Evaluation of Piperacillin-Tazobactam Testing against Enterobacterales by the Phoenix, MicroScan, and Vitek2 Tests Using Updated Clinical and Laboratory Standards Institute Breakpoints

ABSTRACT

In 2022, the Clinical and Laboratory Standards Institute (CLSI) updated piperacillin-tazobactam (TZP) breakpoints for Enterobacterales, based on substantial data suggesting that historical breakpoints did not predict treatment outcomes for TZP. The U.S. Food and Drug Administration (FDA) has not yet adopted these breakpoints, meaning commercial manufacturers of antimicrobial susceptibility testing devices cannot obtain FDA clearance for the revised breakpoints. We evaluated the Phoenix (BD, Sparks, MD), MicroScan (Beckman Coulter, Sacramento, CA), and Vitek2 (bioMérieux, Durham, NC) TZP MICs compared to reference broth microdilution for a collection of 284 Enterobacterales isolates. Phoenix (n = 167 isolates) demonstrated 84.4% categorical agreement (CA), with 4.2% very major errors (VMEs) and 1.8% major errors (MEs) by CLSI breakpoints. In contrast, CA was 85.0% with 4.3% VMEs and 0.8% MEs for the Phoenix with FDA breakpoints. MicroScan (n = 55 isolates) demonstrated 80.0% CA, 36.4% VMEs, and 4.8% MEs by CLSI breakpoints and 81.8% CA, 44.4% VMEs, and 0.0% MEs by FDA breakpoints. Vitek2 (n = 62 isolates) demonstrated 95.2% CA, 6.3% VMEs, and 0.0% MEs by CLSI and 96.8% CA, 0.0% VMEs, and 2.2% MEs by FDA breakpoints. Overall, the performance of the test systems was not substantially different using CLSI breakpoints off-label than using on-label FDA breakpoints. However, limitations were noted with higher-than-desired VME rates (all three systems) and lower-than-desired CA (MicroScan and Phoenix). Laboratories should consider adoption of the revised CLSI breakpoints with automated test systems but be aware that some performance challenges exist for testing TZP on automated systems, regardless of breakpoints applied.

INTRODUCTION

In 2022, the Clinical and Laboratory Standards Institute (CLSI) updated piperacillin-tazobactam (TZP) MIC breakpoints for Enterobacterales, to ≤8 μg/mL (susceptible), 16 μg/mL (susceptible dose-dependent [SDD]), and ≥32 μg/mL (resistant [R]). Disk diffusion (DD) breakpoints were also updated (1). These breakpoints were largely driven by the findings of the MERINO trial, which failed to demonstrate noninferiority of TZP versus meropenem for the treatment of bloodstream infections (2) and found an association between mortality and TZP MICs of ≥32 μg/mL (3). In addition, multiple publications demonstrate poor probability of achieving pharmacokinetic/pharmacodynamic (PK/PD) targets (i.e., greater than 50% time above the MIC), for TZP MICs of ≥16 μg/mL, when intermittent infusion strategies are used (4). The CLSI SDD breakpoint is for extended-infusion TZP strategies, i.e., 4.5 g every 6 h as a 3-h infusion or 4.5 g every 8 h as a 4-h infusion, where a greater time above the MIC is predicted (4). While the U.S. Food and Drug Administration (FDA) recognizes the M100 32nd edition, it lists a specific exception for the Enterobacterales TZP breakpoints and continues to apply M100 31st edition breakpoints for this drug. The rationale for the Enterobacterales TZP breakpoints has been submitted to the FDA by CLSI and is under review.

