UTI in children is an important clinical problem. Renal scarring, which occurs in a small proportion of children (approximately 6%1), is the most important outcome of infection as it is associated with significant future complications2, and ultimately with end stage renal disease3. Young children are considered particularly vulnerable to renal scarring and its consequences4. However, a recently completed 20-year follow-up study suggested that compensatory mechanisms mean no significant changes in overall GFR occur in patients with unilateral scaring1, and the risk of hypertension is low in all patients (regardless of the degree of scarring)5.

Current UK recommendations state that all children under 5 years of age should be investigated after their first confirmed UTI67, although the benefit from this strategy has been questioned8.

Further investigation of children with confirmed UTI has a number of different clinical aims: the localisation of infection, the prediction and detection of renal scarring and the detection of VUR. The current reference standards for these investigations are Tc-99 m-DMSA renal scintigraphy (DMSA scan) for the localisation of infection and for the detection and prediction of renal scarring, and micturating cystourethrography (MCUG) for the detection of VUR. These investigations have the disadvantages of being invasive and involving exposure to ionising radiation. It is desirable to minimise the number of invasive examinations and radiation load to which children are exposed. Alternative tests that offer a potential advantage over the current reference standards, such as ultrasound or laboratory-based tests, are therefore required. An additional aim of the investigation of children with UTI is the detection of anatomical abnormalities that may be amenable to surgery, and a role has been suggested for ultrasound in this context. We did not identify any studies evaluating tests with this objective; therefore it could not be assessed in this review. However, a recently published observational study has suggested that routine ultrasound post-UTI, in children under five years, does not change management9. The role of pre-natal ultrasound is unclear910 and was outside the scope of this review.

We reviewed the diagnostic accuracy of tests evaluated for the further investigation of UTI together with evidence of their long-term effectiveness, with a view to determining the optimum diagnostic pathway. A previous systematic review has evaluated ultrasound for the detection of scarring11. This review was published in 1999 with searches undertaken in 1997 and only included 10 studies. We are unaware of any other systematic reviews in this area. This review therefore represents the most complete review of the area.

We assembled a database of published and unpublished literature from systematic searches of 16 electronic databases (inception to between October 2002 and February 2003), hand searching of 12 journals, consultation with experts in the field, and screening bibliographies of included studies. Update searches were conducted in May 2004. There were no language restrictions. Full details of the search strategy will be reported elsewhere12.

We included controlled trials that compared different imaging strategies and reported patient based outcomes. We also included diagnostic cohort studies that included at least 20 children, some of whom were aged five years or under, and that reported sufficient information to construct 2 × 2 tables. Table 1 presents a summary of the tests evaluated in this review and details the potential advantages that these tests have over the current reference standards. Although other tests have been evaluated, this article will focus on those offering potential advantages over the current reference standard. Studies had to compare one of the index tests listed in this table to the listed reference standard, for each of the different aims.

Two reviewers independently screened titles and abstracts for relevance, any disagreements were resolved by consensus. One reviewer performed inclusion assessment, data extraction and quality assessment; a second reviewer checked this. We extracted 2 × 2 data from each of the studies and used this to calculate estimates of test performance. Where insufficient details were reported, we contacted authors to request further information. We assessed the methodological quality of diagnostic accuracy studies using QUADAS13. Individual QUADAS items were used to investigate heterogeneity and to present a detailed assessment of quality to the reader.

We analysed results grouped by clinical aim. Where studies presented more than one estimate of test performance for the same test, we only included one estimate in the pooled analysis. We aimed to select the most appropriate data set or the one most similar to that used by other studies in terms of population, technique or unit of analysis. For example, data for tests used to localise infection were analysed by patient in preference to kidney or renal unit (kidney and ureter), whereas data for tests used to detect VUR were analysed by renal unit. For each test, or test combination, we calculated the range in sensitivity, specificity, positive (LR+) and negative (LR-) likelihood ratios, and diagnostic odds ratios (DOR). Where sufficient studies were available, pooled likelihood ratios were calculated for each test14. Heterogeneity of likelihood ratios was investigated using the Q statistic15 and through visual examination of forest plots of study results16. We presented individual studies results graphically by plotting estimates of sensitivity and specificity in receiver operating characteristic (ROC) space. Where sufficient data were available, we used regression analysis to investigate heterogeneity. We extended the summary ROC (sROC) model17, estimated by regressing D (log DOR) against S (logit true positive rate – logit false positive rate), weighted according to sample size, to include the variables for patient age (<2 years, <5 years, <12 years and <18 years), geographic region, and QUADAS items18. For ultrasound for the detection of VUR a variable for ultrasound technique (contrast-enhanced or standard) was also included.

We identified more than 10,000 studies. Of these, 73 studies met our inclusion criteria: 72 diagnostic accuracy studies and one RCT of the clinical effectiveness of investigation. Figure 1 shows the flow of studies through the review process. The results of individual included studies are presented [see Additional file 1]. Table 2 presents summary results for each included test.

