Treatment and Epidemiology of Third-Generation Cephalosporin-Resistant Urinary Tract Infections

We performed a retrospective, matched cohort-control study among patients aged 0 to 18 years of age with G3CR UTI at all public acute care facilities in Los Angeles (Harbor-University of California, Los Angeles Medical Center [site A], Los Angeles County + University of Southern California Medical Center [site B], and Olive View-University of California, Los Angeles Medical Center [site C]), and 1 not-for-profit hospital (Miller Children’s and Women’s Hospital [site D]). For sites A and D, the study period was November 1, 2014 to February 28, 2017, and for sites B and C, it was November 1, 2015 to February 28, 2018, reflecting the availability of a fully operational electronic medical record system.

We defined UTI as the presence of an abnormal urinalysis (pyuria and/or bacteriuria) plus a concomitant urine culture of a catheterized or clean-catch/midstream specimen with recovery of ≥10 000 cfu/mL) of an Enterobacterales isolate. Although a threshold of ≥50 000 cfu/mL is recommended for the diagnosis of UTI in children aged 2 to 24 months in the 2011 AAP Clinical Practice Guideline and its 2016 reaffirmation, a threshold of ≥10 000 cfu/mL is permitted for cultures with semiquantitative growth reported as 10 000 to 100 000 cfu/mL.^5,13  At each of our study sites, semiquantitative reporting was often used. We excluded results from specimens found to have squamous epithelial cells on microscopic examination or culture results deemed by study personnel to be consistent with asymptomatic bacteriuria. In addition, the urinalysis criteria were waived in rare circumstances based on clinical context if 2 or more study clinicians agreed unanimously that the incident could be confidently labeled as a UTI. Similarly, polymicrobial culture results could be included if 2 or more study clinicians agreed unanimously that the incident could be confidently labeled as a UTI.

The first identified G3CR UTI was considered the index UTI. For each patient with G3CR, we endeavored to select 2 non-G3CR UTI controls, matched by study site, sex, and age group (0–5 or ≥6 years). To do so, we generated lists of patients with ≥10 000 cfu/mL of a non-G3CR Enterobacterales isolate at each study site and performed chart review to verify inclusion criteria.

Susceptibilities were performed using GN-69 or GN-73 cards on the VITEK 2 automated testing platform (bioMérieux, Inc, Durham, NC). Third-generation cephalosporin-resistance was defined as resistance to ceftriaxone, cefotaxime, or ceftazidime. We defined non-G3CR as susceptible to all (and at least 1) of the third-generation cephalosporins for which a result was reported. Carbapenem-resistant isolates were excluded because of the unique risk factors, treatment approaches, and outcomes among patients with such drug-resistant infections.^14,15  When available, minimum inhibitory concentrations (MICs) were recorded, as were phenotypic resistance mechanisms and method of confirmation. There were differences in reporting processes among study sites related to manual confirmation of extended-spectrum β-lactamase (ESBL) phenotype, cascaded reporting, suppression of MIC values or susceptibility interpretations for β-lactam agents, and VITEK 2 prediction for ampC production. Time to finalization of culture and drug susceptibility results was counted in days; specimen collection date was day 0.

Using a standardized data collection instrument (see Supplemental Information), we collected biosocial, clinical, and health care utilization information for the acute phase and the 6-month period thereafter. To complement other measures of diversity and biosocial determinants, race (Black/non-Black) and ethnicity (Hispanic/non-Hispanic) were recorded according to demographic fields in each chart. International exposure was defined as any recent personal (or current household contact) travel outside the United States. Previous acute health care utilization included any emergency department visit or hospitalization occurring before first date of presentation for the index UTI. Duration of therapy was counted in days, with the date of the first antibiotic prescription considered day 1. Antipseudomonal agents were defined as: ceftazidime, cefepime, piperacillin, meropenem, imipenem, ceftolozane, ciprofloxacin, levofloxacin, gentamicin, amikacin, tobramycin, or aztreonam. Assessment of response to initial antimicrobial therapy was defined as “significant improvement” if there was (1) resolution of fever and other vital sign abnormalities, for hospitalized patients, or (2) documentation of clinical improvement by treating physicians based on their assessment. We defined time to significant improvement as the earlier of these criteria. Use of genitourinary (GU) imaging was recorded, and exposure to GU imaging with ionizing radiation was defined as voiding cystourethrogram, ^99mTc-labeled mercaptoacetyltriglycine renal scan, or dimercaptosuccinic acid scan. Subsequent multidrug-resistant urine culture results included the isolation of Enterobacterales exhibiting G3CR or carbapenem resistance. Subspecialty consultation included care from nephrology, urology, or infectious diseases services. For 5 G3CR and 5 non-G3CR at every site, 2 investigators reviewed data entry for accuracy, and a third adjudicated discrepancies.

