Euroroundups Carbapenem-non-susceptible Enterobacteriaceae in Europe: conclusions from a meeting of national experts

The emergence and global spread of carbapenemase-producing Enterobacteriaceae is of great concern to health services worldwide. These bacteria are often resistant to all beta-lactam antibiotics and frequently co-resistant to most other antibiotics, leaving very few treatment options. The epidemiology is compounded by the diversity of carbapenem-hydrolysing enzymes and the ability of their genes to spread between different bacterial species. Difficulties are also encountered by laboratories when trying to detect carbapenemase production during routine diagnostic procedures due to an often heterogeneous expression of resistance. Some of the resistance genes are associated with successful clonal lineages which have a selective advantage in those hospitals where antimicrobial use is high and opportunities for transmission exist; others are more often associated with transmissible plasmids. A genetically distinct strain of Klebsiella pneumoniae sequence type (ST) 258 harbouring the K. pneumoniae carbapenemases (KPC) has been causing epidemics of national and international proportions. It follows the pathways of patient referrals, causing hospital outbreaks along the way. Simultaneously, diverse strains harbouring New Delhi metallo-beta-lactamase (NDM-1) are repeatedly being imported into Europe, commonly via patients with prior medical exposure in the Indian subcontinent. Since the nature and scale of carbapenem-non-susceptible Entrobacteriaceae as a threat to hospital patients in Europe remains unclear, a consultation of experts from 31 countries set out to identify the gaps in diagnostic and response capacity, to index the magnitude of carbapenem-non-susceptibility across Europe using a novel five-level staging system, and to provide elements of a strategy to combat this public health issue in a concerted manner

The emergence and global spread of carbapenemaseproducing Enterobacteriaceae is of great concern to health services worldwide. These bacteria are often resistant to all beta-lactam antibiotics and frequently co-resistant to most other antibiotics, leaving very few treatment options. The epidemiology is compounded by the diversity of carbapenem-hydrolysing enzymes and the ability of their genes to spread between different bacterial species. Difficulties are also encountered by laboratories when trying to detect carbapenemase production during routine diagnostic procedures due to an often heterogeneous expression of resistance. Some of the resistance genes are associated with successful clonal lineages which have a selective advantage in those hospitals where antimicrobial use is high and opportunities for transmission exist; others are more often associated with transmissible plasmids. A genetically distinct strain of Klebsiella pneumoniae sequence type (ST) 258 harbouring the K. pneumoniae carbapenemases (KPC) has been causing epidemics of national and international proportions. It follows the pathways of patient referrals, causing hospital outbreaks along the way. Simultaneously, diverse strains harbouring New Delhi metallo-beta-lactamase (NDM-1) are repeatedly being imported into Europe, commonly via patients with prior medical exposure in the Indian subcontinent. Since the nature and scale of carbapenem-non-susceptible Entrobacteriaceae as a threat to hospital patients in Europe remains unclear, a consultation of experts from 31 countries set out to identify the gaps in diagnostic and response capacity,

