Abstract
Extended-spectrum β-lactamase-producing Escherichia coli (ESBL-EC) is increasingly identified as a cause of acute pyelonephritis (APN) among patients without recent health care contact, i.e., community-associated APN. This case-control study compared 75 cases of community-associated ESBL-EC APN (CA-ESBL) to 225 controls of community-associated non-ESBL-EC APN (CA-non-ESBL) to identify the risk factors for ESBL-EC acquisition and investigate the impact of ESBL on the treatment outcomes of community-associated APN (CA-APN) caused by E. coli at a Korean hospital during 2007 to 2013. The baseline characteristics were similar between the cases and controls; the risk factors for ESBL-EC were age (>55 years), antibiotic use within the previous year, and diabetes with recurrent APN. The severity of illness did not differ between CA-ESBL and CA-non-ESBL (Acute Physiology and Chronic Health Evaluation [APACHE] II scores [mean ± standard deviation], 7.7 ± 5.9 versus 6.4 ± 5.3; P = 0.071). The proportions of clinical (odds ratio [OR], 1.76; 95% confidence interval [CI], 0.57 to 5.38; P = 0.323) and microbiological (OR, 1.16; 95% CI, 0.51 to 2.65; P = 0.730) cures were similar, although the CA-ESBL APN patients were less likely to receive appropriate antibiotics within 48 h. A multivariable Cox proportional hazards analysis of the prognostic factors for CA-APN caused by E. coli showed that ESBL production was not a significant factor for clinical (hazard ratio [HR], 0.39; 95% CI, 0.12 to 1.30; P = 0.126) or microbiological (HR, 0.49; 95% CI, 0.21 to 1.12; P = 0.091) failure. The estimates did not change after incorporating weights calculated using propensity scores for acquiring ESBL-EC. Therefore, ESBL production did not negatively affect treatment outcomes among patients with community-associated E. coli APN.
INTRODUCTION
The recent increase in extended-spectrum β-lactamase (ESBL) producers among community Escherichia coli isolates has been of great concern worldwide. In previous studies, recent contact with the health care system was identified as a major risk factor for community onset ESBL-producing E. coli (ESBL-EC) infections (1–3). However, ESBL-EC is causing increasing numbers of community onset infections, especially urinary tract infections (UTIs), among patients without health care-associated (HCA) risk factors, i.e., those with community-associated (CA) UTIs (3, 4). These findings emphasize the need for new empirical antibiotic strategies for managing UTIs in the community. However, only a few studies have examined what predispose patients with no discernible HCA risk factors to develop ESBL-EC UTIs and how these patients differ clinically from those with non-ESBL-EC UTIs. Most previous studies have focused on bloodstream infections, which represent the most serious form of E. coli infections, and combined CA and HCA infections in their analyses without analyzing them separately (5, 6). Because HCA infections often resemble hospital-acquired infections and tend to have more severe symptoms (7–10), the impact of ESBL production on treatment outcomes might be overestimated in CA infections, which might potentially lead to overtreatment for less serious infections. Among CA-UTIs, CA acute pyelonephritis (APN) caused by ESBL-EC has rarely been studied, even though APN can pose a dilemma for antibiotic selection because it ranges from clinically benign uncomplicated episodes to severe septic shock.
Therefore, we investigated the risk factors for ESBL-EC acquisition to help identify patients likely to have CA-APN caused by ESBL-EC, and we clarified the impact of ESBL production on the treatment outcomes of CA-APN caused by E. coli by comparing patients with CA-APN caused by ESBL-EC to those with CA-APN caused by non-ESBL-EC.
MATERIALS AND METHODS
Study design and patients.
We conducted a case-control study from a retrospective cohort of adult patients with community onset APN caused by E. coli who were treated at Daejeon St. Mary's Hospital between January 2007 and December 2013. Daejeon St. Mary's Hospital is a 600-bed secondary-care community hospital serving Daejeon city and the surrounding areas, with a population of approximately 1.5 million people. Patients were included in this retrospective cohort if they were >15 years of age and were diagnosed with community onset APN caused by E. coli. This cohort was further divided into two subcohorts, CA-APN and HCA-APN. The case patients were those with CA-APN caused by ESBL-EC (CA-ESBL group), and for patients with multiple episodes, only the first episode was included in the analysis. The control patients were those with CA-APN caused by non-ESBL-EC (CA-non-ESBL group). For each case, three control patients were randomly selected from each risk set composed of all eligible control patients treated within a month of the time when each case was identified. The study was approved by the Daejeon St. Mary's Hospital institutional review board, with a waiver for informed consent (DC14RISI0031).
Data collection and definitions.
Clinical and microbiological data were collected from databases, and medical records were reviewed. The following variables were collected: patient age, gender, comorbidities, and Charlson comorbidity index (CCI), detailed history, including previous use of antibiotic therapy and time and number of APN recurrences, structural and functional urinary tract abnormalities, indwelling urinary instrumentation use, and the microorganisms isolated from urine and blood and their susceptibility patterns. The time to appropriate antibiotic therapy, duration of fever and clinical symptoms, and the length of hospital stay information were also collected. The severity of illness was measured using the presence of shock and Acute Physiology and Chronic Health Evaluation (APACHE) II scores (11). Detailed data on antibiotic therapies were collected, including treatment durations, regimens, and prescribed doses.
APN was diagnosed if patients had (i) pyuria (≥10 white blood cells per high-power field) and a positive urine culture (>105 CFU of E. coli CFU/ml), (ii) fever of ≥38°C on tympanic membrane temperature, (iii) flank tenderness, and (iv) urinary symptoms, such as dysuria, frequency, or urgency. APN was classified as complicated in patients with structural or functional urinary tract abnormalities, urinary instrumentation use, chronic kidney disease, or immunocompromised status (transplant recipients, use of corticosteroids at a dose of ≥20 mg of prednisone or equivalent for ≥2 weeks, or anticancer chemotherapy within the previous 6 months), or if it presented with suppurative complications (12). APN was otherwise classified as uncomplicated.
