Antimicrobial resistance of major clinical pathogens in South Korea, May 2016 to April 2017: first one-year report from Kor-GLASS

The Korean government established an antimicrobial resistance (AMR) surveillance system, compatible with the Global AMR Surveillance System (GLASS): Kor-GLASS. We describe results from the first year of operation of the Kor-GLASS from May 2016 to April 2017, comprising all non-duplicated clinical isolates of major pathogens from blood, urine, faeces and urethral and cervical swabs from six sentinel hospitals. Antimicrobial susceptibility tests were carried out by disk diffusion, Etest, broth microdilution and agar dilution methods. Among 67,803 blood cultures, 3,523 target pathogens were recovered. The predominant bacterial species were Escherichia coli (n = 1,536), Klebsiella pneumoniae (n = 597) and Staphylococcus aureus (n = 584). From 57,477 urine cultures, 6,394 E. coli and 1,097 K. pneumoniae were recovered. Bloodstream infections in inpatients per 10,000 patient-days (10TPD) were highest for cefotaxime-resistant E. coli with 2.1, followed by 1.6 for meticillin-resistant Sta. aureus, 1.1 for imipenem-resistant Acinetobacter baumannii, 0.8 for cefotaxime-resistant K. pneumoniae and 0.4 for vancomycin-resistant Enterococcus faecium. Urinary tract infections in inpatients were 7.7 and 2.1 per 10TPD for cefotaxime-resistant E. coli and K. pneumoniae, respectively. Kor-GLASS generated well-curated surveillance data devoid of collection bias or isolate duplication. A bacterial bank and a database for the collections are under development.


Introduction
Antimicrobial resistance (AMR) is a growing burden in both clinical and socioeconomic context owing to the high morbidity and prolonged hospitalisation of patients that causes elevated medical and societal costs because of loss of productivity [1]. The World Health Organization launched the Global AMR Surveillance System (GLASS) in 2015 [2] as a core global action plan addressing this issue. The standardised GLASS manual allowed an overview of global AMR rates through international comparison. An AMR surveillance system in South Korea, the Korean AMR Monitoring System (KARMS), had been operated between 2002 and 2015 by Korean Centers for Disease Control and Prevention (KCDC) [3,4]. KARMS played an important role in notifying the high AMR rates in South Korea, urging the government to develop a national action plan. However, this system had limitations. Firstly, the antimicrobial susceptibility testing (AST) methods were not well harmonised across the participating clinical laboratories, affecting reliability. Secondly, duplicated isolates were not sufficiently filtered out, which could lead to an overestimation of the national AMR rates. Finally, the epidemiological interpretation of the study was limited because of insufficient clinical data.
From KARMS, the KCDC established an improved AMR surveillance system compatible with the GLASS, named Kor-GLASS [5]. The Kor-GLASS manual was customised from that of GLASS: (i) three bacterial species from blood specimens were added, namely Enterococcus faecalis and Enterococcus faecium to monitor vancomycin resistance and Pseudomonas aeruginosa to monitor carbapenem resistance and (ii) more target antimicrobial agents for AST were included to investigate multidrug resistance by species.
We have operated the Kor-GLASS for one year since May 2016 and report here the first one-year assessment until April 2017.

Collection of isolates and clinical data from sentinel hospitals
The six sentinel hospitals collected bacterial isolates and clinical data, and all tests were performed in a central laboratory [5]. We collected all non-duplicated clinical isolates of Staphylococcus aureus, Streptococcus pneumoniae, Ent. faecalis, Ent. faecium, Acinetobacter spp., and P. aeruginosa from blood, Escherichia coli and Klebsiella pneumoniae from both blood and urine, Salmonella spp. from both blood and faeces, Shigella spp. from faeces and Neisseria  Patients are shown grouped by origin of infections. Detailed numbers can be found in Table 1 in the Supplement.
gonorrhoeae from urethral and cervical swabs. Urine isolates were collected through semi-quantifying culture of urine samples following the criteria: (i) ≥10 4 colony-forming units (CFU)/mL single-species growth of either E. coli or K. pneumoniae and (ii) ≥10 5 CFU/mL of E. coli or K. pneumoniae in growth of mixed species [6]. We recorded epidemiological data including age, sex, infection origin (hospital origin (HO) or community origin (CO)) and admission types (outpatient department (OPD), general ward (GW), intensive care unit (ICU)) of all patients from whom blood, urine, stool or genital cultures were taken during study period. HO was defined when the specimen was taken from a patient hospitalised for two or more calendar days overall, including the hospitalisation days in another healthcare facility before transfer. CO

Figure 2
Occurrence of bloodstream and urinary tract infections per 10,000 patient-days, by target pathogen,     was defined when the specimen was taken either from an outpatient or from a patient hospitalised for less than 2 calendar days.