Approval was obtained from each participating site’s institutional review board, and data usage agreements were in place permitting the entry of data into a secure online project-management platform (RedCAP, Vanderbilt University, Nashville, TN). We performed descriptive statistics, univariate analyses using differences in proportions for categorical variables, and differences in medians with interquartile ranges (IQRs) for continuous variables. Bivariate and multivariate logistic, linear, or Poisson regression analyses were performed, as appropriate. P values < .05 were considered statistically significant. Sensitivity analyses restricted to patients with ≥50 000 cfu/mL were performed for robustness. All analyses were performed using Stata, version 13.1 (2013, StataCorp LP, College Station, TX).

We identified 3408 unique urine cultures with Enterobacterales isolates, of which 131 (3.8%) were G3CR (see Fig 1); 2 carbapenem-resistant isolates were excluded. There were no differences in G3CR prevalence between sites (range, 3.2%–4.1%). One hundred seven children were classified as patients with G3CR; Escherichia coli was the identified pathogen in 89 (83%), and 87 (81%) were ESBL positive (Table 1). We identified 206 non-G3CR UTIs as matched controls; of these, 194 (94%) were because of E. coli.

Pyuria was present for 264 (84.3%) patients. Urinalysis criteria were waived for 20 (6.4%: 10 G3CR and 10 controls) for the following reasons: neutropenia/bone marrow failure and/or quantity not sufficient for urinalysis (but presence of UTI signs/symptoms and concomitant bacteremia with the same uropathogen without another identified focus). Thirty-one (9.9%) patients had >1 uropathogen isolated in urine culture, but with a single predominant Enterobacterales spp. deemed to represent the causative organism by unanimous consensus of clinical study personnel. Two hundred eighty-five (91.1%) patients had ≥50 000 cfu/mL of the uropathogen of interest, and 255 (81.4%) had ≥100 000 cfu/mL.

Susceptibility to noncephalosporin agents is detailed in Table 2. The effect of breakpoint changes on cephalosporin and fluoroquinolone MIC interpretation are provided in Supplemental Tables 5 and 6. Six (5.6%) G3CR and 0 controls were resistant to all oral β-lactam agents, fluoroquinolones, and trimethoprim-sulfamethoxazole (P = .001). Time to culture result was not different between groups: median, 2 days (IQR, 1–3). It was less common, however, for susceptibility results to be reported within 1 day for G3CR isolates versus nonG3CR (16% vs 37.1%; odds ratio [OR], 0.32; 95% confidence interval [CI], 0.18–0.58; P < .001).

Clinical characteristics and risk factors associated with G3CR UTI included previous receipt of antimicrobial therapy, previous health care utilization, and presence of ≥1 underlying medical condition (Table 3). There were no differences when previous drug receipt occurred <30 days versus <60 to 90 days before index UTI (data not shown). Previous hospitalization was more common among patients with G3CR (41%) than control patients (17%; OR, 3.30; 95% CI, 1.77–6.13; P < .001). After adjustment for study site and underlying medical conditions, the odds of G3CR UTI remained significant for previous antimicrobial treatment or prophylaxis (adjusted odds ratio [aOR], 1.9; 95% CI, 1.1–3.4), but not for previous acute health care utilization (aOR, 1.9; 95% CI, 1.0–3.7).

Antimicrobial treatment was prescribed for the index UTI in 92.7% of susceptible patients and 87.9% of patients with G3CR (Table 4). Compared with controls, initial antimicrobial agents for patients with G3CR were less frequently given orally (62% vs 75.4%; OR, 1.90; 95% CI, 1.08–3.34; P = .02) and less frequently exhibited in vitro activity against the patient’s isolate (Table 4). Modifications of the initial antibiotic choice occurred in 55% of patients with G3CR versus 29% of controls (OR, 2.93; 95% CI, 1.70–5.05; P < .001). Initial antibiotics included agents with antipseudomonal activity in 14 (15%) patients with G3CR versus 5 (3%) controls (OR, 6.51; 95% CI, 2.11–23.71; P < .001). Eleven (23%) patients with G3CR who received initial antibiotics without antipseudomonal activity subsequently had their therapy modified to an anti-pseudomonal agent; this did not occur in any control patient (P < .001).

Of patients with documentation of response to initial therapy, there was no difference in the proportion with significant improvement between patients with G3CR (57%) and control patients (52%; OR, 0.81; 95% CI, 0.44–1.50) (Table 4). Improvement was documented within a median of 2 days (IQR, 1–5) in both groups, including 25 patients with G3CR receiving initial therapy without in vitro activity against their isolates (Supplemental Table 7). After adjustment for underlying medical conditions, in vitro susceptibility to initial antimicrobial agents did not increase the odds of significant improvement (aOR, 1.17; 95% CI, 0.42–3.25; P = .77). In patients who were assessed to have had no significant improvement on initial antimicrobial agents, therapy was modified more often for G3CR (42%) than for susceptible UTI (15%; OR, 4.20; 95% CI, 1.61–10.95), however total duration did not differ (Table 4).