Community onset infections were defined as a positive urine culture obtained within 48 h of hospitalization or attendance at an outpatient clinic. Community onset infections were further classified as CA if none of the following HCA risk factors were present: (i) hospitalization in an acute care center for >2 days or residence in a nursing home or long-term care facility during the previous 90 days, (ii) intravenous therapy or specialized home care or invasive procedures during the previous 30 days, and (iii) hemodialysis during the previous 30 days (9). Otherwise, community onset infections were classified as HCA. Antibiotic therapy was considered appropriate if the treatment regimen included at least one active antibiotic to which the E. coli isolate was susceptible in vitro and the recommended dose was administered (13).
Clinical and microbiological outcome measures.
The primary outcomes were clinical and microbiological cure. The secondary outcomes were early clinical success and the length of hospital stay. The clinical outcomes were assessed by a thorough medical record review after 3 days of treatment (day 3) and after 7 days of treatment or at the time of discharge, whichever was earlier (day 7), and during the follow-up period of 4 weeks after the completion of antibiotic therapy. Early clinical success was defined as the resolution of fever with an improvement in urinary symptoms and signs at day 7. Otherwise, patients were regarded as early clinical failures. Clinical cure was defined as the complete resolution of fever and urinary symptoms and signs without recurrence during follow-up. Clinical failure was defined as deaths or a recurrence of symptoms during follow-up after initial clinical improvement. Resolution of fever was defined as an afebrile state with a body temperature of ≤37°C for >1 day. Microbiological cure was defined as the eradication of the baseline pathogen (urine culture, <103 CFU/ml) without microbiological relapse during follow-up. Microbiological failure included microbiological persistence, defined as a positive urine culture (≥104 CFU/ml) of the baseline pathogen after ≥5 days of appropriate antibiotic therapy, and relapse was defined as a positive culture of the baseline pathogen during the follow-up period after microbiological eradication had been confirmed (13). Microbiological cures were assumed for patients seen subsequently in the outpatient clinic without any signs of infection after confirmation of microbiological eradication at the first follow-up visit. Patients were excluded from the microbiological outcome analysis if their urine cultures were not followed up.
Microbiology and susceptibility tests.
Antimicrobial susceptibility tests and species identification were performed with the MicroScan Neg combo panel type 32 (2007 to 2008) or the Neg breakpoint combo panel 44 (2009 to 2013) (Siemens Healthcare Diagnostics, Inc., West Sacramento, CA, USA), according to the manufacturer's instructions. The MIC breakpoint for each antimicrobial agent was determined and ESBL production confirmed with the double disk synergy test performed according to 2007 Clinical and Laboratory Standards Institute guidelines (14).
Statistical analyses.
The demographic and clinical characteristics were summarized as percentages for categorical variables and as means and standard deviations (SD) or medians and interquartile ranges (IQR) for continuous variables, depending on their distributions. Student's t and Wilcoxon rank-sum tests were used to compare the continuous variables. Risk factors were identified using conditional logistic regression analysis to estimate odds ratios (OR) with 95% confidence intervals (CIs). Variables with a P value of <0.2 in the unadjusted models were entered into the multivariable regression model, in which significant variables were selected using a stepwise backward process. Additionally, we included variables that were considered potential confounders based on previous literature or clinical relevance in the model. Interactions between the variables were investigated by including cross-product terms in the multivariable logistic regression. Variables causing substantial confounding, defined as a >10% β-coefficient change, were maintained in the final model. To investigate the impact of ESBL production on treatment outcomes and to identify the prognostic factors associated with treatment outcomes, Cox proportional hazards regression models were used to obtain unadjusted and adjusted hazard ratios (HRs) and 95% CIs with the variable selection process described above. The proportional hazards assumption was tested by graphically inspecting complementary log-log plots for each variable. To further adjust for confounding, we incorporated inverse probability weights into Cox proportional hazards regression models. The inverse probability of treatment weighting was calculated by using a propensity score for each patient's probability for having CA-ESBL-EC (15). The propensity scores were estimated in a multivariable logistic regression in which the covariates were patient age, gender, comorbid conditions, CCIs, concomitant bloodstream infections, indwelling urinary instrumentation, urinary tract abnormalities, history of prior APN, complicated APN, and antibiotic use within the previous year. The weights for the CA-ESBL group were the inverse of the propensity scores, and those for the CA-non-ESBL group were the inverse of 1 − the propensity scores. The stabilized weights were calculated by replacing the numerator with the marginal probability of actually having CA-ESBL-EC (16). After weighting, the variables were tested for balance using the absolute standardized mean difference, whereby differences of >20% represent a meaningful imbalance. The data were analyzed using the Stata statistical software, release 12 (Stata Corp., College Station, TX, USA).
RESULTS
Study population.
We identified 127 adult patients with APN caused by ESBL producers (9.9%) from a retrospective cohort consisting of 1,285 episodes of community onset APN caused by E. coli. Among these 127 patients were 75 (59.1%) with CA-APN. We selected 225 matched control patients with CA-APN caused by non-ESBL-EC. The demographic and baseline characteristics of the patients and the antimicrobial susceptibility results are summarized in Tables 1 and 2, respectively.
TABLE 1.
