Background: Glutamate excitotoxicity might contribute to the pathophysiology of amyotrophic lateral sclerosis. In animal models, decreased excitatory aminoacid transporter 2 (EAAT2) overexpression delays disease onset and prolongs survival, and ceftriaxone increases EAAT2 activity. We aimed to assess the safety and efficacy of ceftriaxone for amyotrophic lateral sclerosis in a combined phase 1, 2, and 3 clinical trial. Methods: This three-stage randomised, double-blind, placebo-controlled study was done at 59 clinical sites in the USA and Canada between Sept 4, 2006, and July 30, 2012. Eligible adult patients had amyotrophic lateral sclerosis, a vital capacity of more than 60% of that predicted for age and height, and symptom duration of less than 3 years. In stages 1 (pharmacokinetics) and 2 (safety), participants were randomly allocated (2:1) to ceftriaxone (2 g or 4 g per day) or placebo. In stage 3 (efficacy), participants assigned to ceftriaxone in stage 2 received 4 g ceftriaxone, participants assigned to placebo in stage 2 received placebo, and new participants were randomly assigned (2:1) to 4 g ceftriaxone or placebo. Participants, family members, and site staff were masked to treatment assignment. Randomisation was done by a computerised randomisation sequence with permuted blocks of 3. Participants received 2 g ceftriaxone or placebo twice daily through a central venous catheter administered at home by a trained caregiver. To minimise biliary side-effects, participants assigned to ceftriaxone also received 300 mg ursodeoxycholic acid twice daily and those assigned to placebo received matched placebo capsules. The coprimary efficacy outcomes were survival and functional decline, measured as the slope of Amyotrophic Lateral Sclerosis Functional Rating Scale-Revised (ALSFRS-R) scores. Analyses were by intention to treat. This study is registered with ClinicalTrials.gov, number NCT00349622. Findings: Stage 3 included 66 participants from stages 1 and 2 and 448 new participants. In total, 340 participants were randomly allocated to ceftriaxone and 173 to placebo. During stages 1 and 2, mean ALSFRS-R declined more slowly in participants who received 4 g ceftriaxone than in those on placebo (difference 0·51 units per month, 95% CI 0·02 to 1·00; p=0·0416), but in stage 3 functional decline between the treatment groups did not differ (0·09, -0·06 to 0·24; p=0·2370). No significant differences in survival between the groups were recorded in stage 3 (HR 0·90, 95% CI 0·71 to 1·15; p=0·4146). Gastrointestinal adverse events and hepatobiliary adverse events were more common in the ceftriaxone group than in the placebo group (gastrointestinal, 245 of 340 [72%] ceftriaxone vs 97 of 173 [56%] placebo, p=0·0004; hepatobiliary, 211 [62%] vs 19 [11%], p<0·0001). Significantly more participants who received ceftriaxone had serious hepatobiliary serious adverse events (41 participants [12%]) than did those who received placebo (0 participants). Interpretation: Despite promising stage 2 data, stage 3 of this trial of ceftriaxone in amyotrophic lateral sclerosis did not show clinical efficacy. The adaptive design allowed for seamless transition from one phase to another, and central venous catheter use in the home setting was shown to be feasible. Funding: National Institute of Neurological Disorders and Stroke.