The epidemiology of carbapenem-non-susceptible Acinetobacter species in Europe: analysis of EARS-Net data from 2013 to 2017

Background Due to limited therapeutic options and their association with high mortality and morbidity, carbapenem-non-susceptible Acinetobacter spp. (CNA) are of significant public health importance. This study aimed to describe current epidemiological trends of CNA proportions in Europe and to identify factors that are associated with carbapenem non-susceptibility of isolates from patients with invasive Acinetobacter spp. infections. Methods Data from routine carbapenem susceptibility testing of 18,412 invasive clinical Acinetobacter spp. isolates from 30 European countries in 2013–2017 were analysed using descriptive statistical analyses and uni- and multivariable regression analyses. These data were obtained from the European Antimicrobial Resistance Surveillance Network (EARS-Net). Results The population-weighted mean proportion of carbapenem-non-susceptible Acinetobacter spp. in Europe is 35.6% (95% confidence interval [CI] 29.7–42.0%). With CNA proportions of 75.5% (95% CI 71.2–79.4%) and 71.5% (95% CI 66.7–75.9%) the burden of CNA is particularly high in Southern and Eastern European regions. In contrast, Northern and Western European regions recorded CNA proportions of 2.8% (95% CI 1.2–6.0%) and 6.3% (95% CI 4.5–8.9%), respectively. Population-weighted mean CNA proportions are especially high in Acinetobacter spp. isolates from intensive care units (54.0% [95% CI 47.6–60.3%]). Male gender, age above 20 years and ICU admission were identified as independent factors associated with an increased likelihood of CNA. Conclusion The burden of carbapenem-non-susceptible Acinetobacter spp. is particularly high in Southern and Eastern Europe. There is a risk that resistance could spread to other parts of Europe. Therefore, increased efforts in infection control and antibiotic stewardship, particularly in Intensive Care Units, are necessary to combat the spread of CNA in Europe.


Background
Acinetobacter species spp. are non-fermenting, largely opportunistic, gram negative bacteria that are ubiquitous in the environment. Acinetobacter baumannii complex (Acinetobacter nosocomialis, Acinetobacter pitti and Acinetobacter baumannii) are the most clinically significant among the over 50 species in the Acinetobacter genus [1]. Of all Acinetobacter spp., Acinetobacter baumannii sensu stricto (referred to Acinetobacter baumannii in this article) is responsible for about 90% of the Acinetobacter spp. clinical infections in humans [1]. Its ability to survive harsh conditions, including desiccation and disinfection, fosters its persistence and spread in hospital environments [2,3]. The Acinetobacter spp. have also evolved multiple resistance strategies, such as overexpression of efflux pumps, reduced permeability and a diverse array of hydrolytic carbapenemases [4], resulting in resistance to carbapenems and other reserve drugs. In addition, Acinetobacter baumannii is naturally resistant to several antibiotics, including priority "Watch group" antibiotics as defined by the WHO, such as cephalosporin. It has acquired many more resistance strategies, leading to poorer clinical outcomes across many healthcare settings [5,6]. Carbapenem-non-susceptible Acinetobacter spp. (CNA) is increasingly recognised as a nosocomial pathogen of significant public health importance worldwide because it presents limited therapeutic options, high treatment costs and is associated with high mortality and morbidity [7][8][9]. In the World Health Organization's (WHO) recent drug research and development prioritisation exercise for pathogens, CNA emerged as the highest weighted antimicrobial-resistant pathogen [10]. Worldwide, CNA have been implicated in several outbreaks of pneumonia, bloodstream, wound and urinary tract infections, especially among patients with severe morbidities like those on Intensive Care Units (ICU) [11,12].
According to the 2017 European Antimicrobial Resistance Surveillance Network (EARS-Net) report, proportions of CNA collected from invasive infections are particularly high in many southern and eastern European countries where resistance proportions often exceed 50% (such as 95% in Greece, 79% in Italy, and 53% in Hungary) [13]. In contrast, western and northern European countries have recorded low proportions of resistance to carbapenems, often less than 5% (such as 3% in the United Kingdom, 4% in Germany and < 1% Norway, Denmark and Sweden).
Despite these European data, a comprehensive epidemiological picture of invasive carbapenem-non-susceptible Acinetobacter spp. in European hospitals is still lacking. While published EARS-Net reports focus on analyses of national trends, until now it has not been systematically assessed how the likelihood of carbapenem-non-susceptibility is impacted by patients' characteristics (such as gender and age) and other factors (such as hospital unit type, season of the year and antibiotic consumption). This study therefore aims to investigate epidemiological trends of CNA proportions in Europe and seeks to identify factors that are associated with carbapenem-non-susceptibility of isolates from patients with invasive Acinetobacter spp. infections using EARS-Net data from 2013 to 2017.

