Urinary tract infections in children: Testing a novel, noninvasive, point-of-care diagnostic marker

Abstract

Objectives:

Urinary neutrophil gelatinase-associated lipocalin (uNGAL) appears highly accurate to identify urinary tract infections (UTIs) when obtained via catheterization. Our primary aim was to determine the agreement in uNGAL levels between paired catheter and bag urine specimens. Our secondary aim was to compare the diagnostic test characteristics of quantitative uNGAL, dipstick uNGAL (a potential point-of-care test), and urinalysis (UA).

Methods:

This was a prospective study of febrile children < 24 months evaluated for UTIs. We evaluated quantitative uNGAL at a previously identified threshold of 39.1 ng/mL, dipstick uNGAL at its built-in threshold of >50 ng/mL, and UA at standard thresholds for leukocyte esterase (LE). A positive urine culture was defined as >100,000 CFUs/mL of a pathogen.

Results:

A total of 211 patients were included (10% with positive urine cultures); 116 had paired catheterized and bagged samples. The agreement between catheterized and bagged samples at a quantitative uNGAL cutoff of ≥39.1 ng/mL was 0.76 (95% confidence interval [CI] = 0.67 to 0.83) and 0.77 (95% CI = 0.68 to 0.84) at a uNGAL dipstick threshold of >50 ng/mL. The area under the receiver operating characteristic curve for uNGAL from a catheterized sample was 0.96 (95% CI = 0.89 to 1.00) compared to 0.93 (95% CI = 0.87 to −0.99) from a bagged sample. The sensitivities of catheterized sample quantitative and dipstick uNGAL (90.5%) were higher than UA at a LE threshold of ≥1+ (57.1%). Bagged-sample uNGAL had lower quantitative and dipstick specificities (both 73.8%) than from catheterized samples (94.3% and 95.3% respectively), similar to UA.

Conclusions:

uNGAL from bagged and catheterized samples showed insufficient agreement to be used interchangeably. The low specificity of uNGAL from bagged samples suggests that sampling technique affects uNGAL levels.

INTRODUCTION

Urinary tract infections (UTIs) are common in children, with more than 400,000 diagnosed annually.¹ The timely diagnosis of UTIs is important, because delays in the initiation of antimicrobial therapy are associated with morbidity, including renal scarring.^2–4

Clinicians rely on the urinalysis (UA) to make a preliminary diagnosis of UTIs. However, the UA has variable and suboptimal sensitivities (78%–88%) and specificities (72%–97%) across studies.^5–12 Moreover, studies have consistently shown that bagged specimens result in falsely positive UAs.^13,14 Therefore, the standard procedure to accurately screen for UTIs in young children is often considered to be bladder catheterization to avoid contamination. Catheterization is painful and leads to adverse events in up to one-fifth of young children, including painful urination, genital pain, gross hematuria, and secondary UTIs.^15,16 There remains a clinical need for a reliable and accurate urine test to make a preliminary diagnosis of UTI that does not require catheterization.

Studies in animals and humans suggest that urine neutrophil gelatinase-associated lipocalin (uNGAL) may be a highly accurate marker for UTIs.^17–25 uNGAL’s potentially improved accuracy compared to other UTI markers may result from its predominant expression in the alpha-intercalated cells in the kidney.¹⁷ uNGAL is significantly elevated with Gram-negative UTIs.^17–23 Investigators have shown that uNGAL has excellent sensitivity and specificity to identify UTIs in children when obtained by catheterization and may be a more accurate screen for UTIs than the standard UA.²⁴ Importantly, small studies in infants and children also suggest that uNGAL levels may be assessed via nonsterile methods such as bagged, which would be a substantial clinical advantage.^{20–22,25–27} However, no studies of children with possible UTIs have directly compared uNGAL levels in bagged and catheterized specimens. Furthermore, research is lacking regarding the accuracy of semiquantitative dipstick uNGAL testing, which could be easily implemented into practice at the point-of-care (POC). Therefore, our primary aim was to determine the agreement in quantitative and dipstick uNGAL levels between paired catheter and bag urine specimens. Our secondary aim was to compare the diagnostic test characteristics of quantitative and dipstick uNGAL and POC UA from catheter and bag specimens.

