Test performance of urinalysis and urine microscopy for the diagnosis of urinary tract infection in children: a meta-analysis.

GJ Williams¹ JC Craig^1,2 L Irwig²

1.Centre for Kidney Research, The Children's Hospital Wesrtmead, Sydney
2.Dept of Public Health and Community Medicine, University of Sydney

Urinary tract infection (UTI) is an important but problematic diagnosis in childhood. The reference standard for diagnosis is urine culture but this takes at least 24 hours to confirm or exclude infection. Because of the clinical importance of early diagnosis, tests such as microscopy for the detection of urinary white cells (pyuria) and reagent strips for leucocyte esterase and nitrite have been developed and used. Many primary studies have assessed the performance of these tests. However the final results, clinical settings, thresholds and definitions have varied substantially, making application difficult. The aim of this study was to obtain an overall estimate of test performances of urine screening tests in children using meta-analytical techniques. Methods included identifying eligible studies by searching MEDLINE (1966-2000) and EMBASE (1988-2000) using terms for urinary tract infection and bacteriuria. These were then combined with many terms for urine microscopy, urinalysis, and diagnostic reagent kits. The search was limited to children and adolescents. The overall test performance of each screening test was analysed using a summary receiver operating curve (sROC) using data points from each primary study. The search found 1412 abstracts, these were reviewed and 249 titles appeared appropriate or ambiguous and were obtained for clarification. Fifty seven papers were eligible and have undergone data extraction. Twenty seven papers assessed nitrite (n = 56583), 16 assessed leucocyte esterase (n = 8885) and 20 papers examined pyuria (n = 56513). The quality of the articles was in general poor. Taking articles that examined nitrite as an example, studies with verification bias over estimated the diagnostic odds ratio (DOR) by 10.9 (0.9-138.2, 95% CI) times compared with studies not having verification bias. Non-blinded studies over estimated the DOR by 83.1 (2.6-2634.8, 95% CI) compared to blinded studies. Thus the quality of a study had a significant impact on the outcome. The summary estimates for each screening test are presented below.

There is substantial information about the test performance of urinalysis and urine microscopy for diagnosing UTI in children. Studies are generally poorly done and this results in substantial over estimation of the true test performance. Nitrite, leucocyte esterase and pyuria are inaccurate indicators of urinary tract infection in children and should not replace urine culture for the diagnosis of UTI

Presented at the Annual Scientific Meeting of the Royal Australasian College of Physicians, Adelaide, May 2000
Presented at a meeting of the Australasian Epidemiological Association, Sydney, September 2001
Presented at a meeting of the Westmead Association, Sydney, 2001
Presented at the World Congress of Nephrology / American Society of Nephrology meeting, San Francisco, October 2001

Correspondence
Dr Gabrielle Williams
GabrieW4@chw.edu.au