Discussion
The surveillance definition of MDR requires the availability of a large number of susceptibility testing results for the correct classification of isolates [
11]. If monitoring and comparison of the prevalence of MDR-GNB is to be an aim for on-going surveillance activities collecting routine microbiology AST data, the optimal strategy for detecting MDR organisms from such data needs to be established. Current surveillance activities tend to request the AST results for a limited subset of antibiotic classes listed by the expert MDR classification algorithm [
12].
In our dataset, the percentage of MDR-GNB isolates was significantly lower (13%) when based on a more limited set of antibiotic classes, such as that used by EARS-Net, compared with the full set available (30%). Utilising the full set of antibiotic classes reportable as part of the ARPEC project, the proportion of paediatric MDR E. coli, K. pneumoniae and P. aeruginosa isolates was around 30% and similar for all three pathogens. Such high levels of isolates with resistance to multiple drugs are concerning and of interest for tracking the epidemiology of resistant GNB over time.
Our study raises several important points regarding the potential of capturing MDR-GNB based on currently available routine microbiology data purely for surveillance:
(1)
Routine reporting of AST data by the 19 European laboratories participating in ARPEC only variably included results for requested antibiotic classes that are part of the classification algorithms for E. coli, K. pneumoniae and P. aeruginosa. A direct application of the MDR algorithms is, therefore, not possible.
(2)
Limited AST result data also cannot be used to reliably estimate the proportion of MDR-GNB. As the ARPEC dataset includes only European isolates, the performance of the current European surveillance system was evaluated. The EARS-Net set of antibiotic classes appeared to lack sensitivity for detecting MDR-GNB. Inclusion of additional frequently tested and reported antibiotic classes increased the detection of MDR E. coli and K. pneumoniae (from 30% detected by the EARS-Net set to 90% based on the Routine set for E. coli and from 46 to 92% for K. pneumoniae). This was in contrast to P. aeruginosa, for which the ARPEC set included only one additional antibiotic class compared with EARS-Net reporting.
(3)
A small number of individual PACCs currently represent the typical method for reporting antimicrobial resistance surveillance internationally. Disappointingly, resistance detected in individual PACCs was not reliable in detecting MDR isolates. This was especially marked for E. coli isolates, for which resistance to higher-generation cephalosporins, for example, had a sensitivity of only 36% for detecting MDR. Escherichia coli is the GNB with the largest number of antibiotic classes in the MDR classification algorithm and in ARPEC reporting. This may increase the detection of many different resistance combinations, especially if multiple different resistance phenotypes occur.
Some of the challenges may be explained by the fact that surveillance collects data primarily generated to inform clinical decision-making: approaches to AST are likely to be guided by the need to optimally inform patient therapy rather than by the need to generate a complete AST dataset for MDR classification. This type of selective AST based on clinical needs could introduce bias when these data are interpreted for public health purposes [
28]. Bias could be magnified when laboratories engage in so-called first- and second-line testing: some antibiotic classes are evaluated only when resistance to antibiotics included in a first-line panel is detected [
16].
Several limitations need to be considered when interpreting the ARPEC data. ARPEC does not cover all antibiotic classes recommended in the recent expert proposal [
11]. It is, therefore, possible that some isolates identified as not MDR in ARPEC would, in fact, be MDR if AST data for all relevant classes were available. It is also possible that antibiotic classes tested for some of the reported isolates were suppressed during ARPEC data entry. This seems unlikely, given the relative uniformity of reporting for each species by each laboratory.
The actual percentages of MDR-GNB reported in this study should be interpreted with caution, as hospitals reporting to ARPEC were tertiary institutions with a patient population not representative of patients in other inpatient settings and potentially at higher risk of MDR-GNB [
20,
21]. Pooling of data prohibits the identification of any differences between individual participating centres, some of which may have had higher or lower than average MDR-GNB percentages. Finally, the burden of MDR-GNB cannot be estimated because data are presented as resistance percentages rather than infection prevalence or incidence [
29].
All isolates represent neonatal or paediatric blood cultures. The antibiotics used to treat bloodstream infections in neonates and children may differ from treatment choices for adults. This could be reflected in the antibiotic classes selected for AST, potentially limiting the transferability of the results to isolates from adults. However, most laboratories process microbiological samples from both adult and childhood patients. It is unlikely that AST strategies will be relevantly different for neonatal and paediatric isolates in these settings.
Surveillance of antimicrobial resistance patterns and trends is necessary to target interventions to reduce the selection and spread of resistant bacteria, and often relies on routine samples collected as part of on-going clinical care. The limitations and biases associated with the use of routine microbiology data in surveillance have been widely discussed [
8,
28,
29]. Resistance percentages of individual PACCs and the EARS-Net set currently in use in Europe do not, on the whole, provide reliable MDR estimates. This study shows that, if MDR surveillance is to be added to the task list of on-going international surveillance, interpretation of the new algorithm will be limited by the variability in AST strategies in microbiological laboratories. MDR-GNB detection could be immediately improved by added surveillance of antibiotic classes already widely tested as part of clinical care. As demonstrated, a larger percentage of MDR-GNB isolates is likely to be identified with such an approach.
