Comparison of ciprofloxacin and aminoglycoside susceptibility testing for ceftriaxone non-susceptible Enterobacterales by disk diffusion and VITEK 2 vs. broth microdilution using updated Clinical and Laboratory Standards Institute breakpoints

Abstract

Background

Fluoroquinolones and aminoglycosides are potential treatment choices in the setting of increasingly multi-drug resistant Enterobacterales. The Clinical & Laboratory Standards Institute (CLSI) breakpoints for fluoroquinolones and aminoglycosides in the Enterobacterales were revised in 2019 and 2022, respectively. However, performance of existing widely used automated systems, such as the VITEK 2 AST-N391 card, has not been extensively tested for MDR isolates at these new breakpoints.

Objective

To assess performance of the new breakpoints for ciprofloxacin, gentamicin, and tobramycin on the VITEK 2 system (bioMérieux, France) and disk diffusion by comparing to broth microdilution for ceftriaxone nonsusceptible Enterobacterales.

Methods

Ninety-four ceftriaxone non-susceptible Escherichia coli and Klebsiella pneumoniae isolates were identified between January 2021-June 2023. Broth microdilution was used as the reference standard against which disk diffusion and VITEK 2 susceptibility testing were compared. For the Vitek 2, we used the AST-N391 card and interpreted the results according to the updated CLSI breakpoints.

Results

Overall, 22.3% of isolates were susceptible to ciprofloxacin by BMD. Compared to BMD, disk diffusion had an overall minor error rate of 7.4% (95%CI 3.0–14.7%) with 0 major or very major errors (97.5% CI 0–3.8%). For the VITEK 2, a minor error rate of 13.8% (95% CI 7.6–22.5%), major error rate 19.0% (95%CI 7.7–40.0%) and very major error rate 0% (97.5%CI 0–3.8%) was noted. By comparison, 69.1% and 56.4% of isolates were susceptible to gentamicin and tobramycin, respectively. Disk diffusion and the VITEK 2 system both correctly categorized 100% of gentamicin susceptible and non-susceptible isolates. For tobramycin, disk diffusion had a 3.2% rate of misclassification (all minor errors) and the VITEK 2 had 2.1% rate of misclassification (all minor errors). There were no major or very major errors.

Conclusions

Our findings suggest that both disk diffusion and to a greater extent the AST-N391 card for the VITEK 2 system will overcall non-susceptibility according to current CLSI breakpoints for ciprofloxacin. By contrast, the existing AST-N391 VITEK 2 card can likely be used to correctly infer susceptibility to gentamicin and tobramycin.

Introduction

Infections with gram negative bacteria producing extended-spectrum β-lactamase (ESBL) enzymes are a global public health threat [1], with recent reports suggesting over 100,000 deaths globally per year attributed to them [2]. Treatment of these infections commonly requires carbapenems, though emerging carbapenem resistance may limit their use in the future. Consequently, significant work has gone into studying the efficacy of carbapenem-sparing agents. Urinary tract infections are important causes of bacteremia caused by Enterobacterales. In this context, aminoglycosides which achieve high concentrations in the urine [3], and fluoroquinolones with high bioavailability [4, 5] represent important carbapenem-sparing agents.

In 2019, the Clinical & Laboratory Standards Institute (CLSI) revised the breakpoints for fluoroquinolones in Enterobacterales, a change precipitated by evidence suggesting that previous breakpoints were too high to detect low-level fluoroquinolone resistance [6]. This low-level resistance was considered clinically relevant as patients with these organisms and suboptimal drug exposures risk potential treatment failure. A retrospective study evaluating the clinical impact of these changes suggested that levofloxacin therapy under the previously accepted higher breakpoints was associated with greater mortality [7].

