Strong correlation between the rates of intrinsically antibiotic-resistant species and the rates of acquired resistance in Gram-negative species causing bacteraemia, EU/EEA, 2016

Background Antibiotic resistance, either intrinsic or acquired, is a major obstacle for treating bacterial infections. Aim Our objective was to compare the country-specific species distribution of the four Gram-negative species Escherichia coli, Klebsiella pneumoniae, Pseudomonas aeruginosa and Acinetobacter species and the proportions of selected acquired resistance traits within these species. Method We used data reported for 2016 to the European Antimicrobial Resistance Surveillance Network (EARS-Net) by 30 countries in the European Union and European Economic Area. Results The country-specific species distribution varied considerably. While E. coli accounted for 31.9% to 81.0% (median: 69.0%) of all reported isolates, the two most common intrinsically resistant species P. aeruginosa and Acinetobacter spp. combined (PSEACI) accounted for 5.5% to 39.2% of isolates (median: 10.1%). Similarly, large national differences were noted for the percentages of acquired non-susceptibility to third-generation cephalosporins, carbapenems and fluoroquinolones. There was a strong positive rank correlation between the country-specific percentages of PSEACI and the percentages of non-susceptibility to the above antibiotics in all four species (rho > 0.75 for 10 of the 11 pairs of variables tested). Conclusion Countries with the highest proportion of P. aeruginosa and Acinetobacter spp. were also those where the rates of acquired non-susceptibility in all four studied species were highest. The differences are probably related to national differences in antibiotic consumption and infection prevention and control routines.


Introduction
Bloodstream infections constitute a major disease burden in Europe. Gram-negative bacteria such as Escherichia coli, Klebsiella pneumoniae, Pseudomonas aeruginosa and Acinetobacter species are the most common organisms involved in these infections. Data from the European Antimicrobial Resistance Surveillance Network (EARS-Net) show that the proportional contribution of each bacterial species to the total number of blood isolates varies considerably between countries [1]. In particular Acinetobacter spp. and P. aeruginosa seem to be proportionally far more commonly isolated in some European countries than in others. Several studies have previously made the same observation [2][3][4].
Antibiotic resistance is a major obstacle for treating these serious infections. Resistance can be both intrinsic, i.e. due to a species' innate ability to resist a particular antibiotic because of structural or functional characteristics, or acquired though a range of resistance mechanisms emerging either by mutation or acquisition of novel genes.
Among Gram-negative bacteria, intrinsic resistance varies markedly between species. For example, even wild-type isolates of K. pneumoniae are intrinsically resistant to penicillins whereas wild-type isolates of E. coli are susceptible to the same antibiotics. More importantly, every wild-type isolate of Acinetobacter spp. and P. aeruginosa is intrinsically resistant to numerous groups of antibiotics (e.g. aminopenicillin-β-lactamase inhibitor combinations, first-and second-generation cephalosporins, cefotaxime/ceftriaxone, ertapenem, trimethoprim, tetracyclines and older quinolones) [5], immediately excluding these as possible treatment alternatives. Acquisition of additional resistance traits can further reduce available treatment options, jeopardise the use of major remaining groups of antibiotics including β-lactams, fluoroquinolones and aminoglycosides. Bacteria can acquire multiple resistance mechanisms, leading to multidrug-resistant (MDR), extensively drug-resistant (XDR) or even pandrugresistant (PDR) isolates [6]. Thus the overall pattern of resistance as presented in the laboratory reports reflects the intrinsic resistance characteristics of the species combined with any additional resistance trait acquired by the isolate.
Similar to the differences in the distribution of Gramnegative species between European countries, EARS-Net has also documented large differences in the percentage of acquired antibiotic resistance [1,2]. However, there is no study addressing the possible link between the ranking of the various species of Gram-negative bacteria and the percentage of acquired resistances in these species. As an example, most publications on resistance in P. aeruginosa and Acinetobacter spp. only focus on acquired resistance mechanisms or on clonal expansion of resistant epidemic clones, and a few others describe the intrinsic resistance characteristics of these species [7][8][9][10]. Correlation between intrinsic and acquired resistance is mentioned only from a mechanistic perspective, such as the link between inducible and de-repressed production of chromosomal AmpC β-lactamase [8], but not from a statistical or epidemiological point of view.
The objective of the present study was to assess the association between the proportion of the two most common intrinsically resistant species (P. aeruginosa and Acinetobacter spp.) among the four major Gram-negative species (E. coli, K. pneumoniae, P. aeruginosa and Acinetobacter spp.) and the percentage of selective acquired resistance traits in these species. As the data source, we used data reported to EARS-Net for countries in the European Union (EU) and European Economic Area (EEA) in the year 2016.

