The European Committee on Antimicrobial Susceptibility Testing (EUCAST) recommends two steps for detecting beta-lactamases in Gram-negative bacteria. Screening for potential extended-spectrum beta-lactamase (ESBL), plasmid-mediated AmpC beta-lactamase, or carbapenemase production is confirmed. We aimed to validate generative pre-trained transformer (GPT)-4 and GPT-agent for pre-classification of disk diffusion to indicate potential beta-lactamases. We assigned 225 Gram-negative isolates based on phenotypic resistances against beta-lactam antibiotics and additional tests to one or more resistance mechanisms as follows: “none,” “ESBL,” “AmpC,” or “carbapenemase.” Next, we customized a GPT-agent with EUCAST guidelines and breakpoint table (v13.1). We compared routine diagnostics (reference) to those of (i) EUCAST-GPT-expert, (ii) microbiologists, and (iii) non-customized GPT-4. We determined sensitivities and specificities to flag suspect resistances. Three microbiologists showed concordance in 814/862 (94.4%) phenotypic categories and were used in median eight words (interquartile range [IQR] 4–11) for reasoning. Median sensitivity/specificity for ESBL, AmpC, and carbapenemase were 98%/99.1%, 96.8%/97.1%, and 95.5%/98.5%, respectively. Three prompts of EUCAST-GPT-expert showed concordance in 706/862 (81.9%) categories but were used in median 158 words (IQR 140–174) for reasoning. Sensitivity/specificity for ESBL, AmpC, and carbapenemase prediction were 95.4%/69.23%, 96.9%/86.3%, and 100%/98.8%, respectively. Non-customized GPT-4 could interpret 169/862 (19.6%) categories, and 137/169 (81.1%) agreed with routine diagnostics. Non-customized GPT-4 was used in median 85 words (IQR 72–105) for reasoning. Microbiologists showed higher concordance and shorter argumentations compared to GPT-agents. Humans showed higher specificities compared to GPT-agents. GPT-agent’s unspecific flagging of ESBL and AmpC potentially results in additional testing, diagnostic delays, and higher costs. GPT-4 is not approved by regulatory bodies, but validation of large language models is needed. IMPORTANCE: The study titled "GPT-4-based AI agents—the new expert system for detection of antimicrobial resistance mechanisms?" is critically important as it explores the integration of advanced artificial intelligence (AI) technologies, like generative pre-trained transformer (GPT)-4, into the field of laboratory medicine, specifically in the diagnostics of antimicrobial resistance (AMR). With the growing challenge of AMR, there is a pressing need for innovative solutions that can enhance diagnostic accuracy and efficiency. This research assesses the capability of AI to support the existing two-step confirmatory process recommended by the European Committee on Antimicrobial Susceptibility Testing for detecting beta-lactamases in Gram-negative bacteria. By potentially speeding up and improving the precision of initial screenings, AI could reduce the time to appropriate treatment interventions. Furthermore, this study is vital for validating the reliability and safety of AI tools in clinical settings, ensuring they meet stringent regulatory standards before they can be broadly implemented. This could herald a significant shift in how laboratory diagnostics are performed, ultimately leading to better patient outcomes.

Interpretative reading of antimicrobial susceptibility test (AST) results allows inferring biochemical resistance mechanisms from resistance phenotypes. For aminoglycosides, however, correlations between resistance pathways inferred on the basis of the European Committee on Antimicrobial Susceptibility Testing (EUCAST) clinical breakpoints and expert rules versus genotypes are generally poor. This study aimed at developing and validating a decision tree based on resistance phenotypes determined by disc diffusion and based on epidemiological cut-offs (ECOFFs) to infer the corresponding resistance mechanisms in Escherichia coli. Phenotypic antibiotic susceptibility of thirty wild-type and 458 aminoglycoside-resistant E. coli clinical isolates was determined by disc diffusion and the genomes were sequenced. Based on well-defined cut-offs, we developed a phenotype-based algorithm (Aminoglycoside Resistance Mechanism Inference Algorithm - ARMIA) to infer the biochemical mechanisms responsible for the corresponding aminoglycoside resistance phenotypes. The mechanisms inferred from susceptibility to kanamycin, tobramycin and gentamicin were analysed using ARMIA- or EUCAST-based AST interpretation and validated by whole genome sequencing (WGS) of the host bacteria. ARMIA-based inference of resistance mechanisms and WGS data were congruent in 441/458 isolates (96·3%). In contrast, there was a poor correlation between resistance mechanisms inferred using EUCAST CBPs/expert rules and WGS data (418/488, 85·6%). Based on the assumption that resistance mechanisms can result in therapeutic failure, EUCAST produced 63 (12·9%) very major errors (vME), compared to only 2 (0·4%) vME with ARMIA. When used for detection and identification of resistance mechanisms, ARMIA resolved >95% vMEs generated by EUCAST-based AST interpretation. This study demonstrates that ECOFF-based analysis of AST data of only four aminoglycosides provides accurate information on the resistance mechanisms in E. coli. Since aminoglycoside resistance mechanisms, despite having in certain cases a minimal effect on the minimal inhibitory concentration, may compromise the bactericidal activity of aminoglycosides, prompt detection of resistance mechanisms is crucial for therapy. Using ARMIA as an interpretative rule set for editing AST results allows for better predictions of in vivo activity of this drug class.

