Discussion
In infants ≤3 months old who presented with their first febrile UTI, elevated CRP and non-E. coli UTI were independently associated with imaging (US KUB, DMSA scan and/or MCU) abnormalities, while bacteraemia and non-E. coli UTI were independently associated with recurrent UTI.
In our study, non-E. coli UTI was an independent risk factor associated with imaging abnormalities and also with recurrent UTI. This finding in our Asian cohort is similar to other international studies. Data from the Finnish nationwide surveillance system among paediatric hospitals showed that obstruction in the urinary tract or VUR occurred more frequently among children with non-E. coli bacteraemic UTI than those with E. coli bacteraemic UTI (89% vs 46%, p<0.01).18 In a review of the medical records of 139 children, 14/18 (77%) with urinary tract anomalies had non-E. coli UTI.19 Larger studies involving 60720 and 414 children21 further confirmed that non-E. coli UTI is associated with higher grades of VUR.
CRP was independently associated with imaging abnormalities in this study, however with only an adjusted OR of 1.01 (95% CI 1.00 to 1.03, p=0.016). With regard to using CRP to diagnose pyelonephritis, a Cochrane meta-analysis showed that the sensitivity and specificity estimates for a CRP level of 20 mg/L were 0.93 (95% CI 0.86 to 0.96) and 0.37 (95% CI 0.24 to 0.53), respectively.22 This suggests that if a child has a CRP <20 mg/L, the probability of this child having pyelonephritis is low. In relation to predicting VUR, a combination of US KUB abnormalities and CRP level of ≥80 mg/L had a sensitivity and specificity of 47.8% and 87.8%, respectively.23 In addition, specific CRP cut-off values were studied: 23.5 mg/L for predicting presence of any VUR (sensitivity 61.9%, specificity 62.8%) and 50 mg/L for predicting high-grade VUR (sensitivity 91%, specificity 71%).24 The cut-off of 23.5 mg/L is similar to our study of 20 mg/L (online supplemental table 1) for predicting MCU abnormalities (sensitivity 92%, specificity 29%) (online supplemental table 2). In another study recruiting 48 children after their first pyelonephritis and diagnosed with unilateral VUR, patients who developed kidney scar on DMSA scan had a higher CRP level on admission (12.9 mg/L±10.9 vs 6.9 mg/L±7.4, p<0.05) compared with those who did not.25 In our study, however, the cut-off of 68.7 mg/L chosen in this study based on the Youden Index did not perform well to predict the primary outcome (sensitivity 30%, specificity 83%). This might be because the primary outcome was a combination of abnormalities of the US KUB, DMSA scan and/or MCU and that different cut-offs for each individual imaging modality might be more relevant. However, the number of patients was too small to perform subgroup analysis based on each imaging result.
Bacteraemia was independently associated with recurrent UTI in the logistic regression model. While VUR has been shown to occur more frequently in children with bacteraemic UTI compared with those who did not have bacteraemia,26 there seems to be no correlation between bacteraemia and kidney scarring.27 In a retrospective cohort study of 257 infants of less than <3 months of age with bacteraemic UTI, 98.8% underwent US KUB and 87.2% underwent MCU, of which 55.1% had US KUB abnormalities and 33.5% had MCU abnormalities.28 In our study, of the 12 infants who had bacteraemia, three (25%) had both abnormal US KUB and MCU and one out of these three had scarring on DMSA scan. This lower percentage of imaging abnormalities might be due to patients defaulting the planned follow-up imaging (12 infants did not undergo any imaging) and MCU was not routinely performed (indications for MCU described in the Materials and methods section).
Among infants with antenatal US abnormalities, about 1%–6% develop UTI postnatally.29–31 Even though a higher proportion of patients with the primary outcome had antenatal US abnormalities in our study, the presence of antenatal US abnormalities was not included in the logistic regression model because of the low numbers—2.1% (4/190) had documented antenatal US abnormalities. Similarly, a prospective observational study of 250 children with first UTI also reported that 1.4% of their patients had antenatal US abnormalities.32
There are several limitations to our study. First, due to small number of patients, the primary outcome was a composite one, combining abnormalities of US KUB, DMSA scan and MCU. Therefore, clinical risk factors associated with abnormalities of the individual scans were not demonstrated in our current study. In addition, as this was a retrospective cohort study, there were significant missing data because there were patients who defaulted routine follow-up. Lastly, the test characteristics of the variables were not externally validated with a separate group of patients.
In conclusion, this is a pilot study in our institution to identify clinical and laboratory risk factors associated with imaging abnormalities after the first febrile UTI in infants ≤3 months old. Non-E. coli UTI was independently associated with imaging abnormalities, while bacteraemia and non-E. coli UTI were independently associated with recurrent UTI. Future studies are required to build a robust prediction model to aid physicians in risk stratifying infants after the first episode of pyelonephritis and therefore avoiding unnecessary investigations and follow-up.