Data source and inclusion criteria
EARS-Net is a surveillance network which collects and analyses data from routine antibiotic susceptibility testing (AST) of bacterial pathogens from all 28 EU countries (Austria, Belgium, Bulgaria, Croatia, Cyprus, the Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, the Netherlands, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden and the United Kingdom) and two EEA countries (Iceland and Norway). The network is coordinated by the European Centre for Disease Prevention and Control (ECDC). Only AST results for selected important antibiotics active against invasive bacterial infections are included in EARS-Net. The AST results are ascertained according to agreed protocols [1] and the general quality and comparability of the data are evaluated through an annual external quality assessment exercise distributed to the participating laboratories.
Data on E. coli, K. pneumoniae, P. aeruginosa, and Acinetobacter spp. isolates reported to EARS-Net for the year 2016 were extracted from The European Surveillance System (TESSy) database at ECDC. Data included isolates from blood and cerebrospinal fluid, both considered as markers of bloodstream infections. Data were de-duplicated to only include the first isolate per species, patient and year. For this study, only data from laboratories reporting observations for at least three of the four above-mentioned species were included. The study was limited to only include antibiotics commonly used for first-line treatment of bacteremia caused by Gram-negative species and routinely included in susceptibility testing in most local clinical laboratories in Europe. The AST information for the following antibiotic-species combinations were included in the study dataset: third-generation cephalosporins (ceftriaxone, ceftazidime or cefotaxime for E. coli, and K. pneumoniae; ceftazidime for P. aeruginosa), carbapenems (meropenem or imipenem, for all four species) and fluoroquinolones (ciprofloxacin, levofloxacin or ofloxacin for E. coli and K. pneumoniae; ciprofloxacin or levofloxacin for P. aeruginosa and Acinetobacter spp.) Isolates were considered as non-susceptible when reported as either intermediately susceptible (I) or resistant (R).
Of the 829 laboratories that reported data on any of the targeted species to EARS-Net for the year 2016 (total: 176,082 isolates), 749 fulfilled the inclusion criterion (total: 173,540 isolates) and were included in the final analysis.

Statistical analysis
The percentages of E. coli, K. pneumoniae, P. aeruginosa, and Acinetobacter spp. isolates among the total of isolates included in the study, as well as the sum of the percentages of P. aeruginosa and Acinetobacter spp. (PSEACI) were calculated for each country. Rank comparison of the percentages of acquired non-susceptibility in each of the four Gram-negative species with the percentage of PSEACI was performed using the Spearman's rank correlation coefficient to assess the monotonic relationships and limit impact of outliers.

Ethical statement
Antimicrobial resistance is listed as a special health issue in the EU case definitions for which ECDC routinely collects, analyses and disseminates surveillance data as stated by the Article 3 of its founding regulation. TESSy data are pseudonymised and processed for public interest in the area of public health. Approval of the study by an ethics committee was therefore not necessary.

Figure 1
Distribution of fully susceptible (S) and non-susceptible (I or R) isolates in four  Figure 1 shows, within each of the four Gram-negative species reported by the 30 countries, the distribution of isolates fully susceptible (S) and non-susceptible (I + R) to broad-spectrum β-lactams (third-generation cephalosporins for K. pneumoniae and E. coli, carbapenems for P. aeruginosa and Acinetobacter spp.) ( Figure 1A) and to fluoroquinolones for the four species ( Figure 1B). The sum of these non-susceptible isolates, expressed as the composite percentage of isolates intermediately susceptible and resistant to broad-spectrum β-lactams, ranged from 4.2% for Iceland and 6.6% for Norway to 15.7% for France and 18.1% for Spain, up to 57.4% for Bulgaria and 60.8% for Greece (Table  2). In Figure 1B,

