Objectives To conduct a systematic review and meta-analysis investigating effects of exercise for people with alcohol use disorders (AUDs) across multiple health outcomes. We also investigated the prevalence and predictors of dropout from exercise studies in AUDs. Design Systematic review and random effects metaanalysis with meta-regression analyses. Data sources 3 major electronic databases were searched from inception until April 2016 for exercise intervention studies in adults with AUDs. Eligibility criteria Studies of acute exercise in people with AUDs; and randomised and non-randomised trials examining effects of long-term (≥2 weeks) exercise.Results 21 studies and 1204 unique persons with AUDs (mean age 37.8 years, mean illness duration 4.4 years) were included. Exercise did not reduce daily alcohol consumption (standardised mean difference (SMD) =−0.886, p=0.24), or the Alcohol Use Disorders Identification Test (AUDIT) total scores (SMD=−0.378, p=0.18). For weekly consumption (n=3 studies), a statistically significant difference was observed favouring exercise (SMD=−0.656, p=0.04), but not after adjustment for publication bias (SMD=−0.16, 95% CI −0.88 to 0.55). Exercise significantly reduced depressive symptoms versus control (randomised controlled trials (RCTs) =4; SMD=−0.867, p=0.006, I 2 =63%) and improved physical fitness (VO 2 ) (RCTs=3; SMD=0.564, p=0.01, I 2 =46%). The pooled dropout rate was 40.3% (95% CI 23.3% to 60.1%) which was no different to control conditions (OR=0.73, p=0.52). Dropouts were higher among men (β=0.0622, p<0.0001, R 2