Multivariate analysis revealed that patients with G3CR were more frequently hospitalized upon presentation than control patients (aOR, 6.60; 95% CI, 2.03–21.44; P = .002) and that presence of an underlying medical condition was an independent predictor of hospitalization (aOR, 2.44; 95% CI, 1.11–5.36; P = .03). Factors not associated with hospitalization in this analysis included highest presenting temperature, in vitro activity of initial antimicrobial agents, and study site. Patients with G3CR were hospitalized longer than control patients (median, 8 days [IQR, 4–15] vs 4 days [IQR, 3–7]; P < .05; Table 3). Controlling for highest presenting temperature and presence of underlying medical condition, G3CR UTI (incidence rate ratio [IRR], 2.27; 95% CI, 1.70–3.05; P < .001), study site (IRR, 1.54; 95% CI, 1.09–2.16; P = .01) and, paradoxically, susceptibility to initial antimicrobial agents (IRR, 1.51; 95% CI, 1.15–1.99; P = .003) were associated with longer length-of-stay.

Seventy-one (66%) patients with G3CR and 134 (65%) control patients had documented follow-up until 6 months after index UTI. No deaths occurred. Patients with G3CR were more likely to have repeat urine cultures (75%) than control patients (53%; OR, 2.61; 95% CI, 1.33–5.24). Among these, patients with G3CR had subsequent multidrug-resistant culture results more frequently than control patients (26% vs 3%; OR, 12.21; 95% CI, 2.63–56.54). There were no significant differences between groups in orders for any imaging studies of the GU tract (Table 4) or exposure to GU imaging with ionizing radiation (data not shown). Subspecialty consultation was more common among patients with G3CR than control patients, and the proportion of patients with G3CR receiving additional courses of antimicrobial therapy for any indication other than treatment of the index UTI was also higher than that of control patients (Table 4). After adjustment for underlying medical conditions (aOR, 2.2; 95% CI, 1.1–4.3; P = .02) and study site (not significant), the odds of patients with G3CR receiving any additional courses of antimicrobial therapy remained significant (aOR, 2.63; 95% CI, 1.41–4.91; P = .002).

Restricting analyses to patients with ≥50 000 cfu/mL, results of univariate (Supplemental Tables 8 and 9) and adjusted/multivariate analyses were robust. Adjusting for presence of underlying conditions, multivariate analysis redemonstrated that in vitro susceptibility to initial antimicrobial agents did not increase the odds of significant improvement on initial therapy (OR, 1.19; 95% CI, 0.40–3.52; P = .75). There were no substantive changes in the direction of results for this or any other analyses.

G3CR and ESBL-producing E. coli are increasingly reported in patients with community-acquired infections, including UTI.^7,8,16,17  The most recent pediatric United States population data (2010–2011) reported a prevalence of G3CR and ESBL Enterobacterales of 3.0% and 0.9%.⁷  Our data revealed a slightly higher prevalence of G3CR, with a majority exhibiting an ESBL phenotype. Our study included only patients with UTI, incorporated a contemporaneous control group of children with susceptible UTI, and selected patients from a diverse population that was not skewed by a preponderance of medically complex children.

There are few pediatric outcomes data on G3CR UTI and even fewer on clinical response to discordant therapy.^18–21  We present the largest number of detailed outcomes data and statistical analyses demonstrating that first-line antimicrobial agents for UTI are equally likely to lead to significant improvement in children with G3CR versus susceptible isolates. Our data are consistent with a recent study in which 186 (80%) of 192 children with G3CR UTI who received discordant therapy had clinical improvement, as well as a European study that found no difference in clinical outcomes among 61 children on initially effective versus 77 on ineffective treatment of febrile G3CR UTI; neither study included a comparator group with susceptible UTI.^21,22 

Observed clinical response to discordant antimicrobial agents for UTI may be superior to predicted in vitro activity for at least 2 reasons. Interpretive criteria are often based upon small numbers of patients with severe, high-inoculum infections including bacteremia limiting generalizability to milder infections such as nonbacteremic UTI.^23–25  Also, the concentration of antimicrobial agents in the renal parenchyma or urine may result in exposures several-fold higher than an organism’s MIC breakpoint, as illustrated by the Clinical and Laboratory Standards Institute’s decision to establish a higher urine-specific breakpoint for cefazolin in 2014.^21,26  “Susceptible” and “resistant” categories reflect probabilities and expert consensus, which may not correlate perfectly with real-world clinical observations.²⁴ 