Discussion
The intrinsic resistance profiles of different bacteria are reflected in the European Committee on Antimicrobial Susceptibility Testing (EUCAST) expert rules [5] as well as in the number of antimicrobial drugs that must be considered to define acquired MDR/XDR/ PDR patterns [6]. As a consequence, for Gram-negative bacteria, identification of the species isolated from a positive blood culture provides immediate information to the microbiologist and the clinician about which antimicrobials should not be used because of intrinsic resistance in that species. This is one of the major reasons for the development of rapid methods for bacterial species identification, such as matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass-spectrometry. In addition, acquired resistance is recognised as a major global public health issue [11,12] because it jeopardises the effectiveness of antimicrobial drugs that are normally active against intrinsically multi-susceptible species such as E. coli and because it further reduces the possibility to treat infections with species that are already intrinsically resistant to many antimicrobial drugs such as P. aeruginosa and Acinetobacter spp.
The results presented here, based on an analysis of EARS-Net data for the year 2016 and focusing on four major Gram-negative species (E. coli, K. pneumoniae, P. aeruginosa and Acinetobacter spp.) isolated from bloodstream infections, clearly show significant statistical association between the distribution of species, and the percentages of acquired nonsusceptibility to major antibiotic groups. All but one correlation were statistically significant, with Spearman's rank correlation coefficient being > 0.75 for 10 of the 11 pairs of variables tested, which is generally considered as indicating a strong correlation [13,14]. In short, the higher the proportion of the two most intrinsically resistant species P. aeruginosa and Acinetobacter spp., the higher the percentages of acquired non-susceptibility in E. coli, K. pneumoniae, P. aeruginosa and Acinetobacter spp., a point that directly indicates the burden of intrinsic resistance. Consequently, there were, at one extremity of the EU/EEA gradient, countries with a very low proportion of bloodstream infections caused by the two most intrinsically resistant species together with low percentages of isolates with acquired non-susceptibility in any of the four Gram-negative species. At the opposite extremity of the EU/EEA gradient, countries with high percentages of bloodstream isolates caused by the two most intrinsically resistant species had high percentages of isolates with acquired non-susceptibility in all the four species. The weak statistical link noted for carbapenem non-susceptibility in E. coli could be due to very low percentages of non-susceptibility reported from a majority of the countries [1].
Statistical association is not equal to causation and the correlations presented in this study do not mean that there is a cause-and-effect relationship between intrinsic and acquired resistance. However, we can hypothesise that the two major driving forces of antibiotic resistance, i.e. the use of antibiotics acting as a selective pressure on resistant bacteria and the spread of the selected antibiotic-resistant bacteria by cross-transmission between humans, animals and the environment, apply to both the intrinsically resistant species such as P. aeruginosa and Acinetobacter spp.
(and to a lesser extent K. pneumoniae), and to strains with acquired resistance (in the present study: resistance to β-lactams or fluoroquinolones). Antibiotic use has a strong selective effect on P. aeruginosa and Acinetobacter spp [15,16] and on strains with acquired resistance traits in many species [17][18][19][20]. Host-to-host cross-transmission of P. aeruginosa and Acinetobacter spp [21,22] and of strains with acquired resistance traits e.g. in enterobacteria [23,24] also plays a crucial role in the spread of resistance. Indeed, the available comparative data suggest that the EU/EEA countries that in our study had the highest proportion of intrinsically resistant species and percentages of isolates with acquired resistance, generally also have the highest antibiotic consumption in humans (in particular broad-spectrum β-lactams and fluoroquinolones) and the lowest levels of preventive measures against cross-transmission of microorganisms in hospitals such as consumption of alcohol hand rub solutions, proportion of rooms with a single bed and staffing of infection control teams [2,25,26]. Differences in healthcare systems or, possibly, climate issues could also be involved in the discrepancies between countries.
This study has several limitations. Firstly, although a very large number of isolates from bloodstream infections (n = 173,540) were analysed, this study only included the four Gram-negative species covered by EARS-Net (E. coli, K. pneumoniae, P. aeruginosa and Acinetobacter spp.) and did not cover other species such as Enterobacter cloacae, Serratia marcescens or Proteus mirabilis. However, the included four species taken together constitute the majority of invasive aerobic Gram-negative isolates in many studies: 78% of bloodstream infections recorded by the SENTRY surveillance programme organised in Europe from 1997 to 1998 [27], 73% in the European point prevalence survey coordinated by ECDC in 2011 and 2012 [2] and 74% in a meta-analysis on infections recorded in developing countries [28]. Secondly, our study did not cover all antibiotics but only a selection of broad-spectrum β-lactams (third-generation cephalosporins and carbapenems) and the fluoroquinolones, widely used for treating bacteraemia caused Gram-negative species. However, we observed in the same data source similar types of correlations with aminoglycosides, another major class of antibiotics used for treating such severe infections (data not shown). Finally, the patient casemix, which depends on the types of included hospitals and on the frequency of blood culture sampling in each country, might have had an impact on the reported resistance percentages. Importantly, in the EARS-Net reports that provide detailed information on the number of laboratories and characteristics of the hospitals included, there were no marked differences between countries with low and high resistance percentages concerning the proportions of tertiary care hospital beds and intensive care unit beds, two types of hospital settings where resistance rates are usually the highest [1,29]. In addition, the representativeness of the population sample for 2016 data has been assessed as high in 23 of the 30 countries [29]. However, there was a trend towards a lower number of blood culture sets taken per 1,000 patient days in some of the countries with the highest percentages of resistance [29], which may have led us to overestimate the percentage of acquired resistance or the percentage of PSEACI in these countries. Concerning the quality of antibiotic susceptibility testing in EARS-Net, the widespread implementation of EUCAST clinical breakpoints in Europe and the high proportion of laboratories that participated in 2016 in the annual EARS-Net external quality assessment exercises with satisfactory results [1] greatly helps to ascertain the ability of the EU/EEA countries to report robust and trustworthy antimicrobial resistance data to EARS-Net.

Conclusion
We observed a strong correlation in bloodstream infections between on the one hand the countries with most intrinsically resistant Gram-negative species, indicating the burden of intrinsic resistance, and on the other hand the percentage of acquired non-susceptibility in these species. This important information adds to the already well-established arguments for a strong reduction in the consumption of antibiotics, particularly those with broad-spectrum activity, which exert a selective pressure on all types of resistant bacteria. It also reinforces the crucial importance of measures to prevent host-to-host cross-transmission of antibioticresistant microorganisms, not only to control acquired resistance in every bacterial species but also to limit the burden of infections caused by species such as P. aeruginosa and Acinetobacter spp., in which intrinsic resistance per se represents a therapeutic problem.