EUCAST expert rules in antimicrobial susceptibility testing

EUCAST expert rules have been developed to assist clinical microbiologists and describe actions to be taken in response to specific antimicrobial susceptibility test results. They include recommendations on reporting, such as inferring susceptibility to other agents from results with one, suppression of results that may be inappropriate, and editing of results from susceptible to intermediate or resistant or from intermediate to resistant on the basis of an inferred resistance mechanism. They are based on current clinical and/or microbiological evidence. EUCAST expert rules also include intrinsic resistance phenotypes and exceptional resistance phenotypes, which have not yet been reported or are very rare. The applicability of EUCAST expert rules depends on the MIC breakpoints used to define the rules. Setting appropriate clinical breakpoints, based on treating patients and not on the detection of resistance mechanisms, may lead to modification of some expert rules in the future.     

Antimicrobial susceptibility testing of Clostridium difficile using EUCAST epidemiological cut-off values and disk diffusion correlates

With the emergence of reduced susceptibility of Clostridium difficile to metronidazole and vancomycin the value of antimicrobial susceptibility testing has increased. The aim of our study was to evaluate disk diffusion for susceptibility testing of C. difficile by comparing disk diffusion results with MICs from gradient tests and to propose zone diameter breakpoint correlates for the EUCAST epidemiological cut-off values (ECOFFs) recently published. We tested 211 clinical isolates of C. difficile, from patients with diarrhoea hospitalized at Aarhus and Odense University Hospitals, Denmark. Furthermore, ten clinical isolates of C. difficile from the Anaerobe Reference Laboratory, University Hospital of Wales, with known reduced susceptibility to either metronidazole or vancomycin, were included. Isolates were tested with Etest gradient strips and disk diffusion towards metronidazole, vancomycin and moxifloxacin on Brucella Blood Agar supplemented with hemin and vitamin K. We found an excellent agreement between inhibition zone diameter and MICs. For each MIC value, the inhibition zones varied from 0 to 8 mm, with 93% of values within 6 mm for metronidazole, 95% of values within 4 mm for vancomycin, and 98% of values within 4 mm for moxifloxacin. With proposed zone diameter breakpoints for metronidazole, vancomycin and moxifloxacin of WT ≥ 23 mm, WT ≥ 19 and WT ≥ 20 mm, respectively, we found no very major errors and only major errors below 2%. In conclusion, we suggest that disk diffusion is an option for antimicrobial susceptibility testing of C. difficile.     

Development of a EUCAST disk diffusion method for the susceptibility testing of rapidly growing anaerobic bacteria using Fastidious Anaerobe Agar (FAA): a development study using Bacteroides species

ObjectivesAntimicrobial resistance among anaerobic bacteria is increasing, leading to a growing demand for inexpensive and reliable susceptibility testing methods. The aim of this study was to determine the suitability of Fastidious Anaerobe Agar (FAA) as a medium for disk diffusion for rapidly growing anaerobic bacteria.MethodsReproducibility of zone diameters and quality of growth were tested using six quality control (QC) strains. We compared four anaerobic incubation systems, two incubation temperatures (35°C and 37°C), and FAA from four manufacturers. The effect of incubation for 16–20 hours instead of 24 hours was tested on ten randomly selected isolates of the Bacteroides fragilis group. The final method was tested on 170 clinical B. fragilis-group isolates and compared to agar dilution MICs.ResultsAfter 24 hours' incubation, all QC strains demonstrated confluent growth. The different anaerobic incubation systems were equal regarding quality of growth and zone diameters. Incubation at 35°C resulted in slightly larger zones (1–2 mm) than at 37°C. Except for Acumedia FAA, the different manufacturers showed good agreement in zone diameters. All B. fragilis-group isolates displayed confluent growth after 16–20 hours. Metronidazole inhibition zones correlated well with the reference MICs. There was an area of poorer separation for meropenem and piperacillin–tazobactam between 19-27 and 14-23 mm respectively. Prolonged incubation (40–44 h) of clindamycin resulted in better separation and the area of overlap was reduced from 13 to 8 mm compared with 16–20 hours' incubation.ConclusionFAA is a suitable medium for disk diffusion of these rapidly growing anaerobic bacteria.     

