The aim of this trial was to evaluate the efficacy of a mandibular advancement appliance (MAA) for obstructive sleep apnoea (OSA). Ninety-three patients with OSA and a mean apnoea–hypopnoea index (AHI) of 34.7 were centrally randomised into three, parallel groups: (a) MAA; (b) mandibular non-advancement appliance (MNA); and (c) no intervention. The appliances were custom made, in one piece. The MAAs had a mean protrusion of the mandible of 74% (range 64–85%). Outcome measures, assessed after continuous use for 4 weeks, were AHI (polysomnography), daytime sleepiness (Epworth) and quality of life (SF-36). Eighty-one patients (87%) completed the trial. The MAA group achieved mean AHI and Epworth scores significantly lower (P<0.001 and P<0.05) than the MNA group and the no-intervention group. No significant differences were found between the MNA group and the no-intervention group. The MAA group had a mean AHI reduction of 14.1 (95% CI 7.4–20.8), and a mean Epworth score reduction of 3.3 (95% CI 1.8–4.8). Eight MAA patients (30%) achieved a reduction in AHI ≥75% ending with an AHI <5, half of them having baseline AHI >30. Sensitivity analyses confirmed these results. MAA had a significant beneficial effect on the vitality domain of SF-36. Four MAA patients (14.8%) and two MNA patients (8%) discontinued interventions because of adverse effects. Our conclusion is that MAA has significant beneficial effects on OSA, including cure in some cases of severe OSA. Protrusion of the mandible is essential for the effect. MNA has no placebo effect. MAA may be a good alternative to CPAP in subsets of OSA patients.