Laces out Dan! The role of tazobactam based combinations for invasive ESBL infections in a post-MERINO world

KEYWORDS:

• Extended-spectrum beta
• lactamase
• beta-lactamase inhibitor
• antimicrobial resistance
• gram-negative infections

1. Introduction

Extended-spectrum beta-lactamase (ESBL) producing Enterobacteriaceae are considered critical priority pathogens for research and development of new antibiotics by the World Health Organization, with rates of infections with ESBL-producing organisms increasing in both inpatient and outpatient settings worldwide [1–4]. Carbapenems have traditionally been regarded as the treatment of choice for serious ESBL-producing infections; however, widespread utilization of carbapenems has driven carbapenem-resistance, which is a more grave threat than ESBLs [5–7].

Identifying alternative, non-carbapenem agents for the treatment of ESBL-producing infections is an emerging strategy in curbing overutilization of carbapenems. Controversy exists regarding the role of beta-lactam/beta-lactamase inhibitor (BLBLI) combination antibiotics as a carbapenem-sparing treatment option for infection due to ESBL producing organisms. The goal of this viewpoint is to discuss the role of tazobactam-containing BLBLI treatment regimens for invasive ESBL-producing infections.

Tazobactam is a beta-lactamase inhibitor that possesses inhibitory activity against most ESBL variants present in Enterobacteriaceae [8]. Tazobactam is available in fixed-dose combinations with the piperacillin (TZP) and ceftolozane (CT). TZP demonstrates in vitro activity against a considerable number of ESBL producing isolates; however, the potential for organisms to have increased expression of the ESBL enzyme, have multiple beta-lactamases or other resistance mechanisms present, and concerns about the ‘inoculum effect’ have called into question the clinical utility of TZP for the treatment of serious ESBL infections [9,10].

2. Pre-MERINO data

Some of the earliest data supporting the use of BLBLIs for ESBL were from Rodriguez-Bano and colleagues who performed a post-hoc analysis of patients with bloodstream infection (BSI) from ESBL-E.coli who received empiric and/or definitive therapy with a BLBLI (amoxicillin-clavulanic acid or TZP) or carbapenem. Patients were analyzed in two cohorts, an empiric therapy cohort (ETC) and a definitive therapy cohort (DTC). Thirty-day mortality in the ETC and DTC for those treated with BLBLI versus carbapenems were 9.7% and 19.4%; and 9.3% versus 16.7%, respectively (P > 0.2), and the various regression models performed demonstrated that empiric or definitive therapy with a BLBLI did not impact mortality. While the findings of this study were promising, there were a few key considerations that limit generalizability. Overall, less than one sixth of the patients were in an ICU and few patients were immunosuppressed. The numerically higher mortality rates in the carbapenem group call into question if there is unaccounted for selection bias between the two groups, and indeed there were higher rates of immunosuppression, severe sepsis or septic shock, and non-biliary or urinary tract sources of infection in the carbapenem group in both the ETC and DTC. Further, roughly half of the patients who received empiric therapy with TZP received definitive therapy with a carbapenem and how this crossover impacted outcomes is unclear. Importantly, the median TZP MIC was low (2/4 mg/L) and more than 90% of patients received ‘high-dose’ of 4.5 grams IV every 6 hours as an intermittent infusion [11]. In fact, in a separately published analysis assessing the impact of source of infection and TZP MIC on outcomes in these patients, the authors demonstrated that 0/11 patients who received TZP in the ETC who had a urinary source of infection died, regardless of TZP MIC value. However, for other sources of BSI, mortality was 0/11 if the MIC was ≤ 2 mg/L, but 7/17 (41%) when the MIC was > 2 mg/L [12].

These analyses were followed by a retrospective cohort from Tamma and colleagues that evaluated 14-day mortality in patients receiving empiric TZP or carbapenems in patients with ESBL bacteremia who all received definitive carbapenem therapy. In converse to the previous data, the adjusted risk of 14-day mortality was 1.92 (95% CI 1.07–3.45) for those who received empiric TZP. Importantly, 99% of isolates had TZP MICs ≥ 4/4 mg/L and the majority of patients received a 3.375 gram every 6-hour (30-minute infusion) dose of TZP. In further contrast to the previous data, the majority of patients had infections from higher-inoculum sources including central line infections and pneumonia [13].

