A total of 73 studies were included in the review: 1 randomised controlled trials (RCT; n=172) and 72 diagnostic cohort studies (number of participants unclear).
Clinical effectiveness.
Details of one RCT comparing routine versus selected imaging (using ultrasound and MCUG) were available in abstract form. Routine investigation led to higher rates of imaging (100% versus 21%), identification of reflux and antibiotic prophylaxis, but did not influence the proportion of children with recurrent UTI or the rate of renal scarring after 2 years' follow-up.
Diagnostic accuracy.
In terms of quality, around half the diagnostic cohort studies reported an appropriate patient spectrum or selection criteria. Similarly, around half did not adequately address incorporation bias, disease progression bias or verification bias.
Localisation of infection.
Ultrasound (20 studies) gave poor performance both for ruling in (pooled LR+ 3.5, 95% confidence interval, CI: 2.5, 4.8) and ruling out (pooled LR- 0.57, 95% CI: 0.47, 0.68) renal involvement. The clinical and laboratory tests investigated gave varied results and showed poor performance in general: for clinical features (5 studies), the LR+ ranged from 1.1 to 26.6 and the LR- from 0.09 to 0.89; for infection markers (10 studies), the LR+ ranged from 1.0 to 8.8 and the LR- from 0.09 to 1.00; for renal function markers (4 studies), the LR+ ranged from 0.7 to 36.7 and the LR- from 0.02 to 1.51; for immunofluorescence detection of bacteria (1 study), the LR+ was 1.8 and the LR- was 0.55.
Detection of reflux.
Standard ultrasound (12 studies) performed poorly (pooled LR+ 1.9, 95% CI: 1.2, 2.9; pooled LR- 0.76, 95% CI: 0.63, 0.93). Contrast-enhanced ultrasound (16 studies) showed much better performance (pooled LR+ 14.1, 95% CI: 9.5, 20.8; pooled LR- 0.20, 95% CI: 0.13, 0.29). Indirect radionuclide cystography (2 studies) showed good LR+ (11.2 and 25.0), but poor LR- (0.41 and 0.68).
Prediction of renal scarring.
The diagnostic accuracies of tests evaluated for this clinical aim (clinical, laboratory-based and imaging techniques) were generally poor. The LR+ ranged from 1.1 to 3.1 for four of the studies, with 12.9 for a single evaluation of IVP. The LR- ranged from 0.44 to 0.88.
Detection of renal scarring.
For ultrasound (7 studies), LR+ ranged from 1.3 to 35.9 and LR- from 0.14 to 0.99. For IVP (4 studies), the LR+ ranged from 10 to 171.3 and the LR- from 0.15 to 0.80. For indirect radionuclide cystography (2 studies), the LR+ ranged from 2.1 to 12.6 and the LR- from 0.15 to 0.75.