Objective Survival rates of extremely premature infants are rising, but changes in neurodevelopmental impairment (NDI) rates are unclear. Our objective was to perform a systematic review of intrainstitutional variability of NDI over time.

Design Systematic review and meta-analysis.

Data sources Ovid MEDLINE, Embase, PubMed, Cochrane Library and Google Scholar.

Study selection Study eligibility: (1) at least two discrete cohorts of infants born <27 weeks’ gestation or <1000 g birth weight, (2) one cohort born after 1990 and at least one subsequent cohort of similar gestational age, (3) all cohorts cared for within the same Neonatal Intensive Care Unit(s) (NICU) and (4) neurodevelopmental outcomes at 18–36 months corrected age.

Main outcome Change in NDI rates. Quality, validity and bias were assessed using Grading of Recommendations, Assessment, Development, and Evaluation and Quality in Prognosis Studies guidelines.

Results Of 203 publications, 15 were eligible, including 13 229 infants. At the first time point, average NDI rate across study groups weighted by sample size was 41.0% (95% CI 34.0% to 48.0%). The average change in NDI between time points was −3.3% (95% CI −8·8% to 2.2%). For each added week of gestation at birth, the rate of NDI declined by 9.7% (95% CI 6.2% to 13.3%). Most studies exhibited moderate–severe bias in at least one domain, especially attrition rates.

Conclusions When comparing discrete same-centre cohorts over time, there was no significant change in NDI rates in infants born <27 weeks’ gestation or <1000 g. Higher survival rates unaccompanied by improvement in neurodevelopment highlight urgency for renewed focus on the causes of NDI and evidence-based strategies to reduce brain injury.