METHODS

Study design and study site

We conducted a single-center, prospective cross-sectional study of a convenience sample of children 0–24 months of age evaluated for fever (≥38.0°C), for whom catheterized urine studies were being obtained to evaluate for UTIs, on the basis of clinician discretion. The study was conducted in the pediatric emergency department (ED) of an urban, university-affiliated tertiary care center, between July 2018 and March 2020.

Population: Inclusion and exclusion criteria

Children were eligible if they presented with fever (≥ 38.0°C) by any method, at home or in the ED, within the preceding 24 h. To enroll otherwise healthy children whose uNGAL expression should not be altered at baseline, we excluded children if they had a major congenital abnormality of any organ system including, but not limited to, inborn errors of metabolism, congenital heart disease, any known urogenital abnormalities at the time of enrollment (e.g., vesicoureteral reflux), chronic lung disease, or immune system disease. We also excluded patients for any of the following: received antibiotics within 48 h of evaluation; presence of indwelling catheters or shunts; evidence of focal infections such as abscess or cellulitis; or definitive sources of fever including, but not limited to, pneumonia, meningitis, or coxsackievirus.

To assess for enrollment bias and understand which patients were having urine obtained for clinical purposes, we retrospectively reviewed 1 random day of every 2-week period during enrollment. We assessed whether patients enrolled were systematically different from otherwise eligible febrile children <24 months of age who (1) had urine obtained but were not enrolled or (2) had no urine obtained.

The local institutional review board approved the study with documentation of informed consent from the participant’s legal guardian. This study was supported in part by an institutional clinical and translational science award (Grant UL1TR001873 from NCATS/NIH).

Urine collection

We obtained a minimum of 1 mL of additional urine via standard bladder catheterization. A sterile, plastic urine collection bag was then placed to collect additional urine (minimum of 1 mL). Urine collection bags were replaced only if soiled or adhesion was lost. The collection bags were checked periodically to remove once the child urinated.

uNGAL measurements and standard diagnostic tests

POC UA was performed immediately after urine collection using the Siemens CLINITEK Status+ analyzer in the pediatric ED. Urine was then immediately refrigerated at 4°C. Within 24 h of collection, the urine was transported to a laboratory where the POC uNGAL dipstick was performed; the remainder was centrifuged at 1000 rpm for 15 min; and the supernatant was aliquoted, frozen at −80°C, and batched for later processing for quantitative uNGAL and urine creatinine (UCr). There were no freeze/thaw cycles prior to biomarker analysis. Data support that uNGAL testing can be delayed with samples processed and stored in this manner for up to 5 years without protein degradation.^28,29 We measured uNGAL quantitatively by ELISA and semiquantitatively in ng/mL using investigational commercial platforms, BioPorto KIT 036 and BioPorto rapid assay device, respectively (BioPorto). For dipstick, semiquantitative uNGAL testing, a color chart differentiated uNGAL levels at intervals of <25, 25–50, 50–100, 100–150, 150–300, 300–600, and >600 ng/mL. uCr was measured by the Integra 400 Plus automated analyzer (Roche Diagnostics). The laboratory technicians performing uNGAL and uCr testing were blinded to clinical data and culture results.

Reference standard

Our reference standard was based on catheterized urine culture results, defined as follows:

1. “Definite” positive urine culture as ≥100,000 colony forming units (CFUs) /mL of up to two pathogens;

2. “Possible” positive urine culture as 10,000–100,000 CFUs/mL of at least one pathogen, without contaminants;

3. “Contaminated” as any growth of lactobacillus, micrococcus, diphtheroids, Bacillus species, or Staphylococcus epidermidis or growth of three or more organisms; and

4. “Negative” as no growth or growth of <10,000 CFUs/mL of a single organism.

Throughout the article, we use the term “positive urine culture” rather than “UTI” as the American Academy of Pediatrics definition of UTI includes both a threshold for CFUs/mL of a pathogen and the presence of inflammation (e.g., positive leukocyte esterase [LE] on UA).¹⁴ We only used culture results because we would not otherwise have been able to compare uNGAL to UA as a diagnostic marker.