Globally, TZP is a heavily relied-upon antimicrobial for the treatment of Gram-negative infections. Most laboratories report the results of TZP on every isolate of Enterobacterales tested in the laboratory, and there is significant desire to use this drug to spare carbapenems, in the face of emerging resistance (5, 6). However, until the FDA recognizes CLSI breakpoints, there is no mechanism for U.S. antimicrobial susceptibility test (AST) device manufacturers to obtain FDA clearance of the CLSI M100 32nd edition Enterobacterales breakpoints. Many laboratories may opt to implement the CLSI TZP breakpoints in advance of FDA recognition, by validating these breakpoints off-label on their automated AST systems. To do so, laboratories must confirm the categorical interpretations against a reference standard, such as reference broth microdilution (BMD) or DD as there are no devices FDA cleared with the updated CLSI breakpoints (7). Few clinical laboratories have access to reference BMD, and DD is associated with high rates of minor errors (MIN) for TZP, making it difficult to determine if errors observed when using DD as the reference are a result of the automated AST device or the reference itself (1). There are few studies available in the literature that laboratories can reference as part of the evaluation of automated AST devices off-label for these breakpoints, and none with the current M100 32nd edition TZP Enterobacterales breakpoints.

In this study, we compared TZP MICs obtained in five clinical laboratories using the Phoenix (BD, Sparks, MD), MicroScan (Beckman Coulter, Sacramento, CA) or Vitek2 (bioMérieux, Durham, NC) to results from reference broth microdilution, performed at a central laboratory. Performance was evaluated using 2022 M100 32nd edition breakpoints for the Enterobacterales. In addition, data were evaluated against current FDA breakpoints (Table 1).

TABLE 1.

Clinical breakpoints utilized in this study

Standard	Breakpoint (μg/mL)b
	S	SDD	I	R
CLSI M100 32nd edition Enterobacterales	≤8	16		≥32
FDA Enterobacteralesa	≤16		32–64	≥128

^ahttps://www.fda.gov/drugs/development-resources/antibacterial-susceptibility-test-interpretive-criteria; accessed 31 October 2022. FDA breakpoints are the same as the 2021 CLSI M100 31st edition breakpoints.

^bS, susceptible; SDD, susceptible dose-dependent; I, intermediate; R, resistant.

MATERIALS AND METHODS

Isolate collection.

Bacteria used in this study were isolated from clinical specimens in 2022, with the request of representation of the species distribution and TZP MICs seen in the United States, as estimated using data published in the SENTRY surveillance system (https://sentry-mvp.jmilabs.com; see Table S1 in the supplemental material). Isolates were identified according to each laboratory’s standard of care method. Isolate and susceptibility distributions are shown in Table 2. Two hundred eighty-four isolates were included in this study. Bacterial isolates were shipped to Vanderbilt University Medical Center (VUMC) on swabs and subcultured onto tryptic soy agar (TSA) plus 5% sheep’s blood agar (BAP; BD) upon receipt. Freezer stocks were made in Brucella broth plus 10% glycerol, and isolates were stored at −70°C. Each isolate was subcultured twice from a frozen state on BAP prior to testing by BMD.

TABLE 2.

Bacterial isolates tested by system and reference broth microdilution results (BMD) for piperacillin-tazobactam

Species	No. tested per system			BMD interpretation with CLSI breakpointsa (%)
	MicroScan	Phoenix	Vitek2	S	SDD	R
Citrobacter freundii complex	1	5	1	7 (100)
Citrobacter koseri	2	3	1	6 (100)
Enterobacter cloacae complex	7	10	4	13 (62)		8 (28)
Escherichia coli	20	66	30	83 (71.5)	11 (9.5)	22 (19.0)
Klebsiella aerogenes	3	18	2	12 (52.2)	1 (4.3)	10 (43.5)
Klebsiella oxytoca	3	7	2	10 (83.3)	1 (8.3)	1 (8.3)
Klebsiella pneumoniae complex	8	39	18	33 (50.8)		32 (49.2)
Morganella morganii	2	2	1	5 (100)
Proteus mirabilis	6	8	3	17 (100)
Serratia marcescens	3	6	0	7 (77.8)		2 (22.2)
Providencia rettgeri	0	1	0	1 (100)
Providencia stuartii	0	2	0	2 (100)
Total	55	167	62	196 (69.0)	13 (4.6)	75 (26.4)

^aClinical and Laboratory Standards Institute 2022, M100, 32nd edition.

