Objective To compare mortality and rates of significant neurosensory impairment (sNSI) at 18–36 months’ corrected age in infants born extremely preterm across three international cohorts. Design Retrospective analysis of prospectively collected neonatal and follow-up data. Setting Three population-based observational cohort studies: the Australian and New Zealand Neonatal Network (ANZNN), the Canadian Neonatal and Follow-up Networks (CNN/CNFUN) and the French cohort Etude (Epidémiologique sur les Petits Ages Gestationnels: EPIPAGE-2). Patients Extremely preterm neonates of <28 weeks’ gestation in year 2011. Main outcome measures Primary outcome was composite of mortality or sNSI defined by cerebral palsy with no independent walking, disabling hearing loss and bilateral blindness. Results Overall, 3055 infants (ANZNN n=960, CNN/CNFUN n=1019, EPIPAGE-2 n=1076) were included in the study. Primary composite outcome rates were 21.3%, 20.6% and 28.4%; mortality rates were 18.7%, 17.4% and 26.3%; and rates of sNSI among survivors were 4.3%, 5.3% and 3.3% for ANZNN, CNN/CNFUN and EPIPAGE-2, respectively. Adjusted for gestational age and multiple births, EPIPAGE-2 had higher odds of composite outcome compared with ANZNN (OR 1.71, 95% CI 1.38 to 2.13) and CNN/CNFUN (OR 1.72, 95% CI 1.39 to 2.12). EPIPAGE-2 did have a trend of lower odds of sNDI but far short of compensating for the significant increase in mortality odds. These differences may be related to variations in perinatal approach and practices (and not to differences in infants’ baseline characteristics). Conclusions Composite outcome of mortality or sNSI for extremely preterm infants differed across high-income countries with similar baseline characteristics and access to healthcare.