Antimicrobial Non-Susceptibility of Escherichia coli from Outpatients and Patients Visiting Emergency Rooms in Taiwan

Longitudinal nationwide surveillance data on antimicrobial non-susceptibility and prevalence of extended-spectrum β-lactamases (ESBLs) as well as AmpC β-lactamases producers among Escherichia coli from different sources in the community settings are limited. Such data may impact treatment practice. The present study investigated E. coli from outpatients and patients visiting emergency rooms collected by the Taiwan Surveillance of Antimicrobial Resistance (TSAR) program. A total of 3481 E. coli isolates were studied, including 2153 (61.9%) from urine and 1125 (32.3%) from blood samples. These isolates were collected biennially between 2002 and 2012 from a total of 28 hospitals located in different geographic regions of Taiwan. Minimum inhibitory concentrations (MIC) were determined using methods recommended by the Clinical Laboratory Standards Institute (CLSI). The prevalence and factors associated with the presence of ESBL and AmpC β-lactamase-producers were determined. Significant increases in non-susceptibility to most β-lactams and ciprofloxacin occurred during the study period. By 2012, non-susceptibility to cefotaxime and ciprofloxacin reached 21.1% and 26.9%, respectively. The prevalence of ESBL- and AmpC- producers also increased from 4.0% and 5.3%, respectively, in 2002–2004, to 10.7% for both in 2010–2012 (P < 0.001). The predominant ESBL and AmpC β-lactamase genes were CTX-M and CMY-types, respectively. Non-susceptibility of urine isolates to nitrofurantoin remained at around 8% and to fosfomycin was low (0.7%) but to cefazolin (based on the 2014 CLSI urine criteria) increased from 11.5% in 2002–2004 to 23.9% in 2010–2012 (P <0.001). Non-susceptibility of isolates from different specimen types was generally similar, but isolates from elderly patients were significantly more resistant to most antimicrobial agents and associated with the presence of ESBL- and AmpC- β-lactamases. An additional concern is that decreased ciprofloxacin susceptibility (MIC 0.12–1 mg/L) was as high as 25% in isolates from all age groups, including those from pediatric patients. Our data indicated that there is a need to re-evaluate appropriate treatment selection for community-acquired infections in Taiwan. Identification of community reservoirs of multidrug-resistant E. coli is also warranted.

as high as 25% in isolates from all age groups, including those from pediatric patients. Our data indicated that there is a need to re-evaluate appropriate treatment selection for community-acquired infections in Taiwan. Identification of community reservoirs of multidrugresistant E. coli is also warranted.

Isolate collection
E. coli isolates were collected as part of the TSAR program conducted biennially between 2002 and 2012 (corresponding to TSAR III-VIII). Isolates were collected from medical centers and regional hospitals, including a total number of 28 hospitals located in the 4 geographic regions of Taiwan following previously described protocol [23]. All isolates were stored at −70°C. Only E. coli clinical isolates from outpatients and patients visiting emergency rooms were enrolled in the present study. The TSAR project was approved by the Research Ethics Committee of National Health Research Institutes, Taiwan (EC960205 and EC1010602-E). Written informed consent was not obtained because the study only used bacterial isolates recovered from clinical samples taken as part of standard care and patient information was anonymized and de-identified prior to analysis.

Isolate identification
Clinical isolates reported to be E. coli by hospitals were first subcultured to blood agar and MacConkey agar plates for purity check. Species confirmation was based on colony morphology, positive spot indole and β-glucuronidase [24,25]. For isolates not typical of E. coli colony morphology or if negative for either of the above biochemical reactions, Vitek II GN cards were used (bioMérieux, Marcy l'Etoile, France).

Antimicrobial susceptibility testing (AST)
Minimum inhibitory concentrations (MICs) were determined using reference broth microdilution method following the guidelines of the manufacturer and CLSI [21]. Sensititre customdesigned plates were used from TSAR III (2002) to TSAR VI (2008), and the standard GNX2F plates were used in TSAR VII (2010) and TSAR VIII (2012) [ThermoFisher Scientific (formerly Trek Diagnostics), East Grinstead, UK]. All isolates were subcultured twice on sheep blood agar plate from -70°C before AST. Quality control was performed using E. coli ATCC 25922, E. coli ATCC 35218, Klebsiella pneumoniae ATCC 700603, and Pseudomonas aeruginosa ATCC 27853.

Detections of ESBL, AmpC β-lactamase, and carbapenemase genes
The CLSI ESBL confirmatory test was performed on all isolates with aztreonam, ceftazidime, or cefotaxime MIC 2 mg/L using cefotaxime and ceftazidime disks with and without clavulanate [21]. These ESBL screening test-positive isolates were also subject to detection of ESBL and/or AmpC β-lactamases genes by multiplex PCR following previously described primers and protocols [26,27]. Isolates non-susceptible to carbapenem were subject to carbapenemase PCR using published primers [28].

