Millions of people worldwide suffer from hepatitis C, which can lead to severe liver disease, liver cancer, and death. Direct-acting antivirals (DAAs) are relatively new and expensive interventions for chronic hepatitis C, and preliminary results suggest that DAAs may eradicate hepatitis C virus (HCV) from the blood (sustained virological response). However, it is still questionable if eradication of hepatitis C virus in the blood eliminates hepatitis C in the body, and improves survival and leads to fewer complications.To assess the benefits and harms of DAAs in people with chronic HCV.We searched for all published and unpublished trials in The Cochrane Hepato-Biliary Group Controlled Trials Register, CENTRAL, MEDLINE, Embase, Science Citation Index Expanded, LILACS, and BIOSIS; the Chinese Biomedical Literature Database (CBM), China Network Knowledge Information (CNKI), the Chinese Science Journal Database (VIP), Google Scholar, The Turning Research into Practice (TRIP) Database, ClinicalTrials.gov, European Medicines Agency (EMA) (www.ema.europa.eu/ema/), WHO International Clinical Trials Registry Platform (www.who.int/ictrp), the Food and Drug Administration (FDA) (www.fda.gov), and pharmaceutical company sources for ongoing or unpublished trials. Searches were last run in October 2016.Randomised clinical trials comparing DAAs versus no intervention or placebo, alone or with co-interventions, in adults with chronic HCV. We included trials irrespective of publication type, publication status, and language.We used standard methodological procedures expected by Cochrane. Our primary outcomes were hepatitis C-related morbidity, serious adverse events, and quality of life. Our secondary outcomes were all-cause mortality, ascites, variceal bleeding, hepato-renal syndrome, hepatic encephalopathy, hepatocellular carcinoma, non-serious adverse events (each reported separately), and sustained virological response. We systematically assessed risks of bias, performed Trial Sequential Analysis, and followed an eight-step procedure to assess thresholds for statistical and clinical significance. The overall quality of the evidence was evaluated using GRADE.We included a total of 138 trials randomising a total of 25,232 participants. The 138 trials assessed the effects of 51 different DAAs. Of these, 128 trials employed matching placebo in the control group. All included trials were at high risk of bias. Eighty-four trials involved DAAs on the market or under development (13,466 participants). Fifty-seven trials administered withdrawn or discontinued DAAs. Trial participants were treatment-naive (95 trials), treatment-experienced (17 trials), or both treatment-naive and treatment-experienced (24 trials). The HCV genotypes were genotype 1 (119 trials), genotype 2 (eight trials), genotype 3 (six trials), genotype 4 (nine trials), and genotype 6 (one trial). We identified two ongoing trials.Meta-analysis of the effects of all DAAs on the market or under development showed no evidence of a difference when assessing hepatitis C-related morbidity or all-cause mortality (OR 3.72, 95% CI 0.53 to 26.18, P = 0.19, I² = 0%, 2,996 participants, 11 trials, very low-quality evidence). As there were no data on hepatitis C-related morbidity and very few data on mortality (DAA 15/2377 (0.63%) versus control 1/617 (0.16%)), it was not possible to perform Trial Sequential Analysis on hepatitis C-related morbidity or all-cause mortality.Meta-analysis of all DAAs on the market or under development showed no evidence of a difference when assessing serious adverse events (OR 0.93, 95% CI 0.75 to 1.15, P = 0.52, I² = 0%, 15,817 participants, 43 trials, very low-quality evidence). The Trial Sequential Analysis showed that the cumulative Z-score crossed the trial sequential boundary for futility, showing that there was sufficient information to rule out that DAAs compared with placebo reduced the relative risk of a serious adverse event by 20%. The only DAA that showed a significant difference on risk of serious adverse events when meta-analysed separately was simeprevir (OR 0.62, 95% CI 0.45 to 0.86). However, Trial Sequential Analysis showed that there was not enough information to confirm or reject a relative risk reduction of 20%, and when one trial with an extreme result was excluded, then the meta-analysis result showed no evidence of a difference.DAAs on the market or under development seemed to reduce the risk of no sustained virological response (RR 0.44, 95% CI 0.37 to 0.52, P < 0.00001, I² = 77%, 6886 participants, 32 trials, very low-quality evidence) and Trial Sequential Analysis confirmed this meta-analysis result.Only 1/84 trials on the market or under development assessed the effects of DAAs on health-related quality of life (SF-36 mental score and SF-36 physical score).Withdrawn or discontinued DAAs had no evidence of a difference when assessing hepatitis C-related morbidity and all-cause mortality (OR 0.64, 95% CI 0.23 to 1.79, P = 0.40, I² = 0%; 5 trials, very low-quality evidence). However, withdrawn DAAs seemed to increase the risk of serious adverse events (OR 1.45, 95% CI 1.22 to 1.73, P = 0.001, I² = 0%, 29 trials, very low-quality evidence), and Trial Sequential Analysis confirmed this meta-analysis result.Most of all outcome results were short-term results; therefore, we could neither confirm nor reject any long-term effects of DAAs. None of the 138 trials provided useful data to assess the effects of DAAs on the remaining secondary outcomes (ascites, variceal bleeding, hepato-renal syndrome, hepatic encephalopathy, and hepatocellular carcinoma).Overall, DAAs on the market or under development do not seem to have any effects on risk of serious adverse events. Simeprevir may have beneficial effects on risk of serious adverse event. In all remaining analyses, we could neither confirm nor reject that DAAs had any clinical effects. DAAs seemed to reduce the risk of no sustained virological response. The clinical relevance of the effects of DAAs on no sustained virological response is questionable, as it is a non-validated surrogate outcome. All trials and outcome results were at high risk of bias, so our results presumably overestimate benefit and underestimate harm. The quality of the evidence was very low.