Microbiological analysis in the central laboratory
Bacterial species were re-checked in the central laboratory using a Bruker Biotyper (Bruker Daltonics GmbH, Bremen, Germany) and/or by nucleotide sequence analysis of the 16S rDNA or rpoB (for Acinetobacter sp p.). AST was carried out by disk diffusion, Etest, and broth microdilution and agar dilution methods following the Clinical and Laboratory Standards Institute guidelines [7]. Antimicrobial susceptibility phenotypes were categorised according to Magiorakos et al. [8] with a few modifications: • fully susceptible (DS): susceptible to all tested drugs; • drug-resistant (DR): non-susceptible to one or two drug classes; • multidrug-resistant (MDR): non-susceptible to three or more antimicrobial classes; • extensively drug-resistant (XDR): susceptible to two or fewer antimicrobial classes; • pandrug-resistant (PDR): not susceptible to any antimicrobial class.
Double-blinded inter-laboratory parallel tests were conducted monthly for randomly selected isolates in order to evaluate proficiency of the results from central and the national KCDC reference laboratories.
Among the 217 Ent. faecium blood isolates, 90.3% (n = 196) were resistant to ampicillin and 29.0% (n = 63) and 18.4% (n = 40) were resistant to vancomycin and teicoplanin, respectively. For these drugs, resistance rates in HO isolates were higher than those in CO isolates (Figures 3 and 4). Forty of the 63 vancomycin-resistant Ent. faecium (VREFM) were also resistant to teicoplanin and 14 were intermediate, while nine remained susceptible to the drug. In addition, 27.6% (n = 60) and 10.1% (n = 22) of all Ent. faecium were high-level resistant to gentamicin and streptomycin, respectively. Two thirds of the Ent. faecium blood isolates were MDR (62.7%, n = 136) and one third was DR (33.2%, n = 72). All VREFM isolates were categorised as MDR.
All Str. pneumoniae blood isolates were susceptible (23/28) or intermediate (n = 5) to penicillin. The five penicillin-intermediate isolates were also nonsusceptible to other drugs. All 28 isolates were susceptible to levofloxacin. Erythromycin resistance was observed in 21 of the 28 isolates.

Enterobacteriaceae
Ampicillin resistance was identified in 65.6% (1,007/1,536) of E. coli blood isolates and decreased to 24.2% (n = 372) when sulbactam was supplemented. Rates of resistance to cefotaxime, ceftazidime and cefepime were 34.7% (n = 533), 11.3% (n = 173) and 21.0% (n = 322), respectively. Carbapenem-nonsusceptible isolates were rarely identified. The resistance rate to ciprofloxacin was 39.5% (n = 606) and that to amikacin was low at 0.7% (n = 11). Colistinresistant isolates were seldom identified, with 0.2% (n = 3). Resistance rates in E. coli urine isolates were similar to those in blood isolates. One isolate was resistant to all three carbapenems and six were nonsusceptible only to ertapenem. Colistin resistance was detected in 14 (0.2%) of the 6,394 E. coli urine isolates. HO isolates had higher resistance rates to most of the tested antimicrobials than CO isolates, and this difference was much greater in blood isolates than in urine isolates. The vast majority (98.8%, 2,616/2,648) of cefotaxime-non-susceptible E. coli isolates were MDR and all XDR isolates (0.2%, 15/7,930) were nonsusceptible to both carbapenems and cefotaxime.
Among 44 Salmonella blood isolates, one isolate was resistant to cefotaxime and intermediate to ceftazidime. All the isolates remained susceptible to imipenem. None was resistant to ciprofloxacin, however, 10 were intermediate to the drug, and colistin resistance was observed in four. Five of the 77 Salmonella stool isolates were cefotaxime-resistant, and four of those five were also resistant to ceftazidime. Similar to the blood isolates, all stool isolates were also susceptible to imipenem and 12 isolates were intermediate to ciprofloxacin. Three isolates were resistant to colistin.
The proportion of AMR in the 188 A. baumannii blood isolates was higher than 70% for all tested drugs (Figures 3 and 4) and the AMR proportion in HO isolates was ca. twice that of CO isolates. Minocycline, tigecycline and colistin were still active in 93.1% (n = 175), 85.6% (n = 161) and 99.5% (n = 187) of isolates, respectively. Among the A. baumanniiblood isolates, 76.1% (n = 143) and 13.8% (n = 26) were XDR and MDR, respectively, and all but two MDR isolates were imipenem-non-susceptible. NBA blood isolates had a markedly lower proportion of AMR than A. baumannii.