Demographics and baseline characteristics of patients with community-associated acute pyelonephritis caused by E. coli, according to ESBL productiona
| Variableb | CA-ESBL group (n = 75)c | CA-non-ESBL group (n = 225) | OR (95% CI)d | P |
|---|---|---|---|---|
| Age >55 yr | 51 (68.0) | 114 (50.7) | 2.08 (1.19–3.63) | 0.010 |
| Male | 9 (12.0) | 14 (6.2) | 2.03 (0.84–4.86) | 0.114 |
| Comorbidities | ||||
| CCI ≥3 | 11 (14.7) | 25 (11.1) | 1.37 (0.64–2.94) | 0.416 |
| CNS condition | 9 (12.0) | 15 (6.7) | 1.92 (0.80–4.63) | 0.146 |
| Cardiovascular condition | 7 (9.3) | 18 (8.0) | 1.20 (0.47–3.06) | 0.710 |
| COPD | 2 (2.7) | 5 (2.2) | 1.22 (0.22–6.94) | 0.819 |
| Hepatobiliary condition | 5 (6.7) | 11 (4.9) | 1.38 (0.47–4.09) | 0.557 |
| Diabetes mellitus | 27 (36.0) | 53 (23.6) | 1.86 (1.05–3.31) | 0.034 |
| Chronic kidney disease | 7 (9.3) | 14 (6.2) | 1.52 (0.60–3.84) | 0.373 |
| Immunocompromisede | 8 (10.7) | 18 (8.0) | 1.40 (0.57–3.46) | 0.469 |
| Bed-ridden status | 1 (1.33) | 7 (3.1) | 0.43 (0.05–3.48) | 0.428 |
| APN characteristics | ||||
| Complicated | 24 (32.0) | 66 (29.3) | 1.14 (0.64–2.01) | 0.660 |
| Recurrent | ||||
| ≥1 event | 25 (33.3) | 38 (16.9) | 2.96 (1.51–5.83) | 0.002 |
| ≥2 events | 14 (18.7) | 11 (4.9) | 6.75 (2.38–19.17) | <0.001 |
| Within 1 yr | 14 (18.7) | 7 (3.1) | 7.60 (2.75–21.50) | <0.001 |
| Within 2 yr | 18 (24.0) | 21 (9.3) | 3.31 (1.59–6.90) | 0.001 |
| Within 3 yr | 18 (24.0) | 27 (12.0) | 2.66 (1.28–5.55) | 0.009 |
| Urinary tract abnormalities | 17 (22.7) | 44 (19.6) | 1.21 (0.64–2.28) | 0.561 |
| Urinary instrumentation | 2 (2.7) | 2 (0.9) | 3 (0.42–21.30) | 0.272 |
| Antibiotic use within the previous year | 24 (32.0) | 21 (9.3) | 5.10 (2.46–10.58) | <0.001 |
| Fluoroquinolones | 9 (12.0) | 13 (5.8) | 2.20 (0.90–5.38) | 0.083 |
| Cephalosporins | 13 (17.3) | 7 (3.1) | 7.10 (2.51–20.07) | <0.001 |
The data represent the no. (%) of patients, unless otherwise specified. ESBL, extended-spectrum β-lactamase.
CCI, Charlson comorbidity index; CNS, cerebrovascular; COPD, chronic obstructive pulmonary disorder; APN, acute pyelonephritis.
CA, community associated.
OR, odds ratio; CI, confidence interval.
Immunocompromised patients included those who received transplantation, corticosteroid therapy (≥20 mg of prednisone or equivalent for ≥2 weeks) or immunosuppressants, and anticancer chemotherapy within 6 months prior to pyelonephritis.
TABLE 2.
Antimicrobial susceptibilities of E. coli isolates in this studya
| Antimicrobial agent(s)b | MIC breakpoint (mg/liter) | No. (%) of isolates nonsusceptible to antimicrobial agents in group: |
|
|---|---|---|---|
| CA-ESBL (n = 75) | CA-non-ESBL (n = 225) | ||
| Amikacin | 16 | 3 (4.0) | 0 (0) |
| Gentamicin | 4 | 32 (42.7) | 35 (15.6) |
| Tobramycin | 4 | 32 (42.7) | 30 (13.3) |
| Ciprofloxacin | 1 | 39 (52.0) | 29 (12.9) |
| Levofloxacin | 2 | 36 (48.0) | 23 (10.2) |
| Amoxicillin-clavulanatec | 8/4 | 24 (37.5) | 24 (18.1) |
| Piperacillin-tazobactam | 16/4 | 4 (5.3) | 2 (0.9) |
| TMP-SMX | 2/38 | 35 (46.7) | 43 (19.1) |
| FQ+TMP-SMX | 22 (29.3) | 15 (6.7) | |
| FQ+TMP-SMX+AG | 14 (18.7) | 8 (3.6) | |
The P values of the odds ratios for each antimicrobial agent between the community-associated extended-spectrum β-lactamase (CA-ESBL) and CA-non-ESBL groups were all <0.01.
FQ, fluoroquinolones; TMP-SMX, trimethoprim-sulfamethoxazole; AG, aminoglycosides.
A total of 191 isolates were tested (64 CA-ESBL and 127 CA-non-ESBL).
Risk factors for ESBL-EC as a cause of CA-APN.
Most baseline characteristics were similar between the CA-ESBL and CA-non-ESBL groups (Table 1). Of the comorbid conditions, only diabetes was significantly associated with ESBL-EC acquisition (OR, 1.86; 95% CI, 1.05 to 3.31; P = 0.034). In terms of APN characteristics, recurrence was the only variable that differed significantly between the cases and controls (OR, 2.96; 95% CI, 1.51 to 5.83; P = 0.002). The number of recurrences and the recentness of recurrent events proportionately increased the risk of ESBL-EC acquisition. Patients who had been exposed to an antibiotic within the previous year were more likely to have CA-APN caused by ESBL-EC (OR, 5.10; 95% CI, 2.46 to 10.58; P < 0.001), and the use of cephalosporins was identified as a significant risk factor (OR, 7.10; 95% CI, 2.51 to 20.07; P < 0.001). In a multivariable analysis, age (>55 years) (OR, 1.97; 95% CI, 1.02 to 3.50; P = 0.043) and antibiotic use within the previous year (OR, 4.61; 95% CI, 1.95 to 10.96; P = 0.001) were independent risk factors for ESBL-EC acquisition (Table 3, model 2). We also identified an interaction between diabetes and APN recurrence, in that the joint effect of diabetes and APN recurrence on the risk of CA-ESBL-EC acquisition was significantly higher (OR, 4.24; 95% CI, 1.31 to 16.92; P = 0.041) than was the effect of diabetes without APN recurrence (OR, 1.36; 95% CI, 0.60 to 3.11; P = 0.461) or that of APN recurrence without diabetes (OR, 1.34; 95% CI, 0.62 to 2.56; P = 0.545) (Table 3, model 2; see also the supplemental material). Multivariable analysis including specific antibiotic classes (Table 3, model 3) showed that previous cephalosporin therapy was significantly associated with CA-ESBL-EC acquisition (OR, 6.33; 95% CI, 2.01 to 19.92; P = 0.002).
TABLE 3.