Similarly, in 2022, CLSI lowered the breakpoints for aminoglycosides in the Enterobacterales [8]; susceptibility breakpoint for gentamicin and tobramycin were lowered from ≤ 4 to ≤ 2 mg/L [9], intermediate from 8 to 4 mg/L, and resistant from ≥ 16 to ≥ 8 mg/L. Principally, this change brought the CLSI breakpoints into alignment with those of the European Committee on Antimicrobial Susceptibility testing (EUCAST) [10] and took into consideration epidemiologic cut-off data presented from the SENTRY Antimicrobial Surveillance Program.

An essential step in proposing the use of ciprofloxacin and aminoglycosides in the treatment of ESBL bacteremia is to ensure that the commonly employed antimicrobial susceptibility testing methods adequately measure susceptibility to these agents. Therefore, we assessed performance of the VITEK 2 system (bioMérieux, France) and disk diffusion against the gold standard broth microdilution for the detection of ciprofloxacin and aminoglycoside susceptibility in ceftriaxone non-susceptible Escherichia coli and Klebsiella pneumoniae, the most encountered third generation cephalosporin resistant Enterobacterales, under the new CLSI breakpoints.

Materials and methods

Isolate collection

Ceftriaxone non-susceptible E. coli and K. pneumoniae, defined as MIC ≥ 2 µg/mL, were retrospectively identified from positive blood cultures between January 2021-June 2023 obtained from hospitalized patients at the Royal Victoria Hospital, Montreal General Hospital, and Jewish General Hospital, in Montréal, Québec. These cultures were unique clinical specimens with no more than one culture per patient included in this study. Blood isolates were chosen specifically to enrich for pathogens associated with life-threatening infection. A total of 94 bacterial isolates were included: E. coli (76 isolates) and K. pneumoniae (18 isolates). Each isolate was sub-cultured in duplicate from glycerol broth to blood agar and shipped to a reference laboratory at Element/JMI Laboratories (Iowa, USA) for broth microdilution testing (BMD) and disk diffusion testing.

Broth microdilution

BMD testing was carried out at JMI Laboratories according to CLSI M07-Ed11 [11] using cation-adjusted Mueller–Hinton broth (BD BBL; B12322). Susceptibility testing against ciprofloxacin (Sigma-Aldrich; 17,850), gentamicin (USP; 1289003) and tobramycin (Sigma-Aldrich; T4014) was performed. BMD was used as the reference standard against which disk diffusion and the VITEK 2 were compared. Quality control was performed using E. coli ATCC 25922 and Pseudomonas aeruginosa ATCC 27853 on each day of testing.

Disk diffusion

Disk diffusion testing was performed by JMI Laboratories on Muller-Hinton agar (Remel; R04052) using disks containing 10 μg gentamicin, disks containing 10 μg tobramycin, and disks containing 5 μg of ciprofloxacin. Inocula were prepared by suspending colonies from overnight blood agar plates in sterile saline to a 0.5 McFarland standard. BMD and disk diffusion testing was performed using the same inoculum. The ciprofloxacin and aminoglycoside-containing disks were dispensed onto the surfaces of inoculated agar plates and incubated at 35 °C for 16 to 18 h. Testing and interpretation were performed according to CLSI M02-A12 and M100-ED34 [12, 13].

VITEK 2

Automated antimicrobial susceptibility testing was performed using VITEK 2 AST-N391 cards (bioMérieux, France) according to the manufacturer’s instructions. Testing was performed at the McGill University Health Centre (MUHC) Clinical Microbiology Laboratory. The 2019 breakpoints of MIC ≤ 0.25 μg/ml, susceptible; MIC > 0.25 and < 1 μg/ml intermediate; and MIC ≥ 1 μg/ml, resistant were used for ciprofloxacin and 2022 interpretive breakpoints of MIC ≤ 2 μg/ml, susceptible; MIC 4 μg/mL, intermediate; MIC ≥ 8 μg/ml, resistant were used for gentamicin and tobramycin. Interpretation of MICs was performed using breakpoints established in the CLSI M100-Ed34 [13].