Neither AAP guidelines or a recent state-of-the-art review provides treatment recommendations for drug-resistant UTI.^5,27  Interestingly, despite recommending avoidance of cefepime and piperacillin-tazobactam for ESBL isolates, Infectious Diseases Society of America permits continuation of either drug if initiated empirically when the patient improves clinically.²⁸ 

We provide new information on health care utilization during the acute care phase and the first usage data during the 6 months after a G3CR UTI in children. Patients with G3CR had a more than sixfold adjusted odds of hospitalization at presentation for their index UTI, and length-of-stay was twice as long as that for controls. Total treatment duration correlated with study site and not study group, likely reflecting clinician preference rather than clinical indication. Similarly, patients with G3CR were more likely to have subsequent courses of antimicrobial therapy and repeat urine culture.

Subsequent multidrug-resistant isolates were recovered in approximately one-quarter of patients with G3CR who had repeat urine cultures, similar to a molecular epidemiology study at 4 pediatric centers which found that 14% had subsequent resistant infections.²⁹  A common clinical practice is to use previous susceptibility results to guide the choice of empirical therapy for subsequent incidents of suspected UTI.³⁰  This may result in exposure to inappropriately broad spectrum antimicrobial agents and may warrant reconsideration.

We identified several risk factors for G3CR UTI, some novel and others previously described.^{7,18,20,31–35}  Risk factors have included underlying medical conditions, previous acute health care utlization, and previous antibiotic exposure.^{16,32,34,36–40}  Our data reproduce these findings in a diverse and underserved population in the United States. Although >50% of patients in either group had at least 1 underlying medical condition, the majority reflected common conditions including obesity, metabolic syndrome, and developmental delay. Unlike in prior studies, international exposure was not a significant risk factor despite being present in 22% of patients with G3CR compared with 11% of control patients.^41,42 

Our findings are subject to several limitations. We used dichotomized age ranges that some may consider too broad, but this was decided a priori to ensure identification of enough matched controls. Data were missing for several variables, particularly biosocial risk factors previously unexamined for this condition. MICs were also missing owing to variable suppression of results, although application of changing interpretive criteria would not have resulted in meaningful changes in study group classification (Supplemental Table 5). Generalizability may be limited because of our predominantly Medicaid patients in Los Angeles County, California, who face many barriers to health care access and continuity. We did not have complete data on initial treatment response, and documentation may have varied for nonhospitalized patients. We did not attempt to ascertain UTI relapse, as there is no widely accepted standard for distinguishing relapse from recurrence.¹  Patients may have sought care outside of our hospital systems, but it is highly likely that they would have been seen at 1 of our study hospitals.

Because we accepted semiquantitative cultures ≥10 000 cfu/mL (a minority of our study population), some patients may have had asymptomatic bacteriuria and not true UTI. Such cultures were included only if sufficient data were available to confidently label an incident as UTI according to guidelines and common clinical practice.^5,13,27  Sensitivity analyses restricted to patients with ≥50 000 cfu/mL did not reveal substantive changes in results (Supplemental Tables 7 and 8).

Because AAP guidelines target children aged 2 to 24 months, our study shares a limitation of pediatric UTI studies: the absence of a national consensus diagnostic definition outside of this age range.^5,13  Indeed, the definition of pyuria remains an area of active research, and 1 commentary has problematized the use of a strict cfu/mL cutoff.^43–45  We used a composite of urinalysis and urine culture, each performed in response to signs/symptoms attributable to UTI.^5,13,27  We could not reliably distinguish children with pyelonephritis beyond initial working diagnosis, as we did not collect sufficiently detailed physical examination data, and no child had a dimercaptosuccinic acid scan.

We describe clinical, epidemiologic, and usage characteristics of G3CR UTI in an underserved pediatric population. Our data highlight the need for randomized controlled trials on safe and optimal treatment approaches for G3CR UTI. Future investigations should evaluate if such infections alter subsequent practice behaviors, reflecting priming by a previous drug-resistant UTI.

Members of the Resistant UTI in Children Study Team include Dr Thomas Tarro and Dr Mikhaela Cielo. The authors thank Alan Navarro, J.R. Caldera, Tiffany Pham, Karen S. White, Meghan Regalado, Annabelle Sasu, and Drs Michael Bolaris, Dong Chang, Ruey-Kang Chang, Diana Etcola, and Loren Miller.

AAP

American Academy of Pediatrics

CI

confidence interval

ESBL

extended-spectrum β-lactamase

G3CR

third-generation cephalosporin-resistant

GU

genitourinary

IQR

interquartile range

MIC

minimum inhibitory concentration

OR

odds ratio

UTI

urinary tract infection