The estimated prevalence of major AMR pathogens
In general, the AMR proportion of major pathogens for ICU patients was higher than for GW and OPD patients (Supplement: supplementary figure 3). BSI occurrence by major AMR pathogen in inpatients per 10TPD ( Figure  5) had the highest mean value of 2.1 for cefotaximeresistant E. coli (CXREC), followed by 1.6 for MRSA, 1.1 for imipenem-resistant A. baumannii, 0.8 for cefotaxime-resistant K. pneumoniae (CXRKP), 0.4 for VREFM, 0.2 for penicillin-resistant Ent. faecalis (PREFA) and 0.2 for imipenem-resistant P. aeruginosa. As shown in Figure 5, the ratio of BSI occurrence of the major AMR pathogens was much higher in ICU than on GW. Mean UTI occurrence in inpatients per 10TPD was higher for CXREC at 7.7 than for CXRKP at 2.1. The ratio of UTI occurrence in ICU vs GW was higher for CXRKP at 2.4 than for CXREC at 1.5.

Discussion
De Kraker et al. [9] described an increasing trend in the overall incidence of bacteremia caused by the top five pathogens (E. coli, Sta. aureus, Str. pneumoniae, Ent. faecalis and Ent. faecium) between 2002 and 2008, based on the European Antimicrobial Resistance Surveillance Network (EARS-Net; https://ecdc. europa.eu/en/about-us/networks/disease-networksand-laboratory-networks/ears-net-about) database. Predominance of E. coli among pathogens causing BSIs is a common phenomenon worldwide. We assessed the relative incidence of BSIs caused by E. coli compared with other species: The relative ratio of E. coli to Sta. aureus in Kor-GLASS was moderate at 2.6, which was lower than the 3.2 observed in Norway [10], similar to the ratios of 2.4 in Taiwan [11] and 2.2 in Netherlands [12], and higher than the ratios of 1.2 in Japan [13], 1.2 in Vietnam [14], 1.0 in Greece [15] and 1.1 in Malawi [16]. The relative ratio of E. coli to K. pneumoniae in South Korea was also moderate at 2.6, lower than in Northern European countries (5.4 in the Netherlands and 5.0 in Norway), similar to 2.5 in Japan and 3.0 in Taiwan, and higher than 1.0 in Vietnam. Str. pneumoniae was frequently isolated from blood in European countries (relative ratio of E. coli to Str. pneumoniae: 3.2 in the Netherlands, 3.1 in Norway and 4.1 in Spain), but rarely identified in South Korea (ratio: 54.9) which was similar to Asian countries (ratio: 18.3 in Vietnam and 31.8 in Taiwan). The Salmonella spp. is still a major pathogen causing BSI in developing countries such as Vietnam (relative ratio of E. coli to Salmonella spp.: 1.4) and Malawi (ratio: 0.2), however, this species was seldom identified in South Korea (ratio: 34.9), similar to the ratios 28.1 in Spain and 14.3 in Taiwan.
Sta. aureus was the second most common pathogen causing BSI following E. coli. BSI caused by Sta. aureus occurred in 2.8 inpatients per 10TPD in our study, which is 1.8-fold more than the 1.6 inpatients per 10TPD measured in 2011 in a previous prospective multi-center study in South Korea [17]. Since the surveillance system of that study is not compatible to Kor-GLASS, the potential increase in Sta. aureus-BSI needs to be further followed up. Similarly, we observed an increased incidence of MRSA-BSI at 1.6 inpatients per 10TPD compared with 1.2 inpatients per 10TPD in 2011. This incidence was 2.5 times higher than the 0.62 inpatients per 10TBD reported in a Canadian surveillance study in 2014 [18]. The high prevalence of MRSA in blood isolates in South Korea has decreased from 72% in 2013 to 66% in 2015 as measured by KARMS [4] and further to 54.3% in 2016 according to Kor-GLASS data. EARS-Net reported similarly that the percentage of MRSA has decreased from 18.1% in 2013 to 13.7% in 2016 [9]. In addition, MRSA occurred more frequently in ICU patients than in GW patients (5.8 vs 1.1), indicating that MRSA is a problem in ICUs.
