Forty-eight studies (92,766 samples) were included: 31 used consecutive patients and 17 used nonconsecutive patients.
The validity scores ranged from 6 to 16 (maximum score not stated). The index test and the standard were independently measured in all patients, and the clinical spectrum of patients was wide in almost all of the studies.
The primary studies revealed heterogeneity in terms of, among others, the age groups, sampling technique, whether the urine samples were centrifuged and Gram stained, definition of standard according to sampling technique used, positivity threshold values and clinical features. Most of the studies included uncentrifuged, unstained urine samples.
Pyuria and bacteriuria were reported, respectively, as the number of leukocytes and number of bacteria per high-power field (hpf). The bivariate summary ROC curves showed that pyuria of at least 10/hpf and bacteriuria of any/hpf (P10 and B), and bacteriuria of at least 10/hpf (B10) had the best diagnostic performances. The other tests were of intermediate or low performance.
After the final model of multivariate analysis was run, P10 and B remained the best combination, being better than when urine was collected through suprapubic aspiration or catheter, irrespective of age group and whether the urine was centrifuged or not. P10 was consistently the test with the lowest performance for all age groups.
The validity assessment score was tested in a separate stepwise regression (appears to be post-hoc) along with all the other individual variables. It was not found to have a significant influence on the diagnostic performance of the index tests.
The results of the dipstick tests were not reported.