Latent class analysis to assess whole-genome sequencing versus broth microdilution for monitoring antimicrobial resistance in livestock

Ayla Hesp; Kees Veldman; Michael S M Brouwer; Jaap A Wagenaar; Dik Mevius; Gerdien van Schaik

doi:10.1016/j.prevetmed.2021.105406

Latent class analysis to assess whole-genome sequencing versus broth microdilution for monitoring antimicrobial resistance in livestock

Ayla Hesp
, Kees Veldman
, Michael S M Brouwer
, Jaap A Wagenaar
, Dik Mevius
, Gerdien van Schaik

Clinical Infectiology

Utrecht University, University Medical Center Utrecht
Host Pathogen Interaction and Diagnostics Development

Research output: Contribution to journal › Article › Academic › peer-review

Abstract

Antimicrobial resistance (AMR) monitoring in animals is performed in commensal Escherichia coli, and other microorganisms relevant for human or veterinary health. Due to advances in the field and major reductions in cost, it is expected that whole-genome sequencing (WGS)-based antimicrobial susceptibility testing (AST) will (partly) replace culture-based AST. So far, no studies have been performed without using culture-based AST as the gold standard. Our aim was to use Bayesian latent class analysis to evaluate the accuracy of susceptibility testing of commensal E. coli by WGS-based AST versus culture-based AST as this test does not assume a gold standard. OpenBUGS was used to model two independent tests in three animal populations (N = 150, 50 bacterial isolates per population): veal calves, pigs, and broilers. This resulted in the first estimation of sensitivity and specificity of WGS-based AST versus culture-based AST to detect AMR without a gold standard. Both methods had high sensitivity (>0.92, lowest limit probability interval: 0.76) and specificity was generally high for both methods for all antimicrobial classes except for aminoglycosides and macrolides. We compared WGS results for different length and identity settings (%) of gene alignment and found few differences between the 60/90, 90/90 and 95/95 settings. We recommend to further investigate sensitivity and specificity of WGS-based AST by means of latent class analysis, especially for low-prevalent resistance.

Original language	English
Article number	105406
Pages (from-to)	1-6
Number of pages	6
Journal	Preventive Veterinary Medicine
Volume	193
Early online date	4 Jun 2021
DOIs	https://doi.org/10.1016/j.prevetmed.2021.105406
Publication status	Published - Aug 2021

Bibliographical note

Funding Information:
We thank the Ministry of Agriculture, Nature and Food Quality for funding the WOT-O project (WOT-01-002-003) for the sequencing of the bacterial isolates, and the EFFORT consortium for funding the susceptibility testing of the bacterial strains. Roosmarijn Luiken and Liese van Gompel we would like to thank for organizing the sampling of the EFFORT strains. We thankfully acknowledge Yvon Geurts, Alieda van Essen, Joop Testerink, Marga Japing and Albert de Boer for their technical assistance. We also acknowledge the contribution of Jeanet van der Goot to this work and Daniela Ceccarelli for analysing the MIC dataset within EFFORT, and Haitske Graveland for the project management within EFFORT. Finally, we would like to thank Bas Engel for his aid in optimizing the OpenBUGS coding.

Funding Information:
This research was supported by the Ministry of Agriculture, Nature and Food Quality ( WOT-01-002-003 ) and Wageningen Bioveterinary Research . The susceptibility testing of the bacterial strains was part of the Ecology from Farm to Fork Of microbial drug Resistance and Transmission (EFFORT) project, funded by the European Commission, 7th Framework Program for Research and Innovation ( FP7-KBBE-2013–7 , grant agreement: 613754).

