Advisory Committee on the Microbiological Safety of Food recommendations for standardisation of antimicrobial resistance terminology

THE use of antimicrobial resistance (AMR) terminology in Food Standards Agency (FSA) research and surveillance publications was reviewed by the AMR subgroup of the FSA's independent Advisory Committee on the Microbiological Safety of Food (ACMSF). This article details the recommendations made by the ACMSF (referred to as ‘we’ throughout). The ACMSF's AMR subgroup was convened with the aim ‘to consider defining specific AMR-related terms, including their applicability in different situations’. A key consideration was the use of the terms ‘resistant’ or ‘resistance’ to describe bacteria when epidemiological cut-off values (ECOFFs) and/or clinical breakpoints (CBPs) are used to determine susceptibility. Consideration was also given to using the term ‘less susceptible’ rather than ‘resistant’ unless a CBP has been used or the ECOFF has been shown to be the same as, or lower than, the CBP. In addition, AMR from animals, definitions of intrinsic resistance, multidrug resistance and genotypic versus phenotypic resistance, particularly in light of the increasing use of sequencing-based methods such as whole genome sequencing (WGS) and metagenomics, were considered. While this article sets out the recommendations regarding AMR terminology that will be incorporated into FSA project reports, these recommendations may also provide a basis for the harmonisation of AMR terminology within the wider agri-food sector. Antimicrobial resistance data derived using epidemiological cut-off values should be interpreted cautiously, as these values may not align with clinical breakpoints The EU summary reports on antimicrobial resistance in zoonotic and indicator bacteria from people, animals and food also follow this approach, and in the case of quantitative data, define an isolate as ‘resistant’ to a selected antimicrobial when its minimum inhibitory concentration (MIC) value is above the cut-off value or the disc diffusion diameter is below the cut-off value. The Swedish Veterinary Antimicrobial Resistance Monitoring (SVARM) reports follow a slightly different approach, stating that ‘ECOFFS classify isolates with acquired reduced susceptibility as non-wild type’. In SVARM reports, non-wild type isolates are called ‘resistant’. This classification is relevant for monitoring purposes, but resistance defined in this manner does not always imply clinical resistance. ECOFFs are determined by a different approach than CBPs and do not consider the results of clinical efficacy studies, dosing and route of administration of the antimicrobial agents, nor the agent's pharmacokinetic and pharmacodynamic parameters in people or the animal species in which it is used. For some combinations of bacteria and antimicrobials, the ECOFF and CBP lie at the same MIC value despite the difference in their derivation. However, there are differences in the ECOFF and CBP for some clinically important combinations of bacteria and antimicrobials. For example, although the ECOFF of ciprofloxacin (an antimicrobial considered ‘critically important’ by the World Health Organization) is low for Salmonella species (MIC <0.06 mg/l), there is evidence indicating a poor clinical response to ciprofloxacin treatment in people with systemic Salmonella infections.1 Regarding the suggestion that the term ‘less susceptible’ should be used instead of ‘resistant’ unless a CBP has been used or the ECOFF has been shown to be the same as, or lower than, the CBP, we consider that this may lead to confusion as the term ‘resistant’ is widely used in reports involving foodborne bacteria. It should also be noted that, in Europe, it has been suggested that the term ‘zero resistance’ should replace ‘full susceptibility’ as this may be more easily understood by non-specialist readers of AMR reports. However, we consider that introducing the term ‘zero resistance’ could in itself be confusing, particularly when genes encoding resistance to several antibiotics may be present but not expressed. As such, we recommend that the term ‘full susceptibility’ should still be used for bacteria that do not show resistance to any of the antimicrobials tested. Although AMR associated with the use of antimicrobials in food-producing animals is not primarily the responsibility of the FSA, we believe this is important to consider within the One Health context, as some antimicrobials that are used exclusively in animals can influence the development of resistance to antibiotics used in human medicine. For example, the veterinary use of apramycin can result in resistance to gentamicin – an antibiotic used to treat severe infections caused by Gram-negative bacilli in people.2 Due to this potential for veterinary antimicrobial use to drive AMR in medically important bacteria, it is critical that approved standards are used for antimicrobial agents and bacteria in veterinary medicine. To this end, some authors have recommended that standards from the USA Clinical Laboratory Standards Institute Veterinary Antimicrobial Susceptibility Testing