When considering further testing following confirmation of UTI, it is important to bear in mind clinical aim. If the information derived cannot be used to prevent renal disease there is little benefit in testing. Tests should only be conducted if (a) the results will lead to a change in management and (b) this change is likely to lead to an improved outcome. For this reason, the ideal study would be a randomised controlled trial of different testing strategies, or no testing. We identified only one such study, and this was only available as an abstract reporting limited information, and we were unsuccessful in obtaining further details19. Our results are therefore primarily derived from diagnostic accuracy studies, which assume the validity of the clinical aims of current testing.

Localisation of infection can be considered a first step in the investigation of UTI. Lower UTI does not involve the kidneys and so cannot lead to renal scarring. Children with lower UTI are therefore unlikely to benefit from immediate investigation. Given that therapeutic delay is thought to be associated with renal damage92, the possibility that they may benefit from monitoring for recurrence remains open to question. The ideal test to localise infection would be non-invasive, inexpensive, and quick to perform. Further investigation of children with infections of the lower urinary tract could thus be avoided. Our results do not support the use of any of the minimally invasive tests evaluated as alternatives to renal scintigraphy for the localisation of infection. However, the available evidence was limited, and further primary research in this area would be useful. Testing with the specific aim of localisation of infection is not common in current practice. Baseline renal scintigraphy in all children with confirmed UTI, in whom further investigation is planned, may be beneficial. This approach would eliminate a substantial proportion of children from further invasive investigations.

The detection of VUR has historically been considered an important element in the investigation of UTI, as it has been thought to indicate an increased risk of scarring. This idea is currently the subject of considerable debate. The only study of the effectiveness of imaging identified by our review compared routine and selective imaging, using US and MCUG for the detection of VUR19. This study found increased rates of VUR detection and prophylaxis with routine imaging, but no reduction in scarring or recurrent UTIs. Other studies have shown that the presence of VUR, as determined by MCUG, correlates poorly with the presence of renal scarring78939495. A recent systematic review also found that VUR is a weak predictor for renal damage in children hospitalised with UTI96. The management of VUR and how this impacts on the risk of future renal disease is also the subject of debate. A clinical trial comparing surgical and medical management found no difference in outcome97, and a systematic review evaluating antimicrobial prophylaxis for the prevention of UTI in children found a lack of data for children with VUR98. Given the considerable doubts surrounding both the link between VUR and renal scarring, and the benefits to be derived from treating VUR, it appears difficult to justify the routine use of MCUG. This is an invasive, and costly test, involving considerable exposure to ionising radiation; its use should therefore be minimised where possible. Should the evaluation of VUR be considered clinically necessary, the available evidence supports the use of contrast-enhanced ultrasound techniques as an accurate alternative. Although not currently in widespread use in the UK, these techniques have the advantage of not involving exposure to ionising radiation. In addition, standard ultrasound forms part of the examination, allowing simultaneous screening for anatomical abnormalities and some types of calculi.

A test predicting risk of renal scarring would be useful were a treatment available to prevent its development. Anti-microbial therapy is usually initiated in children with confirmed UTI prior to investigation, as there is some evidence that treatment delay may affect scarring37. Predicting risk of renal scarring as a result of a current infection would, therefore, appear to be of academic interest alone. We identified no acute tests that were able to accurately predict the development of renal scarring. The prediction of risk of future infection is of potential interest in guiding the initiation of prophylactic antimicrobial therapy, but is outside the scope of this paper.

Although the presence of renal scarring represents the initial stages of renal disease, there is little that can be done to treat patients with scarring in order to prevent complications. If progressive scarring is assumed to be the consequence of repeat infections then anti-microbial prophylaxis may be initiated, although the effectiveness of this strategy remains open to debate. Imaging for the detection of renal scarring may be seen as a means of monitoring disease progression. If repeat examination is required then a less invasive alternative to the reference standard (renal scintigraphy), and one which avoids the use of ionising radiation, would seem particularly desirable. We found standard ultrasound examination to be a potentially useful test for ruling in scarring, but poor for ruling it out. This fits with anecdotal opinion that ultrasound is good at identifying gross scarring, but poor at detecting minor lesions. It may be that ultrasound images are insufficiently subtle to enable their use in monitoring disease progression. Further research on the accuracy of ultrasound in grading scarring is therefore required. Indirect radionuclide cystography is sometimes advocated as an alternative test for renal scarring, on the grounds that it may combine detection of VUR and scarring in a single test. We found it to be an accurate test for scarring, but poor for ruling out VUR.

There is no evidence to support the clinical effectiveness of routine investigation of children with confirmed UTI. There is limited evidence that routine imaging to detect VUR, following first UTI in older children, has no effect on recurrence or renal scarring.

High quality primary research on the effectiveness, in terms of improved patient outcome, of testing at all stages in the investigation of confirmed UTI is urgently required.

The author(s) declare that they have no competing interests.

All authors contributed towards the conception and design of the study and the interpretation of the data. They also read and approved the final manuscript. MEW and PFW participated in data extraction, the analysis of data, and drafted the article.