Risk Factors for Amoxicillin-Clavulanate Resistance in Community-Onset Urinary Tract Infections Caused by Escherichia coli or Klebsiella pneumoniae: The Role of Prior Exposure to Fluoroquinolones

Background: High rates of amoxicillin-clavulanate (AMC) resistance among Enterobacterales isolated from urinary tract infections (UTIs) were observed in our area. The aim of this study was to identify risk factors associated with AMC resistance in patients with community-onset UTI in emergency departments (EDs). Methods: A retrospective study was performed of all ED patients with positive urine cultures for Escherichia coli or Klebsiella pneumoniae in a Spanish tertiary-care hospital. Results: 330 urine cultures in all were included: 261 (79.1%) for E. coli and 69 (20.90%) for K. pneumonia. Rates of AMC resistance were 14.94% and 34.78%, respectively. UTI was clinically confirmed in 212 (64.24%) cases. Previous antimicrobial exposure was independently associated with AMC resistance development in E. coli and K. pneumoniae urinary isolates (OR = 2.94, 95% CI = 1.55–5.58). Analyses of infected patients revealed that previous exposure to fluoroquinolones (OR = 3.33, 95% CI = 1.10–10.12, p = 0.034) and to AMC (OR = 5.68, 95% CI = 1.97–16.44, p = 0.001) was significantly associated with isolation of AMC-resistant strains. Conclusions: Prior antibiotic exposure, particularly to AMC or fluoroquinolones, was the only independent risk factor associated with development of AMC resistance in E. coli and K. pneumoniae urinary isolates from patients attending the ED.


Introduction
Urinary tract infection (UTI) is a common reason for attending hospital emergency departments (ED). In the United States, UTIs represented more than 3 million visits to ED [1], making it one of the most common reasons for prescribing empirical antibiotics [2,3]. In Spain, medical emergency visits due to infectious diseases represent more than 14% of all attended emergencies. UTIs are the second most frequent infection with 2517 visits/year (22.1%), surpassed only by respiratory tract infections with 3678 (32.3%) [2]. Escherichia coli (E. coli) is the most frequent (>60%) uropathogen responsible for this clinical condition, followed by Klebsiella pneumoniae (K. pneumoniae) (10%) [4,5].
There is general concern about increasing antibiotic resistance among uropathogens, partly related to the extensive use of antimicrobials [5]. This has reduced the empirical therapeutic options in many settings [6]. In our area, current guidelines no longer recommend quinolones or trimethoprim/sulfamethoxazole for empirical use [4]. Amoxicillin/clavulanate (AMC) is the most prescribed antimicrobial (in 26.5% of the cases) to treat community-onset UTIs (CO-UTI) [2]; in Spain, a striking increase in resistance of almost 15% has been observed in UTI isolates in just 5 years [7]. Taking into account that; following the EUCAST guidelines, the MIC established to define Enterobacterales as sensitive or resistant to AMC depends on whether it is a complicated or uncomplicated UTI; this data should be analyzed with caution before recommending no to use AMC in this setting [8].
E. coli and K. pneumoniae are of particular concern due to their high capacity to acquire extended-spectrum β-lactamases (ESBLs) that confer bacterial resistance to most β-lactam antibiotics [6,9]. ESBLs production often show cross-resistance to other group of antibiotics, being particularly worrisome the close relationship between ESBL production and fluoroquinolone resistance [10]. Conversely, the association between the use of fluoroquinolones and the development of ESBLs has been less evident.
Our institution periodically updates its antimicrobial guidelines, based on local microbiological data, within the framework of the Antimicrobial Stewardship Program. We noticed an upwards trend in AMC resistance in Enterobacterales urinary isolates, which could jeopardize its empirical use. Since AMC was one of the recommended empirical antimicrobial agents at our hospital for treatment of CA-UTI, we made it a priority to review the appropriateness of its empirical use in the ED setting. The aim of the present work was to investigate risk factors associated with AMC resistance in community-onset, ED-diagnosed UTIs caused by E. coli and K. pneumoniae.