Statistical analysis

Using BMD as reference standard, we compared the performance of disk diffusion and the VITEK 2 at classifying susceptible and non-susceptible isolates. For the VITEK 2, we used the existing cards and used the updated breakpoints to interpret susceptibility based on the reported MICs. We reported the proportions correctly identified as susceptible and non-susceptible with 95% binomial exact confidence intervals. Errors were categorized as minor (miE) (intermediate reported as susceptible or resistant; susceptible or resistant reported as intermediate), major (ME) (erroneous categorization of true-susceptible isolates as resistant; falsely resistant), or very major (VME) (erroneous categorization of true-resistant isolates as susceptible; falsely susceptible) [14]. All analyses were conducted on STATA version 17.

Results

Twenty-one of 94 (22.3%) isolates were characterized by BMD as susceptible to ciprofloxacin and 73 (77.7%) as non-susceptible [composed of 2 (2.1%) intermediate and 71 (75.5%) resistant] (Fig. 1A). Disk diffusion correctly categorized 14/21 (66.6%; 95% CI 43.0–85.4%) of susceptible identified by BMD (all 7 were miE) and 73/73 (100%; 95% CI 95.1–100%) of non-susceptible. Therefore, the overall miE rate was 7.4% (95%CI 3.0–14.7%) with 0 ME or VME (95% CI 0–3.8%). Categorical agreement by disk diffusion was 92.6%.

Fig. 1

Ciprofloxacin MIC by broth microdilution vs (A) disk diffusion zone in mm and (B) VITEK 2 MIC. The gray lines represent the breakpoint for susceptibility and the red lines represent the breakpoint for resistance. Items are shaded based on BMD results: green for susceptible, yellow for intermediate, and red for resistant. Numbers represent the count of isolates with corresponding values

By comparison, VITEK 2 only correctly categorized 6/21 (28.6%; 95%CI 11.3% to 52.2%) isolates which were susceptible by BMD (11 miE, 4 ME) while correctly categorizing 73/73 (100%; 95% CI 95.1–100%) of those non-susceptible by BMD (2 miE) (Fig. 1B). Therefore, the minor error rate was 13.8% (95%CI 7.6–22.5%), ME rate 19.0% (95% CI 7.7–40.0%) and VME rate 0% (97.5%CI 0–3.8%). Categorical agreement by VITEK 2 was 81.9%.

For aminoglycosides, 65 of 94 (69.1%) isolates were susceptible to gentamicin by BMD and 53 of 94 (56.4%) were susceptible to tobramycin. Disk diffusion and the VITEK 2 correctly categorized 65/65 (100%; 95% CI 94.5–100%) of BMD classified gentamicin susceptible and 29/29 (100%; 95%CI 88.1–100%) of non-susceptible isolates without any errors. Figure 2 shows the relationship for gentamicin between BMD and disk diffusion (panel A) and BMD and VITEK 2 (panel B).

Fig. 2

Gentamicin MIC by broth microdilution vs (A) disk diffusion zone in mm and (B) VITEK 2 MIC. The gray lines represent the breakpoint for susceptibility and the red lines represent the breakpoint for resistance. Items are shaded based on BMD results: green for susceptible, yellow for intermediate/susceptible dose dependent, and red for resistant. Numbers represent the count of isolates with corresponding values

Figure 3 depicts the relationship for tobramycin between BMD and disk diffusion (panel A) and BMD and VITEK 2 (panel B). Disk diffusion correctly categorized 52/53 (98.1%; 95%CI 90.0–100%) of BMD susceptible isolates (1 miE) and 39/41 (95.1%; 95%CI 83.5–99.4%) non-susceptible isolates (2 miE). Within the disk diffusion intermediate zone, there were an additional 9 minor errors with organisms testing resistant by BMD. The VITEK 2 correctly categorized 52/53 (98.1%; 95%CI 90.0–100%) of BMD susceptible isolates (1 miE) and 40/41 (97.6%; 95%CI 87.1–99.9%) of non-susceptible isolates (1 miE). Two additional minor errors were noted for MIC in the intermediate range with organisms testing resistant by BMD; an additional 2 minor errors with organisms classified as ‘resistant’ by VITEK 2 testing intermediate by BMD.