Acknowledgements
The authors thank all contributors to the ARPEC bacteraemia antimicrobial resistance data collection. The contributors are as follows: C. Berger, MD, University Children’s Hospital Zurich, Zürich, Switzerland; S. Esposito, MD, PhD, E. Danieli, MBiol and R. Tenconi, MD, Pediatric Clinic 1, Department of Pathophysiology and Transplantation, Fondazione IRCCS Ca’ Granda Ospedale Maggiore Policlinico, Milan, Italy; L. Folgori, MD, Department of Pediatrics (DPUO), University of Rome Tor Vergata, Bambino Gesù Children’s Hospital, IRCCS, Rome, Italy; P. Bernaschi, MD, Unit of Microbiology, Laboratory Department, Bambino Gesù Children’s Hospital, IRCCS, Rome, Italy; B. Santiago, MD and J. Saavedra, MD, PhD, Pediatric Infectious Diseases Division, Gregorio Marañón Hospital, Madrid, Spain; E. Cercenado, PharmD, Servicio de Microbiologia y Enfermedades Infecciosas, Gregorio Marañón Hospital, Madrid, Spain; A. Brett, MD and F. Rodrigues, MD, Infectious Diseases Unit and Emergency Service, Hospital Pediátrico de Coimbra, Centro Hospitalar e Universitário de Coimbra, Coimbra, Portugal; M. Cizman, MD, PhD, Department of Infectious Diseases, UMC Ljubljana, Ljubljana, Slovenia; J. Jazbec, MD, PhD, Children’s Hospital, UMC Ljubljana, Ljubljana, Slovenia; J. Babnik, MD and Maja Pavcnik, MD, PhD, UMC Ljubljana, Ljubljana, Slovenia; M Pirš (Pirs), MD, PhD and M. Mueller Premrov, MD, PhD, Institute of Microbiology and Immunology, Medical Faculty, University of Ljubljana, Ljubljana, Slovenia; M Lindner, PhD and M. Borte, MD, Hospital St. Georg, Leipzig, Germany; N. Lippmann, MD and V. Schuster, MD, University Hospital Leipzig, Leipzig, Germany; A. Thürmer, MD and F. Lander, MD, University Hospital Dresden, Dresden, Germany; J. Elias, MD and J. Liese, MD, MsC, University Hospital Würzburg, Würzburg, Germany; A. Durst, MD and S. Weichert, MD, University Hospital Mannheim, Mannheim, Germany; C. Schneider, MD and M. Hufnagel, MD, University Medical Center Freiburg, Freiburg im Breisgau, Germany; A. Rack, MD and J. Hübner, MD, University Hospital München, Munich, Germany; F. Dubos, MD, PhD and M. Lagree, MD, Pediatric Emergency Unit and Infectious Diseases, Université Lille Nord-de-France, UDSL, CHRU Lille, Lille, France; R. Dessein, Laboratory of Microbiology, Pathology-Biology Center, Lille-2 University, UDSL, CHRU Lille, Lille, France; P. Tissieres, MD, PhD, Pediatric and Neonatal ICU, AP-HP, Bicêtre Hospital, France; G. Cuzon, MD, PhD, Department of Bacteriology, AP-HP, Bicêtre Hospital, France; V. Gajdos MD, PhD, Pediatric Department, Antoine Béclère Hospital, Assistance Publique–Hôpitaux de Paris, Paris Sud University, Clamart, France; F. Doucet-Populaire, Laboratory of Microbiology and Infection Control, Antoine-Béclère Hospital, Assistance Publique–Hôpitaux de Paris, Paris Sud University, Clamart, France; V. Usonis, MD, PhD, Vilnius University Clinic of Children Diseases, Vilnius, Lithuania and Children’s Hospital, Affiliate of Vilnius University Hospital Santariskiu Klinikos, Vilnius, Lithuania; V. Gurksniene, MD, and Genovaite Bernatoniene, MD, Children’s Hospital, Affiliate of Vilnius University Hospital Santariskiu Klinikos, Vilnius, Lithuania; M. Tsolia, MD, PhD and N. Spyridis, MD, PhD, Second Department of Paediatrics, National and Kapodistrian University of Athens School of Medicine, Athens, Greece; E. Lebessi, MD, PhD and A. Doudoulakakis, MD, Department of Microbiology, “P. and A. Kyriakou” Children’s Hospital, Athens, Greece; I. Lutsar, MD, PhD and S. Kõljalg, MD, PhD, University of Tartu, Tartu, Estonia; T. Schülin, MD, PhD, Department of Medical Microbiology, Radboud University Medical Center, Nijmegen, The Netherlands; A. Warris, MD, PhD, Department of Paediatric Infectious Diseases, Radboud University Medical Center, Nijmegen, The Netherlands.