Introduction
Enterobacteriaceae are among the most abundant commensal microorganisms in humans. They are also the most frequent cause of bacterial infections in patients of all ages [1]. Their ubiquity and frequent acquisition of mobile genetic elements means that their human hosts are regularly exposed to new strains with novel genetic repertoires -including antibiotic resistance -through food and water, or from other animate and inanimate sources in the community, hospitals and during travel.
Since the 1950s and 60s -when broad-spectrum antibiotics became available for the treatment of Gramnegative infections -Enterobacteriaceae have acquired a growing range of mechanisms to evade these agents. In particular, beta-lactam antibiotics such as penicillins and cephalosporins are vulnerable to hydrolysis by enzymes called beta-lactamases. In the mid 1970s two new beta-lactamase-stable cephalosporin compounds, cefamandole and cefuroxime were marketed [2,3], soon followed by related analogues such as cefotaxime and ceftriaxone [4]. However, novel extendedspectrum beta-lactamases (ESBL) soon emerged in Enterobacteriaceae, compromising these new compounds [5]. The first hospital outbreaks caused by ESBL producers occurred in France in the mid-1980s [6], soon followed by large outbreaks in the United States (US) [7,8]. ESBL producers, are now widespread worldwide, and often are multidrug-resistant (MDR) also to fluoroquinolones and aminoglycosides [9].
A further class of beta-lactam antibiotics, the carbapenems, came into clinical use in 1985 [10]. These drugs combine exceptional intrinsic antibacterial activity with stability to most of the prevalent beta-lactamases, including ESBLs and have become the treatment of choice for infections due to the ESBL-producing strains, which are increasingly diagnosed in European hospitals. Regrettably, it has become clear that bacteria also can acquire carbapenem-hydrolysing beta-lactamases (carbapenemases). Such enzymes have emerged in various parts of the world, including Europe, the Indian subcontinent and the US [11]. In Europe and countries that were covered by the European Antimicrobial Resistance Surveillance System (EARSS, now EARS-Net) large nationwide outbreaks with carbapenemaseproducing Klebsiella pneumoniae have occurred in Israel and Greece, and problems of variable scale are unfolding elsewhere in Europe [12][13][14]. Acknowledging the ineffectiveness of almost all alternative antibiotics and resistance even to those under development, there is a growing awareness that carbapenem-non-susceptible Enterobacteriaceae (CNSE) may thwart the ability to treat many life-threatening infections in the future.
The need for a European-wide consultation on this matter was recognised during the 2009 annual EARSS meeting, and thus a workshop of scientists involved in the surveillance of antibiotic resistance in Enterobacteriaceae from 31 European countries was hosted at the Netherlands' National Institute for Public Health and the Environment (RIVM) on 29 and 30 April 2010. These scientists already participated in the EARSS network and were selected on the basis of their expertise in the epidemiology of carbapenem resistance. This workshop aimed (i) to identify the gaps in diagnostic and response capacity, (ii) to index the magnitude of carbapenem-non-susceptibility across Europe using a novel five-level staging system, and (iii) to provide elements of a strategy to combat this public health issue in a concerted manner This report summarises the discussion and outlines the complexity of the diagnostic issues and also provides information on the epidemiologic situation in European countries. The experts' conclusions aim to solidify diverse country-specific experiences into a coherent plan of action on surveillance and response to prevent the endemic establishment of carbapenemase-producing Enterobacteriaceae in European hospitals.

The emergence of carbapenem-nonsusceptible Enterobacteriaceae
In contrast to the increasing prevalence of ESBLproducing Enterobacteriaceae in Europe [15], CNSE were extremely rare during the 1990s and early 2000s, and mostly comprised K. pneumoniae, and Enterobacter spp. with a permeability deficit that reduced drug uptake. This was associated with the inactivation of genes coding for outer membrane proteins (in K. pneumoniae OmpK35 and OmpK36) that function as major porins, allowing solutes to enter the bacterial cell [16,17]. This impermeability reduces carbapenem susceptibility when combined with ESBLs or AmpC beta-lactamases [18] which have trace carbapenemhydrolysing activity. Ertapenem is the carbapenem most affected by this mechanism, and several cases of emerging ertapenem resistance during treatment have been reported [19]. Most isolates are unique, with limited clonal dissemination, perhaps because the impermeability is detrimental to the bacteria, making them less competitive in the absence of antibiotics.
Carbapenemases, which readily hydrolyse carbapenems, became an international health issue 15 years after the introduction of carbapenems, and pose a greater threat. They have been described in all four classes of beta-lactamases, but the epidemiologically most relevant carbapenemases fall into three of these [20]: Class B includes the metallo-beta-lactamases (MBLs) IMP (imipenemase)*, and VIM (Verona integronencoded metallo-beta-lactamase) and the recently described New Delhi metallo-beta-lactamase (NDM-1). In class A, KPC (K. pneumoniae carbapenemases) is clinically and epidemiologically the most important enzyme, whereas SME (Serratia marcescens enzyme), NMC-A/IMI (not metalloenzyme carbapenemase/imipenem-hydrolysing beta-lactamase) and GES (Guiana extended spectrum) pose minor problems. Class D includes the OXA-type carbapenemases which are mostly found in Acinetobacter spp., although OXA-48 occurs in Enterobacteriaceae.
The first transferable carbapenemase identified in Gram-negative bacteria was an IMP-like MBL in the Far East [21], followed by VIM types in Europe [22]. In early 2003, VIM-producing Enterobacteriaceae began to spread in Greek hospitals [23]. Enterobacteriaceae with VIM MBLs have also caused some hospital outbreaks in Spain [24] and have been observed sporadically in other countries. In some cases, MBL-positive isolates were associated with travel, such as importation from Greece of K. pneumoniae producing VIM-1 and -2 carbapenemases [25]. Since 2008, Enterobacteriaceae producing NDM-1 metallo-beta-lactamases have been imported repeatedly into Europe from the Indian subcontinent [26], particularly into the United Kingdom (UK) [27] but also to Austria, Belgium, France, Germany, the Netherlands, Norway and Sweden. There were also importations to Australia, Canada, Japan and the US, again largely in patients with recent hospitalisation in India, Pakistan or Bangladesh [27,28]. Most were susceptible only to colistin and, more variably, tigecycline.
NDM-1 producers are mainly K. pneumoniae but also include Escherichia coli and Enterobacter spp. Most isolates with MBLs, particularly those with NDM types, also contain ESBL and acquired ampC genes, which makes them resistant to all antibiotics except polymyxins, tigecycline and, occasionally, certain aminoglycosides.
K. pneumoniae with KPC carbapenemase were first detected in 1996 in North Carolina, then spread along the east coast of the US [29,30], and finally across the whole country [31], posing a significant threat with 70% or higher mortality in bacteraemic patients [32,33].  [38,39], and, more recently, also in Pseudomonas spp. and Acinetobacter baumannii [40,41]. As with the MBL producers, few treatment options remain, although some isolates remain susceptible to few aminoglycosides (gentamicin, isepamicin) as well as to polymyxins (such as colistin) or tigecycline.
OXA-48 was first described in Turkey during an outbreak of K. pneumoniae in Istanbul but has since attained international distribution not only among K. pneumoniae but also E. coli [42,43] cephalosporins and monobactams but express resistance or decreased susceptibility to carbapenems [47].
Carbapenem-non-susceptibility thus displays a very diverse picture, in geographical occurrence and enzyme types; it also challenges conventional diagnostic abilities because the presence of carbapenemase genes does not always translate into clinical resistance as defined by the current guidelines and breakpoints, as discussed below.