Data analysis and sample size

Primary aim: Agreement in uNGAL levels between paired samples

Only patients with paired catheter and bag samples were included in this analysis. We dichotomized quantitative uNGAL levels, with ≥39.1 ng/mL being positive based on the potentially optimal cutoff for the diagnosis of UTIs.²⁴ At this threshold, we estimated the proportion of agreement between catheter and bag specimens and its 95% confidence interval (CI). Additionally, we separately estimated the agreement for patients with and without positive urine cultures.

Although studies have suggested that uNGAL levels will not be substantially altered by urine concentration,^25,30 we conducted similar analyses of agreement for the uNGAL/uCr ratio. This analysis accounted for the variable time intervals between catheterized and bagged urine sample collection and possible resulting changes in urine concentration. We identified the point on the ROC curve for uNGAL/uCr at which the Youden’s Index (J = Sensitivity + Specificity − 1) was closest to 1 and calculated agreement at this threshold value. The threshold value at which the Youden’s Index is closest to 1 represents the optimized tradeoff in sensitivity and specificity.

For dipstick uNGAL, we dichotomized positivity at >50 ng/mL because it was the closest value on the dipstick to the 39.1 ng/mL threshold from quantitative uNGAL. We estimated agreement between catheter and bag specimens at this >50 ng/mL threshold.

Secondary aim: Overall accuracy and diagnostic test characteristics of quantitative uNGAL, dipstick uNGAL, and POC UA from catheter and bag specimens

Quantitative uNGAL, dipstick uNGAL, and POC UA test results were analyzed as positive or negative. As above, we defined quantitative uNGAL as positive if >39.1 ng/mL and, dipstick uNGAL as positive if >50 ng/mL. POC UA was defined as positive if either the LE test was positive (≥trace) or the nitrite test was positive.

Urine cultures were analyzed as positive or negative. For the aim 1 subanalysis of agreement based on culture results, we considered cultures as positive only if pathogen count was ≥100,000 CFUs/mL (the definite criterion above). For the main analysis of aim 2 (test characteristics), we similarly used the ≥100,000 CFUs/mL criterion to define a positive culture; all others were considered negative. In a sensitivity analysis for aim 2, we combined possible (10,000–100,000 CFUs/mL) and definite (≥100,000 CFUs/mL) results to define a positive culture. We determined the accuracy of the uNGAL tests using the area under the receiver operating characteristic curve (AUC) with 95% CIs. We also determined the sensitivities and specificities for the urine tests with 95% CIs and compared the test characteristics using McNemar’s test for paired samples. The exact McNemar’s test was used when the sum of the discordant pairs was less than 10.³¹ All study samples were used for the comparison of test characteristics within collection method (catheter or bag) and only paired samples were used for the comparison of test characteristics between collection methods (catheter vs. bag).

Clinical and demographic variables were described using summary statistics and compared using parametric and nonparametric tests. Statistical analyses were performed using IBM SPSS Statistics, version 26 (IBM Corp.) and R, version 4.0.5 (R Core Team, 2021).

We aimed to enroll 300 patients but closed enrollment early due to the COVID pandemic. We based the sample size on the ability to have narrow CIs around the estimate of agreement between uNGAL levels. Assuming an expected agreement of 95% between uNGAL levels in paired catheter and bag specimens, a sample size of 300 would have resulted in a 95% CI of 91.9%–97.2%.

RESULTS

We enrolled 211 patients, all of whom had catheterized samples; 116 of the 211 (55.0%; constituting the sample analyzed for agreement) had all diagnostic studies obtained from both catheterized and bagged samples (Figure 1). Of the 211 enrolled, 21 (10.0%) had definite and 7 (3.3%) had possible positive urine cultures; 26 of the 28 (92.9%) were caused by Gram-negative organisms. For the 211 with catheterized samples, median uNGAL levels were 210.1 ng/mL (interquartile range [IQR] = 95.8–100.0) for definite positive urine cultures, 16.5 ng/mL (IQR = 12.0–162.1) for possible positive urine cultures and 0.01 ng/mL (IQR = 0.01–11.6) for the culture-negative group. Median uNGAL levels from patients with bagged samples (n = 116) were 318.7 (IQR = 70.2.1–411.1) for definite positive urine cultures, 193.9 (IQR = 113.8–273.9) for possible positive urine cultures and 21.4 (IQR = 11.3–41.1) in the culture-negative group.