Publisher Copyright:
© 2021 The Authors

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 3 Good Health and Well-being

Keywords

Antimicrobial resistance
Bayesian
Broth microdilution
Escherichia coli
Livestock
Monitoring
Surveillance
Whole-genome sequencing

Access to Document

10.1016/j.prevetmed.2021.105406Licence: CC BY

1-s2.0-S0167587721001501-mainFinal published version, 405 KBLicence: CC BY

Cite this

@article{dcc106490a1f465a9fbd690f342204a4,

title = "Latent class analysis to assess whole-genome sequencing versus broth microdilution for monitoring antimicrobial resistance in livestock",

abstract = "Antimicrobial resistance (AMR) monitoring in animals is performed in commensal Escherichia coli, and other microorganisms relevant for human or veterinary health. Due to advances in the field and major reductions in cost, it is expected that whole-genome sequencing (WGS)-based antimicrobial susceptibility testing (AST) will (partly) replace culture-based AST. So far, no studies have been performed without using culture-based AST as the gold standard. Our aim was to use Bayesian latent class analysis to evaluate the accuracy of susceptibility testing of commensal E. coli by WGS-based AST versus culture-based AST as this test does not assume a gold standard. OpenBUGS was used to model two independent tests in three animal populations (N = 150, 50 bacterial isolates per population): veal calves, pigs, and broilers. This resulted in the first estimation of sensitivity and specificity of WGS-based AST versus culture-based AST to detect AMR without a gold standard. Both methods had high sensitivity (>0.92, lowest limit probability interval: 0.76) and specificity was generally high for both methods for all antimicrobial classes except for aminoglycosides and macrolides. We compared WGS results for different length and identity settings (\%) of gene alignment and found few differences between the 60/90, 90/90 and 95/95 settings. We recommend to further investigate sensitivity and specificity of WGS-based AST by means of latent class analysis, especially for low-prevalent resistance.",

keywords = "Antimicrobial resistance, Bayesian, Broth microdilution, Escherichia coli, Livestock, Monitoring, Surveillance, Whole-genome sequencing",

author = "Ayla Hesp and Kees Veldman and Brouwer, \{Michael S M\} and Wagenaar, \{Jaap A\} and Dik Mevius and \{van Schaik\}, Gerdien",

note = "Funding Information: We thank the Ministry of Agriculture, Nature and Food Quality for funding the WOT-O project (WOT-01-002-003) for the sequencing of the bacterial isolates, and the EFFORT consortium for funding the susceptibility testing of the bacterial strains. Roosmarijn Luiken and Liese van Gompel we would like to thank for organizing the sampling of the EFFORT strains. We thankfully acknowledge Yvon Geurts, Alieda van Essen, Joop Testerink, Marga Japing and Albert de Boer for their technical assistance. We also acknowledge the contribution of Jeanet van der Goot to this work and Daniela Ceccarelli for analysing the MIC dataset within EFFORT, and Haitske Graveland for the project management within EFFORT. Finally, we would like to thank Bas Engel for his aid in optimizing the OpenBUGS coding. Funding Information: This research was supported by the Ministry of Agriculture, Nature and Food Quality ( WOT-01-002-003 ) and Wageningen Bioveterinary Research . The susceptibility testing of the bacterial strains was part of the Ecology from Farm to Fork Of microbial drug Resistance and Transmission (EFFORT) project, funded by the European Commission, 7th Framework Program for Research and Innovation ( FP7-KBBE-2013–7 , grant agreement: 613754). Publisher Copyright: {\textcopyright} 2021 The Authors",

year = "2021",

month = aug,

doi = "10.1016/j.prevetmed.2021.105406",

language = "English",

volume = "193",

pages = "1--6",

journal = "Preventive Veterinary Medicine",

issn = "0167-5877",

publisher = "Elsevier BV",

}

TY - JOUR

T1 - Latent class analysis to assess whole-genome sequencing versus broth microdilution for monitoring antimicrobial resistance in livestock

