OBJECTIVE We investigated the accuracy of fetal umbilical artery Doppler to predict the risk of compromise of fetal/neonatal wellbeing in a high-risk population. METHODS Searches in MEDLINE, Embase, The Cochrane Library and Medion (from inception to March 2009) were carried out, together with hand searching of relevant journals, reference list checking of included articles and contact with experts. Criteria for selection were observational studies with umbilical artery Doppler used in a high-risk pregnant population with an outcome measure for compromise of fetal/neonatal wellbeing. Data on study design, quality and results were extracted to construct 2 × 2 tables. Bivariate meta-analysis was performed. Likelihood ratios (LRs) were used as the summary measure of accuracy. RESULTS One-hundred and four studies met the selection criteria (19 191 fetuses). In a high-risk population, umbilical artery Doppler predicted small-for-gestational age with a pooled LR+ of 3.76 (2.96, 4.76) and pooled LR- of 0.52 (0.45, 0.61), and compromise of fetal/neonatal wellbeing with a pooled LR+ of 3.41 (2.68, 4.34) and pooled LR- of 0.55 (0.48, 0.62). In this group it was also possible to predict, with accuracy, intrauterine death (pooled LR+ = 4.37 (0.88, 21.8); pooled LR- = 0.25 (0.07, 0.91)) and acidosis (pooled LR+ = 2.75 (1.48, 5.11); pooled LR- = 0.58 (0.36, 0.94)). CONCLUSIONS In a high-risk population, fetal umbilical artery Doppler is a moderately useful test with which to predict mortality and risk of compromise.