Identification of carbapenem-nonsusceptible Enterobacteriaceae by routine susceptibility testing methods
The standard approach for the testing of the antimicrobial susceptibility of bacteria in routine diagnostic practice is based on measuring bacterial growth in the presence of the drugs; either by classical agar disk diffusion assays (Kirby Bauer technique) or with commercially available automated test systems that expose bacterial suspensions to a limited range of antimicrobial concentrations. The goal is to predict clinical outcomes by classifying bacterial isolates as susceptible (S), intermediate (I) or resistant (R) on the basis of agreed breakpoints. These breakpoints take into account (i) the range of antimicrobial susceptibility in a natural bacterial population in the absence of resistance mechanisms (the so-called wildtype distribution), (ii) the pharmacology with regards to the time course of the drug concentration in the human body (pharmacokinetics) and the biological effect of the drug at these concentrations on the bacteria (pharmacodynamics), and (iii), whenever available, information on clinical outcomes in relation to the minimum inhibitory concentration (MIC). Clinical results provide the most important information but are also the most difficult to acquire and to evaluate [48].
International breakpoint committees such as the Clinical Laboratory Standards Institute (CLSI, formerly NCCLS) in the US and the European Committee on Antimicrobial Susceptibility Testing (EUCAST) set and modify breakpoints by well defined decision processes [49,50]. Importantly, breakpoints as defined by these committees are applied by automated diagnostic test systems provided by various manufacturers or, after conversion into inhibition zone diameters, by agar disk diffusion assays. EUCAST also provides crucial guidance for the European Medicines Agency (EMA, formerly EMEA) when approving the clinical indications for new antimicrobial agents.
The CLSI modified its clinical breakpoints for carbapenems after an expert consultation meeting in January 2010, reducing their previous values two-fold in order to better identify the carbapenemase-producing K. pneumoniae (mainly with KPC enzymes) that have attained considerable prevalence in US hospitals in the last 10 years. EUCAST had previously, in 2008, decided to set their breakpoints for the purpose of clinical, i.e. therapeutic decision-making and not for optimal detection of carbapenemase production per se. As a consequence, the clinical breakpoints adopted by EUCAST remain one dilution step higher than the modified CLSI values (Table 1). Since the updated CLSI breakpoints came into use in June 2010, laboratories using these standards have been defining susceptibility more conservatively than those using EUCAST [51].
Both committees recommend reporting susceptibility testing results at face value, and performing phenotypic tests for carbapenemase production only for epidemiological or infection control purposes. This will have consequences for clinical diagnosis, routine surveillance and public health.