Data analysis
Susceptibility interpretation analysis was made using the WHONET software [29]. Intermediate susceptibility and resistance were grouped together as "non-susceptibility". Duplicate isolates were excluded before analysis. TSAR III and IV (2002)(2003)(2004), TSAR V and VI (2006)(2007)(2008), and TSAR VII and VIII (2010-2012) were grouped as Periods I, II, and III, respectively, in the related analysis. Categorical variables were compared using chi-square test or Fisher's exact test (if the number was less than 10). If statistical difference was obtained when the tested categorical variables with three different levels were compared, post-hoc analysis was performed to identify which level was significantly different from the others. Trends were analyzed using chi-square for trend analysis. Multivariable logistic regression analysis was performed to assess the variables (including study year, specimen type, and patient age group) among ESBL or AmpC β-lactamase-producers vs.-non-producers. SAS 9.2 (SAS Institute, Cary, NC, USA) was used for the above analyses. A 2-tailed P value less than 0.05 was considered statistically significant.  (2012), respectively. By geographic regions, a total of 1091, 979, 1004, and 407 isolates were from the northern, central, southern, and eastern parts of Taiwan, respectively. The most common specimen type was urine (2153, 61.9%), followed by blood (1125, 32.3%), and others (203, 5.8%). The age of the source patients was missing in 74 patients. The mean ± standard deviation age of the remaining 3407 patients was 57.4 ± 23.6 years (y), with 6.9% (234) being pediatric patients ( 18 y, mean 4.3 y), of whom the majority (195) were younger than 10 years of age, 45.9% being adult (19-64 y, mean 45.0 y), and 47.3% being elderly ( 65 y, mean 77 y).

Non-susceptibility to different antimicrobial agents over the study period
For ease of comparison, we grouped the 6 study years into 3 periods, with period I, II, III corresponding to 2002-2004, 2006-2008, and 2010-2012, respectively. The rates of overall nonsusceptibility of the isolates to various antimicrobial agents from the 3 studied periods are listed in Table 1. Significant increases in non-susceptibility to ciprofloxacin and several β-lactams, including first to fourth generation cephalosporins, aztreonam, and cefoxitin, were noted. In contrast, non-susceptibility to tetracycline and trimethoprim/sulfamethoxazole (SXT) decreased but still remained high, at 59.1% (2006-2008) and 47.2% (2010-2012), respectively. Rates of non-susceptibility to all 4 generations of cephalosporins and ciprofloxacin by each study year are delineated in Fig 1, which highlights the increasing rates (all P < 0.001) of nonsusceptibility over the study years. For example, the overall non-susceptibility to cefotaxime and ciprofloxacin increased from 8.2% and 15% in 2002 to 21.1% and 26.9% in 2012, respectively. Non-susceptibility to cefazolin of isolates from urine also increased from 11.5% in 2002 to 23.9% in 2012. The activities of the different antimicrobial agents against all 3481 E. coli isolates are presented in S1 File.
Isolates from different specimen types generally had similar rates of non-susceptibility, with except that isolates from blood had lower non-susceptibility than urine isolates to piperacillin (66.0% vs. 75.2%, P = 0.028), piperacillin/tazobactam (2.4% vs. 5.5%, P < 0.001), and ciprofloxacin (18.2% vs. 21.9%, P = 0.039) ( Table 2). However, significant difference in rates of non-susceptibility to most tested agents existed in isolates from the three age groups, with isolates from   Table 2). Although isolates from elderly patients had the highest rates of non-susceptibility to most agents, the increase of non-susceptibility to extended-spectrum cephalosporins occurred in isolates from all age groups. Of note, rates of non-susceptibility of pediatric isolates were either higher than or similar to those from adult patients, except ciprofloxacin, which differed slightly (10.7% vs. 15.2%, P = 0.067). In addition, isolates from pediatric patients had significantly higher rate of non-susceptibility to amoxicillin/clavulanate than those from the adult patients (30.7% vs. 22.4%, P = 0.014).
Rates of non-susceptibility also differed significantly between isolates from the 4 geographic regions, with those from central and southern regions having the higher rates of non-susceptibility while those from the northern and eastern regions were lower ( Table 2). For example, non-susceptibility to cefotaxime was 11.5%, 19.1%, 16.1%, and 10.6%, and to gentamicin was 21.9%, 27.4%, 27.6%, and 23.6% in isolates from the northern, central, southern, and eastern regions, respectively. These differences are also reflected in the prevalence of ESBL and AmpC-producers (Table 2).