The reference method for cefiderocol antimicrobial susceptibility testing is broth microdilution (BMD) with iron-depleted-Mueller-Hinton (ID-MH) medium, whereas breakpoints recommended for disk diffusion (DD) are based on MH-agar plates. We aimed to compare the performance of the commercial BMD tests ComASP (Liofilchem) and UMIC (Bruker), and DD and E-test using MH- and ID-MH-agar plates with the reference BMD method using 100 carbapenem-resistant-A. baumannii isolates. Standard BMD was performed according to the EUCAST guidelines; DD and E-test were carried out using two commercial MH-agar plates (BioMérieux and Liofilchem) and an in-house ID-MH-agar plate, while ComASP and UMIC were performed according to the manufacturer’s guidelines. DD performed with the ID-MH-agar plates led to a higher categorical agreement (CA, 95.1%) with standard BMD and fewer categorization errors compared to the commercial MH-agar plates (CA BioMérieux 91.1%, Liofilchem 89.2%). E-test on ID-MH-agar plates exhibited a significantly higher essential agreement (EA, 75%) with standard BMD compared to the two MH-agar plates (EA BioMérieux 57%, Liofilchem 44%), and showed a higher performance in detecting high-level resistance than ComASP and UMIC (mean log2 difference with standard BMD for resistant isolates of 0.5, 2.83, and 2.08, respectively). In conclusion, DD and E-test on ID-MH-agar plates exhibit a higher diagnostic performance than on MH-agar plates and the commercial BMD methods. Therefore, we recommend using ID-MH-agar plates for cefiderocol susceptibility testing of A. baumannii.

Antimicrobial resistance is a growing public health threat of broad concern worldwide. Timely detection of antibiotic resistance mechanisms can help to monitor and to curb the spread of resistant bacteria within the hospital setting as well as in the environment.

Pseudomonas aeruginosa can acquire carbapenem resistance through various mechanisms, including genomic mutations leading to the overexpression of efflux pumps, intrinsic AmpC-β-lactamase, and/or reduced permeability, and/or through the acquisition of plasmid-mediated carbapenemases and/or extended-spectrum-β-lactamases (ESBLs). Unfortunately, carbapenemase/ESBL-producing-carbapenem-resistant- P. aeruginosa (CP/ESBL-CRPA) cannot be differentiated from non-CP/ESBL-CRPA based solely on susceptibility testing results of conventional β-lactam (BL)-antibiotics. Knowing that these two groups display different activity profiles toward novel BL/β-lactamase-inhibitor (BLI) combinations, we developed and verified a cost-effective and easy-to-use diagnostic algorithm for screening and differentiation of carbapenemase and ESBL production in CRPA. We determined disc diffusion inhibition zones and gradient strip minimal inhibitory concentration values of 136 whole-genome-sequenced CRPA (70 metallo-β-lactamase-[MBL-], 1 GES-5-, 1 KPC-2-, 12 ESBL-, and 53 AmpC-hyper-producing isolates). We used the following BL-BLI combinations: ceftolozane-tazobactam (C-T), ceftazidime-avibactam, imipenem-relebactam (I-R), meropenem-vaborbactam, cefepime-enmetazobactam (C-E), and aztreonam-avibactam. We also included a lateral flow immunoassay (Carba-5, NG-Biotech) for confirmation of MBL production and double disc synergy testing (DDST) to improve ESBL detection. C-T was the most effective screening antibiotic for distinguishing MBL and ESBL producers from AmpC-hyperproducing CRPA, achieving a sensitivity of 100% for both MBL and ESBL producers. I-R reliably confirmed MBL production in C-T positive screened CRPAs, with a sensitivity of 92.8% and specificity of 100%. Incorporating Carba-5 into the phenotypic algorithm improved sensitivity for confirming MBL production to 100%. For the remaining C-T positive but I-R negative isolates, C-E showed 75% sensitivity and 78.6% specificity in detecting ESBL production. The DDST further confirmed ESBL production in six out of nine ESBL producers (66.6%). In conclusion, we established a simple and cost-effective diagnostic algorithm, enabling screening and confirmation of carbapenemase and ESBL production in CRPA. IMPORTANCE: Carbapenem-resistant Pseudomonas aeruginosa (CRPA) is a major global health threat, and rapid identification of its resistance mechanisms is crucial for effective treatment and infection control. Differentiating between carbapenemase-producing (CP), extended-spectrum β-lactamase-producing (ESBL), and AmpC-hyperproducing CRPA is challenging, as conventional susceptibility testing cannot reliably distinguish these resistance mechanisms. Our study presents a simple, cost-effective, and easy-to-implement phenotypic diagnostic algorithm that enables accurate screening and confirmation of CP and ESBL production in CRPA. This method is particularly valuable for laboratories lacking access to molecular diagnostics, as it provides a practical alternative for routine testing. By facilitating the early detection of resistant P. aeruginosa strains, this approach has the potential to improve patient outcomes, optimize antimicrobial therapy, and enhance global surveillance efforts against multidrug-resistant pathogens.