Clinical laboratory diagnosis
On the basis of available evidence and simulated target attainment probabilities, EUCAST decided that Enterobacteriaceae should be regarded as clinically susceptible to imipenem at an MIC of ≤2 mg/L when treated with the standard recommended adult dose of 500 mg four times a day intravenous administration. However, the maximum dose of 1g four times a day for severe infections was taken into consideration in setting the I/R breakpoint >8 mg/L [52]. EUCAST adds a note to the breakpoint table that "some strains that produce carbapenemase are categorized as susceptible with these breakpoints and should be reported as tested, i.e. the presence or absence of a carbapenemase does not in itself influence the categorization of susceptibility. In many areas, carbapenemase detection and characterization is recommended or mandatory for infection control purposes" [53].
The more conservative modified CLSI breakpoint addresses the wide demand in the US for simplification of phenotypic characterisation of K. pneumoniae isolates in the wake of the KPC epidemic. The adoption of the new breakpoints is intended to render the modified Hodge test (see below) unnecessary, whereas this test was recommended in previous updates to the guidelines and had to be used frequently in many laboratories. The adoption of these breakpoints by the CLSI clearly improves the ability of microbiological laboratories to detect carbapenemase-producing Enterobacteriaceae, but with an unknown trade-off in specificity because more strains with combinations of impermeability and ESBL or AmpC are likely to be scored as resistant. Moreover, even these breakpoints will fail to detect carbapenemase-producing Enterobacteriaceae with very low MICs [54]. For clinical purposes, a breakpoint-guided therapeutic decision algorithm, as favoured by EUCAST, may be sufficient if additional molecular events such as mutational porin losses that reduce susceptibility during treatment are rare enough for bacteriological treatment failures to remain uncommon [16,55].

Routine surveillance and the role of the clinical laboratory in public health
The detection of carbapenem resistance within passive surveillance systems such as EARS-Net is complicated by breakpoint changes. Until recently most laboratories within EARS-Net have used the old CLSI breakpoint (S if MIC ≤4 mg/L imipenem or meropenem), but reported resistance rates will artificially rise when European countries/laboratories shift to use the lower EUCAST breakpoints routinely in 2010-11. Some laboratories may continue with CLSI methodology but by adopting the now lower breakpoint prescribed by that organisation will cause a resistance shift in the same direction. This will lead to some minor discrepancies and it remains to be seen if passive surveillance in its current format provides sufficient information for infection control practitioners and public health experts about the extent of the problem at national or international levels.
The contribution of diagnostic laboratories to infection control and public health is often underappreciated, underfunded and increasingly compromised by the streamlining of hospital budgets along tight service lines, which often results in the outsourcing of diagnostic services [56]. The conundrum of carbapenemase-producing Enterobacteriaceae with resistance below the radar of routine surveillance but relevant enough to cause concern, exposes the lack of consensus on the precise role of microbiological diagnostic laboratories in European countries. If isolates with lower-level resistance are worth monitoring for infection control and public health purposes then a simple laboratory tool is needed for detection -as simple as Etest, or double-disk synergy test (DDST), combination disk tests (CDTs) or an expert rule integrated into automated test systems.

Detection of carbapenemaseproducing Enterobacteriaceae
Various selective agar media can be used for preliminary screening and are convenient, especially if they are chromogenic, allowing different species or resistance types to be recognised easily [57]. Different selective media may be necessary to detect carbapenemase-producers with very low carbapenem MICs. The addition of a selective agar that contains extendedspectrum cephalosporins would improve sensitivity in the detection of carbapenemases, but would also have lower specificity (regular ESBLs are also detected) [58].
Comparing growth on agars containing cephalosporins and carbapenems might also help in the detection of OXA-48 producers.
For confirmation of carbapenemase production, two inexpensive types of tests can be deployed in routine as well as reference laboratories: The first are disk diffusion synergy tests, where the potential carbapenemase-producer is tested against a carbapenem antibiotic in the presence of carbapenemase-inhibiting compounds, including dipicolinic acid for MBLs and boronic acid for KPC enzymes (Figure). These tests may be performed in disk-approximation, i.e. DDST, or disk-combination, i.e. CDT formats [59]. Combination discs can be prepared locally by applying defined amounts of inhibitors to routine (meropenem 10 μg) antibiotic disks, or can be purchased [60]. Inhibition, and carbapenemase presence, is indicated by zone expansion. The respective inhibitors achieve reasonable specificity for MBLs and KPC enzymes [60,61] but no specific inhibitor of OXA-48 has been identified so far.
The other type of test exploits the leakage of carbapenemases from the producer into the surrounding agar and its ability to protect susceptible strains on the same plate. These tester-reporter assays consist of various modifications of the cloverleaf or Hodge test. They lack specificity, are difficult to standardise, labour-intensive and require a certain degree of experience to provide confident interpretation of results [61,62].

