OBJECTIVE The aim was to evaluate the effect of duration of therapy (DOT) on mortality and relapse for patients with Staphylococcus aureus bacteraemia (SAB). METHODS We performed a retrospective single-centre cohort study including adult patients with SAB. We determined the association between DOT (≤14 days versus >14 days) and mortality by adjusted hazard ratios (aHR) and 95% confidence intervals through Cox regression adjusted for immortal-time bias and confounding by indication, stratified by presence of complicated SAB (any of: endocarditis, implant, duration of SAB >2 days, fever >3 days). The primary outcome was 90-day all-cause mortality, and the secondary outcome was 90-day relapse. RESULTS Between January 2010 and December 2015, we included 530 patients, of whom 94 out of 530 (17.7%) had methicillin-resistant SAB and 305 out of 530 (57.6%) had complicated SAB. Ninety-day mortality was 27.0% (143/530), with no significant trend across the study period; median time to death was 17 days (interquartile range (IQR) 8-30) after onset of SAB. Median DOT was 20 days (IQR 13-39). Patients with complicated SAB had significantly reduced mortality with DOT >14 days (aHR 0.32, 95% CI 0.16-0.64). DOT was not associated with mortality in patients with uncomplicated SAB (aHR 0.85; 0.41-1.78). Eighteen (18/530) patients (3.4%) relapsed; on univariate analysis, DOT was not associated with relapse (HR 1.01; 0.97-1.06). CONCLUSIONS DOT >14 days is associated with higher survival in patients with complicated SAB, but not for patients with uncomplicated SAB. No association was found for relapse, but 90-day relapse was very low in this cohort. Importantly, 90-day mortality remained high across the study period.