Numerous other retrospective studies have been published evaluating the question of BLBLIs versus carbapenems including the largest cohort; a multinational, observational study by Gutierrez-Gutierrez and colleagues which showed that BLBLIs and carbapenems had no difference in 30-day mortality rates. However, similar to the Rodriguez-Bano study, few patients were critically ill, the majority of patients had infections from urinary or biliary sources, and most patients received a 4.5 gram dose of TZP [14]. These conflicting studies left many questions regarding the differences seen in outcomes. Were the differences in findings related to the differences in infection source? Study Design? TZP MIC distributions or TZP dosing? The MERINO trial was set to be the definitive answer.

2.1. Post-MERINO and beyond

The MERINO trial was a prospective, multicenter, international, open-label randomized controlled noninferiority trial comparing meropenem (1 gram every 8 hours as a 30-minute infusion) versus TZP (4.5 grams every 6 hours as a 30-minute infusion) for definitive treatment of ceftriaxone-nonsusceptible E. coli or K. pneumoniae BSI in adult patients. Patients had to be randomized within 72 hours of the collection of the positive blood culture and have isolates susceptible (per local laboratory testing) to both study drugs. Randomized patients received 4–14 days of study treatment at the discretion of the treating clinician and were stratified based on severity of illness, site of infection, and causative pathogen. The primary outcome was all cause 30-day mortality. Secondary outcomes included time until resolution of signs and symptoms of infection and the incidence of secondary infection with a meropenem- or TZP-resistant organism or C. difficile infection within 30 days post randomization.

The study was designed to enroll 454 patients in order to demonstrate non-inferiority of TZP, however it was stopped early when an interim analysis demonstrated increased mortality in the TZP arm (23/187 (12.3%) versus 7/191 (3.7%) in the meropenem arm, risk difference 8.6%; p = 0.004.) Further, although underpowered to demonstrate significance, secondary outcomes also tended to favor meropenem. Clinical and microbiological success at day 4 was demonstrated in 121/177 (68.4%) of TZP patients and 138/185 (74.6%) of meropenem patients (- 6.2 95% CI −15.5 to 3.1). Additionally, secondary infection with a TZP or meropenem-resistant isolate or C. difficile infection occurred more frequently in TZP (15/187 (8.0%)) than meropenem (8/191 (4.2%)) treated patients; difference 3.8 (- 1.1 to 9.1), including numerically higher (3.2% vs 2.1%) carbapenem-resistant organisms in the TZP arm. These advantages of meropenem were in spite of similar baseline characteristics between the two groups. Similar to the majority of the previous studies the source of bacteremia was primarily the urine (61%) or the abdomen (16%) [15].

When mortality was assessed in TZP treated patients as a function of TZP MIC (determined on site by MIC test strips), no association was demonstrated and mortality rates were similar in patients with isolates where the MIC was ≤ 2 mg/L (10/69, 14.5%) and those with isolates with MICs > 2 mg/L (8/63, 12.7%.) Of note, a post-hoc analysis was recently presented at a scientific conference. In this analysis, the central lab performed broth microdilution MIC testing for TZP for the 157 patients receiving TZP therapy with isolates saved. In this analysis it was demonstrated that mortality was higher in TZP treated patients with non-susceptible isolates (MIC > 16 mg/L, 5/10, 50%) than those with susceptible isolates (MIC 13/147, 8.8%); p = 0.002 [16].

So what is a clinician to take from the failure of TZP to demonstrate noninferiority in MERINO? While the increased number of deaths with TZP is certainly troubling it is worth considering that none of the deaths appeared to be directly related to the index infection. Common causes were advanced metastatic cancer, re-infection with other pathogens, or progression of longstanding comorbidities. While it is impossible to ascertain the impact of the index infection on the progression of these other illnesses, the lack of direct correlation warrants mention. With that said, it is important to note that the secondary outcomes also tended to favor meropenem, and there was no signal that development of resistance, the primary rationale for wanting carbapenem alternatives, occurred less commonly in TZP patients.

