Back to 2014 Fall Congress Meeting Abstracts

BACKGROUND:
Renal and bladder ultrasound (RBUS) is often used as an initial screening test for children after urinary tract infection (UTI), but prior research has suggested that RBUS is neither sensitive no specific for VCUG findings such as vesicoureteral reflux (VUR). We sought to evaluate the association of specific renal, ureteral, and bladder findings on RBUS with VCUG findings, and to determine whether predictive models can be constructed that accurately identify patients at high risk for VCUG abnormalities.
Methods: We identified 3995 clinical encounters from 1/1/06-12/31/10 during which VCUG and renal and bladder ultrasound (RBUS) were performed on the same day. RBUS and VCUG reports were reviewed and findings classified. Analysis was limited to patients aged 0-60 months with no prior postnatal genitourinary imaging and no history of prenatal hydronephrosis. We investigated associations between a large number of specific RBUS findings with abnormalities seen on VCUG. Both multivariate logistic models and a neural network machine learning algorithm were built to evaluate predictive power of RBUS for VCUG abnormalities. Sensitivity, specificity, predictive values and area under receiving operating curves (AUROC) of RBUS for VCUG abnormalities were determined.
Results:
We identified 2259 patients age 0-60 months with UTI as the indication for imaging. RBUS was reported as “normal” in 75.0%. On VCUG, any VUR was identified in 41.7%, VUR grade >II in 20.9%, and VUR grade >III in 2.8%. Many individual RBUS findings were significantly associated with VUR on VCUG, including hydronephrosis [any grade VUR (OR=1.36, p=0.0086), VUR grade > II (OR=2.32, p<0.0001), and VUR grade > III (OR=4.67, p<0.0001)] and ureteral dilation [any grade of VUR (OR=1.43, p=0.0967), VUR grade > II (OR=2.97, p<0.0001), and VUR grade > III (OR=7.41, p<0.0001).] Certain parenchymal abnormalities (dysplasia, duplication, atrophy, urothelial thickening) were associated with VUR [any grade of VUR (OR=1.61, p=0.0005) , VUR grade > II (OR=2.10, p<0.0001), and VUR grade > III (OR=2.75, p<0.0001), although renal cysts, agenesis, and stones were not significant predictors. Bladder abnormalities (trabeculation, wall thickening, debris, ureterocele) were also strongly associated with any grade of VUR (OR=1.99, p=0.0089) , VUR grade > II (OR=1.98, p=0.0110), and VUR grade > III (OR=5.43, p<0.0001). Despite these strong univariate associations, multivariate modeling did not result in a highly accurate predictive model (Table 1). Multivariate logistic regression built via stepwise selection had AUROC=0.57, sensitivity=86% and specificity=25% for any VUR; AUROC=0.60, sensitivity=5% and specificity=99% for VUR grade>II; and AUROC=0.67, sensitivity=6% and specificity=99% for VUR grade > III. The best predictive model constructed via neural networks had AUROC=0.69, sensitivity=64% and specificity=60% for any VUR; AUROC=0.67, sensitivity=18% and specificity=98% for VUR grade>II; and AUROC=0.79, sensitivity=32% and specificity=100% for VUR grade > III.
Conclusions: Even with the state-of-the-art predictive modelling, abnormal findings on RBUS provide a poor screening test for VCUG findings such as VUR. RBUS and VCUG should be considered complementary as they provide important, but different, information.

Test characterisitcs of Optimized Models for predicting VUR (any, >II, or >III)

Predictive Model	Sensitivity	Specificity	PPV	NPV	AUROC
Logistic Regression
Any VUR	86.3%	24.7%	53.7%	64.1%	0.57
VUR grade > II	5.1%	99.1%	70.6%	71.9%	0.60
VUR grade > III	6.0%	99.9%	77.8%	93.2%	0.67
Neural Networks
Any VUR	64.2%	59.6%	61.6%	62.2%	0.69
VUR grade > II	17.8%	97.8	76.4%	74.5%	0.67
VUR grade > III	31.6%	99.8%	92.5%	95%	0.78