AU - Hesp, Ayla

AU - Veldman, Kees

AU - Brouwer, Michael S M

AU - Wagenaar, Jaap A

AU - Mevius, Dik

AU - van Schaik, Gerdien

N1 - Funding Information: We thank the Ministry of Agriculture, Nature and Food Quality for funding the WOT-O project (WOT-01-002-003) for the sequencing of the bacterial isolates, and the EFFORT consortium for funding the susceptibility testing of the bacterial strains. Roosmarijn Luiken and Liese van Gompel we would like to thank for organizing the sampling of the EFFORT strains. We thankfully acknowledge Yvon Geurts, Alieda van Essen, Joop Testerink, Marga Japing and Albert de Boer for their technical assistance. We also acknowledge the contribution of Jeanet van der Goot to this work and Daniela Ceccarelli for analysing the MIC dataset within EFFORT, and Haitske Graveland for the project management within EFFORT. Finally, we would like to thank Bas Engel for his aid in optimizing the OpenBUGS coding. Funding Information: This research was supported by the Ministry of Agriculture, Nature and Food Quality ( WOT-01-002-003 ) and Wageningen Bioveterinary Research . The susceptibility testing of the bacterial strains was part of the Ecology from Farm to Fork Of microbial drug Resistance and Transmission (EFFORT) project, funded by the European Commission, 7th Framework Program for Research and Innovation ( FP7-KBBE-2013–7 , grant agreement: 613754). Publisher Copyright: © 2021 The Authors

PY - 2021/8

Y1 - 2021/8

N2 - Antimicrobial resistance (AMR) monitoring in animals is performed in commensal Escherichia coli, and other microorganisms relevant for human or veterinary health. Due to advances in the field and major reductions in cost, it is expected that whole-genome sequencing (WGS)-based antimicrobial susceptibility testing (AST) will (partly) replace culture-based AST. So far, no studies have been performed without using culture-based AST as the gold standard. Our aim was to use Bayesian latent class analysis to evaluate the accuracy of susceptibility testing of commensal E. coli by WGS-based AST versus culture-based AST as this test does not assume a gold standard. OpenBUGS was used to model two independent tests in three animal populations (N = 150, 50 bacterial isolates per population): veal calves, pigs, and broilers. This resulted in the first estimation of sensitivity and specificity of WGS-based AST versus culture-based AST to detect AMR without a gold standard. Both methods had high sensitivity (>0.92, lowest limit probability interval: 0.76) and specificity was generally high for both methods for all antimicrobial classes except for aminoglycosides and macrolides. We compared WGS results for different length and identity settings (%) of gene alignment and found few differences between the 60/90, 90/90 and 95/95 settings. We recommend to further investigate sensitivity and specificity of WGS-based AST by means of latent class analysis, especially for low-prevalent resistance.

AB - Antimicrobial resistance (AMR) monitoring in animals is performed in commensal Escherichia coli, and other microorganisms relevant for human or veterinary health. Due to advances in the field and major reductions in cost, it is expected that whole-genome sequencing (WGS)-based antimicrobial susceptibility testing (AST) will (partly) replace culture-based AST. So far, no studies have been performed without using culture-based AST as the gold standard. Our aim was to use Bayesian latent class analysis to evaluate the accuracy of susceptibility testing of commensal E. coli by WGS-based AST versus culture-based AST as this test does not assume a gold standard. OpenBUGS was used to model two independent tests in three animal populations (N = 150, 50 bacterial isolates per population): veal calves, pigs, and broilers. This resulted in the first estimation of sensitivity and specificity of WGS-based AST versus culture-based AST to detect AMR without a gold standard. Both methods had high sensitivity (>0.92, lowest limit probability interval: 0.76) and specificity was generally high for both methods for all antimicrobial classes except for aminoglycosides and macrolides. We compared WGS results for different length and identity settings (%) of gene alignment and found few differences between the 60/90, 90/90 and 95/95 settings. We recommend to further investigate sensitivity and specificity of WGS-based AST by means of latent class analysis, especially for low-prevalent resistance.

KW - Antimicrobial resistance

KW - Bayesian

KW - Broth microdilution

KW - Escherichia coli

KW - Livestock

KW - Monitoring

KW - Surveillance

KW - Whole-genome sequencing

UR - https://www.scopus.com/pages/publications/85108233494

U2 - 10.1016/j.prevetmed.2021.105406

DO - 10.1016/j.prevetmed.2021.105406

M3 - Article

C2 - 34147959

SN - 0167-5877

VL - 193

SP - 1

EP - 6

JO - Preventive Veterinary Medicine

JF - Preventive Veterinary Medicine

M1 - 105406

ER -

Latent class analysis to assess whole-genome sequencing versus broth microdilution for monitoring antimicrobial resistance in livestock

Abstract

Bibliographical note

UN SDGs

Keywords

Access to Document

Other files and links

Fingerprint

Cite this