Multivariable analysis for risk factors associated with acquisition of ESBL-producing E. coli as a cause of community-associated acute pyelonephritisa
| Variable | Model 1: without interaction term (OR [95% CI])b | P | Model 2: with interaction term (OR [95% CI])c | P | Model 3: with interaction term and cephalosporin use (OR [95% CI])c | P |
|---|---|---|---|---|---|---|
| Age >55 yr | 1.99 (1.02–3.86) | 0.043 | 1.97 (1.02–3.5) | 0.043 | 1.81 (0.94–3.48) | 0.074 |
| Male | 2.19 (0.71–6.75) | 0.174 | 2.26 (0.74–6.94) | 0.152 | 2.36 (0.79–7.05) | 0.123 |
| Urinary tract abnormalities | 0.68 (0.30–1.53) | 0.352 | 0.71 (0.31–1.60) | 0.403 | 0.76 (0.34–1.73) | 0.519 |
| Diabetesd | 1.69 (0.78–3.44) | 0.193 | 1.36 (0.60–3.11) | 0.461 | 1.41 (0.62–3.22) | 0.409 |
| APN recurrenced,e | 1.69 (0.74–3.90) | 0.215 | 1.34 (0.52–3.42) | 0.545 | 1.88 (0.76–4.61) | 0.170 |
| Diabetes and APN recurrence | 4.24 (1.06–16.92) | 0.041 | 7.40 (1.96–27.95) | 0.003 | ||
| Antibiotic use within the previous yr | 4.69 (1.97–11.13) | <0.001 | 4.61 (1.95–10.96) | 0.001 | ||
| Cephalosporin use within the previous yr | 6.33 (2.01–19.92) | 0.002 |
ESBL, extended-spectrum β-lactamase.
OR, odds ratio; CI, confidence interval.
An interaction term between diabetes and APN recurrence was included in models 2 and 3 to estimate the joint effect of diabetes and APN recurrence.
APN, acute pyelonephritis.
In models 2 and 3, these variables (diabetes and APN recurrences) refer to diabetic patients without a history of APN recurrence and nondiabetic patients with a history of APN recurrence, respectively.
Clinical features and antibiotic therapies for CA-ESBL-EC APN and CA-non-ESBL-EC APN.
The severity of illness did not differ significantly between the CA-ESBL and CA-non-ESBL groups. More patients in the CA-ESBL group experienced symptom improvement within 3 days than did those in the CA-non-ESBL group (OR, 2.11; 95% CI, 1.16 to 3.85; P = 0.015). When stratified by APACHE II scores, the estimate did not change significantly (OR, 2.43; 95% CI, 1.30 to 4.57; P = 0.005). However, the proportions of early clinical failure at day 7 were similar between the cases and controls (OR, 0.91; 95% CI, 0.39 to 2.29; P = 0.909) (Table 4).
TABLE 4.
Clinical features and treatment of community-associated acute pyelonephritis caused by E. coli, according to ESBL productiona
| Clinical and treatment characteristicsb | CA-ESBL group (n = 75)c | CA-non-ESBL group (n = 225) | OR (95% CI)d | P |
|---|---|---|---|---|
| Clinical features | ||||
| Concomitant BSI | 24 (32.0) | 97 (43.1) | 0.53 (0.28–1.01) | 0.056 |
| APACHE II score (mean [SD]) | 7.7 (5.9) | 6.4 (5.3) | – | 0.071 |
| Septic shock | 6 (8.0) | 9 (4.0) | 2.08 (0.71–6.06) | 0.179 |
| Febrile days before hospitalization (median [SD]) | 2 (1–3) | 3 (1–3) | 0.047 | |
| Symptom improvement at day 3 | 57 (76.0) | 136 (60.4) | 2.11 (1.16–3.85) | 0.015 |
| Treatment | ||||
| Days to appropriate antibiotic therapy (median [SD]) | ||||
| From admission | 3 (0–4) | 0 (0–2) | <0.001 | |
| From symptom onset | 5 (3–7) | 3 (2–5) | 0.002 | |
| Appropriate antibiotics in <48 h | 35 (46.7) | 181 (80.4) | 0.21 (0.12–0.38) | <0.001 |
| Days of antibiotic therapy (median [SD]) | 15 (13–18) | 14 (13–16) | 0.088 | |
| Days of i.v. antibiotic therapy (median [SD]) | 8 (3–15) | 5 (4–6) | <0.001 | |
| Days of hospitalization (median [SD]) | 11 (6–16) | 7 (6–9) | <0.001 | |
| Treatment outcomes | ||||
| Early clinical success at day 7 | 68 (90.7) | 205 (91.1) | 0.91 (0.39–2.29) | 0.909 |
| Clinical cure | 71 (94.7) | 205 (74.3) | 1.76 (0.57–5.38) | 0.323 |
| Microbiological cure | 66 (88) | 151 (85.3) | 1.16 (0.51–2.65) | 0.730 |
The data represent the no. (%) of patients, unless otherwise specified. ESBL, extended-spectrum β-lactamase.
BSI, bloodstream infection; APACHE II, Acute Physiology and Chronic Health Evaluation II; IQR, interquartile range; i.v., intravenous.
CA, community associated.
OR, odds ratio; CI, confidence interval.
The empirical antibiotic regimens of the groups were also similar: ampicillin-sulbactam was the most commonly used agent in both the CA-ESBL and CA-non-ESBL groups (62.7% versus 59.1%, respectively), followed by aminoglycosides (17.3% versus 21.8%, respectively) and fluoroquinolones (5.3% versus 4.9%, respectively). Carbapenems were empirically used in 4 patients, with 2 in each group. The antibiotics used as definitive therapy in the CA-ESBL group included carbapenems (n = 42; 37 meropenem and 5 imipenem), fluoroquinolones (n = 11), aminoglycosides (n = 9), β-lactam/β-lactamase inhibitors (n = 9; 6 amoxicillin-clavulanate and 3 piperacillin-tazobactam), and trimethoprim-sulfamethoxazole (n = 4). In the CA-non-ESBL group, second- or third-generation cephalosporins (n = 64), ampicillin-sulbactam (n = 46), aminoglycosides (n = 54), and fluoroquinolones (n = 37) were commonly used as definitive therapy.
