Interobserver variation in the interpretation of the leucocyte-esterase test by general practitioners

Abstract

Background: Urinary tract infections are frequent in general practice and the search for a rapid and reliable diagnostic aid continues. Test strips are attractive but need more evaluation in the general practice context. Objective: To determine the interobserver variation between general practitioners in the interpretation of the leucocyte-esterase test (LE test) to detect leucocytes in urine. Methods: 37 General practitioners (of whom 85% use these sticks more than once a week) were asked to interpret the LE test (Nephur-test® + leuco Boehringer Mannheim) in 11 urine samples. There were four types of samples: four plainly positive (>500 wbc/μl), one containing 75 wbc/μl, three between 10 and 20 wbc/μl and three negative samples. Evaluation with a spectrophotometric analyser was done twice (MEDITRON® Boehringer): before and after the visual interpretations. The main outcome measure was κ for interobserver agreement. Results: In one of the three negative samples, 26/37 (72%) judged the sample correctly as negative, in the two other negative samples 36/37 (97%) judged them correctly as negative. One in 37 (3%) and 25/37 (68%) of the two slightly positive samples (10-25 wbc) were wrongly interpreted as negative. The κ of interobserver agreement in all the observations was 0.68; in the clinically most relevant cut-off point between negative and slightly positive samples (10 wbc/μl), κ was 0.54, indicating a modest agreement. In 3 to 68% of the observations, the GP will consider a slightly positive specimen as non-infected. Three to 28% of the negative specimens will be considered as infected. In the four clearly positive specimens the interobserver agreement was between 98 and 100%. In the sample with 'nearly 75' leucocytes 24/37 participants (65%) judged that it was 'more than 500'. Conclusion: In a general practice setting the reliability of the LE test should be questioned because of an important interobserver variation in some slightly positive and some negative samples, which are frequently encountered in clinical practice.