Study design and the European antimicrobial resistance surveillance database
We conducted a retrospective observational study on Acinetobacter spp. using data retrieved from the EARS-Net database. EARS-Net is a network of European surveillance systems collecting routine clinical antimicrobial susceptibility (AST) data on invasive isolates (blood and cerebrospinal fluid [CSF]) from the 28 European Union countries as well as Norway and Iceland [13]. Detailed information about the EARS-Net methodology is provided in their surveillance reports and protocols [14]. In EARS-Net, no information on the specific species of the Acinetobacter spp. isolates are reported. Acinetobacter spp. are classified by EARS-Net as sensitive (S), intermediate (I), or resistant (R) to antimicrobial drugs, based on the standards used in the participating laboratories, such as the guidelines of the European Committee on Antimicrobial Susceptibility Testing (EUCAST), Clinical and Laboratory Standards Institute (CLSI) or other national guidelines. In our data set, more than 98% of all isolates with information on the used guideline have their AST interpreted with EUCAST or CLSI. As reported by the ECDC, there is a general consensus between the EUCAST and CLSI interpretation for carbapenem susceptibility testing for Acinetobacter spp. [15].
In addition to S-I-R data, individual laboratories provide further epidemiological information. This includes the date of specimen collection, specimen type (i. e. blood and CSF), care type (such as inpatient or outpatient care), patient gender and patient age as well as the hospital unit (such as ICU or internal medicine unit), where the sample was collected.

Selection of Acinetobacter spp. isolates
We extracted data for Acinetobacter spp. from the TESSy database in February 2019 with the approval of the European Centre for Disease Prevention and Control. Only the first isolate from a given patient in the respective year is included in the TESSy database of EARS-Net. We excluded isolates with multiple AST tests against the same antibiotic as well as isolates with the same unique identification number. In the next steps, we excluded isolates from outpatient care and isolates that were neither tested against meropenem nor imipenem. Since the initial pilot study of EARS-Net was conducted in 2012 and included data from only 18 countries, we exclusively analysed data on Acinetobacter spp. from 2013 to 2017, when all 30 countries provided data to EARS-Net.

Outcomes and variables
The primary outcome of interest is the populationweighted proportion of carbapenem-non-susceptible Acinetobacter spp. isolates among all Acinetobacter spp. isolates, expressed as a percentage (%) and with 95% confidence intervals (95% CI). An Acinetobacter spp. isolate was defined as carbapenem-non-susceptible if it was tested resistant or intermediate against meropenem and/ or imipenem. Clinical specimens were grouped by sampling site into either blood or cerebrospinal fluid. Patient ages were categorized into six age categories (< 1, 1-19, 20-39, 40-59, 60-79 and ≥ 80 years). Patient genders were classified as female or male. The isolates' country of origin were grouped into four major European regions (Northern: Denmark, Finland, Iceland, Ireland, Norway, Sweden, United Kingdom; Western: Austria, Belgium, France, Germany, Luxembourg, Netherlands; Southern: Croatia, Cyprus, Greece, Hungary, Italy, Malta, Portugal, Slovenia, Spain; Eastern: Bulgaria, Czech Republic, Estonia, Latvia, Lithuania, Poland, Romania, Slovakia). Hospital units were categorised into "ICU", "Internal medicine", "Surgery", "Oncology" and "Others". To determine potential seasonal trends in the occurrence of CNA, the months in which the sample were collected were categorised into "Warm months" (May -September) and "Cold months" (October -April). In order to investigate the populationweighted proportion of co-resistance to gentamicin and ciprofloxacin among CNA isolates, only isolates that were separately tested against gentamicin and ciprofloxacin were selected. An Acinetobacter spp. isolate was defined as ciprofloxacin-or gentamicin-non-susceptible if it was tested intermediate or resistant against ciprofloxacin and gentamicin, respectively.

Statistical analyses
All statistical analyses were performed using R version 3.5.1 [16] and the "survey" package (version 3.35). Proportions and 95% confidence intervals of carbapenemnon-susceptibility were estimated accounting for clustering at the hospital level and stratified by country level. For all calculations in all strata, country populationbased weighting was applied. The population data were obtained from the Eurostat database (https://ec.europa. eu/eurostat/data/database). Weightings were used for each country's population in the calculation of nonsusceptibility proportions in order to ensure that the data contributed proportionally to their population sizes. This was done to minimise biases from significant differences in isolate numbers between the countries. The potential association between different variables and carbapenem-non-susceptibility was analysed using univariable and multivariable logistic regression analyses accounting for clustering and using weights proportional to each country's population sizes as described above. For the univariable analyses, the following predictors for carbapenem-non-susceptibility were considered: Year of sampling, gender, age group, specimen type, European region, hospital unit type and season. These variables were selected before conducting the analysis and were based on the availability of the data and our prior hypotheses on which variables may be associated with carbapenem-non-susceptibility in Acinetobacter spp. We included all variables from the univariable analyses in the model for the multivariable analysis.