Molecular confirmation tests
Molecular confirmation tests are largely the realm of reference laboratories, but can be used for rapid screening under epidemic conditions, as demonstrated by experiences with faecal screening in Israel [63]. PCR assays can be designed to seek different target genes in single or multiplex formats with different modes of amplicon detection [64]. A commercial one-day test utilising ligase chain reaction and a microarray hybridisation format provides a versatile platform for the identification of ESBLs and KPCs [65][66][67], whilst a new test that also detects the genes for AmpC, VIM, IMP and NDM-1 enzymes is currently under clinical evaluation.

Carbapenem-non-susceptible Enterobacteriaceae in European countries
Although first seen sporadically in the Far East, CNSE are now established in Europe. K. pneumoniae is the species that most often hosts the resistance and, depending on the country, assumes various epidemiological patterns. The current knowledge of the epidemiology of carbapenem-non-susceptible Enterobacteriaceae in Europe is summarised below.  [36]. By the end of 2009, 10 hospitals in Warsaw and its surroundings were affected. Each reported between one and nine cases, and indicating continuing and widening spread. By April 2010, cases had been reported from more than 30 hospitals, and from six outpatient clinics in 16 cities. The situation is still dynamic, with some hospitals seemingly in control of the problem whilst others report newly emerging cases or have stopped reporting new cases altogether.

Italy
In Italy, various Enterobacteriaceae producing VIM enzymes have been reported from different regions since 2002. These isolates were mostly sporadic [74,75], although a single outbreak involving nine patients with bloodstream infections arose due to the clonal spread of a VIM-1-positive Klebsiella [76]. VIM carbapenemases were also found in 36 of 5,500 routinely collected Enterobacteriaceae from acute care hospitals and longterm-care facilities in Bolzano (Alto Adige region). These isolates also had various other resistance traits (qnrS, bla SHV , bla CTX-M ) [77]. In late 2008, the first patient with KPC-3-positive K. pneumoniae ST258 was identified in Tuscany [78], but by early 2010, indistinguishable strains had already been reported from at least 11 locations in seven regions, and in some hospital settings they have already reached remarkable levels [Rossolini, unpublished results]. Part of this spread was associated with patient transfers between hospitals.

Germany
In Germany, the first outbreak of KPC-2-producing K. pneumoniae was reported in 2008 [79,80]. In 2009 and 2010 two outbreaks with KPC-3-producing K. pneumoniae (ST 258) and more than 40 single cases of KPC-2/3 in E. coli and K. pneumoniae were observed. Identified index patients came from Greece or Israel. Regional spread of two distinct multidrug-resistant clones over two years due to movement of colonised patients between hospitals was shown in Bavaria. Moreover several small regional clusters of OXA-48 producing K. pneumoniae were identified in 2009 [unpublished results]. Currently, OXA-48 is the most frequent carbapenemase in Germany, occasionally in patients with connection to Turkey. NDM-1 MBL was identified in three E. coli and one Acinetobacter baumannii isolates from epidemiologically unrelated patients.

France
In France, five monoclonal (single hospital or regional) outbreaks with carbapenemase-producing Enterobacteriaceae have been reported to health authorities since 2004 through the national early warning system set up at the beginning of the 2000s. In three of these outbreaks, involving strains with VIM-1 or KPC-2 enzymes, the index patient had been transferred from a Greek hospital [25]. A national programme initially designed to contain the spread of vancomycinresistant Enterococcus spp. (VRE) was applied to each outbreak. This consisted of the rapid implementation of a step-by-step containment plan within the affected hospital, constant support by local infection control teams, regional experts and health authorities, and feedback to the medical community at the national level. The hospital containment strategy has the following components: (i) stopping transfer of cases and contacts within and between hospitals, (ii) cohorting separately case and contact patients with dedicated healthcare workers, (iii) screening all contact patients, and (iv) continuous vigilance through surveillance.

Hungary
In Hungary, nine K. pneumoniae isolates with nonsusceptibility to carbapenems carrying the KPC-2 carbapenemase and SHV-12 ESBL were isolated in three centres in the North Eastern Region between October 2008 and April 2009. All belonged to the ST258 international clone, were indistinguishable by pulsed-field gel electrophoresis (PFGE), and were extensively drugresistant. Eight were resistant even to colistin although none of the source patients had received this drug, which had never been used in any of the affected hospitals in that period. All infected patients died. The index patient had a history of hospitalisation in Greece [37]. Since then, only one further carbapenemase producer has been recorded, a K. pneumoniae ST11 strain with VIM-4 enzyme. Interestingly, ST11 is a single-locus variant of ST258 and, in Hungary, is commonly seen producing the CTX-M-15 ESBL but normally lacking VIM genes [81].