subcommittee (CLSI-VAST) should be applied.3 However, we do not consider that using CLSI-VAST breakpoints is, at present, appropriate for the epidemiological surveillance of AMR in bacterial pathogens associated with infections primarily spread by contaminated food. Furthermore, it should be noted that some of the CLSI-VAST interpretive criteria may not be appropriate to the treatment of certain pathogens causing infections in animals in the UK and the EU, where dosing regimens differ from those normally used in the USA. Intrinsic antimicrobial resistance is a naturally occurring phenomenon that predates antimicrobial therapy and can be observed in almost all bacterial species. Intrinsic resistance is defined as a significant proportion (typically more than 90 per cent) of a bacterial species being naturally resistant to a certain antimicrobial or family of antimicrobials without the need for mutation or gain of further genes.4 This means that these antimicrobials can never be used at normal therapeutic doses to treat infections caused by that species of bacteria. Although EUCAST has decided to replace the term ‘intrinsic’ with the terms ‘expected susceptible phenotype’ and ‘expected resistant phenotype’, these terms are designed for clinical microbiology laboratories reporting to physicians dealing with medical patients. We consider that ‘intrinsic resistance’ will be more readily understood by relevant stakeholders in the agri-food sector. It is also our opinion that ‘intrinsic resistance’ is meaningful when applied to bacteria associated with foodborne infections, and the term should thus be retained for FSA-funded projects. However, consideration may be given to replacing the term ‘intrinsic’ with ‘natural’, as the general public may better understand this term. Following a joint initiative by the European Centre for Disease Prevention and Control and the USA Centers for Disease Control and Prevention to create a standardised international terminology with which to describe acquired resistance profiles,5 we recommend that MDR is considered as non-intrinsic resistance to three or more different classes of antimicrobials. Indeed, this terminology is now widely used. WGS and metagenomics are increasingly being used in the field of AMR, either independently or in conjunction with culture-dependent phenotypic methods. However, there is an additional challenge in interpreting and presenting the outcome of methods used to detect the presence of AMR genes. WGS of single bacteria or metagenomics of mixed communities offers the potential to predict antimicrobial susceptibility (predictive MICs) from a single assay and may be particularly useful in identifying genes conferring resistance by small changes in susceptibility, as is the case with some multidrug resistance genes. These methods could also potentially identify the presence of resistance determinants for a wider number of antimicrobial classes than might be screened by phenotypic testing. However, this may be problematic as it could result in bacteria being classified as ‘resistant’ or ‘multi-resistant’ when some genes identified by genotypic methods are either not relevant or not expressed in the resistance phenotype. The use of these tools is likely to increase, especially in combination with artificial intelligence and machine learning algorithms. Therefore, it will be important to find a way to explain what the detection of resistance genes using these tools means and how the findings can be described in terms of resistance and multiple resistance, with appropriate caveats. It is worth noting that in 2017, EUCAST established a subcommittee to review the current development status of WGS for bacterial antimicrobial susceptibility testing. They concluded that the published evidence for using WGS as a tool to infer antimicrobial susceptibility accurately was either poor or non-existent, and the evidence and knowledge base required significant expansion.6 A suggested way forward for AMR surveillance activities is to list and describe genes identified by metagenomics with an indication of the antimicrobial resistances potentially conferred, whether such resistances are clinically significant in the organism(s) in which they are identified and whether the genes have the capacity to be transferred to other organisms (including key foodborne pathogens). However, metagenomics is a rapidly developing technology, and its use in relation to the detection of AMR genes should be regularly reviewed. This review sets out ACMSF's position regarding AMR terminology in FSA reports and provides recommendations to support clarity in AMR reports in the wider agri-food sector. FSA has accepted these recommendations and agreed that a statement should be included as a precursor to all FSA reports involving AMR surveillance activities to indicate that when ECOFF terminology is used to indicate ‘resistance’, such resistance is not necessarily at clinical levels. Furthermore, as AMR data may be analysed by applying both CBPs and ECOFFs, the agri-food community needs to ensure they clearly indicate which interpretive criteria have been used in their reports.