Dataset characteristics
The baseline characteristics of the analysed Acinetobacter spp. isolates are outlined in Table 1. In total, 18,412 isolates of Acinetobacter spp. from 18,167 patients were collected in 1191 hospitals across Europe between 2013 and 2017. The majority of isolates originated from elderly patients (median age: 64 years). For the isolates with a reported patient gender (n = 12,678), the female/male ratio was 0.72. The clinical samples were predominantly from blood specimens (97%). Isolates were most frequently collected from patients treated on ICUs (41%) and internal medicine units (25%), followed by surgical units (11%) and oncology units (5%). About threequarters of the isolates (73%) were from the Southern and Eastern regions of Europe, which represented 45% of the mean population of the 30 countries included in the study.

Age and gender
In order to understand the possible influence of the age and gender of patients on CNA infections, these two factors were tested for any association with carbapenemnon-susceptibility. The population-weighted mean proportions for the whole of Europe show that isolates from infants and babies (0-1 years) as well as from children and adolescents (1-20 years) exhibited markedly lower   Table 1). This finding is supported by univariable and multivariable analyses, which confirmed that Acinetobacter spp. isolates from patients aged 0-1 years and 1-20 years are less likely to be carbapenemnon-susceptible than isolates from older patients ( Table  2). Importantly, similar patterns of relative age differences were observed in Western, Eastern and Southern European regions (Additional File 1: Additional Table 1), irrespective of the overall mean CNA proportion reported in these regions. There was an overall population-weighted mean difference of 3% in the proportion of CNA among isolates between female ( Table 1). However, only 3% of the isolates were from CSF, and no statistically significant association between specimen type and the likelihood of CNA infection was found in univariable and multivariable regression analyses (Table 2).

Hospital unit type
Analyses of various hospital unit types where the samples were collected reveal substantial differences in CNA proportions ( Fig. 3 and Additional File 1: Additional