Antimicrobial susceptibility testing of Kingella kingae with broth microdilution and disk diffusion using EUCAST recommended media

ObjectivesIncreasing use of improved culture techniques and sensitive nucleic acid amplification assays have resulted in recognition of Kingella kingae as an important cause of invasive infections in young children, especially in septic arthritis, osteomyelitis, bacteraemia, and endocarditis. In 2016, EUCAST established clinical MIC breakpoints for K. kingae (published in EUCAST Clinical Breakpoint Tables v 7.0, 2017). The present study was carried out to produce MIC-zone diameter correlations for K. kingae on an international collection of isolates, with the aim of suggesting zone diameter breakpoints corresponding to the clinical MIC breakpoints.MethodsAntimicrobial susceptibility testing was performed for 18 clinically relevant agents on a collection of 159 clinical isolates of K. kingae. Broth microdilution MIC determination and disk diffusion were performed according to EUCAST recommendations for fastidious organisms.ResultsThe correlation between MICs and zone diameters was good for all agents with EUCAST breakpoints for K. kingae. β-lactamase was detected in 41 isolates (26%) and these isolates were resistant to aminopenicillins. These isolates were also resistant to trimethoprim-sulfamethoxazole. Resistance to tetracyclines was detected in 8% of all isolates. All resistant isolates were correctly categorized for these agents with the proposed zone diameter breakpoints. One isolate, resistant to erythromycin but susceptible to other macrolides, was categorized as susceptible with erythromycin disk diffusion. No resistance was detected for the cephalosporins, carbapenems, and fluoroquinolones tested.ConclusionBased on the results in this study, zone diameter breakpoints for K. kingae calibrated to EUCAST clinical MIC breakpoints were proposed and approved by EUCAST.     

Calibration of fusidic acid disk diffusion susceptibility testing of Staphylococcus areus

Single strain regression analysis, SRA, was used to calibrate disk diffusion fusidic acid susceptibility testing of Staphylococcus aureus in two laboratories using different standard methods but the same interpretative MIC limits. SRA equation constants were calculated using five different fusidic acid disk contents (1.5, 5, 15, 50, 150 microg). These disks were tested on five separate occasions against quality control strain S. aureus ATCC 29213. The National Committee for Clinical Laboratory Standards (NCCLS) method was employed in Tartu, Estonia (TE) and the Swedish Reference Group for Antibiotics (SRGA) method in Sweden at the Karolinska Hospital (KS). SRA constants obtained were used for calculating zone breakpoints corresponding to MIC breakpoints recommended by the SRGA (S < or = 0.5 mg/L, R > or = 1 mg/L). Zone diameter histograms from KS, performed with a 50 microg disk, and from TE, using a 10 microg disk, showed a clustering of wild type strains around 41 mm and 30 mm, respectively, reflecting differences in methodology. Zone breakpoints calculated from the equations were validated by comparison with the histograms. Breakpoints were also calculated for a suggested lower disk content in Sweden, 10 microg, and validated in tests of clinical isolates and by histogram analysis.     

Evaluation of ten brands of pre-poured Mueller-Hinton agar plates for EUCAST disc diffusion testing

ObjectivesMueller-Hinton agar (MHA) is recommended by EUCAST and CLSI for disc diffusion antimicrobial susceptibility testing (AST). We have previously investigated the quality of dehydrated MHA from several manufacturers. In this study, we evaluated the performance of ten commercial brands of pre-poured MHA plates.MethodsAST was performed according to EUCAST methodology and results analyzed against targets and ranges in EUCAST quality control (QC) tables. MHA plates from different brands were tested in triplicate against four non-fastidious QC strains. The agar depth and pH were measured for all products.ResultsThe best performance was observed for MHA from Becton Dickinson (BBL MHA II), bioMérieux (MHE agar) and Hardy Diagnostics, for which >97% of zone diameters were within QC ranges and >60% on target ±1 mm. The poorest performance was seen for plates from HiMedia (MHA and MHA no. 2), where 20% and 18% of readings were outside the QC ranges, respectively. The differences in pH and agar depth of the products were small and mostly within EUCAST specifications.DiscussionThe accuracy and reproducibility of disc diffusion AST depends on standardised procedures and high-quality discs and media. The performance among ten brands of pre-poured MHA plates differed significantly. The results indicate a poorer performance for pre-poured commercial plates as compared to in-house prepared plates from dehydrated powder of corresponding brands in our previous study. Manufacturers and clinical laboratories have a shared responsibility for the quality of AST. EUCAST provides QC criteria to be used both by manufacturers and laboratories.     

