Twenty-one patients' blood samples, each with modified haematocrit concentrations of about 20%, 30%, 40%, 50% and 60%, were assayed for the presence of glucose by two reagent-strip methods--the Ames Glucometer with BG reagent strips and the Reflolux reflectance meter with BM-test strips. The reference method was the YSI blood glucose analyser. Both Ames and BM systems were found to have good precision. Ames results were falsely high (mean, 15%) at a haematocrit value of 20% and falsely low (mean, 21%) at a haematocrit value of 60%. BM results also overestimated (mean, 4%) and underestimated (mean, 5%) at the corresponding haematocrit levels. The results showed good agreement between Ames and BM methods and the YSI method for normal haematocrit concentrations of 40% to 50%. Clinicians should be aware of inaccuracies in glucose results by test-strip methods in patients with abnormal haematocrit concentrations.