Background Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disorder associated with respiratory muscle weakness and respiratory failure. Non-invasive ventilation alleviates respiratory symptoms and prolongs life, but is a palliative intervention. Slowing the deterioration of diaphragm function before respiratory failure would be desirable. We aimed to assess whether early diaphragm pacing could slow down diaphragm deterioration and would therefore delay the need for non-invasive ventilation. Methods We did a multicentre, randomised, controlled, triple-blind trial in patients with probable or definite ALS in 12 ALS centres in France. The main inclusion criterion was moderate respiratory involvement (forced vital capacity 60–80% predicted). Other key eligibility criteria were age older than 18 years and bilateral responses of the diaphragm to diagnostic phrenic stimulation. All patients were operated laparoscopically and received phrenic stimulators. Clinicians randomly assigned patients (1:1) to receive either active or sham stimulation with a central web-based randomisation system (computer-generated list). Investigators, patients, and an external outcome allocation committee were masked to treatment. The primary outcome was non-invasive ventilation-free survival, analysed in the intention-to-treat population. Safety outcomes were also assessed in the intention-to-treat population. This trial is registered with ClinicalTrials.gov, number NCT01583088. Findings Between Sept 27, 2012, and July 8, 2015, 74 participants were randomly assigned to receive either active (n=37) or sham (n=37) stimulation. On July 16, 2015, an unplanned masked analysis was done after another trial showed excess mortality with diaphragm pacing in patients with hypoventilation (DiPALS, ISRCTN 53817913). In view of this finding, we analysed mortality in our study and found excess mortality (death from any cause) in our active stimulation group. We therefore terminated the study on July, 16, 2015. Median non-invasive ventilation-free survival was 6·0 months (95% CI 3·6–8·7) in the active stimulation group versus 8·8 months (4·2–not reached) in the control (sham stimulation) group (hazard ratio 1·96 [95% CI 1·08–3·56], p=0·02). Serious adverse events (mainly capnothorax or pneumothorax, acute respiratory failure, venous thromboembolism, and gastrostomy) were frequent (24 [65%] patients in the active stimulation group vs 22 [59%] patients in the control group). No treatment-related death was reported. Interpretation Early diaphragm pacing in patients with ALS and incipient respiratory involvement did not delay non-invasive ventilation and was associated with decreased survival. Diaphragm pacing is not indicated at the early stage of the ALS-related respiratory involvement. Funding Hospital Program for Clinical Research, French Ministry of Health; French Patients' Association for ALS Research (Association pour la Recherche sur la Sclérose Latérale Amyotrophique); and Thierry de Latran Foundation.