BACKGROUND Low blood glucose in newborns is difficult to detect clinically. Hence a reliable "point of care" device (glucometer) for early detection and treatment of low glucose is needed. OBJECTIVE To evaluate the performance of five readily available glucometers for the detection of low blood glucose in newborn infants. METHODS Glucostix measurements were taken for newborns with risk factors using a Reflolux S (Boehringer) glucometer. If the initial reading was low (< 2.6 mmol/l), further measurements were taken with two other glucometers (phase I, Advantage and Glucotrend (Roche); phase II, Elite XL (Bayer) and Precision (Abbott)), and plasma glucose was measured in the laboratory (Aeroset; Abbott). RESULTS Over 10 months, 101 specimens were collected from 71 newborns (57 in phase I; 44 in phase II). The Advantage glucometer usually overestimated blood glucose with a mean difference of 1.07 mmol/l (p < 0.01) at all low glucose ranges. The Glucotrend, Precision, and Elite XL glucometers performed better; the mean differences were not significantly different from the laboratory measured value (0.17 mmol/l (p = 0.37); -0.12 mmol/l (p = 0.13), and 0.24 mmol/l (p = 0.13) respectively). For detection of glucose concentrations < 2.6 mmol/l, the Precision glucometer had the highest sensitivity (96.4%) and negative predictive value (90%). For lower glucose concentrations (< 2.0 mmol/l), the Glucotrend glucometer performed even better (sensitivity 92.3%, negative predictive value 96.3%). CONCLUSIONS Point of care devices should have good precision in the low glucose concentration range, sensitivity, and accuracy for early detection of neonatal hypoglycaemia. None of the five glucometers was satisfactory as the sole measuring device. The Glucotrend and Precision glucometers have the greatest sensitivity and negative predictive value. However, confirmation with laboratory measurements of plasma glucose and clinical assessment are still of the utmost importance.