Interpretive criteria of ceftibuten disk diffusion susceptibility tests according to the DIN 58 940 method

Objective   This study aimed to establish interpretive criteria for agar diffusion tests with ceftibuten disks according to DIN standards.
Methods   Minimal inhibitory concentrations (MICs) and inhibition zones produced by ceftibuten in the disk diffusion test were determined for 275 recent bacterial isolates, including 11 species with 25 strains each. Regression analysis was performed for two disk loads (10 µg and 30 µg).
Results   Correlation of MICs and zone diameters was good, with correlation coefficients of r  = − 0.97 for both tested disk loads. Evaluation of the calculated zone size criteria for all species showed no very major discrepancies or no major discrepancies. The 30-µg disks, however, produced unacceptably large inhibition zones for very susceptible strains, so that usage of 10-µg disks must be recommended when testing according to DIN standards.
Conclusion   Based on the MIC breakpoints recommended by the DIN (≥8 mg/L and ≤ 1 mg/L), the following interpretive breakpoints for disk diffusion susceptibility tests with 10-µg ceftibuten disks were calculated using regression line analysis: ≤19 mm for resistance and ≥ 27 mm for susceptiblity. Proposed inhibition zone diameters for the reference strain Escherichia coli ATCC 25922 are between 31 and 36 mm.     

Candida species distribution and antifungal susceptibility testing according to European Committee on Antimicrobial Susceptibility Testing and new vs. old Clinical and Laboratory Standards Institute clinical breakpoints: a 6-year prospective candidaemia survey from the fungal infection network of Switzerland

We analyzed the species distribution of Candida blood isolates (CBIs), prospectively collected between 2004 and 2009 within FUNGINOS, and compared their antifungal susceptibility according to clinical breakpoints defined by the European Committee on Antimicrobial Susceptibility Testing (EUCAST) in 2013, and the Clinical and Laboratory Standards Institute (CLSI) in 2008 (old CLSI breakpoints) and 2012 (new CLSI breakpoints). CBIs were tested for susceptiblity to fluconazole, voriconazole and caspofungin by microtitre broth dilution (Sensititre^® YeastOne™ test panel). Of 1090 CBIs, 675 (61.9%) were C. albicans, 191 (17.5%) C. glabrata, 64 (5.9%) C. tropicalis, 59 (5.4%) C. parapsilosis, 33 (3%) C. dubliniensis, 22 (2%) C. krusei and 46 (4.2%) rare Candida species. Independently of the breakpoints applied, C. albicans was almost uniformly (>98%) susceptible to all three antifungal agents. In contrast, the proportions of fluconazole- and voriconazole-susceptible C. tropicalis and F-susceptible C. parapsilosis were lower according to EUCAST/new CLSI breakpoints than to the old CLSI breakpoints. For caspofungin, non-susceptibility occurred mainly in C. krusei (63.3%) and C. glabrata (9.4%). Nine isolates (five C. tropicalis, three C. albicans and one C. parapsilosis) were cross-resistant to azoles according to EUCAST breakpoints, compared with three isolates (two C. albicans and one C. tropicalis) according to new and two (2 C. albicans) according to old CLSI breakpoints. Four species (C. albicans, C. glabrata, C. tropicalis and C. parapsilosis) represented >90% of all CBIs. In vitro resistance to fluconazole, voriconazole and caspofungin was rare among C. albicans, but an increase of non-susceptibile isolates was observed among C. tropicalis/C. parapsilosis for the azoles and C. glabrata/C. krusei for caspofungin according to EUCAST and new CLSI breakpoints compared with old CLSI breakpoints.     

How to isolate,identify and determine antimicrobial susceptibility of anaerobic bacteria in routine laboratories

BackgroundThere has been increased interest in the study of anaerobic bacteria that cause human infection during the past decade. Many new genera and species have been described using 16S rRNA gene sequencing of clinical isolates obtained from different infection sites with commercially available special culture media to support the growth of anaerobes. Several systems, such as anaerobic pouches, boxes, jars and chambers provide suitable anaerobic culture conditions to isolate even strict anaerobic bacteria successfully from clinical specimens. Beside the classical, time-consuming identification methods and automated biochemical tests, the use of matrix-assisted laser desorption/ionization time-of-flight mass spectrometry has revolutionized identification of even unusual and slow-growing anaerobes directly from culture plates, providing the possibility of providing timely information about anaerobic infections.AimsThe aim of this review article is to present methods for routine laboratories, which carry out anaerobic diagnostics on different levels.SourcesRelevant data from the literature mostly published during the last 7 years are encompassed and discussed.ContentThe review involves topics on the anaerobes that are members of the commensal microbiota and their role causing infection, the key requirements for collection and transport of specimens, processing of specimens in the laboratory, incubation techniques, identification and antimicrobial susceptibility testing of anaerobic bacteria. Advantages, drawbacks and specific benefits of the methods are highlighted.ImplicationsThe present review aims to update and improve anaerobic microbiology in laboratories with optimal conditions as well as encourage its routine implementation in laboratories with restricted resources.     