FIGURE 1.

Flow diagram

Table 1 notes that enrolled patients with positive urine cultures were younger than those in the culture-negative group. Data Supplement S1, Table S1 (available as supporting information in the online version of this paper, which is available at http://onlinelibrary.wiley.com/doi/10.1111/acem.14402/full), presents the demographics and clinical characteristics of the patients with paired catheter and bag specimens (n = 116). Table S2 notes that children who were not enrolled but had urine cultures obtained (n = 59) had longer durations of fever and higher maximal ED temperatures compared to those enrolled. In contrast, those who were otherwise eligible but did not have urine studies obtained (n = 266) were older, typically had other sources of fever, and had shorter durations of fever (<72 h) compared to study patients.

Table 1.

Patient Demographics and Clinical Characteristics

	Culture Negative (N=190)	Culture Positivea (N=21)
Median age, d (IQR)	295 (150, 456)	171 (62, 261)
Boys, n (%)	87 (45.8)	10 (47.6)
Uncircumcised, n/N (%)	55/87 (63.2)	10/10 (100.0)
Hispanic ethnicity, n/N (%)	163/185 (88.1)	17/21 (81.0)
History of previous UTI, n (%)	6 (3.2)	2 (9.5)
History of urologic abnormality, n (%)	1 (0.5)	0 (0)
Fever duration at home, n (%)
< 24 h	55 (28.9)	10 (47.6)
24–72 h	61 (32.1)	7 (33.3)
> 72 h	74 (38.9)	4 (19.0)
Maximum temperature at home °C, mean (SD)	39.3 (0.8)	38.8 (0.6)
Maximum temperature in ED °C, mean (SD)	38.6 (1.0)	38.8 (1.0)

^aPositive urine culture defined as ≥100,000 CFUs/mL of ≤ 2 uropathogens.

IQR-Interquartile Range; SD-Standard deviation; ED-Emergency department

Primary aim: Agreement in uNGAL levels between paired catheter and bag specimens

For the 116 patients with paired specimens, the overall proportion of agreement between catheterized and bagged samples at a uNGAL cutoff of ≥39.1 ng/mL was 0.76 (95% CI = 0.67–0.83). In a subanalysis of patients with definite positive urine cultures (defined as ≥100,000 CFUs/mL), the proportion of agreement at this quantitative uNGAL threshold was 0.92 (95% CI = 0.62–1.00). Among patients with negative cultures (negative and possible positive cultures combined), the proportion of agreement between catheterized and bagged samples at a uNGAL cutoff of ≥39.1 ng/mL was 0.74 (95% CI = 0.64–0.82). When we measured the uNGAL/uCr ratio, the agreement at a cutoff of 190.0 ng/mg was 0.77 (95% CI = 0.68–0.84).

For the 116 patients with paired samples, the agreement between catheterized and bagged samples at a uNGAL dipstick threshold of >50 ng/mL was 0.77 (95% CI = 0.68–0.84). Among patients with definite positive urine cultures, the proportion of agreement between catheterized and bagged samples at this dipstick threshold was 0.77 (95% CI = 0.46–0.94). Among patients with negative cultures, the proportion of agreement between catheterized and bagged samples at a uNGAL cutoff of >50.0 ng/mL was 0.77 (95% CI = 0.67–0.84).

Secondary aim: Overall accuracy and test characteristics of quantitative uNGAL, dipstick uNGAL, and POC UA from catheter and bag specimens

The AUC for quantitative uNGAL from all patients with catheterized samples (n = 211) was 0.96 (95% CI = 0.89–1.00) and 0.93 (95% CI = 0.87–0.99) from patients with bagged samples (n = 116). Table 2 displays the tests characteristics for all study patients with catheterized urine samples (n = 211) and bagged samples (n = 116). Quantitative uNGAL and dipstick uNGAL from catheterized samples had higher sensitivities (p = 0.02 for quantitative and dipstick uNGAL) but lower specificities (p = 0.008 for quantitative uNGAL and p = 0.03 for dipstick uNGAL) compared to POC UA from catheterized samples at a LE ≥ 1+ threshold. When compared to UA at a threshold of either nitrite positive or ≥ trace LE, quantitative and dipstick uNGAL from catheterized samples were similarly sensitive (p = 1 for quantitative and dipstick uNGAL) and specific (p = 0.5 and p = 1, respectively).