Automated antimicrobial susceptibility testing.

MIC results were obtained by the clinical laboratories submitting isolates for this study, using their laboratory’s automated AST system and routine standard operating procedures. Four sites performed testing using the Phoenix test, two using the MicroScan test, and two using the Vitek2 test. Phoenix panels tested included Emerge NMIC-306 (n = 135 isolates; MIC calling range, ≤2 to >128 μg/mL) and NMIC-300 (n = 32 isolates; MIC calling range, ≤2 to >64 μg/mL) panels; MicroScan included NM56 panels, using the Prompt inoculation system (n = 55; MIC calling range, ≤8 μg/mL to >64 μg/mL); Vitek2 included GN70 (n = 62 isolates; MIC calling range, ≤4 μg/mL to >128 μg/mL). MicroScan panels were read by the instrument software, not manually. Personnel performing BMD testing were blind to the automated AST system results. All protected health information was stripped from all isolates, and the VUMC Institutional Review Board (IRB) deemed this study exempt as non-human subject research.

BMD testing.

BMD was performed according to standards described by CLSI (8). Briefly, panels were prepared in-house using an HTF plate dispenser system (MDZ Automation, Chicago, IL) into 96-well polystyrene plates. Antimicrobials on each panel included piperacillin ranging from 128 μg/mL to 2 μg/mL in a 2-fold dilution series in cation-adjusted Mueller-Hinton broth (CA-MHB). Tazobactam was held at 4 μg/mL in each well. Piperacillin and tazobactam were obtained from U.S. Pharmacopeia (Rockville, MD), and CA-MHB was obtained from BD. Panels were frozen at −70°C prior to use, and a single panel lot was used for the entire study. Quality control was performed using Escherichia coli ATCC 25922, E. coli ATCC 35218, and Pseudomonas aeruginosa ATCC 27853 on each day of testing (9).

Isolates were tested by making a suspension equivalent to a 0.5 McFarland standard in water, which was diluted and inoculated to the BMD panels using a multiprong inoculator according to M07 protocols (8). Plates were incubated at 35°C in ambient air for 16 to 18 h and read by two readers.

Data analysis.

The BMD MIC was used as the reference MIC in analyses. Essential agreement (i.e., MIC ± 1 log₂ dilution from the reference [EA]) and categorical agreement between BMD and the automated AST systems were evaluated. For isolates with off-scale MICs (e.g., ≤2 μg/mL), EA was assumed unless the BMD MIC was >1 log₂ dilution from the highest MIC (e.g., 8 μg/mL or above). Breakpoints evaluated in these analyses are in Table 1. Error rates were defined as very major error (VME; number of false susceptible), major error (ME; number of false resistant), and minor error (MIN; SDD or intermediate by one method but either susceptible or resistant by the other), compared to BMD interpretation. VME rates were scored using the number that were resistant by BMD as the denominator, and ME rates were scored using the number of isolates with BMD results that were susceptible as the denominator.

RESULTS

Phoenix.

One hundred sixty-seven isolates were tested by BMD and on the Phoenix (Table 2). This included 111 isolates susceptible by M100 32nd edition breakpoints, 8 SDD, and 48 resistant, by reference BMD. Thirty-eight (22.8%) isolates had BMD MICs within a dilution of the clinical breakpoints (i.e., MICs of 8, 16, or 32 μg/mL). Overall EA between BMD and Phoenix MICs was 83.2%, with 28 isolates displaying MICs out of EA with the reference result (Table 3). Six of these resulted in a lower MIC by the Phoenix than by BMD (data not shown). When CLSI breakpoints were applied to both Phoenix and BMD MICs, categorical agreement between the two was 84.4% with 2 VME (4.8%), 2 ME (1.8%), and 22 MIN (13.2%). VMEs were for Escherichia coli isolates with an MIC of 8 μg/mL by Phoenix and 32 or >128 μg/mL by BMD (data not shown). MEs included one Klebsiella aerogenes (MIC of 32 μg/mL by the Phoenix but 8 μg/mL by BMD) and one Klebsiella pneumoniae (MIC of >64 μg/mL by Phoenix but ≤2 μg/mL by BMD). Among the MINs, 10 were more susceptible by the Phoenix than by BMD, including 2 isolates that tested susceptible by Phoenix but SDD by BMD and 8 that tested SDD by Phoenix but resistant by BMD. Fifteen MINs were for isolates with MICs in EA with the reference.