Epidemiological Data
A total of 330 urine cultures positive for E. coli and K. pneumoniae were collected during the study period. The median age of patients was 75.0 (47-85) years, and most were female (237 (71.82%)). Almost two-thirds of patients had at least one comorbidity (201 (60.91%)); of these, 56.72% (114/201) had more than two comorbidities. The most prevalent being diabetes and renal failure, present in 85 (25.76%) and 76 (23.03%) patients, respectively.
Most of the cases were community-acquired (229 (69.39%)). Fifty-nine (17.88%) patients had a history of recurrent UTI. Fifty-four (16.36%) patients carried an indwelling urinary device, of which 31 (57.41%) were urinary catheters. Overall, 49 (14.85%) patients had been admitted to hospital and 110 (33.33%) had received a course of antibiotics in the 3 months prior to ED admission.

Risk Factors for AMC Resistance in Infected Patients
Factors related to AMC resistance were analyzed in the 'confirmed UTI' patient subset, and shown in Table 3. Age, Charlson index, SAPS II score, diabetes, COPD, malignant disease, immunosuppression, indwelling urinary catheter use, previous hospital stay, previous use of AMC, carbapenems or any other antibiotic, complicated UTI, and CO-HCA were significantly associated with AMC resistance.  Multivariate analysis showed that the two variables independently associated with AMC resistance in this subgroup of patients were: previous exposure to fluoroquinolones (OR = 3.33, 95% CI = 1.10-10.12, p = 0.034) and previous exposure to AMC (OR = 5.68, 95% CI = 1.97-16.44, p = 0.001) ( Table 4).