The patients in the CA-ESBL group were less likely to receive appropriate antibiotics within 48 h than were those in the CA-non-ESBL group (OR, 0.21; 95% CI, 0.12 to 0.38; P < 0.001). However, among those who initially received inappropriate antibiotics, the time to switch to an appropriate antibiotic was similar between the CA-ESBL group (median, 3 days [IQR, 2 to 4 days]) and CA-non-ESBL group (median, 2.5 days [IQR, 2 to 3 days]) (P = 0.280). The duration of intravenous antibiotic therapy and length of hospital stay were significantly longer in the CA-ESBL group than those in the CA-non-ESBL group (Table 4). In the CA-ESBL group, the duration of intravenous therapy was considerably longer in patients who received carbapenems (median, 14 days [IQR, 11 to 16 days]) than in those who received noncarbapenem antibiotics (median, 4 days [IQR, 2 to 7 days]) (P < 0.001). Subgroup analysis with nonbacteremic CA-APN caused by ESBL-EC showed that carbapenems were used for a longer duration than were intravenous noncarbapenem antibiotics (median, 13 days [IQR, 9 to 14 days] versus 3 days [IQR, 2 to 5 days], respectively; P < 0.001), even though the severity of illness (median APACHE II score, 6 [IQR, 3 to 9] versus 5 [IQR, 2 to 7], respectively; P = 0.152) and duration of symptoms (median, 2 days [IQR, 2 to 4 days] versus 3 days [2 to 4 days], respectively; P = 0.427) were not significantly different between the patients treated with carbapenems and those treated with noncarbapenem antibiotics.
Prognostic factors for treatment outcomes of CA-APN caused by E. coli.
With regard to final treatment outcomes, there were 24 clinical failures, including 5 deaths (1 CA-ESBL and 4 CA-non-ESBL) and 19 clinical relapses (3 CA-ESBL and 16 CA-non-ESBL). Among the 252 patients (75 CA-ESBL and 177 CA-non-ESBL) available for assessment of their microbiological outcomes during follow-up, 35 microbiological failures were observed (9 CA-ESBL and 26 CA-non-ESBL).
In the unadjusted analysis, the risks for clinical (HR, 0.69; 95% CI, 0.24 to 2.02; P = 0.496) and microbiological (HR, 0.89; 95% CI, 0.41 to 1.89; P = 0.753) failures were not significantly different between the groups (see the supplemental material). The multivariable analysis also showed that ESBL production did not negatively affect clinical and microbiological outcomes. The independent prognostic factors for adverse outcomes of CA-APN caused by E. coli were septic shock (HR, 6.16; 95% CI, 2.08 to 18.26; P = 0.001) and immunocompromised status (HR, 4.59; 95% CI, 1.72 to 12.33; P = 0.002) for clinical failure (Table 5) and the use of antibiotics within the previous year (HR, 4.84; 95% CI, 2.22 to 10.55; P < 0.001) and chronic kidney disease (HR, 2.51; 95% CI, 1.01 to 6.24; P = 0.048) for microbiological failure (Table 6). There was no significant interaction between septic shock and the receipt of appropriate antibiotics within 48 h. A delay of >48 h in the appropriate antibiotic therapy was not significantly associated with final clinical failure (HR, 1.73; 95% CI, 0.63 to 4.76; P = 0.288).
TABLE 5.
Multivariable analysis for prognostic factors associated with clinical failure in patients with community-associated acute pyelonephritis caused by E. coli
| Variable | Adjusted analysis |
Weighted adjusted analysis |
||
|---|---|---|---|---|
| HR (95% CI)a | P | HR (95% CI) | P | |
| ESBL productionb | 0.39 (0.12–1.30) | 0.126 | 0.35 (0.09–1.38) | 0.133 |
| Septic shock | 6.16 (2.08–18.26) | 0.001 | 8.29 (2.11–32.45) | 0.002 |
| Immunocompromisedc | 4.59 (1.72–12.33) | 0.002 | 3.63 (1.19–11.10) | 0.023 |
| Delay in appropriate antibiotic treatment >48 h | 1.73 (0.63–4.76) | 0.288 | 1.89 (0.55–6.52) | 0.316 |
| Age > 65 yr | 0.69 (0.29–1.68) | 0.420 | 0.52 (0.19–1.36) | 0.181 |
| Antibiotic use within the previous yr | 1.61 (0.60–4.37) | 0.347 | 1.23 (0.33–4.86) | 0.738 |
HR, hazards ratio; CI, confidence interval.
ESBL, extended-spectrum β-lactamase.
Immunocompromised patients included those who received transplantation, corticosteroid therapy (≥20 mg of prednisone or equivalent for ≥2 weeks) or immunosuppressants, and anticancer chemotherapy within 6 months prior to pyelonephritis diagnosis.
TABLE 6.
Multivariable analysis for prognostic factors associated with microbiological failure in patients with community-associated acute pyelonephritis caused by E. coli
| Variablea | Adjusted analysis |
Weighted adjusted analysis |
||
|---|---|---|---|---|
| HR (95% CI)b | P | HR (95% CI) | P | |
| ESBL productionb | 0.49 (0.21–1.12) | 0.091 | 0.51 (0.23–1.11) | 0.107 |
| Antibiotic use within the previous yr | 4.84 (2.22–10.55) | <0.001 | 5.42 (2.72–10.83) | <0.001 |
| Chronic kidney disease | 2.51 (1.01–6.24) | 0.048 | 2.91 (1.26–6.75) | 0.014 |
| Diabetes | 1.91 (0.91–4.02) | 0.089 | 1.88 (0.86–4.15) | 0.115 |
| Age >65 yr | 1.60 (0.62–3.37) | 0.215 | 1.39 (0.60–3.27) | 0.439 |
| Concomitant BSI | 1.30 (0.62–2.70) | 0.489 | 1.21 (0.57–2.58) | 0.621 |
| Urinary tract abnormalities | 1.13 (0.51–2.52) | 0.769 | 0.87 (0.39–1.94) | 0.726 |
ESBL, extended-spectrum β-lactamase; BSI, bloodstream infection.
HR, hazards ratio; CI, confidence interval.
To further control for potential confounding, we created a pseudopopulation by applying weights so that the demographic and baseline characteristics were well balanced between the groups (see the supplemental material). When weights were incorporated into a multivariable regression model, ESBL production was not associated with clinical (HR, 0.35; 95% CI, 0.09 to 1.38; P = 0.133) or microbiological failure (HR, 0.51; 95% CI, 0.23 to 1.11; P = 0.107) (Tables 5 and 6).