Colistin Susceptibility Testing of Gram-Negative Bacilli: Better Performance of Vitek2 System than E-Test Compared to Broth Microdilution Method as the Gold Standard Test

Introduction: Unavailability of optimal susceptibility testing (ST) challenges the clinical use of colistin. Broth microdilution (BMD), which is the reference for colistin ST, is inconvenient for diagnostics. Vitek2 and E-test although technically easier, are no longer recommended. Materials and Methods: For the evaluation of Vitek2 and E-test in reference with BMD, a total of 138 Gram-negative bacilli (GNB) especially carbapenem-resistant isolates from Tata Medical Center, Kolkata, India, were included during 2017–2018. The evaluation was performed only for Enterobacteriaceae (n = 102), but not for non-fermentative GNB (n = 36) due to lack of colistin-resistant (COL^R) isolates. Results and Conclusion: Of 138 isolates, meropenem, colistin and dual resistance were detected in 110 (79.7%), 31 (22.5%) and 21 (15.2%) of isolates, respectively. Using the European Committee on Antimicrobial Susceptibility Testing guidelines (susceptible, ≤2 μg/ml), Vitek2 performed better than E-test (essential agreement, 92.2% vs. 63.7%; categorical agreement, 94.1% vs. 93.1%; very major error [VME], 10% vs. 23.3%). However, Vitek2 overcalled resistance than E-test (major error, 4.2% vs. 0%). Considering Chew et al. proposed breakpoints (susceptible, ≤1 μg/ml), VMEs declined for both test (6.7% vs. 10%), but still remained unacceptable. Of eight colistin-heteroresistant isolates, two VME were categorised by Vitek2, one VME was by E-test, and two were uninterpretable. Both Vitek2 and E-test are unreliable. Further studies correlating minimum inhibitory concentrations with clinical outcome are needed to determine the accurate breakpoints for better patient management.     

Comparison of the Copan WASPLab incorporating the BioRad expert system against the SIRscan 2000 automatic for routine antimicrobial disc diffusion susceptibility testing

ObjectivesThis study investigated the agreement at the categorical level between the Copan WASPLab incorporating the BioRad expert system against the SIRscan 2000 automatic for antimicrobial disc diffusion susceptibility testing.MethodsThe 338 clinical strains (67 Pseudomonas aeruginosa, 19 methicillin-resistant Staphylococcus aureus, 75 methicillin-sensitive S. aureus and 177 Enterobacterales isolates) analysed in this study were non-duplicate isolates obtained from consecutive clinical samples referred to the clinical bacteriology laboratory at Geneva University Hospitals between June and August 2019. For the WASPLab the inoculum suspension was prepared in strict accordance with the manufacturer's instruction (Copan WASP srl, Brescia, Italy) by adding 2 mL of the 0.5 McFarland primary suspension used for the SIRscan analysis into a sterile tube filled with 4 mL of sterile saline (1:3 dilution). The inoculum (2 × 30 μL loop/spreader) was spread over the entire surface of Mueller–Hinton agar plates according to the AST streaking pattern defined by Copan. The antibiotic discs were dispensed by the WASP and inoculated media were loaded on conveyors for transfer to the automatic incubators. The plates were incubated for 16 h, and several digital images were acquired. Inhibition zone diameters were automatically read by the WASPLab and were adjusted manually whenever necessary. For the SIRscan 2000 automatic, the antimicrobial disc diffusion susceptibility testing was performed according to the EUCAST guidelines. The gradient strip method was used to resolve discrepancies.ResultsThe overall categorical agreement between the compared methods reached 99.1% (797/804; 95% CI 98.2%–99.6%), 99.5% (1029/1034; 95% CI 98.9%–99.8%), and 98.8% (2798/2832; 95% CI 98.3%–99.1%) for P. aeruginosa, S. aureus and the Enterobacterales, respectively.ConclusionsWASPLab incorporating the BioRad expert system provides a fully automated solution for antimicrobial disc diffusion susceptibility testing with equal or better accuracy than other available phenotypic methods.     