Table 2.

Diagnostic Test Characteristics of uNGAL and POC Urinalysis

	Catheterized Samples (N=211; 21 positive urine cultures)		Bagged Samples (N=116; 13 positive urine cultures)
Urine Test	Sensitivity % (95% CI)	Specificity % (95% CI)	Sensitivity % (95% CI)	Specificity % (95% CI)
Absolute uNGAL ≥39.1 ng/mL	90.5 (69.6–98.8)	94.3 (89.9–97.1)	92.3 (64.0–99.8)	73.8 (64.2–82.0)
Dipstick uNGAL >50 ng/mL	90.5 (69.6–98.8)	95.3 (91.2–97.8)	76.9 (46.2–95.0)	73.8 (64.2–82.0)
POC Urinalysis
Either nitrite (+) or ≥ trace LE	95.2 (76.2–99.9)	95.8 (91.9–98.2)	92.3 (64.0–99.8)	67.0 (57.0–75.9)
Nitrite (+)	38.1 (18.1–61.6)	100.0 (98.1–100.0)	30.8 (9.1–61.4)	100.00 (96.5–100.0)
≥ trace LE	90.5 (69.6–98.8)	95.8 (91.9–98.2)	92.3 (64.0–99.8)	67.0 (57.0–75.9)
LE ≥1+	57.1 (34.0–78.2)	98.4 (95.5–99.7)	61.5 (31.6–86.1)	84.5 (76.0–90.9)
LE ≥2+	33.3 (14.6–57.0)	100.0 (98.1–100.0)	53.9 (25.1–80.8)	91.3 (84.1–95.9)

Positive urine culture defined as ≥100,000 CFUs/mL of ≤ 2 uropathogens.

uNGAL-urinary neutrophil gelatinase-associated lipocalin; POC-point-of-care; LE-leukocyte esterase

From bagged specimens, although the point estimate for quantitative uNGAL had higher sensitivity than LE at a threshold of ≥1+ (92.3% vs. 61.5%, respectively), this difference was not statistically significant (p = 0.1). Quantitative uNGAL from a bagged specimen had lower specificity compared to bagged POC UA at LE ≥1+ (p = 0.04). When compared to UA at a threshold of nitrite positive or ≥trace LE, quantitative uNGAL from bagged samples was similarly sensitive (p = 1) and specific (p = 0.25). Bagged specimens for quantitative and dipstick uNGAL were also found to be similarly sensitive (p = 0.5) and specific (p = 1).

In similar analyses of test characteristics but only of paired samples, we noted no meaningful differences in results. Urine testing from bagged samples had significantly lower specificities across all urine tests (p < 0.05), with the exception of nitrite (Table S3). In subanalyses based on age groups (0–3 months vs. 3–24 months) and sex, we also found similar findings (Table S4 and data by sex available upon request).

When we repeated the accuracy and test characteristic analyses and considered possible positive urine cultures as definite, the AUC for quantitative uNGAL from a catheterized sample was 0.92 (95% CI = 0.85–0.99) compared to 0.92 (95% CI = 0.86–0.99) from a bagged sample. Quantitative and dipstick uNGAL from catheterized samples again had higher sensitivities with similar specificities compared to POC UA at a threshold of ≥1+ LE (data available upon request). When compared to UA at a threshold of nitrite positive or ≥trace LE, quantitative and dipstick uNGAL from catheterized samples were similarly sensitive and specific (p > 0.05).