TABLE 3.

Piperacillin-tazobactam results, tested by Phoenix, MicroScan, and Vitek2, compared to reference broth microdilution interpreted using CLSI or FDA breakpoints^a

Test	Breakpoint	EA (%)	CA (%)b	VME (%)c	ME (%)d	MIN (%)e
Phoenix	CLSI	139/167 (83.2)	141/167 (84.4)	2/48 (4.2)	2/111 (1.8)	22/167 (13.2)
	FDA		142/167 (85.0)	1/23 (4.3)	1/119 (0.8)	23/167 (13.7)
MicroScan	CLSI	46/55 (83.6)	44/55 (80.0)	4/11 (36.4)	2/42 (4.8)	5/55 (9.1)
	FDA		45/55 (81.8)	4/9 (44.4)	0/44 (0.0)	6/55 (10.9)
Vitek2	CLSI	60/62 (96.7)	59/62 (95.2)	1/16 (6.3)	0/43 (0.0)	2/62 (3.2)
	FDA		60/62 (96.8)	0/8 (0.0)	1/46 (2.2)	1/62 (1.6)

^aCLSI, Clinical and Laboratory Standards Institute 2022 (Enterobacterales); FDA, U.S. Food and Drug Administration susceptibility test interpretive criteria (STIC) website breakpoints, accessed 31 October 2022; EA, essential agreement; CA, categorical agreement; VME, very major error; ME, major error; MIN, minor error.

^bNumber of isolates in CA/total isolates.

^cNumber of VME/resistant isolates.

^dNumber of ME/susceptible isolates.

^eNumber of MIN/total isolates.

When evaluated using the FDA breakpoints for both BMD and Phoenix, categorical agreement was 85.0%, with 1 VME (4.3%), 1 ME (0.8%), and 23 MINs (13.7%, Table 3). Differences in categorical agreement between FDA and CLSI breakpoints included one VME by CLSI that scored as a MIN by FDA, for an E. coli isolate with a Phoenix MIC of 8 μg/mL (susceptible by both breakpoints) and BMD MIC of 32 μg/mL (resistant by CLSI but intermediate by FDA breakpoints). One of the MEs by CLSI breakpoints (for K. aerogenes with an MIC of 32 μg/mL by Phoenix and 8 μg/mL by BMD) was converted to a MIN when interpreted by FDA breakpoints.

MicroScan.

Fifty-five isolates were evaluated by the MicroScan. Forty-two isolates were susceptible by CLSI breakpoints, 2 were SDD, and 11 were resistant, by BMD. Seven isolates had reference BMD MICs within a doubling dilution of the CLSI breakpoints (12.7%). Overall EA between MicroScan and BMD was 83.6%. Among the 9 isolates not in EA, six yielded a more susceptible MIC by MicroScan (data not shown). When evaluated by CLSI breakpoints, 44 isolates were in categorical agreement (80.0%). Four VMEs (36.4%), 2 MEs (4.8%), and 5 MINs (9.1%) were observed (Table 3). Three VMEs were for isolates that displayed MICs of ≤8 μg/mL by the MicroScan but >128 μg/mL by BMD and included two E. coli isolates and one Enterobacter cloacae isolate. The fourth VME was for an isolate of E. coli with a MicroScan MIC of ≤8 μg/mL but a BMD MIC of 64 μg/mL (data not shown). MEs included one isolate of E. coli and one isolate of Serratia marcescens, with MicroScan MICs of 32 μg/mL and 64 μg/mL, respectively, but BMD MICs of 2 μg/mL. Three of the 5 MINs were due to a more susceptible result by the MicroScan (1 SDD by MicroScan but R by BMD and 2 susceptible by MicroScan but SDD by BMD), and 2 were more resistant by MicroScan (both SDD by MicroScan but susceptible by BMD).