Discussion
Increasing antibiotic resistance among uropathogens [5] can lead to inappropriate empirical treatment of UTI. In this challenging scenario, guidelines based on local antimicrobial susceptibility profiles are essential to select appropriate empirical antibiotic treatments. Given the high rates of AMC resistance detected in Enterobacterales (EUCAST criteria) while updating our local UTI guidelines, we conducted a retrospective study to assess risk factors for AMC resistance in urinary enterobacterial isolates from adult patients attending the ED. As expected, the isolates were mainly E. coli and K. pneumoniae.
AMC nonsusceptibility was detected in 14.94% of E. coli and in 28.96% of K. pneumoniae. A multicentric study performed in Spain showed that AMC-resistance in E. coli was due to OXA-1 β-lactamase production in one fourth of the isolates [11], followed by hyperproduction of penicillinases, hyperproduction of the chromosomic AmpC β-lactamase, production of acquired AmpC, and production of inhibitor-resistant TEM. Currently, carbapenemases may also contribute, especially OXA-48 type in our environment [9]. Thus, AMC resistance may be acquired by several mechanisms that interplay in a complex epidemiological background, including clonal spread, dissemination of different bla genes, and mutations in individual isolates as a response to selective antimicrobial pressure [11].
Risk factors for antibiotic resistance in UTI isolates in community settings have been evaluated previously [10,12,13]. None of these reports focused specifically on risk factors for AMC resistance. In our report, previous antibiotic exposure was the only independent risk factor associated with AMC resistance in both E. coli and K. pneumoniae isolates. When analyzed only the subgroup of patients with symptomatic UTI, previous use of either AMC or fluoroquinolones were significantly associated.
It has been widely reported that antimicrobial use exerts selective pressure that leads to an increase in antimicrobial resistance [10,13,14]. It is reasonable to suppose that individual exposure to AMC results in selection of AMC-resistant bacteria. V. Leflon-Guibot et al. specifically studied this association in a recent study in hospitalized patients with confirmed UTI [15]. They conclude that exposure to AMC was a risk factor for selection of AMC-resistant E. coli isolates. They did not however evaluate previous use of other antibiotics [15].
The association between fluoroquinolones and the development of resistance in β lactams may not be as straightforward as in the case of AMC. Fluoroquinolones have been associated with the clonal expansion of different multidrug-resistant (MDR) bacteria, ranging from methicillin-resistant Staphylococcus aureus to ESBL-carrying Enterobacterales such as the E. coli ST131-H30 subclone [10,12,16,17]. Double-serine fluoroquinolone resistance mutations in the DNA gyrase and topoisomerase IV genes are present in the major successful clones [16,17]. These specific mutations seem to impact fitness favorably, providing clones with increased spreading capacity in fluoroquinolone environments. Other mutations in minor clones do not seem to contribute to increased fitness [16,17]. Fluoroquinolone resistance has also been involved in the spread of other enterobacterial clones, regardless of their resistance profile to other antibiotics [17,18]. Hence, the association between fluoroquinolone resistance and AMC observed in our study may be partly driven by clonal spread, despite the low number of ESBL carriers and other MDR isolates. The high rates of fluoroquinolone resistance in our isolates may support the previously reported role of these broad-spectrum antibiotics in bacterial evolutionary success [16,17,19].
Mechanisms other than chromosomal mutations may also have contributed. Plasmidmediated quinolone-resistance can carry other antimicrobial resistance genes such as ESBLs [16,17,20], while multidrug chromosomal efflux pumps have the ability to actively remove different families of antibacterial drugs, including fluoroquinolones [14,17,21,22]. Indeed, previous fluoroquinolone use has been associated with the emergence of multidrug efflux pumps [14,23], which cause cross-resistance between sublethal concentrations of fluoroquinolones and other antibiotic families, including AMC [24].
Moreover, β-lactamase production and porin decrease are well-recognized mechanisms of resistance against β-lactam antibiotics among Gram-negative bacteria. However, we did not perform molecular studies to analyze the presence of any of these fluoroquinolone resistance determinants.
According to the EUCAST guidelines, the MIC breakpoints to categorize AMC susceptibility in Enterobacterales causing UTIs are either ≤8 mg/L or ≤32 mg/L, depending on whether the infection leads to a complicated or an uncomplicated UTI, respectively [8]. However, microbiologists usually do not have enough clinical information in this respect and the episode is often considered a complicated UTI. In the present study, the AMC MIC value was above 8 mg/L in 17.88% of enterobacterial isolates and were consequently all categorized as resistant. However, 10.30% (34/330) had MICs of either 16 or 32 mg/L, and could have been categorized as susceptible or resistant to AMC depending on whether the physician considered the UTI episode as complicated or uncomplicated. AMC doses of 250 mg/125 mg have been related to maximum urine concentrations between 647 and 1547 mg/L and 150 and 439 mg/L for amoxicillin and clavulanate, respectively [25]. These values would explain why current AMC doses of 1 g/125 mg could be sufficient for the treatment of most UTIs caused by these strains. Unfortunately, these "in-between" MIC values may overestimate AMC resistance and lead many clinicians to refrain from using this antibiotic in certain clinical situations where it could be used safely. Replacing AMC with broader spectrum antibiotics would have a negative impact on the ecological background and therefore on stewardship programs. Clinical uncertainty is further increased by the absence of consensus on definitions of complicated UTI in the IDSA and ESCMID guidelines. The medical literature uses many different definitions of complicated UTI [26]. Such heterogeneity makes it even more difficult to apply the EUCAST criteria, although this committee has recently included definitions of these processes [27]. In this context, it is essential to reach international consensus on the standard definitions for establishing complicated and uncomplicated UTI. Furthermore, we consider that having two different AMC breakpoints for enterobacterial isolates in UTI can be misleading in some clinical settings. This EUCAST criterion would probably benefit from revision.
The present study has some limitations. First, it was a retrospective data analysis, and dependent on the accuracy of the clinical histories collected. Second, we did not conduct a clonal study, which would have been useful to confirm a possible association between AMC resistance and certain clones. Molecular studies would also have been useful to study resistance determinants. Third, it was a single-centre study, conducted in the ED setting and focused only on E. coli and K. pneumoniae isolates; our results may not be generalizable to other geographical areas, healthcare settings or infections caused by other bacteria. Fourth, prior amoxicillin/clavulanate use in the subgroup of infected patients, variable that is independently associated with AMC resistance, show a rather large 95% CI. Therefore, the association should be verified in new studies with a larger sample size. Finally, it was conducted over a short period (2 months) with no follow-up of the clinical consequences of the observed resistance. Large-sample, prospective multicenter studies are needed to better understand the extent and outcomes of infections caused by AMC-resistant Gram-negative bacteria.

Study Design
This was a retrospective observational cohort study performed over a two-month period (November-December 2017) at the Hospital del Mar, a 420-bed tertiary-care university teaching hospital in Barcelona (Spain). Using computer-generated microbiological data, all adult patients (>18-years-old) attending the ED during this period with positive urine cultures for E. coli or K. pneumoniae were included. Cases were defined as patients with AMC-resistant isolates; controls were those with AMC-susceptible isolates. This paper was written following the STROBE guidelines for observational studies [28].