DISCUSSION
This study showed that older patient age, previous antibiotic use, and diabetes with recurrent APN were independent risk factors for acquiring ESBL-EC as a cause of CA-APN. The CA-ESBL and CA-non-ESBL groups did not differ in severity of illness or treatment outcomes. Even after adjusting for differences in the baseline characteristics between the groups, ESBL production was not associated with treatment outcomes of CA-APN caused by E. coli, indicating that ESBL production has no impact on the treatment outcomes of E. coli APN among patients without discernible HCA risk factors.
Despite the worldwide increase in ESBL-EC APN among patients without recent health care contact (2–4), the acquisition of ESBL-EC is difficult to predict in this patient population because it is mainly based on indicators of recent contact with health care environments (2). Similarly, the present study showed that the presence of ESBL-EC was difficult to predict based solely on the characteristics of patients with CA-APN because of the similarities of the baseline characteristics and clinical features. However, this study demonstrates that more concern should be directed toward diabetic patients with previous APN. Diabetes increases the risk of UTIs (17, 18) and their recurrence (19), which might place diabetic patients at risk of acquiring ESBL-EC as a result of repeated antibiotic exposure (1, 20–22). In this study, we found that the risk of acquiring ESBL-EC was markedly increased in diabetic patients with a history of APN, whereas either factor alone was not an independent risk factor. Therefore, we advise that physicians be aware of the possibility of ESBL-EC when diabetic patients present with recurrent APN, even those with no discernible HCA risk factors.
ESBL-producing Enterobacteriaceae are usually considered to cause more severe infections than non-ESBL producers (23–25), and the delayed administration of appropriate antibiotic therapy is often believed to lead to adverse outcomes. However, given that most previous studies focused on bloodstream infections from diverse sources and included heterogeneous populations encompassing patients with both CA and HCA infections in their analyses, the impact of ESBL on treatment outcomes might have been overemphasized. The present study demonstrates that ESBL production does not negatively affect the treatment outcomes of CA-APN caused by E. coli. Instead, immunocompromised status and septic shock were independent prognostic factors of clinical failure. A delay of >48 h in appropriate antibiotic therapy was not associated with clinical failure. This study also shows that patients with ESBL-EC are more likely to have early clinical improvement than are those with non-ESBL-EC, despite the delay in appropriate antibiotic therapy. This can partly be explained by the discrepancy between in vitro susceptibility and clinical efficacy due to the high antibiotic concentrations in urine. In addition, ESBL-EC strains have been reported to harbor multiple resistance genes, possibly causing a fitness cost leading to a decrease in virulence potential (26–28). In this study, since concomitant bacteremia was more frequently observed in the non-ESBL-EC group, patients with non-ESBL-EC infections might have longer symptom duration than those with ESBL-EC. Based on the study results, it is suggested that carbapenems should not be considered an initial empirical therapy for CA-APN when E. coli is suspected, unless patients present with septic shock. Given that noncarbapenem antibiotics are reportedly as efficacious as carbapenems for ESBL-EC APN if they are active in vitro (13), it would be prudent to choose noncarbapenem antibiotics to which ESBL-EC is susceptible as empirical therapy for CA-APN in cases where ESBL-EC is suspected, such as in diabetic patients with recent and multiple episodes of recurrent APN.
Although the clinical implications of ESBL-EC were less serious in this patient population, ESBL production might have caused a considerable economic burden because of longer hospitalizations in the CA-ESBL group. In this study, lengthened hospitalization appeared to be due to both delayed appropriate antibiotic therapy and prolonged carbapenem intravenous therapy without careful consideration of the severity of illness and clinical response to therapy. Antibiotic therapy could have been de-escalated from carbapenems to noncarbapenem antibiotics shown to be active in vitro. This finding implies that the therapeutic strategy for ESBL-EC infections is uncertain and that physicians are still reluctant to adopt short-term therapy for ESBL-EC APN. Additionally, the limited availability of oral antibiotic agents against ESBL-EC might have led to the prolonged use of carbapenems. Therefore, it is important to continue the search for oral antibiotics that are effective against ESBL-EC among the currently available oral agents, and more efforts should be made to develop new oral agents against ESBL-EC.
This study identified previous use of antibiotics as an independent prognostic factor for microbiological failure. Recent reports have also demonstrated that previous antibiotic exposure is associated with adverse outcomes in Gram-negative bacteremia (29–31), suggesting that the risk of developing antibiotic resistance to multiple antibiotics and the consequent delay in appropriate therapy might lead to adverse outcomes. In addition, previous antibiotic use can be a surrogate for the severity of comorbid conditions. Thus, its association with treatment outcome might have been confounded. However, in addition to these possible explanations, there seemed to be another pathway between previous antibiotic use and treatment outcomes. In this study, the estimate of the association did not change even after adjusting for potential confounding variables. The effect of previous antibiotic exposure on treatment outcome requires further elucidation.
This study has several limitations. This was a retrospective study at a single center; thus, information bias is inevitable and the generalizability of our findings might be limited. At this center, it is part of standard clinical practice to collect detailed information about past medical history and medications in patients with APN, so the medical records should have contained complete data on past medical histories and medications. However, there may still have been information bias, because patients may not have been able to recall all antibiotics used during the preceding year. Consequently, the effect of previous antibiotic use might have been underestimated in this study. Because of the limited generalizability of our findings, the study results should be interpreted in the context of ESBL-EC epidemiology in the community as well as the population characteristics. Despite these limitations, our study clarifies the characteristics of CA-APN caused by ESBL-EC and will help guide treatment strategies for this infection.
In conclusion, this study shows that it is difficult to identify patients likely to have CA-APN caused by ESBL-EC based solely on patient characteristics. Only old age, previous antibiotic use, and diabetes with recurrent APN are causes for suspicion. However, ESBL production by itself is not associated with adverse treatment outcomes, regardless of a delay in appropriate antibiotic therapy. Therefore, the empirical use of carbapenems can be deferred in this patient population until susceptibility test results are available. In an effort to initiate appropriate antibiotic therapy in a timely manner, physicians should be keenly aware of the local epidemiology of E. coli and be up to date with the current susceptibility patterns. More importantly, health care-associated risk factors should carefully be assessed in patients in the community presenting with APN to identify patients with CA-APN in the first place, and therapeutic strategies should be tailored based on risk assessment. Furthermore, concerted efforts should be made to obtain more evidence about the antibiotic treatment for CA-APN caused by ESBL-EC, such as the effective duration of therapy or the efficacy of noncarbapenem antibiotics, including oral agents.
