Anaemia affects about a quarter of the world's population. An estimated 50% of anaemic people have anaemia due to iron deficiency. To assess the safety and efficacy of iron therapies for the treatment of adults with anaemia who are not pregnant or lactating and do not have chronic kidney disease. We ran the search on 11 July 2013. We searched the Cochrane Central Register of Controlled Trials (CENTRAL), PubMed, EMBASE (Ovid SP), the Cumulative Index to Nursing and Allied Health Literature (CINAHL) Plus (EBSCO Host), the Institute for Scientific Information Web of Science (ISI WOS) Scientific Citation Index (SCI)-EXPANDED (1970) and Conference Proceedings Citation Index (CPCI)-Science (1990) and Clinicaltrials.gov; we also screened reference lists. An updated search was run on 24 November 2014 but the results have not yet been incorporated into the review. Two review authors independently selected references for further assessment by going through all titles and abstracts. Further selection was based on review of full-text articles for selected references. Two review authors independently extracted study data. We calculated the risk ratio (RR) with 95% confidence interval (CI) for binary outcomes and the mean difference (MD) or the standardised mean difference (SMD) with 95% CI for continuous outcomes. We performed meta-analysis when possible, when I(2) was less than or equal to 80% using a fixed-effect or random-effects model, using Review Manager software. The range of point estimates for individual studies is presented when I(2) > 80%. We included in this systematic review 4745 participants who were randomly assigned in 21 trials. Trials were conducted in a wide variety of clinical settings. Most trials included participants with mild to moderate anaemia and excluded participants who were allergic to iron therapy. All trials were at high risk of bias for one or more domains. We compared both oral iron and parenteral iron versus inactive controls and compared different iron preparations.The comparison between oral iron and inactive control revealed no evidence of clinical benefit in terms of mortality (RR 1.05, 95% CI 0.68 to 1.61; four studies, N = 659; very low-quality evidence). The point estimate of the mean difference in haemoglobin levels in individual studies ranged from 0.3 to 3.1 g/dL higher in the oral iron group than in the inactive control group. The proportion of participants who required blood transfusion was lower with oral iron than with inactive control (RR 0.74, 95% CI 0.55 to 0.99; three studies, N = 546; very low-quality evidence). Evidence was inadequate for determination of the effect of parenteral iron on mortality versus oral iron (RR 1.49, 95% CI 0.56 to 3.94; 10 studies, N = 2141; very low-quality evidence) or inactive control (RR 1.04, 95% CI 0.63 to 1.69; six studies, N = 1009; very low-quality evidence). Haemoglobin levels were higher with parenteral iron than with oral iron (MD -0.50 g/dL, 95% CI -0.73 to -0.27; six studies, N = 769; very low-quality evidence). The point estimate of the mean difference in haemoglobin levels in individual studies ranged between 0.3 and 3.0 g/dL higher in the parenteral iron group than in the inactive control group. Differences in the proportion of participants requiring blood transfusion between parenteral iron and oral iron groups (RR 0.61, 95% CI 0.24 to 1.58; two studies, N = 371; very low-quality evidence) or between parenteral iron groups and inactive controls (RR 0.84, 95% CI 0.66 to 1.06; eight studies, N = 1315; very low-quality evidence) were imprecise. Average blood volume transfused was less in the parenteral iron group than in the oral iron group (MD -0.54 units, 95% CI -0.96 to -0.12; very low-quality evidence) based on one study involving 44 people. Differences between therapies in quality of life or in the proportion of participants with serious adverse events were imprecise (very low-quality evidence). No trials reported severe allergic reactions due to parenteral iron, suggesting that these are rare. Adverse effects related to oral iron treatment included nausea, diarrhoea and constipation; most were mild.Comparisons of one iron preparation over another for mortality, haemoglobin or serious adverse events were imprecise. No information was available on quality of life. Thus, little evidence was found to support the use of one preparation or regimen over another.Subgroup analyses did not reveal consistent results; therefore we were unable to determine whether iron is useful in specific clinical situations, or whether iron therapy might be useful for people who are receiving erythropoietin. • Very low-quality evidence suggests that oral iron might decrease the proportion of people who require blood transfusion, and no evidence indicates that it decreases mortality. Oral iron might be useful in adults who can tolerate the adverse events, which are usually mild.• Very low-quality evidence suggests that intravenous iron results in a modest increase in haemoglobin levels compared with oral iron or inactive control without clinical benefit.• No evidence can be found to show any advantage of one iron preparation or regimen over another.• Additional randomised controlled trials with low risk of bias and powered to measure clinically useful outcomes such as mortality, quality of life and blood transfusion requirements are needed.