Estimating the precision of disc diffusion antibiotic susceptibility data published by the European committee on antimicrobial susceptibility testing

Analysis of 163 disc diffusion data sets, 115 for bacterial species groups and 48 for types strains, published by EUCAST, was used to optimize the setting of the parameters of a standardized protocol for normalized resistance interpretation of these data. The standard deviations of the normalized distributions of these data sets, calculated using this standardized protocol, were shown to be independent of the means of their respective distributions. These standard deviations could, therefore, be used as a metric to quantify the precision of disc diffusion data sets. The median value of the standard deviations for 115 EUCAST data sets produced in multiple laboratory studies of bacterial species groups was 2.3 mm and that for the 48 EUCAST data sets produced in multiple laboratory studies of type strains was 1.5 mm. It is argued that this standardized method for estimating the precision of disc diffusion data provides a tool by which individual laboratories can assess the quality of the disc diffusion data they produce.     

Evaluation of susceptibility testing methods for Burkholderia cepacia complex: a comparison of broth microdilution,agar dilution,gradient strip and EUCAST disc diffusion

ObjectivesTo evaluate the accuracy and reproducibility of antimicrobial susceptibility testing methods in Burkholderia cepacia complex (BCC).MethodsMinocycline, ciprofloxacin, trimethoprim/sulphamethoxazole, meropenem, ceftazidime and chloramphenicol were tested against 155 BCC strains using broth microdilution at 35 ± 1°C (BMD₃₅) in triplicate, then BMD at 30 ± 1°C (BMD₃₀), agar dilution at 30°C and 35°C (AD₃₀ and AD₃₅), gradient strip (GS) and EUCAST standardized disc diffusion (DD) testing methods once.ResultsBMD₃₅ reproducibility ranged from 70% to 84.5% for all agents. Correlations of MICs from BMD₃₅ with BMD₃₀ ranged from 63% to 85%, with AD₃₅ from 32.9% to 87% and with GS methods from 36% to 83.9%. Essential agreement (EA) of MICs by GS with BMD₃₅ ranged from 62.6% (trimethoprim-sulphamethoxazole) to 83.9% (minocycline). EA of EUCAST DD zone diameters using CLSI breakpoint criteria was between 85.8% and 97.4%, however Very Major Errors (VME) for trimethoprim/sulphamethoxazole were 31%.ConclusionsBMD at 35 ± 1°C was poorly reproducible for most agents and no method showed acceptable performance. Of particular concern were the GS results. Although this is the most commonly used method for determining MICs in laboratories, there was poor correlation with BMD₃₅ for meropenem and trimethoprim/sulphamethoxazole. EUCAST DD correlated poorly with BMD₃₅ MICs. This study confirms that no susceptibility method is capable of providing reproducible and accurate MICs when testing BCC.     

In-vitro susceptibility testing by agar dilution method to determine the minimum inhibitory concentrations of amphotericin B, fluconazole and ketoconazole against ocular fungal isolates

Purpose: To standardize in-vitro antifungal susceptibility testing by agar dilution method to find out the minimum inhibitory concentration (MIC) of amphotericin B, fluconazole and ketoconazole on ocular fungal isolates. Methods: A total of 180 ocular fungal isolates (130 filamentous fungi and 50 yeasts) were included. The antifungal drugs such as amphotericin B (0.0625-8 μg/mL), fluconazole (0.2-819.6 μg/mL) and ketoconazole (0.025-6.4 μg/mL) were incorporated in doubling dilutions in the yeast nitrogen base medium. The MIC was determined as the lowest concentration of the antifungal drug preventing growth of macroscopically visible colonies on drug containing plates when there was visible growth on the drug - free control plates. Results: All 50 ocular isolates of yeast were susceptible to amphotericin B, while two (4%) and five (10%) strains were resistant to fluconazole and ketoconazole respectively. Of the 130 filamentous fungi tested, six (4.6%) were resistant to amphotericin B, 49 (37.7%) and 10 (7.6%) were resistant to fluconazole and ketoconazole respectively. Percentile 50 (MIC 50) and Percentile 90 (MIC 90) for all the three antifungal agents were calculated. Aspergillus niger, Aspergillus terreus and Candida krusei were found to be resistant to fluconazole and ketoconazole. Conclusion: This technique was found to be reliable, cost effective and easy to perform with consistent results.     