When FDA breakpoints were applied to both MicroScan and BMD, categorical agreement was 81.8%, with 4 VMEs (44.4%), 0 ME (0.0%), and 6 MINs (10.9%, Table 3). One of the VMEs by M100 32nd edition breakpoints converted to a MIN, for an E. coli isolate that tested ≤8 μg/mL by MicroScan but 64 μg/mL by BMD, which is resistant by CLSI but intermediate by FDA breakpoints. One MIN by the CLSI breakpoints converted to a VME, for an E. coli isolate with a MicroScan MIC of 16 μg/mL (susceptible by FDA and SDD by CLSI) and BMD MIC of >128 μg/mL. Similarly, two of the four MEs by M100 32nd edition breakpoints became MINs when evaluated by FDA breakpoints, for isolates of Enterobacterales with MicroScan MICs of 32 μg/mL or 64 μg/mL (intermediate by FDA but resistant by CLSI). Among the 6 MINs, 4 were more susceptible by MicroScan than by BMD (data not shown).

Vitek2.

Sixty-two isolates were evaluated by the Vitek2. Forty-three isolates were susceptible, 3 were SDD, and 16 were resistant by M100 32nd edition breakpoints and BMD. Fifteen isolates had BMD MICs within a doubling dilution of the CLSI breakpoints (24.2%). Only two isolates (both E. coli) were out of EA, leading to an EA of 96.7%—one isolate had a higher MIC by Vitek2 and one a lower MIC by Vitek2, than by BMD. When both BMD and Vitek2 were evaluated using M100 32nd edition breakpoints, categorical agreement was 95.2% with one isolate yielding a VME (6.3%) and 2 isolates yielding a MIN (3.2%) (Table 3). The single VME was for an isolate of E. coli with a Vitek2 MIC of ≤4 μg/mL but a BMD MIC of 32 μg/mL. The two MINs were both for isolates of E. coli, one of which was outside EA, with an MIC of ≥128 μg/mL by Vitek 2 but 16 μg/mL (SDD) by BMD. When both BMD and Vitek2 were evaluated by FDA breakpoints, categorical agreement was 96.8%. The VME observed by CLSI breakpoints resolved a MIN by FDA breakpoints (Table 3). One of the MINs by CLSI breakpoints became an ME (isolate of E. coli with an MIC of ≥128 μg/mL by Vitek2 and 16 μg/mL by BMD), and one MIN was resolved (Vitek2 MIC, 8 μg/mL, and BMD MIC, 16 μg/mL—both susceptible by FDA breakpoints).

Impact of not adopting CLSI breakpoints on commercial AST systems.

