Building a Better Test for Piperacillin-Tazobactam Susceptibility Testing: Would that It Were So Simple (It's Complicated)

TEXT

Piperacillin-tazobactam (TZP) has long been a workhorse for the empirical and directed treatment of common infections involving Gram-negative pathogens, such as complicated urinary tract infections, intra-abdominal infections, and febrile neutropenia. However, the MERINO trial illustrated that TZP may be a poor choice for some infections, such as those caused by extended-spectrum-β-lactamase (ESBL)-producing Enterobacterales. Specifically, MERINO found that TZP was not noninferior to meropenem for the treatment of ceftriaxone-not-susceptible (intermediate or resistant) but TZP-susceptible bloodstream infections caused by Escherichia coli or Klebsiella. Furthermore, mortality did not correlate to the initial TZP MIC when testing was performed by batch testing at a central laboratory by Etest (1). These clinical woes are echoed in the laboratory, with analytical performance challenges being reported for several commercially distributed TZP susceptibility tests. A warning statement was posted to the European Committee on Antimicrobial Susceptibility Testing (EUCAST) website regarding the poor performance of TZP gradient diffusion strips in 2015, a U.S. FDA recall of the TZP Etest in 2015, and a U.S. recall of TZP on Vitek2 cards in 2011. In this issue of the Journal of Clinical Microbiology (JCM), García-Fernandez and colleagues present the outcomes of a new study that evaluated the analytical performance of a reformulated Etest for TZP (2). The data are excellent, with >95% essential agreement (EA) for Enterobacterales, 98.3% EA for Pseudomonas aeruginosa, and 91.6% EA for the Acinetobacter baumannii complex being found. Categorical agreement was 93.0%, 93.3%, and 89.2% for these organism groups, respectively, by the use of Clinical and Laboratory Standards Institute (CLSI) breakpoints and 94.8% and 95.8% for Enterobacterales and P. aeruginosa, respectively, by the use of EUCAST breakpoints. However, in light of the tarnished reputation for TZP, what is the true diagnostic certainty of a “susceptible” result for this drug on a laboratory report?

TZP MIC values for wild-type Enterobacterales typically range from 0.25 to 8 μg/ml (www.eucast.org), and the epidemiological cutoff (ECOFF and ECV by EUCAST and CLSI language, respectively) is 8 μg/ml. This value reflects the upper MIC limit of the normal MIC distribution range for the wild-type population. For wild-type isolates of P. aeruginosa, TZP MIC values are ≤16 μg/ml. Pharmacokinetic-pharmacodynamic (PK-PD) data suggest that a susceptible breakpoint of ≤16 μg/ml is appropriate for both organism groups, based primarily upon piperacillin PK-PD modeling, as little is currently known about the PK-PD indices of tazobactam (3). In vitro testing for TZP utilizes a fixed concentration of tazobactam at 4 μg/ml for both the EUCAST and CLSI methods. Importantly, non-wild-type Enterobacterales isolates harboring ESBL and those with AmpC genes have TZP MICs that overlap the wild-type TZP distribution (4, 5). Prior to the MERINO trial, over 20 observational studies suggested that TZP may be a suitable carbapenem-sparing agent for the treatment of these infections (6). Similar to other β-lactam combination agents, the in vitro activity of TZP can be affected by inhibitor-resistant TEMs, TEM hyperproduction, an altered penicillin binding protein 3 (PBP 3), porin mutations, and/or efflux pumps.