Statistical analysis to identify factors associated with ESBL and AmpCβ-lactamase producers
By univariate analysis, factors associated with carriage of ESBL and AmpC included age, geographic origin of bacterial isolates, and TSAR study period (Table 4). By multivariate analysis, age (elderly vs. pediatric patients), geographic origin (central Taiwan vs. eastern Taiwan) and later study years (2010 and 2012 vs. 2002) remained independent factors associated with ESBL gene carriage (Table 4). Age (elderly vs. pediatric patients), geographic origin (southern, central Taiwan vs. eastern Taiwan) and study period (2006 to 2012 vs. 2002) were also independent factors associated with carriage of AmpC β-lactamase (Table 4).

Discussion
The present study analyzed data from a nationwide surveillance program and revealed that over the past decade (2002 to 2012), significant increase in non-susceptibility to most β-lactam agents and ciprofloxacin occurred in E. coli isolates from community settings (outpatients and patients visiting emergency rooms) in Taiwan. In addition, non-susceptibility of our 2010-2012 isolates to several agents were higher than those from similar time periods in western countries, including cefazolin (59.7% against 27.7% in Canada), cefotaxime (19.6% against lower than 10% in Canada, USA, and Europe, and lower than 7.4% in UK based on non-susceptibility to 3 rd generation cephalosporins, gentamicin (24,1% against 9.8% in Canada and lower than 6% in UK), and trimethoprim/sulfamethoxazole (47.2% against lower than 27.4% in Canada and Europe), even though some of those studies included isolates from inpatients [8,19,20,30].
The increased non-susceptibility to nearly all broad-spectrum β-lactams (except imipenem) in our isolates correlated with the increased ESBL-and AmpC β-lactamase-producers, which increased to 10.7% for both, respectively, in 2010-2012. Although increased prevalence of ESBL-producing E. coli has been noted in several countries [31], the 10.7% prevalence of ESBL Table 3. Non-susceptibility a (%) to key agents among E. coli isolates with and without ESBL or AmpC β-lactamases genes.
In the present study, most ESBL genes were CTX-M types (96.3%), and the predominant AmpC β-lactamases genes was the CMY type (90.7%). Prior studies have shown that the most prevalent ESBL and AmpC enzymes types worldwide, including Taiwan, were CTX-M and CMY types, respectively [34][35][36]. Since only a few isolates had low level non-susceptibility to ertapenem and/or imipenem. Carbapenemase (IMP-type) was detected only in 1 isolate (data not shown). Thus carbapenemase producing E. coli appeared not to be a major problem in community settings in Taiwan yet.
CLSI lowered the susceptibility breakpoints of third-generation cephalosporins for Enterobacteriaceae in 2010 to facilitate the identification of ESBL-and/or AmpC-β-lactamases Table 4. Factors associated with carriage of ESBL and AmpC β-lactamases genes in E. coli.