The impact of not adopting CLSI breakpoints on the commercial systems described here was evaluated by comparing the interpretations as reported by the test systems (i.e., using FDA breakpoints) against the reference BMD results interpreted using CLSI breakpoints (Table 4). Overall, among the 75 isolates resistant by CLSI breakpoints, 21.3% would be reported as susceptible by the commercial systems (n = 16). By the Phoenix test, this amounted to 20.8% VMEs (n = 10), 0.9% ME (n = 1), and 10.8% MINs (n = 18, 17 of which were more susceptible by the Phoenix interpreted by FDA breakpoints than by reference BMD interpreted by CLSI breakpoints). For isolates tested by MicroScan, 45.5% VME (n = 5) would occur, with 0.0% ME (n = 0) and 14.5% MINs (n = 8, 6 more susceptible by MicroScan interpreted by FDA than by reference BMD interpreted by CLSI). For isolates tested by Vitek2, it was 6.3% VME (n = 1), 0% ME (n = 0), and 16.1% MINs (n = 10, 9 more susceptible by Vitek2 interpreted using FDA breakpoints than by reference BMD interpreted by CLSI breakpoints). In all cases, the number of VMEs (i.e., false susceptibility) was higher in this analysis than when the commercial AST data were interpreted using the M100 32nd edition Enterobacterales breakpoints. Similarly, the overall categorical agreement with reference BMD interpreted with M100 32nd edition breakpoints was improved when adopting the CLSI breakpoints compared to not adopting them. For the Phoenix test, categorical agreement with BMD was 82.6% versus 85.0%, for MicroScan, categorical agreement was 76.4% versus 80.0%, and for Vitek2, categorical agreement was 82.3% versus 95.2%, for no update to breakpoints (Table 4) versus updating to M100 32nd edition (Table 3), respectively.

TABLE 4.

Piperacillin-tazobactam results, as reported using FDA-cleared software with FDA breakpoints by Phoenix, MicroScan, and Vitek2, compared to reference broth microdilution interpreted using current CLSI breakpoints

Test	CA (%)a	VME (%)b	ME (%)c	MIN (%)d
Phoenix	138/167 (82.6)	10/48 (20.8)	1/111 (0.9)	18/167 (10.8)
MicroScan	42/55 (76.4)	5/11 (45.5)	0/42 (0.0)	8/55 (14.5)
Vitek2	51/62 (82.3)	1/16 (6.3)	0/43 (0.0)	10/62 (16.1)

^aNumber of isolates in CA/total isolates.

^bNumber of VME/resistant isolates.

^cNumber of ME/susceptible isolates.

^dNumber of MIN/total isolates.

DISCUSSION

The use of TZP for the treatment of infections caused by Enterobacterales has been a subject of much recent debate. Of primary concern is the risk of inaccurate results generated by clinical laboratory testing. Indeed, in the MERINO trial, several patients who failed TZP therapy were infected by isolates that were erroneously reported as susceptible by the clinical laboratory using commercial systems. When the MERINO trial data were reevaluated using reference BMD, a mortality signal for isolates with MICs of ≥32 μg/mL was observed (3). While anecdotes suggest reproducible TZP MICs are elusive even by the reference BMD method, we recently found >90% categorical agreement across brands of Mueller-Hinton broth and drug powders for TZP using a challenging set of isolates with MICs near the clinical breakpoint (9).

Overall, the test systems did not meet traditionally accepted performance specifications, as outlined by the CLSI M52 guideline (10). The Phoenix test displayed EA and categorical agreement below 90%, with VMEs of >3% (4.2%) by CLSI breakpoints. In addition, while not specified by M52, MIN rates of <10% are generally considered acceptable, but the Phoenix test displayed a MIN rate of 13.2% (Table 3). Of note, the categorical agreement was only marginally better when both Phoenix and BMD were evaluated using FDA breakpoints (85.0% categorical agreement by FDA versus 84.4% by CLSI), and VME and MIN rates remained above 3% and 10%, respectively. The Phoenix test does not have any FDA-imposed limitations for TZP, so laboratories should be aware that adopting CLSI breakpoints yields performance no worse than that by FDA breakpoints. In contrast, 20.8% of isolates resistant by the CLSI breakpoints would be reported as susceptible using the FDA-cleared software on the Phoenix test, which is a substantial patient safety concern for laboratories not updating to current CLSI breakpoints, as BMD MICs of ≥32 μg/mL are associated with increased risk of mortality for patients treated with TZP (2, 3).

