register (July 2012), the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE (1948 to March 2011), and EMBASE (1947 to March 2011) without language restriction. We included randomised controlled trials (RCTs) that compared high-flux haemodialysis with low-flux haemodialysis in people with end-stage kidney disease (ESKD) who required long-term haemodialysis. Data were extracted independently by two authors for study characteristics (participants and interventions), risks of bias, and outcomes (all-cause mortality and cause-specific mortality, hospitalisation, health-related quality of life, carpal tunnel syndrome, dialysis-related arthropathy, kidney function, and symptoms) among people on haemodialysis. Treatment effects were expressed as a risk ratio (RR) or mean difference (MD), with 95% confidence intervals (CI) using the random-effects model. We included 33 studies that involved 3820 participants with ESKD. High-flux membranes reduced cardiovascular mortality (5 studies, 2612 participants: RR 0.83, 95% CI 0.70 to 0.99) but not all-cause mortality (10 studies, 2915 participants: RR 0.95, 95% CI 0.87 to 1.04) or infection-related mortality (3 studies, 2547 participants: RR 0.91, 95% CI 0.71 to 1.14). In absolute terms, high-flux membranes may prevent three cardiovascular deaths in 100 people treated with haemodialysis for two years. While high-flux membranes reduced predialysis beta-2 microglobulin levels (MD -12.17 mg/L, 95% CI -15.83 to -8.51 mg/L), insufficient data were available to reliably estimate the effects of membrane flux on hospitalisation, carpal tunnel syndrome, or amyloid-related arthropathy. Evidence for effects of high-flux membranes was limited by selective reporting in a few studies. Insufficient numbers of studies limited our ability to conduct subgroup analyses for membrane type, biocompatibility, or reuse. In general, the risk of bias was either high or unclear in the majority of studies. High-flux haemodialysis may reduce cardiovascular mortality in people requiring haemodialysis by about 15%. A large well-designed RCT is now required to confirm this finding.