Assessment of laboratory methods for detection of unsuspected diabetes in primary health care.

In order to assess different methods for early detection of unsuspected diabetes, urine and venous blood samples were collected at random from 1082 patients visiting a primary health care centre in southern Sweden. Blood glucose was analysed by the hexokinase method along with the Dextrostix-Eyetone reflectance meter. Urine glucose was determined by Clinistix, Diastix, Neostix, Rediatest, Clinitest and quantitatively by the hexokinase method. Patients fulfilling the criteria of a positive screen were subjected to a diagnostic investigation with an oral glucose tolerance test. Out of 89 positive screenees, 37 patients were classified as diabetics, showing a prevalence of diabetes in the study population of 3.4% according to the WHO criteria. Impaired glucose tolerance was found in 14 patients. In a control group of 56 patients, randomly selected among negative screenees, no cases of diabetes were found. Random blood glucose measurement by the hexokinase method, using 7 mmol/l as a screening level, had a significantly higher sensitivity (95%) than all urine glucose methods (59-30%) with comparable specificity (97-99%). Use of the Dextrostix-Eyetone reflectance meter resulted in a decrease in sensitivity to 75% without any change in specificity or predictability, compared with the hexokinase method. Urine testing for glucose was found to be a suboptimal method for early case finding of diabetes among patients receiving primary health care.