Evaluation of the WIDER I system for antimicrobial susceptibility testing of clinical isolates of Haemophilus influenzae and Streptococcus pneumoniae

The aim of this study was to evaluate the WIDER I system for susceptibility testing of Haemophilus influenzae and Streptococcus pneumoniae . MICs of 12 antimicrobials against 42 H. influenzae and 58 S. pneumoniae strains were determined using 1W MIC panels and compared with those obtained by microdilution. Overall essential agreements were >99%. Very major errors were not detected. Major errors occurred with ampicillin (1.7% H. influenzae ). Minor errors were 2.3% (amoxicillin–clavulanate, cefuroxime, chloramphenicol), 7.1% (ampicillin) and 16.7% (clarithromycin) for H. influenzae , and 1.7% (chloramphenicol, erythromycin, meropenem), 3.4% (amoxicillin–clavulanate, cefuroxime, tetracycline) and 8.6% (levofloxacin) for S. pneumoniae . The WIDER I system is a reliable method for susceptibility testing of H. influenzae and S. pneumoniae .     

The quality of antimicrobial discs from nine manufacturers—EUCAST evaluations in 2014 and 2017

Objectives

Antimicrobial discs for susceptibility testing can be obtained from many manufacturers. We evaluated the quality of discs from nine manufacturers in 2014 and 2017.

Comparison of antifungal MICs for yeasts obtained using the EU-CAST method in a reference laboratory and the Etest in nine different hospital laboratories

In routine laboratory practice, the determination of MICs of antifungals for yeasts often relies on the Etest, because of a good correlation with reference methods. However, this correlation was established through predesigned studies, rather than prospective testing. The surveillance programme of fungaemia (YEASTS programme), implemented since 2003, facilitated our comparison of the Etest and the EU-CAST results, obtained on a routine basis in nine different hospitals and in a reference laboratory, respectively. The analysis included 690 isolates recovered from blood culture (362 Candida albicans, 113 Candida glabrata, 69 Candida parapsilosis, 55 Candida tropicalis, 31 Cryptococcus neoformans, and 60 other yeast species) that were tested for their susceptibility to amphotericin B (n = 655), fluconazole (n = 669), itraconazole (n = 198), voriconazole (n = 588), flucytosine (n = 314), and caspofungin (n = 244). Agreement between the Etest and EU-CAST datasets was calculated and categorized on the basis of previously published breakpoints. The level of agreement at ±2 dilutions was 75% for amphotericin B and 90% for flucytosine; for the azoles, it ranged from 71% for itraconazole to 87% for voriconazole. No significant difference was observed among the yeast species, except for Cryptococcus neoformans and flucytosine, with an agreement <40. Categorical agreement ranged from 60% for itraconazole to 90% for flucytosine. Major and very major discrepancies occurred in <12% and 6%, respectively. The Etest, even when performed on a routine basis, shows a ≥71% agreement with the EU-CAST reference method.     

Antilogic,a new supervised machine learning software for the automatic interpretation of antibiotic susceptibility testing in clinical microbiology: proof-of-concept on three frequently isolated bacterial species

ObjectiveAntibiotic susceptibility testing (AST) is necessary in order to adjust empirical antibiotic treatment, but the interpretation of results requires experience and knowledge. We have developed a machine learning software that is capable of reading AST images without any human intervention and that automatically interprets the AST, based on a database of antibiograms that have been clinically validated with European Committee on Antimicrobial Susceptibility Testing rules.MethodsWe built a database of antibiograms that were labelled by senior microbiologists for three species: Escherichia coli, Klebsiella pneumoniae, and Staphylococcus aureus. We then developed Antilogic, a Python software based on an original image segmentation module and supervised learning models that we trained against the database. Finally, we blind tested Antilogic against a validation set of 5100 photos of antibiograms.ResultsWe trained Antilogic against a database of 18072 pictures of antibiograms. Overall agreement against the validation set reached 97% (16 855/17 281) regarding phenotypes. The severity rate of errors was also evaluated: 1.66% (287/17 281) were major errors and 0.80% (136/17 281) were very major errors. After implementation of uncertainty quantifications, the rate of errors decreased to 0.80% (114/13 451) and 0.42% (51/13 451) for major and very major errors respectively.DiscussionAntilogic is the first machine learning software that has been developed for AST interpretation. It is based on a novel approach that differs from the typical diameter measurement and expert system approach. Antilogic is a proof of concept that artificial intelligence can contribute to faster and easier diagnostic methods in the field of clinical microbiology.