Spain
In Spain, VIM-positive Enterobacteriaceae have, as of April 2010, been reported from 15 different hospitals, IMP-positive strains from another two, and KPC-positive Enterobacteriaceae (K. pneumoniae and Citrobacter freundii) from two university hospitals in Madrid [24]. The KPC-positive K. pneumoniae strains did not belong to the ST258 epidemic clone, but to the ST384 and ST388 clones. ST388 K. pneumoniae had previously persisted as a carbapenem-susceptible CTX-M-10 betalactamase-producing clone [82]. A structured survey in 2008 covering 38 hospitals across Spain, collected 100,132 isolates of Enterobacteriaceae but only identified 43 with carbapenemases, mainly VIM types (76%), whilst none had KPC types [83]. However, 45 of 245 carbapenem-non-susceptible Enterobacteriaceae isolates submitted for reference testing between January 2009 and April 2010 harboured VIM (18%), 15 IMP (6%) and 15 were non-carbapenemase-producing strains (6%). Two outbreaks involving VIM-1 enzyme-producing K. pneumoniae were reported at two hospitals in Madrid [84] and three small outbreaks were reported with carbapenemase-producing Enterobacter spp. [85].

United Kingdom
In the United Kingdom NDM seems to de the dominant carbapenemase in Enterobacteriaceae, although producers of KPC (increasingly), VIM and OXA-48 carbapenemases are also recorded. Patients infected with producers of this NDM-1 enzyme have a history of hospitalisation on the Indian subcontinent, where producer strains of K. pneumoniae, E. coli and other Enterobacteriaceae appear to be in wide circulation [27]. The dominance of this enzyme in the UK may reflect the country's historic links with India, and the consequent population flows to and from the subcontinent.

Other countries
Import of CNSE by travel has also been detected in other countries such as Sweden, Denmark, Norway, Finland, and Belgium but the spread was limited, most likely thanks to infection control [35,70,86,87].

Proposing a staging scheme for the epidemiology of carbapenem resistance in European countries
The experiences reported above suggest that the epidemiology of CNSE and especially carbapenemase-producing K. pneumoniae in European countries follows a pattern typical for hospital-acquired pathogens.
Initially there is sporadic occurrence and stochastic extinction, followed by single hospital outbreaks and then spread along the regional and national hospital patient referral routes. This also means that hospitals that share the same patients are at a high risk of importing colonised or infected individuals, providing the sources for the next outbreaks. An intuitive way of assessing the degree to which carbapenemase-producing Enterobacteriaceae have become established in national hospital networks is by indexing these stages, and we therefore suggest a simple numerical staging system ( Table 2).
Applying this staging scheme and data provided by 31 reference laboratories for the period up to July 2010 allowed us to categorise the European countries (Table 3). Most countries that reported early stage events mentioned documented introduction by travel and many were concerned about likely underreporting owing to a lack of detection or lack of communication (not shown). In clonal outbreak situations, KPC is the dominant resistance mechanism, mainly linked to the spread of K. pneumoniae international clone ST258.