DISCUSSION

In this prospective study, uNGAL had suboptimal agreement when comparing bagged and catheterized samples from children evaluated for UTIs. The suboptimal agreement was noted for both quantitative and semiquantitative dipstick uNGAL tests. Additionally, although the overall accuracy was high for both quantitative and dipstick uNGAL using both urine sampling techniques, the test characteristics were similar to those of the UA when the UA was considered positive if either any LE was present or the nitrite test was positive. Although our study had a small population of children with positive urine cultures, our results suggest that bagged and catheterized uNGAL testing should not be used interchangeably and that there appears to be little benefit of uNGAL over UA by either catheterization or bagged specimen urine collection in the evaluation of UTIs in young, febrile children.

The potential promise of uNGAL as a more accurate screen for UTIs compared to UA is based on animal models showing that, while neutrophils contribute to the uNGAL pool, the kidney intercalated cells are a predominant source of uNGAL.^17,32 In a mouse UTI model devoid of neutrophils, uNGAL was still expressed.³² Additionally, mice devoid of the kidney medullary cells responsible for uNGAL expression had markedly reduced uNGAL levels and impaired bacterial clearance of their UTIs,¹⁷ supporting the role of uNGAL as a marker for UTIs independent of neutrophil activation. This independence from neutrophil activation would potentially improve sensitivity to detect UTIs compared to UA by having a direct renal response to inciting organisms, leading to elevated uNGAL levels. However, we did not see higher sensitivity unless an LE of ≥1+ was used as the threshold for preliminary diagnosis. The optimal threshold for clinical use of LE is unclear and the definition of LE positivity has varied across studies.^10,12,33 For uNGAL to be used across clinical settings, uNGAL dipstick, particularly via noninvasive bag sampling, would need to be similarly sensitive with superior specificity, compared to bagged dipstick UA at all typically used LE thresholds (trace or 1+). We speculated that a uNGAL test would improve the specificity for UTI evaluation by minimizing the effect of neutrophil contamination, which is particularly problematic with bagged samples. The widely varying specificities of the UA (72%–97%) across studies reflects in part the array of techniques used for sample collection that alter the likelihood of contamination.^2,6–12 We did not, however, find a clear signal for meaningful improvement in specificity using uNGAL, suggesting that contamination by neutrophils contributes to falsely positive uNGAL results, similar to that expected for LE.

Our findings of similar sensitivity and specificity for uNGAL and UA, for a given collection method, support prior studies noting a relationship between LE activity and uNGAL.^34,35 The effect of neutrophils may also explain our finding that the median quantitative uNGAL levels were higher in the bagged specimens across categories of culture results (negative, possible UTI, definite UTI) than the median levels from catheterized samples. In a study evaluating sampling technique and uNGAL levels in healthy adult female volunteers, they found that uNGAL levels were elevated in the first stream compared to midstream urine samples, even in cases where LE was negative, suggesting contributing factors other than pyuria.³⁴

LIMITATIONS

Our study had limitations, including studying a convenience sample of patients. However, our surveillance of patients who were otherwise eligible but not enrolled demonstrated that this group was similar to our study patients. Our sample was predominantly Hispanic (88.0%), limiting the potential generalizability to other populations. Our sample was relatively small, with only 55% of study patients having paired bagged sample for comparison of agreement, thereby widening the CIs around our point estimates. We also did not compare uNGAL to pyuria on microscopy (WBCs/hpf) given that, in our lab, as in others, microscopy is only performed for UAs that are positive for blood, LE, or nitrite. Finally, our laboratory does not further differentiate urine colony counts between 10,000 and 100,000 CFUs/mL, so we were unable to use the threshold for UTI recommended by the American Academy of Pediatrics of ≥50,000 CFUs/mL.¹⁴ Given this limitation, the cases categorized as having possible UTIs may have included patients with true UTIs, asymptomatic bacteriuria, or contaminants. The sensitivities of all preliminary urine screening tests, including quantitative and dipstick uNGAL, may be underestimated due to false-positive urine culture results.

CONCLUSION

In this prospective study, urine neutrophil gelatinase-associated lipocalin agreement from bagged and catheterized samples had insufficient agreement to be used interchangeably. Furthermore, quantitative and dipstick urine neutrophil gelatinase-associated lipocalin had decreased specificity from a bagged compared to a catheterized sample, suggesting that sampling technique may affect urine neutrophil gelatinase-associated lipocalin levels, similar to leukocyte esterase testing on urinalysis.