Supplementary Material
Footnotes
Supplemental material for this article may be found at http://dx.doi.org/10.1128/AAC.04821-14.
REFERENCES
- 1.Rodríguez-Baño J, Alcala JC, Cisneros JM, Grill F, Oliver A, Horcajada JP, Tortola T, Mirelis B, Navarro G, Cuenca M, Esteve M, Pena C, Llanos AC, Canton R, Pascual A. 2008. Community infections caused by extended-spectrum beta-lactamase-producing Escherichia coli. Arch Intern Med 168:1897–1902. doi: 10.1001/archinte.168.17.1897. [DOI] [PubMed] [Google Scholar]
- 2.Ben-Ami R, Rodriguez-Bano J, Arslan H, Pitout JD, Quentin C, Calbo ES, Azap OK, Arpin C, Pascual A, Livermore DM, Garau J, Carmeli Y. 2009. A multinational survey of risk factors for infection with extended-spectrum beta-lactamase-producing Enterobacteriaceae in nonhospitalized patients. Clin Infect Dis 49:682–690. doi: 10.1086/604713. [DOI] [PubMed] [Google Scholar]
- 3.Doi Y, Park YS, Rivera JI, Adams-Haduch JM, Hingwe A, Sordillo EM, Lewis JS Jr, Howard WJ, Johnson LE, Polsky B, Jorgensen JH, Richter SS, Shutt KA, Paterson DL. 2013. Community-associated extended-spectrum β-lactamase-producing Escherichia coli infection in the United States. Clin Infect Dis 56:641–648. doi: 10.1093/cid/cis942. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Park SH, Choi SM, Lee DG, Kim J, Choi JH, Kim SH, Kwon JC, Yoo JH. 2011. Emergence of extended-spectrum β-lactamase-producing Escherichia coli as a cause of community-onset bacteremia in South Korea: risk factors and clinical outcomes. Microb Drug Resist 17:537–544. doi: 10.1089/mdr.2011.0072. [DOI] [PubMed] [Google Scholar]
- 5.Kim B, Kim J, Seo MR, Wie SH, Cho YK, Lim SK, Lee JS, Kwon KT, Lee H, Cheong HJ, Park DW, Ryu SY, Chung MH, Ki M, Pai H. 2013. Clinical characteristics of community-acquired acute pyelonephritis caused by ESBL-producing pathogens in South Korea. Infection 41:603–612. doi: 10.1007/s15010-013-0441-z. [DOI] [PubMed] [Google Scholar]
- 6.Kung CH, Ku WW, Lee CH, Fung CP, Kuo SC, Chen TL, Lee YT. 2013. Epidemiology and risk factors of community-onset urinary tract infection caused by extended-spectrum beta-lactamase-producing Enterobacteriaceae in a medical center in Taiwan: a prospective cohort study. J Microbiol Immunol Infect pii:S1684-1182(13)00151-5. doi: 10.1016/j.jmii.2013.08.006. [DOI] [PubMed] [Google Scholar]
- 7.Aguilar-Duran S, Horcajada JP, Sorli L, Montero M, Salvado M, Grau S, Gomez J, Knobel H. 2012. Community-onset healthcare-related urinary tract infections: comparison with community and hospital-acquired urinary tract infections. J Infect 64:478–483. doi: 10.1016/j.jinf.2012.01.010. [DOI] [PubMed] [Google Scholar]
- 8.Cheong HS, Kang CI, Kwon KT, Heo ST, Wi YM, Kim ES, Lee JS, Ko KS, Chung DR, Lee NY, Song JH, Peck KR. 2007. Clinical significance of healthcare-associated infections in community-onset Escherichia coli bacteraemia. J Antimicrob Chemother 60:1355–1360. doi: 10.1093/jac/dkm378. [DOI] [PubMed] [Google Scholar]
- 9.Friedman ND, Kaye KS, Stout JE, McGarry SA, Trivette SL, Briggs JP, Lamm W, Clark C, MacFarquhar J, Walton AL, Reller LB, Sexton DJ. 2002. Health care-associated bloodstream infections in adults: a reason to change the accepted definition of community-acquired infections. Ann Intern Med 137:791–797. doi: 10.7326/0003-4819-137-10-200211190-00007. [DOI] [PubMed] [Google Scholar]
- 10.Horcajada JP, Shaw E, Padilla B, Pintado V, Calbo E, Benito N, Gamallo R, Gozalo M, Rodriguez-Bano J, ITUBRAS Group, Grupo de Estudio de Infección Hospitalaria (GEIH), Sociedad Española de Enfermedades Infecciosas y Microbiología Clínica (SEIMC) . 2013. Healthcare-associated, community-acquired and hospital-acquired bacteraemic urinary tract infections in hospitalized patients: a prospective multicentre cohort study in the era of antimicrobial resistance. Clin Microbiol Infect 19:962–968. doi: 10.1111/1469-0691.12089. [DOI] [PubMed] [Google Scholar]
- 11.Knaus WA, Draper EA, Wagner DP, Zimmerman JE. 1985. APACHE II: a severity of disease classification system. Crit Care Med 13:818–829. doi: 10.1097/00003246-198510000-00009. [DOI] [PubMed] [Google Scholar]
- 12.Foxman B. 2010. The epidemiology of urinary tract infection. Nat Rev Urol 7:653–660. doi: 10.1038/nrurol.2010.190. [DOI] [PubMed] [Google Scholar]
- 13.Park SH, Choi SM, Chang YK, Lee DG, Cho SY, Lee HJ, Choi JH, Yoo JH. 2014. The efficacy of non-carbapenem antibiotics for the treatment of community-onset acute pyelonephritis due to extended-spectrum β-lactamase-producing Escherichia coli. J Antimicrob Chemother 69:2848–2856. doi: 10.1093/jac/dku215. [DOI] [PubMed] [Google Scholar]
- 14.Clinical and Laboratory Standards Institute. 2007. Performance standards for antimicrobial susceptibility testing; 17th informational supplement. Approved standard M100-S17. Clinical and Laboratory Standards Institute, Wayne, PA. [Google Scholar]
