The performance and practicability of 2 blood glucose meters (Glucocard Memory 2 and Accutrend sensor) were evaluated. Both glucose meters produced acceptably precise results in the hyper- and normoglycaemic concentration ranges. In the hypoglycaemic concentration range, the imprecision of Accutrend sensor was much higher than recommended by the American Diabetes Association. Within-run coefficients of variation for Glucocard Memory 2 were 6.3%, 3.9% and 2.4% at glucose concentrations of 1.7 mmol/l, 5.8 mmol/l and 11.7 mmol/l, respectively: for Accutrend sensor these were 15.2%, 5.0% and 1.2% at respective concentrations of 0.9 mmol/l, 4.2 mmol/l and 19.6 mmol/l. Between-day coefficients of variation for Glucocard Memory 2 were 4.8% and 3.5% at glucose concentrations of 3.9 mmol/l and 17.2 mmol/l, respectively and for Accutrend sensor they were 3.8% and 2.9% at glucose concentrations of 3.8 mmol/l and 18.7 mmol/l, respectively. Results were linear over a range of 1.6 mmol/l -29.7 mmol/l for Glucocard Memory 2 and 1.6 mmol/l -33.3 mmol/l for Accutrend sensor. Results of both blood glucose meters correlated closely with the hexokinase/glucose-6-phosphate dehydrogenase laboratory method. Ninety-eight percent of both Glucocard Memory 2 and Accutrend sensor results were within 20% of the comparison method values. Ninety-three percent of the Glucocard Memory 2 and 96% of the Accutrend sensor results were within 15% of the comparison method results. An inverse relation between the glucose readings and haematocrit values was observed for both blood glucose meters in the hyperglycaemic range and this effect was more pronounced for Accutrend sensor. In the normo- and hypoglycaemic ranges the effect was insignificant and absent, respectively. Minimum sample volume for Glucocard Memory 2 was 3 microliters and for Accutrend sensor it was 9 microliters. Lower sample volumes gave erroneous results. Presenting more than the required volume had no effect on results.