Data Collection and Definitions
Demographic, clinical, and epidemiological data were collected by examining the hospital and primary care medical and nursing records. Comorbidities considered were diabetes mellitus, chronic obstructive pulmonary disease (COPD), cardiovascular disease, chronic kidney disease, chronic liver disease, neurological disorders, malignant disease, human immunodeficiency virus infection, and immunosuppression. Patients were considered to have malignant disease when malignancy was diagnosed within the previous 5 years or were receiving specific oncological therapy. Patients were considered immunosuppressed when they received chemotherapy, radiotherapy, systemic corticosteroids at a dose higher than 10 mg of prednisone per day or equivalent, or other immunosuppressive agents in the 3 months prior to ED admission. The Charlson comorbidity index [29] and the simplified acute physiologic score (SAPSII) [30] were recorded upon ED arrival. Variables related to the patient's urinary history, such as urinary tract abnormalities, recurrent UTI, or urinary device use in the preceding month were also recorded. Patients with at least 3 UTI episodes in the previous 12 months were considered recurrent UTI. Regarding site of acquisition, community-onset healthcare-associated (CO-HCA) episodes were defined as those fulfilling any of the Friedman criteria [31], and otherwise as community-acquired. Prior antibiotic exposure was defined as administration of antibiotics for more than 48 h in the previous 3 months, and prior hospital stay as previous hospitalization in the last 3 months before ED admission.
Definition of infection was established according to the Centers for Disease Control and Prevention criteria [32]. All episodes were reviewed retrospectively by two authors (J.M.-C. and A.D.M.) and classified as confirmed UTI or asymptomatic bacteriuria. Controversies were double-checked by a third investigator (S.G.-Z.). Among confirmed UTIs, cystitis was considered to have uncomplicated UTI; other infections (prostatitis, pyelonephritis, and orchiepididymitis) were considered as complicated UTI.

Microbiological Data
Urine cultures were performed as part of clinical routine following standard laboratory procedures. Cultures with growth yielding ≥10 5 colony-forming units/mL of a single bacterial type in a urine sample collected midstream were considered positive. Pyuria was defined as ≥10 white blood cells/mm 3 . Only one urine culture per patient was included.
Antibiotic susceptibility testing was performed by microdilution using MicroScan panels (Beckman Coulter, Brea, CA, USA) in an automated WalkAway system (Beckman Coulter). Results were interpreted following European Committee for Antimicrobial Susceptibility Testing (EUCAST) 2019 guidelines [8]. Isolates with AMC minimum inhibitory concentration (MIC) values >8 mg/L were considered resistant. Phenotypic detection of ESBLs and carbapenemases was in accordance with EUCAST recommendations [33]. Molecular confirmation of carbapenemases was by real time PCR, using LightMix ® modular carbapenemase kits (TIB Molbiol, Berlin, Germany) in a LightCycler 480 II instrument (Roche Diagnostics, Rotkreuz, Switzerland).

Statistical Analysis
Continuous variables were expressed as median and interquartile range (IQR); categorical variables were expressed as counts and percentages. Continuous variables were compared using the Mann-Whitney U-test and categorical variables by the Fisher's exact test.
An analysis of all patients was made for AMC resistance, and a separate one for the 'confirmed UTI' subgroup. Univariate analysis was performed first. Variables showing statistically significant differences between susceptible and resistant microorganisms were included in the multivariate model. Clinically relevant variables considered by the literature to be potential confounders were also studied in the multivariate model, even if they did not show significant differences in the previous step. Multivariate analysis was carried out using a logistic regression model to estimate the risk factors involved in the development of antibiotic resistance. Strength of association was expressed as odds ratios (ORs). A two-sided p-value <0.05 was statistically significant. All analyses were performed using STATA v. 15.1.

Conclusions
Prior to antibiotic exposure, particularly to AMC and fluoroquinolones, within the previous three months, was independently associated with development of AMC-resistant E. coli and K. pneumoniae strains in urinary isolates of patients attending the ED. These results could be used to assess the appropriateness of AMC as empirical treatment in similar UTI cases attending the emergency department A general reduction in the use of AMC and fluoroquinolones should be encouraged to reduce antimicrobial resistance.