BACKGROUND: Patients with N2-3 nasopharyngeal carcinoma have a high risk of treatment being unsuccessful despite the current practice of using a concurrent adjuvant cisplatin-fluorouracil regimen. We aimed to compare the efficacy and safety of concurrent adjuvant cisplatin-gemcitabine with cisplatin-fluorouracil in N2-3 nasopharyngeal carcinoma. METHODS: We conducted an open-label, randomised, controlled, phase 3 trial at four cancer centres in China. Eligible patients were aged 18-65 years with untreated, non-keratinising, stage T1-4 N2-3 M0 nasopharyngeal carcinoma, an Eastern Cooperative Oncology Group performance status score of 0-1, and adequate bone marrow, liver, and renal function. Eligible patients were randomly assigned (1:1) to receive concurrent cisplatin (100 mg/m intravenously) on days 1, 22, and 43 of intensity-modulated radiotherapy followed by either gemcitabine (1 g/m intravenously on days 1 and 8) and cisplatin (80 mg/m intravenously for 4 h on day 1) once every 3 weeks or fluorouracil (4 g/m in continuous intravenous infusion for 96 h) and cisplatin (80 mg/m intravenously for 4 h on day 1) once every 4 weeks, for three cycles. Randomisation was done using a computer-generated random number code with a block size of six, stratified by treatment centre and nodal category. The primary endpoint was 3-year progression-free survival in the intention-to-treat population (ie, all patients randomly assigned to treatment). Safety was assessed in all participants who received at least one dose of chemoradiotherapy. This study was registered at ClinicalTrials.gov, NCT03321539, and patients are currently under follow-up. FINDINGS: From Oct 30, 2017, to July 9, 2020, 240 patients (median age 44 years [IQR 36-52]; 175 [73%] male and 65 [27%] female) were randomly assigned to the cisplatin-fluorouracil group (n=120) or cisplatin-gemcitabine group (n=120). As of data cutoff (Dec 25, 2022), median follow-up was 40 months (IQR 32-48). 3-year progression-free survival was 83·9% (95% CI 75·9-89·4; 19 disease progressions and 11 deaths) in the cisplatin-gemcitabine group and 71·5% (62·5-78·7; 34 disease progressions and seven deaths) in the cisplatin-fluorouracil group (stratified hazard ratio 0·54 [95% CI 0·32-0·93]; log rank p=0·023). The most common grade 3 or worse adverse events that occurred during treatment were leukopenia (61 [52%] of 117 in the cisplatin-gemcitabine group vs 34 [29%] of 116 in the cisplatin-fluorouracil group; p=0·00039), neutropenia (37 [32%] vs 19 [16%]; p=0·010), and mucositis (27 [23%] vs 32 [28%]; p=0·43). The most common grade 3 or worse late adverse event (occurring from 3 months after completion of radiotherapy) was auditory or hearing loss (six [5%] vs ten [9%]). One (1%) patient in the cisplatin-gemcitabine group died due to treatment-related complications (septic shock caused by neutropenic infection). No patients in the cisplatin-fluorouracil group had treatment-related deaths. INTERPRETATION: Our findings suggest that concurrent adjuvant cisplatin-gemcitabine could be used as an adjuvant therapy in the treatment of patients with N2-3 nasopharyngeal carcinoma, although long-term follow-up is required to confirm the optimal therapeutic ratio. FUNDING: National Key Research and Development Program of China, National Natural Science Foundation of China, Guangdong Major Project of Basic and Applied Basic Research, Sci-Tech Project Foundation of Guangzhou City, Sun Yat-sen University Clinical Research 5010 Program, Innovative Research Team of High-level Local Universities in Shanghai, Natural Science Foundation of Guangdong Province for Distinguished Young Scholar, Natural Science Foundation of Guangdong Province, Postdoctoral Innovative Talent Support Program, Pearl River S&T Nova Program of Guangzhou, Planned Science and Technology Project of Guangdong Province, Key Youth Teacher Cultivating Program of Sun Yat-sen University, the Rural Science and Technology Commissioner Program of Guangdong Province, and Fundamental Research Funds for the Central Universities.