Enterococci have become a major cause of BSIs globally, owing to their intrinsic resistance to various antibiotics and their enormous ability to acquire resistance to antimicrobials. Ent. faecium and Ent. faecalis were the fourth and sixth most common pathogens causing BSI in inpatients in our study, with 1.2 and 0.8 BSI occurrences per 10TPD, respectively, and they caused over four times more BSI in ICUs than on GWs. Ent. faecium exhibited remarkably higher resistance rates than Ent. faecalis to ampicillin (90.3% vs 0.6%) and vancomycin (29.1% vs 0.6%), resulting in the identification of 1.3-fold more Ent. faecium than Ent. faecalis, similar to the 1.8-fold difference found in Spain [19]. A reversed ratio was observed in the Netherlands (0.8:1) [14], and Japan (0.7:1) [12], countries that have low rates of AMR to these drugs. The vancomycin resistance rate (29.1%) in Ent. faecium was higher than that of teicoplanin (18.4%) in South Korea, although all isolates were vanA-positive. This might be caused by the dissemination of clones with a vanA genotype-VanA phenotype along with a vanA genotype-VanD phenotype following inactivation of vanYand vanZ in the vanA operon by rearrangement of Tn1546 [20]. PREFA was also a common BSI-causative AMR pathogen, especially in ICUs. Clinical impacts of the penicillin resistance need to be further investigated.
Both E. coli and K. pneumoniae exhibited higher resistance rates to cefotaxime than to ceftazidime owing to the dissemination of CTX-M-type extended-spectrum β-lactamases (ESBLs). The difference in the resistance rates to these drugs was greater in E. coli than in K. pneumoniae, which could have two possible causes: higher prevalence K. pneumoniae compared with E. coli (i) of CTX-M group 1 ESBLs, which have an expanded hydrolytic activity to ceftazidime, and (ii) of SHV-type ESBLs, which hydrolyse both cefotaxime and ceftazidime [21,22]. Identification of carbapenemase-producing Enterobacteriaceae (CPEs) in our surveillance study seemed the tip of an iceberg of the notorious AMR pathogen disseminated in clinical settings. All CPEs we identified were K. pneumoniae carbapenemase (KPC) producers. Continuous monitoring and action plans for CPEs are required because various types of carbapenemases, including KPC, New Delhi metalloβ-lactamase and OXA-48-likes, have been introduced to South Korea during the last decade [23][24][25]. Colistin resistance in Enterobacteriaceae was rare in our study and none contained the mobile colistin resistance (mcr) genes mcr-1 to mcr-5 [26][27][28], although clinical Enterobactericeae isolates carrying the mcr-1 gene have already been reported in South Korea [29].
Amikacin non-susceptibility was a better indicator for XDR P. aeruginosa than imipenem-non-susceptibility. The carbapenemase-producing P. aeruginosa strains in South Korea harboured a class 1 integron carrying several resistance gene cassettes coding not only for a carbapenemase, but also for aminoglycoside-modifying enzymes [30]. Thus, most of the amikacin-nonsusceptible isolates were XDR. However, carbapenem resistance in P. aeruginosastrains was conferred not only through acquisition of a gene for carbapenemase, but also through loss or alteration of the intrinsic OprD porin or overproduction of the efflux pumps [31], which rarely confer resistance to amikacin to the bacterial hosts.
BSIs caused by A. baumannii strains mostly occurred in ICUs rather than on GWs [32]. The likely reason is that the A. baumannii BSIs originated predominantly from pulmonary infections associated with ventilators. The rate of carbapenem resistance in A. baumannii strains was markedly high at 89.9% and most of these strains were XDR. This result is in line with previous reports from South Korea, and the resistance was mostly associated with OXA-23 carbapenemase production [33].

Conclusion
The Kor-GLASS, which performs strain collection and centralised analysis, was launched and operated. It provided well-curated surveillance data devoid of collection bias or isolate duplication, including patient data associated with the bacterial cultures. In addition, frequency of infection occurrence by patient-days was estimated for comparison with foreign countries. The results of this surveillance helped plan national action in response to the high rate of drug resistance. A bacterial bank and a database for the collections are under development.