MicroScan data, while for a smaller number of isolates, yielded high rates of VMEs, regardless of use of CLSI or FDA breakpoints. Upon starting this study, the error rates were so high that the first set of data from one site was discarded, and testing was repeated (by both reference and MicroScan for all isolates), due to concern of isolate mislabeling. However, the data did not change outside the expected MIC variability, and the repeat results were not used for analysis (data not shown). No limitations are listed in the package insert for the MicroScan for TZP. Errors were observed for isolates of E. coli (n = 3) and E. cloacae (n = 1) and were apparent by both CLSI and FDA breakpoints. We cannot exclude the possibility that the three VMEs for E. coli were due to a clonal isolate, as all three were from the same clinical laboratory. Further evaluations of TZP on the MicroScan test, using more isolate collections, will help to resolve this issue. Nonetheless, implementing the CLSI breakpoints on the MicroScan test would not substantially change performance compared to continued use of FDA-cleared software and breakpoints.

Finally, the Vitek2 test had fewer errors than the other commercial systems evaluated here with the revised CLSI breakpoints, with only 1 VME and 2 MINs (Table 3). While the VME rate is technically above 3% accepted by M52, 1 error is acceptable in the M52 guidance when updating breakpoints. This excellent performance of the Vitek2 test may be in part due to recent reformulations of TZP on the Vitek2 test following FDA recalls in 2010; these reformulations were approved by the FDA in 2013 (11).

It should be noted that the isolates selected in this study, while pragmatic based on laboratory availability, are enriched for those with MICs near the CLSI breakpoint. In particular, 22.8% and 24.2% of isolates tested by the Phoenix and Vitek2 tests, respectively, had reference MICs at the clinical breakpoint, which would be expected to yield a higher rate of MINs. We note this proportion, while high, did not meet the 40% threshold of use of the error-rate bounded method of analysis (7). Nonetheless, this is one challenging aspect of TZP testing, as the epidemiological cutoff for wild-type MICs abuts the clinical breakpoints of ≤8 μg/mL for Enterobacterales (4). Thus, by the expected variability of MIC testing, the majority of the wild-type population may yield MINs when tested repeatedly. As such, it is encouraging that MIN rates were near the acceptance criteria. In our experience, testing errors are often as equally attributable to the initial “reference” BMD result as they are to commercial test inaccuracies (11, 12).

This study has several limitations. Isolates were tested on the commercial devices and by reference BMD in different laboratories, using different inocula. Inoculum variability, even when adjusted to within the acceptable CLSI ranges, can impact MIC results (13, 14). However, the method used here (i.e., testing at a centralized reference laboratory) is commonly employed during clinical trials for AST devices. Since most clinical laboratories do not have access to reference BMD, our study provides a reasonable estimation of the type of data individual clinical laboratories may expect to observe when validating CLSI TZP breakpoints. Furthermore, the study was enriched with isolates tested on the Phoenix test, due to instrument availability in the laboratories performing testing. Finally, we did not assess underlying resistance mechanisms for the isolates evaluated, although this is not routinely required for AST device evaluations.

A CDC-FDA AR Bank Enterobacterales isolate panel will soon be available for validation of the current CLSI breakpoints for TZP. Laboratories may opt to use these isolates to validate their devices or wait for FDA clearance. Clearance may take years to achieve if test manufacturers must repeat clinical trials and/or reformulate their devices to accommodate the revised breakpoints (15). Laboratories may consider discussing these changes with their infectious disease physicians and antibiotic stewardship team, to determine the best path for their patients. Importantly, we found 20.8% (Phoenix), 45.5% (MicroScan), and 6.3% (Vitek2) of isolates with MICs of ≥32 μg/mL by reference BMD were reported as susceptible by current software, which poses a significant patient safety risk, as these MICs are associated with mortality if TZP is used for therapy. The M100 32nd edition CLSI breakpoints are the best predictors of patient outcome, and serious consideration should be given to implementing these as soon as possible. The analytical errors identified in this study are mitigated when the CLSI breakpoints are implemented.