Fig. 3

Tobramycin MIC by broth microdilution vs (A) disk diffusion zone in mm and (B) VITEK 2 MIC. The gray lines represents the breakpoint for susceptibility and the red lines represents the breakpoint for resistance. Items are shaded based on BMD results: green for susceptible, yellow for intermediate/susceptible dose dependent, and red for resistant. Numbers represent the count of isolates with corresponding values

Discussion

We report that in comparison to BMD, disk diffusion for ciprofloxacin yields no ME or VMEs, whereas VITEK 2 was found to have a ME rate of 19.0% and VME rate 0%. For aminoglycosides, testing by disk diffusion or VITEK 2 did not lead to any ME or VMEs. Therefore, we demonstrate that in the setting of a high pretest probability of ciprofloxacin resistance (e.g., in multidrug resistant organisms), both disk diffusion and to a much greater extent the AST-N391 card for the VITEK 2 system will overcall non-susceptibility but is unlikely to label isolates as falsely susceptible. Errors for disk diffusion were typically in the intermediate diameters and are likely acceptable. While overcalling resistance may impact oral step-down options and antibiotic stewardship, such as overuse of carbapenems, the tests reliably identified non-susceptibility and therefore patient safety with the use of these methods is likely not impacted. In contrast to ciprofloxacin, the current AST-N391 card for the VITEK 2 system performs adequately at categorizing gentamicin and tobramycin susceptibility.

Commercial systems such as VITEK 2 have streamlined AST determination from culture and have a much faster turnaround time compared to classical techniques of broth microdilution or disk diffusion. Given the widespread use of these commercial methods, their performance should be independently assessed. Few studies have sought to undertake this; one notable report by Huang and colleagues evaluated AST for levofloxacin determined by VITEK 2 compared to agar dilution in 253 isolates from bacteremia caused by Enterobacterales. Their study reports an overall minor error rate of 11.9%, ME rate of 0%, and VME rate of 50% [15]. While no studies have formally assessed the performance of VITEK 2 in ESBL producing gram negatives, similar observational study was undertaken for bacteremia with K. pneumoniae KPC isolates. The authors report a VME rate of 1.6% and ME rate of 21% for gentamicin [16]. Overall, our observations as well as reports from the literature underscore the importance of rapid and accurate AST in the context of multi-drug-resistant infections with few therapeutic options. These should be independently assessed by individual laboratories.

Strengths of our work include the use of the two most common Enterobacterales found in the blood and our use of invasive specimens with high pre-test probability of harboring ciprofloxacin and aminoglycoside resistance by virtue of already being ceftriaxone resistant. This is critical because when the prevalence of resistance in a sample is low, susceptibility testing may appear to perform better simply by having fewer opportunities to be incorrect. Limitations include the absence of genetic characterization of the isolates, the use of only one model of VITEK 2 AST card, the use of blood isolates only, and reliance on specimens from one city. Additionally, different biological replicates were used for BMD and disk diffusion testing compared to VITEK 2 (conducted at MUHC laboratories) which may pose a minor source of error. Lastly, while we a priori decided to systematically include isolates of both E. coli and K. pneumoniae, the total number of K. pneumoniae was relatively limited and the results interpreted with caution.

In summary, we believe that disk diffusion should be preferred to the AST-N391 VITEK 2 card when attempting to correctly infer susceptibility to ciprofloxacin. While both approaches will correctly identify resistance, disk diffusion will more readily allow for fluoroquinolone therapy to be safely employed, particularly for multidrug resistant organisms. These results should be corroborated by a larger international collaboration and with other automated testing systems.