Conclusions
Care of hospitalised patients throughout Europe is threatened by the spread of carbapenem-non-susceptible Enterobacteriaceae. There are very few therapeutic options left to treat these patients, and invasive infections are associated with disturbingly high mortality rates. Little is known about the patient-related risk factors other than hospitalisation abroad, but the description of outbreaks indicates that producer strains seem to benefit from selective advantages in hospitals where antimicrobial use is much higher and opportunities for transmission more frequent than in the community [88]. The association of KPC (and occasionally VIM) enzymes with an internationally successful clonal lineage of K. pneumoniae indicate that hospital outbreaks are local expansions following long transmission chains. This is also supported by the frequent international introduction of sporadic or primary cases. Consistent with the spread of hospital-adapted lineages is the repeated observation that outbreaks, especially of KPC-positive ST258 K. pneumoniae, follow patient referral patterns, with initial local spread and occasional regional and nationwide dissemination. The fact that transmission of these clones is mainly confined to healthcare settings provides an opportunity for targeted prevention and control. Israel has shown that national consensus approaches with agreed screening protocols and mandatory reporting can reduce the incidence of resistance during a nationwide Klebsiella ST258 epidemic.  Single hospital outbreak Outbreak defined as two or more epidemiologically related cases in a single institution 2a Sporadic hospital outbreaks Unrelated hospital outbreaks with independent, i.e. epidemiologically unrelated introduction or different strains, no autochthonous inter-institutional transmission reported 2b Regional spread More than one epidemiologically related outbreak confined to hospitals that are part of a regional referral network, suggestive of regional autochthonous inter-institutional transmission 3 Inter-regional spread Multiple epidemiologically related outbreaks occurring in different health districts, suggesting interregional autochthonous inter-institutional transmission 4 Endemic situation Most hospitals in a country are repeatedly seeing cases admitted from autochthonous sources 5 Plasmid-borne spread within and between species also can occur for bla KPC and is the dominant mode of dissemination for NDM and VIM genes. This epidemiology is more complex and harder to interrupt, complicating national intervention strategies.
Regardless of whether it is the spread of the strain or the gene that is dominant, the key to success in preventing the establishment of carbapenemase-producing Enterobacteriaceae is early detection and good diagnostic practice. Recent decisions by international breakpoint committees are taking into account the fact that MICs for carbapenemase-producing Enterobacteriaceae may represent a continuum close to or even overlapping with the top end of the wildtype distribution. Even the most conservative breakpoints will not assign all carbapenemase-producing Enterobacteriaceae into the non-susceptible class and may also classify strains with porin loss and presence of ESBL or AmpC lactamases as resistant. The epidemiological consequences of underdetection of carbapenemase producers with MICs in the sub-breakpoint range are still unclear. Current MIC breakpoints are set to guide treatment, but more clinical studies on the effectiveness of carbapenems against carbapenemase-producing Enterobacteriaceae with relatively low MICs are needed, as are simple, inexpensive, phenotypic tests to recognise producers, irrespective of MIC, with adequate specificity and sensitivity. Molecular confirmation tests are useful for reference purposes and are conveniently rapid when screening faecal samples during outbreaks.
Hospital outbreaks, defined as more than two epidemiologically related cases, need to be brought to the attention of regional health authorities as well as to all hospitals receiving referred patients. A close collaboration between the microbiological laboratory and the local and regional infection control team(s) is decisive for the prevention and control at local hospital level.

Areas for improvement
The workshop identified ten areas for improvement, displayed in the list below. Areas 1 to 6 recognise the need for better laboratory-based detection and surveillance, whereas areas 7 to 10 address infection control and clinical research needs.

Area 1: Ad hoc case ascertainment with existing laboratory capacity
• All routine diagnostic laboratories must test the susceptibility of all isolated Enterobacteriaceae, from all anatomical sites in each patient, with at least one carbapenem, specifically meropenem or imipenem*. Resistance to ertapenem is more prone to arise through combinations of impermeability and AmpC or ESBL activity, especially in Enterobacter spp., reducing specificity. • Laboratories should inform their local infection control teams of their tentative findings and report non-susceptibility of blood culture isolates to the national EARS-Net data manager.
• They should forward isolates to national reference centres for confirmation and molecular testing.

Area 2: Standardisation of detection and reporting
• Agreement needs to be achieved on the minimum test requirements for detection and data reporting of CNSE within the national and international reporting structures. • A panel of highly characterised CNSE isolates should be made available to all laboratories for test calibration. • For laboratories that wish to participate in national or international surveillance initiatives (EARS-Net) participation in regular external quality assessment exercises should be mandatory, and should include carbapenemases. • An improvement to EARS-Net would be a recommendation to report carbapenem-non-susceptibility rather than resistance, similar to the reporting of susceptibility for penicillin in Streptococcus pneumoniae. • A list of national reference and other centres with the required skills to identify molecular mechanisms of carbapenem resistance should be available to all clinical laboratories in each country.

Area 3: Need for consistent capacity building of reference diagnostics
• Workshops should be organised to train reference laboratory personnel on a set of phenotypic and genetic test methods to allow exhaustive characterisation of carbapenemase-producing isolates submitted by routine diagnostic laboratories. • The workshops shall follow a 'training the trainers approach' in order to provide European reference laboratories with the means to train peripheral laboratories. • This will increase the coherence of reporting and strengthen national surveillance and diagnostic capacity. • The training should ideally take place in an endemic country to provide course participants with handson experience of the workload, and to appraise the challenges in task management, procurement and costing.

Area 4: Need for structured surveys to determine sensitivity and specificity of defined breakpoints or other inclusion criteria
• A group of experts shall develop the protocol for a structured survey aiming to optimise a diagnostic algorithm to identify CNSE with a high degree of accuracy and a minimum number of false positives. • This will require agreement on a set of selection criteria (e.g. overall resistance profile and a meropenem MIC ≥ 0.5 mg/L) [54]. • Furthermore the sampling frame needs to be defined, a sample size estimated, and a design for roll-out to all European countries developed. • The results should not only provide the best sensitivity and specificity but also reveal the true prevalence of CNSE in a representative cross-sectional sample of the population.