Strata
ESBL AmpC β-lactamase -producing isolates [37]. By applying the revised breakpoints, nearly all (98.7%) of our ESBL and/or AmpC positive isolates were non-susceptible to cefotaxime, yet only 77.6% of them were non-susceptible to ceftazidime. Thus cefotaxime is a better predictor for ESBLs and/or AmpC β-lactamase-producers than ceftazidime in terms of sensitivity (98.7% vs. 77.6%). These results are similar to prior studies [23,38,39], and reflects the predominance of CTX-M since this enzyme hydrolyzes cefotaxime more efficiently than ceftazidime [40]. Of noteworthy is that 16 of the 52 PCR ESBL-pos/AmpC-pos isolates in the present study tested negative by the CLSI phenotypic ESBL confirmatory test so the overall false-negative phenotypic ESBL confirmatory test rate was 6.6% (16/242). Co-carriage of ESBL and AmpC βlactamase genes has become common among Enterobacteriaceae in our region [23,36]. It has been demonstrated that the presence of AmpC β-lactamases could mask the underlying ESBLs by traditional phenotypic confirmatory test [41]. Our result echoes this finding and lends support to using the revised CLSI breakpoints to avoid missing ESBL-producers by the phenotypic ESBL confirmatory test.
Based on the 2014 revised cefepime breakpoints, over 90% of the ESBL producers are nonsusceptible to cefepime including 23.6% with cefepime MIC in the susceptible dose dependent range, indicating that higher dosing regimens are needed if cefepime was used for these isolates [21]. Significant difference on non-susceptibility to cefazolin in the non-urine (60.4%) vs. urine isolates (23.9%) from the 2010-2012 isolates exists using the 2014 cefazolin non-urine and urine breakpoints. This information is important in antibiotic selection for uncomplicated UTI also.
Non-susceptibility to ciprofloxacin in our 2010-2012 isolates (25.1%) was much higher than that of community E. coli isolates from the 15.5% (UTI hospital and community isolates combined from 2013) reported in UK [19] and other countries in Europe (0.5-7.6% in 2007-2008) [20], but was lower than the rate (32.4%) found by a multicenter surveillance of 2009-2011 E. coli from inpatients with community-acquired UTI in USA [8]. Whether such differences were dependent on patient age is unknown since isolates from elderly patients in our study had the highest rate of ciprofloxacin non-susceptibility, which reached 34.5% in 2010-2012.
Isolates from recent years were more likely to carry ESBL and AmpC β-lactamase genes, and these isolates were most likely to be found in elderly patients as well as those from central or southern Taiwan (compared to those from eastern Taiwan). Elderly patients probably had more recent exposure to medical care and antibiotics, resided in long-term care facilities, which have been associated with acquisition of drug-resistant bacteria [15,[42][43][44]. This in turn would explain why isolates from elderly patients were more resistant than those from adult and pediatric patients. The reason for isolates from central and southern Taiwan being more likely to carry ESBL and AmpC β-lactamase genes, and thus more non-susceptible to most agents, needs further investigation.
We also showed that isolates from pediatric patients had higher rates of non-susceptibility to amoxicillin/clavulanic acid than those from adults and elderly. In addition, the pediatric isolates had similarly high rate (near 25%) of decreased ciprofloxacin susceptibility (DCS, ciprofloxacin MIC 0.12-1 mg/L), probably due to single mutation in the quinolone-resistance-determining region of the drug target or the presence of plasmid-mediated quinolone resistance determinants [45][46][47]. This is a cause for additional concern since the effectiveness of FQ in treating patients infected with DCS isolates may be compromised [47][48][49]. The reasons for the high rates of DCS and amoxicillin/clavulanic acid in the pediatric population needs further study.
Prior reports have pointed out that inappropriate empirical therapy was associated with added mortality among critically ill patients with severe infection [4,50]. Therefore it has been suggested that the empirical antibiotic therapy for patients with severe infection should cover more than 90% spectrum of all the possible bacterial etiology [51]. For community uncomplicated UTI, the suggested prevalence of resistance threshold above which the agent is not recommended was 20% for SXT and 10% for fluoroquinolones [5]. Given our study results, cefotaxime might no longer be a reasonable empirical antibiotic for patients with severe community-acquired infections in Taiwan if E. coli is a probable etiologic agent since around 20% of the isolates from 2010-2012 were cefotaxime-resistant. Similarly, fluoroquinolones are also not appropriate empirical therapy option since non-susceptibility to ciprofloxacin reached 25% in 2010-2012. Instead, ertapenem or cefepime should be considered as the drugs of choice in such situation.
A recent study indicated that nitrofurantoin, an old antimicrobial agent, retained its activity against multidrug resistant E. coli isolates from the USA, with resistance observed ranging 2.1% to 24.1% of isolates resistant to three to five antimicrobial agents, respectively, with an overall resistance of only 1.2% [9]. Our data showed that nitrofurantoin non-susceptibility in urinary E. coli isolates from Taiwan remained at lower than 8% over the years. Non-susceptibility to fosfomycin was low (lower than 1%) in our 2012 urinary E. coli isolates. Therefore these 2 older antimicrobial agents could be options for the treatment of uncomplicated cystitis as suggested [5].
One limitation of the present study is that since we only have minimum clinical information on the source patients, we could not determine if other independent factors, such as prior antimicrobial use or hospitalization, exist for ESBL and/or AmpC-producers [15,42,43]. However, we did show age (elderly) to be an independent factor for ESBL and AmpC-producers, thus prior hospitalization and more antimicrobial use are likely associated factors. Another limitation is that our isolates were collected biennially during a three months period. However, these isolates were from a total of 28 hospitals located in all four geographical regions of Taiwan, and 25 of them, including 11 medical centers and 14 regional hospitals, participated in all 6 rounds of TSAR between 2002 and 2012. Therefore the results described here can be a representation of the E. coli from community settings in Taiwan.
In conclusion, this decade-long multicenter surveillance revealed significant increase in non-susceptibility to several broad spectrum antimicrobial agents, including cefotaxime and ciprofloxacin, occurred in E. coli from community settings in Taiwan. Therefore, multidrug resistance is prevalent not only in healthcare settings but also in the community in our region. The prevalence of ESBL-and AmpC β-lactamase-producing E. coli also increased significantly, especially in elderly patients. Whether such increase was due to clonal spread or horizontal gene transfer requires further study. The additional high rates of decreased ciprofloxacin susceptibility in E. coli isolates from all age groups are also a cause for concern. Identification of community reservoirs of the resistant bacteria may help to halt their spread. Finally, our data indicated that there is a need to re-evaluate appropriate treatment selection for communityacquired infections.