Introduction Postoperative delirium (POD) is one of the most frequent complication after surgery in elderly patients, and is associated with increased morbidity and mortality, prolonged length of stay, cognitive and functional decline leading to loss of autonomy, and important additional healthcare costs. Perioperative inflammatory stress is a key element in POD genesis. Melatonin exhibits antioxidative and immune-modulatory proprieties that are promising concerning delirium prevention, but in perioperative context literature are scarce and conflicting. We hypothesise that perioperative melatonin can reduce the incidence of POD. Methods and analysis The DELIRLESS trial is a prospective, national multicentric, phase III, superiority, comparative randomised (1:1) double-blind clinical trial. Among patients aged 70 or older, hospitalised and scheduled for surgery of a severe fracture of a lower limb, 718 will be randomly allocated to receive either melatonin 4 mg per os or placebo, every night from anaesthesiologist preoperative consultation and up to 5 days after surgery. The primary outcome is POD incidence measured by either the French validated translation of the Confusion Assessment Method (CAM) score for patients hospitalised in surgery, or CAM-ICU score for patients hospitalised in ICU (Intensive Care Unit). Daily delirium assessment will take place during 10 days after surgery, or until the end of hospital stay if it is shorter. POD cumulative incidence function will be compared at day 10 between the two randomised arms in a competing risks framework, using the Fine and Grey model with death as a competing risk of delirium. Ethics and dissemination The DELIRLESS trial has been approved by an independent ethics committee the Comité de Protection des Personnes (CPP) Sud-Est (ref CPP2020-18-99 2019-003210-14) for all study centres. Participant recruitment begins in December 2020. Results will be published in international peer-reviewed medical journals. Trial registration number NCT04335968, first posted 7 April 2020.