Like all antimicrobial susceptibility tests, TZP MIC test results are precise only within 1 log₂ dilution, i.e., within EA. CLSI and EUCAST breakpoints differ for TZP, but both abut the values for the wild-type population; this, combined with the imprecision of MIC tests, can lead to categorical errors. An important difference between CLSI and EUCAST is the approach to the challenge of test imprecision. CLSI uses the intermediate category to define both test variability and the possibility of treatment success through increased exposure, whereas EUCAST utilizes “I” only to indicate the possibility of successful treatment if increased antimicrobial exposure can be used, such as by increasing the dose (7). EUCAST also recently introduced a separate concept, called the area of technical uncertainty (ATU), which accounts for the challenge of test variability. However, an ATU has not been established for all drug-bug combinations with clinical breakpoints. For instance, the TZP ATU for the Enterobacterales is >8 μg/ml to ≤16 μg/ml, and no ATU has been defined for P. aeruginosa (7). Enterobacterales isolates with an MIC of 16 μg/ml (i.e., the ATU) typically express an ESBL or an ampC gene (with 16 μg/ml being the high end of the MIC range for these phenotypes) or a narrow-spectrum oxacillinase gene. In particular, the OXA-1 gene in the presence of an ESBL has been shown to increase the modal MIC from 2 μg/ml to 8 to 16 μg/ml (4). Again, because of the variability of MIC testing, isolates with MIC values of 16 μg/ml on initial testing may yield MIC values of 8 to 32 mg/liter on repeat testing (4). Evidence of this imprecision is found in the work of García-Fernandez et al. (2), where 85 of the 239 (35%) MICs generated as a part of the precision study did not yield the same MIC when the isolate was retested. In fact, the authors had to exclude two isolates (one E. coli isolate and one Klebsiella isolate) from the analysis because a reference MIC could not be established by broth microdilution (BMD). A similar finding of no reference MIC (in this case, no mode from three replicate, parallel MICs) was found for 3.4% of 3,724 TZP results in the PhenoTest BC clinical trial (8). Two instances of false resistance (major errors) and two instances of false susceptibility (very major errors) were observed in the Etest study by García-Fernandez et al. (2) due to organisms for which MICs were within EA with the MICs obtained by the reference BMD methods but for which the MICs straddled the susceptible/resistant EUCAST breakpoint. Admittedly, these numbers are small in the context of the overall study (11 out of 977 total tests [1.1%] had categorical issues). However, only the minority of isolates evaluated in the study had MICs near the breakpoint. For instance, 24 of the 775 (3.1%) Enterobacterales isolates evaluated had a BMD MIC of 16 μg/ml and 28 (3.6%) had an MIC of 32 μg/ml. Among these isolates, 26 (50%) yielded a different categorical interpretation between BMD and Etest by the use of CLSI breakpoints, a finding to be expected of any antimicrobial susceptibility testing (AST) device. Such data highlight that close scrutiny of MICs by the treating clinician to inform dosing and/or therapy duration can be problematic, in particular, when decisions are made over MICs within EA (9). Of note, these data also illustrate an interesting point regarding allowable trial design between the U.S. Food and Drug Administration (FDA) requirements and those of the International Standards Organization (ISO). ISO allows the resolution of discrepancies between the reference AST method (BMD) and the device under investigation (in this case, Etest), whereas FDA does not allow such discrepancy testing. However, knowledge of the test performance for isolates with values in these difficult analytical areas (i.e., isolates with MICs close to the breakpoint) is valuable to the clinical laboratory and the treating physician. Allowing discrepancy resolution is one way to ensure that a relevant bacterial population is evaluated in the device trial while not penalizing manufacturers that identify such challenging isolates with an acceptance criterion more stringent than is possible by the reference standard BMD method.

When searching for the value of diagnostic certainty for AST devices, the MERINO trial provides an excellent case in point (1). Participants from 26 hospitals in 9 countries were randomized to either meropenem or TZP treatment, after confirming that their isolate was ceftriaxone not susceptible but TZP and meropenem susceptible. This testing was done by the local laboratories, using their standard methodologies, as the trial was designed to be pragmatic. A post hoc analysis was performed on most isolates (320/379 primary blood culture isolates) in a central laboratory, which found that many of the isolates were in fact TZP not susceptible by broth microdilution. Furthermore, MICs and categorical interpretations differed significantly between the Etest (performed with the older Etest version) and BMD methods (1, 10). Isolates coharboring OXA-1 and ESBL genes accounted for a significant proportion of the strains that tested nonsusceptible to TZP by BMD but susceptible by the routine laboratory methods. These isolates typically had MICs of 8 to 32 μg/ml, i.e., MICs straddling the breakpoint. Additional analysis is under way to assess the true association between the TZP MIC, the β-lactamase resistome, and mortality in the MERINO trial, with preliminary results suggesting a strong mortality association for isolates with MIC values of >16 mg/liter, as would be predicted by current breakpoints (8).

It is without question that the well-informed use of antimicrobials is critical to the future utility of these agents. Clinicians place tremendous weight on MIC results as a part of clinical decision making, and unfortunately, the accuracy of these results is too often taken for granted. The consequences of inaccurate MIC results cannot be underscored, as evidenced by the MERINO trial. As such, antimicrobial susceptibility tests must be evaluated against contemporary microorganisms and updated as needed in studies such as the one presented in this issue of JCM for the TZP Etest (2). However, the work does not end with a new regulatory clearance. Rather, mechanisms are needed to ensure the frequent and robust evaluation of the AST devices used by clinical laboratories—by both the academic community and the device manufacturers—to ensure that these continue to perform acceptably in the face of changing epidemiology, breakpoints, and treatment paradigms.