Background: The erosion of the early mortality advantage of elective endovascular aneurysm repair (EVAR) compared with open repair of abdominal aortic aneurysm remains without a satisfactory explanation. Methods: An individual-patient data meta-analysis of four multicentre randomized trials of EVAR versus open repair was conducted to a prespecified analysis plan, reporting on mortality, aneurysm-related mortality and reintervention. Results: The analysis included 2783 patients, with 14 245 person-years of follow-up (median 5·5 years). Early (0–6 months after randomization) mortality was lower in the EVAR groups (46 of 1393 versus 73 of 1390 deaths; pooled hazard ratio 0·61, 95 per cent c.i. 0·42 to 0·89; P = 0·010), primarily because 30-day operative mortality was lower in the EVAR groups (16 deaths versus 40 for open repair; pooled odds ratio 0·40, 95 per cent c.i. 0·22 to 0·74). Later (within 3 years) the survival curves converged, remaining converged to 8 years. Beyond 3 years, aneurysm-related mortality was significantly higher in the EVAR groups (19 deaths versus 3 for open repair; pooled hazard ratio 5·16, 1·49 to 17·89; P = 0·010). Patients with moderate renal dysfunction or previous coronary artery disease had no early survival advantage under EVAR. Those with peripheral artery disease had lower mortality under open repair (39 deaths versus 62 for EVAR; P = 0·022) in the period from 6 months to 4 years after randomization. Conclusion: The early survival advantage in the EVAR group, and its subsequent erosion, were confirmed. Over 5 years, patients of marginal fitness had no early survival advantage from EVAR compared with open repair. Aneurysm-related mortality and patients with low ankle : brachial pressure index contributed to the erosion of the early survival advantage for the EVAR group. Trial registration numbers: EVAR-1, ISRCTN55703451; DREAM (Dutch Randomized Endovascular Aneurysm Management), NCT00421330; ACE (Anévrysme de l'aorte abdominale, Chirurgie versus Endoprothèse), NCT00224718; OVER (Open Versus Endovascular Repair Trial for Abdominal Aortic Aneurysms), NCT00094575.