Area 5: Need for a harmonised typing tool/initiative
• Molecular typing of CNSE is complex due to the multifaceted nature of their spread. Plasmid spread occurs among different species, and repeated introduction to Europe via travel may challenge typing laboratories with a near-random sample of strains circulating in other countries. • Nevertheless, the potential for rapid spread of single lineages, as seen with K. pneumoniae ST258 in Greece, Israel and recently in Poland and Italy, underscores the need for rapid assessment of the spread of such clones, as these are of particular public health importance. • It is therefore highly desirable to invest in the development of typing systems with a better resolution for strains and plasmids. • Only sequence-based data will provide the robustness and portability required for modern decentralised approaches. • In any case the practicability and applicability of typing methods must be considered.

Area 6: Need for central data collection on the dissemination and introduction of strains with particular public health importance
• With harmonised test methods, detection, typing and reporting criteria comes the ability to network the data collected at local and national levels into international databases that would be freely accessible and searchable for hospitals and public health agencies. • This will allow for early recognition of temporalspatial tends, outbreaks and importation by travel. Systems of this kind have been developed for infections caused by Legionella pneumophila and Staphylococcus aureus [89,90]. • We believe that the spread of CNSE and the consequent treatment problems create an urgent need for the construction of a similar IT platform to prevent these traits from becoming endemic in the European region.

Area 7: Need for guidelines for graded approaches to infection control
• Appropriate infection control measures need to be guided by epidemiological staging, which can be defined at national level as described above (See Tables 2 and 3). • If CNSE have not yet been reported, highly sensitive detection criteria coupled with an early warning system and preparedness should be in place. • For countries with sporadic outbreaks, infection control teams should be trained to implement measures to contain spread at the local level following a ready-to-use stepwise approach.
• Reporting the occurrence and the outcome of outbreaks will inform health authorities on the epidemiology and success of national strategies. • Countries with advanced-stage epidemiology should resort to screening and isolation in accordance with epidemiological and geographic extent of the cases reported. Such policies were successful in Israel. • In such settings, national health authorities should inform other EU Member States on the prevailing epidemiology, so that safe policies for patient transfer from their countries can be established.

Area 8: Antibiotic policy
• Antibiotic overuse and misuse are the main factors that select multidrug-resistant organisms such as CNSE from the commensal flora. • Diversification and de-escalation of antibiotic treatment, particularly carbapenems, fluoroquinolones but also third-generation cephalosporins, are key to the response to the CNSE emergence. • This should include guidelines on antibiotic use for non-severe infections (e.g. urinary tract infections) and an intensified dialogue with prescribers across Europe.

Area 9: Treatment and clinical research
• Clinical trials about the effectiveness of remaining alternative treatment strategies for CNSE infections are urgently needed. • Incentives need to be provided for the development of new antibiotics active against CNSE.

Area 10: Political commitment
• Importantly, the political commitment at national governmental as well as European level is critical. • The European Centre for Disease Prevention and Control should play a role in harmonising European surveillance, detection and identification strategies. • The World Health Organization (WHO) should address this issue in a proactive manner globally, possibly through the International Health Regulations which are an international legal instrument that is binding for 194 countries across the globe, including all the WHO member states. • Their aim is to help the international community prevent and respond to acute public health risks that have the potential to cross borders and threaten people worldwide [91].

Funding
The consultation and the work leading up to it was funded by the European Commission Directorate for Research and Development through the collaborative project Translational Research on Combating Antimicrobial Resistance (TROCAR) project number 223031.

*Authors' correction:
At the request of the authors, the following changes were made on 25 November 2010.

•
In the the top box of the Figure, 'imipenem ≥1 mg/L (nonwildtypea)' was changed to 'meropenem ≥0.5 mg/L [54]', and footnote a was deleted.

•
In the first bullet point of Area 1 in the section entitled 'Areas for improvement', 'specifically imipenem or meropenem' was changed to 'specifically meropenem or imipenem' In addition, on 10 December 2010, the following change was made in Table 2 at the request of the authors: 'Outbreak defined as more than two epidemiologically related cases in a single institution' should read 'Outbreak defined as two or more epidemiologically related cases in a single institution'.
On 17 January 2011, the names of Aurora Garcia-Fernandez and Maurine A. Leverstein -van Hall were corrected in the members' list of the CNSE Working Group.