Validity and interobserver agreement of reagent strips for measurement of glucosuria

Background. Measurement of glucosuria by means of a visually readable reagent test strip is frequently used in a wide variety of clinical settings. The aim of this study was to evaluate the validity and reliability of this semi-quantitative measurement of glucosuria compared to laboratory measurements of glucose concentrations in urine. Methods. Reagent test strips (Combur(3)Test (R) Roche) from 375 artificially supplemented samples of urine, covering a wide range of glucose concentrations, were independently read by three different observers. Scores of the strips were categorized as 0, 1+, 2+, 3+ or 4+, in ascending degree of glucosuria. Results of the test-strips were compared to the quantitative measurement of urinary glucose concentration in the laboratory. Results. 21.7% of reagent strips readings were discordant with the laboratory measurements ( p < 0.001). Under-or overestimating the degree of glucosuria occurs predominantly in category 1+ and 2+. In category '0' only 5.1% of the readings were incorrect. The interobserver-agreement was very good with 85% overall agreement and multirater Kappa 0.81. Interobserver-scores of the reagent strips never deviated more than one category from each other. Conclusion. The validity as well as the interobserver-agreement for the semi-quantitative measurement of glucosuria using reagent strips is moderate, but sufficient for excluding glucosuria. However it is too imprecise for an accurate quantitative measurement. It might only be valuable in settings where automated readings are not available or suitable.