114 venous blood samples (plasma glucose ranging between 2.6 and 30.7 mmol/l) were measured by a new pen-sized glucose meter designed for blood glucose self-monitoring working with an electro-chemical method. Glucose readings of three pen-meters were compared with plasma glucose measurements obtained from a standard glucose-oxidase-method. Precision, accuracy and clinical relevance were determined by assessment of the agreement between the two methods and error grid analysis. The mean differences between the pen-meters' blood glucose readings and plasma glucose were -1.35, -1.43 and -1.56 mmol/l, with limits of agreement (+/- 2 SD) of 2.2 and -4.9, 2.1 and -5.0 and 2.0 and -5.1 mmol/l, respectively. The 57 samples in the clinically relevant range, i.e., with plasma glucose concentrations below 13 mmol/l showed mean differences of -0.04, -0.10, and -0.04 mmol/l, with limits of agreement between -1.08 and 1.00 mmol/l, respectively. Error grid analysis showed that 90.7, 95.4, and 91.9% of the respective pen-meter readings fell in the zone A, i.e., gave clinically accurate results, the remaining values fell in zone B. One pen-meter broke during the study and had to be replaced. The results confirm that this new device gives accurate and reproducible measurements--faultless technical function provided--and, compares favorably with the well-established reagent strips for blood glucose self-monitoring.