- 15.Stuart EA. 2010. Matching methods for causal inference: a review and a look forward. Stat Sci 25:1–21. doi: 10.1214/09-STS313. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Xu S, Ross C, Raebel MA, Shetterly S, Blanchette C, Smith D. 2010. Use of stabilized inverse propensity scores as weights to directly estimate relative risk and its confidence intervals. Value Health 13:273–277. doi: 10.1111/j.1524-4733.2009.00671.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.de Lastours V, Foxman B. 2014. Urinary tract infection in diabetes: epidemiologic considerations. Curr Infect Dis Rep 16:389. doi: 10.1007/s11908-013-0389-2. [DOI] [PubMed] [Google Scholar]
- 18.Hirji I, Guo Z, Andersson SW, Hammar N, Gomez-Caminero A. 2012. Incidence of urinary tract infection among patients with type 2 diabetes in the UK General Practice Research Database (GPRD). J Diabetes Complications 26:513–516. doi: 10.1016/j.jdiacomp.2012.06.008. [DOI] [PubMed] [Google Scholar]
- 19.Mody L, Juthani-Mehta M. 2014. Urinary tract infections in older women: a clinical review. JAMA 311:844–854. doi: 10.1001/jama.2014.303. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Soraas A, Sundsfjord A, Sandven I, Brunborg C, Jenum PA. 2013. Risk factors for community-acquired urinary tract infections caused by ESBL-producing Enterobacteriaceae–a case-control study in a low prevalence country. PLoS One 8:e69581. doi: 10.1371/journal.pone.0069581. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Kim Y, Wie SH, Chang UI, Kim J, Ki M, Cho YK, Lim SK, Lee JS, Kwon KT, Lee H, Cheong HJ, Park DW, Ryu SY, Chung MH, Pai H. 2014. Comparison of the clinical characteristics of diabetic and non-diabetic women with community-acquired acute pyelonephritis: a multicenter study. J Infect 69:244–251. doi: 10.1016/j.jinf.2014.05.002. [DOI] [PubMed] [Google Scholar]
- 22.Rodríguez-Baño J, Navarro MD, Romero L, Martinez-Martinez L, Muniain MA, Perea EJ, Perez-Cano R, Pascual A. 2004. Epidemiology and clinical features of infections caused by extended-spectrum beta-lactamase-producing Escherichia coli in nonhospitalized patients. J Clin Microbiol 42:1089–1094. doi: 10.1128/JCM.42.3.1089-1094.2004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Rottier WC, Ammerlaan HS, Bonten MJ. 2012. Effects of confounders and intermediates on the association of bacteraemia caused by extended-spectrum β-lactamase-producing Enterobacteriaceae and patient outcome: a meta-analysis. J Antimicrob Chemother 67:1311–1320. doi: 10.1093/jac/dks065. [DOI] [PubMed] [Google Scholar]
- 24.Lee S, Song do Y, Cho SH, Kwon KT. 2014. Impact of extended-spectrum beta-lactamase on acute pyelonephritis treated with empirical ceftriaxone. Microb Drug Resist 20:39–44. doi: 10.1089/mdr.2013.0075. [DOI] [PubMed] [Google Scholar]
- 25.Marchaim D, Gottesman T, Schwartz O, Korem M, Maor Y, Rahav G, Karplus R, Lazarovitch T, Braun E, Sprecher H, Lachish T, Wiener-Well Y, Alon D, Chowers M, Ciobotaro P, Bardenstein R, Paz A, Potasman I, Giladi M, Schechner V, Schwaber MJ, Klarfeld-Lidji S, Carmeli Y. 2010. National multicenter study of predictors and outcomes of bacteremia upon hospital admission caused by Enterobacteriaceae producing extended-spectrum beta-lactamases. Antimicrob Agents Chemother 54:5099–5104. doi: 10.1128/AAC.00565-10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Lavigne JP, Blanc-Potard AB, Bourg G, Moreau J, Chanal C, Bouziges N, O'Callaghan D, Sotto A. 2006. Virulence genotype and nematode-killing properties of extra-intestinal Escherichia coli producing CTX-M beta-lactamases. Clin Microbiol Infect 12:1199–1206. doi: 10.1111/j.1469-0691.2006.01536.x. [DOI] [PubMed] [Google Scholar]
- 27.Andersson DI. 2006. The biological cost of mutational antibiotic resistance: any practical conclusions? Curr Opin Microbiol 9:461–465. doi: 10.1016/j.mib.2006.07.002. [DOI] [PubMed] [Google Scholar]
- 28.Beceiro A, Tomas M, Bou G. 2013. Antimicrobial resistance and virulence: a successful or deleterious association in the bacterial world? Clin Microbiol Rev 26:185–230. doi: 10.1128/CMR.00059-12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Ku NS, Kim YC, Kim MH, Song JE, Oh DH, Ahn JY, Kim SB, Kim HW, Jeong SJ, Han SH, Kim CO, Song YG, Kim JM, Choi JY. 2014. Risk factors for 28-day mortality in elderly patients with extended-spectrum β-lactamase (ESBL)-producing Escherichia coli and Klebsiella pneumoniae bacteremia. Arch Gerontol Geriatr 58:105–109. doi: 10.1016/j.archger.2013.07.002. [DOI] [PubMed] [Google Scholar]
- 30.Johnson MT, Reichley R, Hoppe-Bauer J, Dunne WM, Micek S, Kollef M. 2011. Impact of previous antibiotic therapy on outcome of Gram-negative severe sepsis. Crit Care Med 39:1859–1865. doi: 10.1097/CCM.0b013e31821b85f4. [DOI] [PubMed] [Google Scholar]
- 31.Micek S, Johnson MT, Reichley R, Kollef MH. 2012. An institutional perspective on the impact of recent antibiotic exposure on length of stay and hospital costs for patients with Gram-negative sepsis. BMC Infect Dis 12:56. doi: 10.1186/1471-2334-12-56. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.