Discussion
In this study, we assessed epidemiological trends of invasive carbapenem-non-susceptible Acinetobacter spp. isolates from hospital patients in the European Union and the European Economic Area (EAA) using data from EARS-Net. Similar to the conclusions reached previously by the ECDC [1], our study demonstrated the persisting challenge of CNA in Europe as the country-population weighted mean proportion of non-susceptible isolates exceeds 30% for the period between 2013 and 2017. The CNA proportion observed in our study is similar to the CNA proportion (35%) reported in a recently published nationwide survey done in the United States among hospitalized patients [18]. A recent global study showed that carbapenem-non-susceptibility in Acinetobacter baumannii complex isolates from both invasive and noninvasive infections was considerably higher in the Asia-Pacific region (~79%), Latin America (~85%) and North America (~45%) compared to the mean CNA proportion in Europe described in our study [19]. Importantly, in Europe the Southern-Eastern to Northern-Western gradient persists. In the Southern and Eastern regions, more than 70% of all Acinetobacter spp. isolates from invasive infections were carbapenem-non-susceptible, while in the North and the West, CNA proportions of less than 10% were observed. This gradient seen in the non-susceptibility of invasive Acinetobacter spp. to carbapenems was also observed for other pathogens in Europe, including Pseudomonas aeruginosa and Klebsiella pneumoniae [13], suggesting a systematic higher burden of invasive infections with drug-resistant Gramnegative pathogens in Southern and Eastern European regions. High CNA proportions in the Southern and Eastern regions is of great clinical concern, particularly in light of growing resistance against other antimicrobial drugs used as alternative therapeutic options to treat CNAs infections, such as fluoroquinolones and aminoglycoside antibiotics. Our analyses show that on average, > 70% of CNA are also non-susceptible to ciprofloxacin or gentamicin in Southern and Eastern Europe.
Differences in antimicrobial use, infection control and health care utilisation practices between the countries are possible reasons for the variation in the reported proportion of drug-resistant Acinetobacter spp. isolates [20]. Previous studies have shown limited organizational support and practice of infection control measures like standard precautions among health workers that work in the medical ward as well as in critical areas like the ICU in Eastern (e.g. Poland) and Southern European (e.g. Greece) countries [21][22][23]. The 2017 ESAC-Net report showed an increasing trend of carbapenem consumption between 2013 and 2017 in many of the Southern and Eastern region countries, including Latvia, Lithuania, Romania, Poland and Hungary [24]. This might partly explain the surge in carbapenem-non-susceptibility seen in these regions. In fact, an association between hospital carbapenem consumption and carbapenem-resistance rates of Acinetobacter have been described by multiples studies [25][26][27][28][29]. Beside antibiotic usage, other possible explanations for the observed gradient in CNA proportion have been suggested, such as differences in climate and socio-economic conditions across these European regions [30,31].
In addition to regional differences in CNA proportions in Europe, this study identified factors that are associated with a higher likelihood of acquisition of carbapenemnon-susceptible Acinetobacter spp. Adult patients (> 20 years) show markedly higher CNA proportions than both babies and infants (< 1 years) and children and adolescents between 1 and 20 years. Importantly, the differences in non-susceptibility proportion between adults and babies/ adolescents is also observed in regional analyses in Western, Southern and Eastern European regions, suggesting that the higher proportion of CNA in adult patients is a common feature in invasive Acinetobacter spp. infections, independent of inter-regional differences in CNA proportions. This finding is consistent with other studies that found children and adolescents are less susceptible to several infections with drug-resistant bacteria, including carbapenem-non-susceptible Klebsiella pneumoniae [32], Pseudomonas aeruginosa [33], as well as methicillinresistant Staphylococcus aureus [34] and vancomycinresistant Enterococcus faecium [35,36]. Several factors may explain these age differences. It has been shown that elderly patients are more likely to be colonised with drugresistant pathogens due to more frequent exposure to antibiotics throughout their lives. Moreover, elderly patients have more comorbidities than younger patients and are more likely to reside in nursing homes or other healthcare facilities, both factors that have been shown to be associated with increased AMR [37]. This study also shows that male patients are associated with an increased likelihood of CNA compared to female patients. Male gender is also associated with higher proportions of other carbapenemnon-susceptible pathogens, including of Klebsiella pneumoniae and other Gram-negative rods [32,38].
Unsurprisingly, most of the isolates in this study are from the ICUs, which is a known hotbed for nosocomial transmission of Acinetobacter spp. [11,39]. This study shows that the proportion of CNA is highest among isolates collected in the ICU where many patients tend to have severe comorbidities and are exposed to antibiotics and devices that enable the transmission of pathogens [40][41][42]. Studies have also reported that admission to the ICU is a risk factor for the acquisition and isolation of CNA infections in other regions of the world, such as Northern Africa [43,44].

Strengths and limitations
To our knowledge, this study is the largest and most comprehensive analysis of CNA infections in Europe, with more than 18,000 clinical isolates from 30 countries in the EU/EAA. EARS-Net data are based on routine clinical antimicrobial susceptibility data from national surveillance programmes. The accuracy and validity of the AMR data has been assured through annual external quality assessments on all laboratories from all countries, which showed they all performed generally well [45]. However, it is important to consider the limitations of EARS-Net data. Firstly, participating national laboratories and hospitals might not be representative of an individual country, although according to the EARS-Net report 2017 [13], many countries across all regions reached a high isolate sample representativeness in their national surveillance systems. Secondly, population coverage varied among reporting countries. To address this issue, all statistical analyses used weightings based on the population sizes of the individual countries similar to that done in the EARS-Net reports. Thirdly, Acinetobacter spp. are not identified to the species level in the EARS-Net database. This is important in terms of carbapenem resistance, as A. baumannii isolates are more frequently carbapenem-resistant that other Acinetobacter species. Therefore, the differential proportion of A. baumannii among all Acinetobacter species across countries and regions might impact the CNA proportions recorded in EARS-Net. Last, different sampling routines in different healthcare settings such as due to limited resources can result in biased estimates of CNA proportions across European regions. However, due its clinical significance, inclusion of only invasive clinical specimen limits the bias that may result from some of the inconsistencies in sampling, although frequency of blood sampling varies between hospitals and countries.

Conclusion
Carbapenem-non-susceptible Acinetobacter spp. pose a significant threat to public health in Europe, particularly in the Southern and Eastern regions where more than 70% of all invasive Acinetobacter isolates were non-susceptible to carbapenems. The proportion of CNA is particularly high on ICUs and among older patients. These findings raise concern for inter-regional spread to areas with traditionally low CNA proportions from contiguous European countries with high CNA proportions. In order to prevent the spread of CNA infections, continued surveillance and implementation of effective infection prevention and control measures are needed, particularly in high-risk settings and through targeting high risk groups.