Then there is the MIC analysis. The data presented at ECCMID have raised eyebrows that when broth microdilution MICs were considered, mortality was significantly higher in patients with TZP non-susceptible isolates, supporting that the appropriate dose might be able to salvage TZP therapy. While it is true that patients treated with TZP with non-susceptible isolates had higher rates of mortality there are a few important points to note with regards to this line of discussion. First and foremost, mortality in TZP patients with ‘susceptible’ isolates (MIC ≤ 16 mg/L) was still higher 13/147 (8.8%) than in patients who received meropenem 9/191 (3.7%). Secondly, mortality rates were actually quite high in patients with the most susceptible isolates (3/10, 30% if MIC was 1 mg/L; 6/51, 11% if MIC was 2 mg/L), and if one was to parse data to force an MIC outcome, mortality would be significantly higher in patients with MICs ≤ 2 mg/L (9/61, 14.8%) than those with isolates of 4–16 mg/L (4/86, 4.7%); p = 0.04.

Thirdly, pharmacokinetic data from hospitalized patients demonstrate that even this ‘high dose’ of piperacillin/tazobactam, when administered as a 30-minute infusion fails to reliably achieve a time > MIC goal of 50% or higher for piperacillin when the MIC is > 4 mg/L. In fact, only a 4.5 gram every six-hour dose as a 3-hour infusion would be able to attain the targeted piperacillin exposure at the susceptibility breakpoint [17].

However, to be honest, that PK/PD analysis is largely irrelevant because it only takes into consideration the piperacillin concentration. Given that we are discussing restoration of piperacillin by tazobactam, identifying the tazobactam exposures necessary to achieve this restoration is the more relevant question. While very little evidence currently exists assessing tazobactam PK/PD, what is out there should worry clinicians. VanScoy and colleagues, in data presented at a scientific conference, assessed optimal tazobactam exposures required to ‘restore’ bacteriostatic and 1 log reduction bactericidal activity of a simulated 4 gram every six-hour dosing regimen of piperacillin in a one compartment in vitro infection model [18]. In this analysis the investigators demonstrated that tazobactam’s ability to restore piperacillin’s activity was associated with optimizing the time that a free tazobactam concentration was above a threshold concentration. For TZP the most predictive threshold tazobactam concentration was the TZP MIC. In order to restore stasis free tazobactam concentrations had to remain above the TZP MIC for 64% of the dosing interval, and for bactericidal activity 77% time above this concentration was required [18].

While robust tazobactam pharmacokinetic analyses with 500 mg every six-hour dosing (the dose given with ‘high-dose’ piperacillin) have not been performed, simply piecing everything together clearly shows the need for pause. The average free Cmax of tazobactam, from the TZP package insert is ~ 24 mg/L, which is a value barely above the susceptibility breakpoint of 16/4 mg/L [19]. With an estimated half-life of roughly one hour, tazobactam concentrations would be expected to quickly drop and given the exposures needed even treating highly ‘susceptible’ isolates would become a concern. This concern is highlighted in data presented at a scientific conference where a Monte Carlo Simulation was performed based on tazobactam PK data from 44 healthy volunteers [20]. These data demonstrated that a higher tazobactam dose of 1 gram every 8 hours as a longer 90-minute infusion was only able to achieve a 79% probability of a tazobactam concentration above 0.5 mg/L for 60% of the dosing interval. Let that sink in. It comes just short of the stasis target at an MIC of 0.5 mg/L (the breakpoint is 16 mg/L) with a higher dose/longer infusion strategy. Similarly, even if we apply data presented from tazobactam (1 gram every 8 hours as a 60-minute infusion) given in combination with ceftolozane in a critically ill population where the half-life (2.24 hours) was over twice as high as those derived from healthy volunteers (allowing a longer time above a threshold concentration,) the median (range) % time above a threshold concentration of 1 mg/L of tazobactam was 82% (35–100) [21]. These data would suggest that while the average patient would achieve the 1 log kill target at an MIC of 1/4 mg/L, there is little doubt that if we applied the normal 90% probability of target attainment standard that we would fail to reliably achieve either the stasis or bactericidal target at a TZP MIC of 1/4 mg/L!!

So the bottom line when it comes to TZP is that before we get hung up on the findings and/or limitations from MERINO and why they differed from the previous retrospective data perhaps we should step back and focus our efforts on figuring out what exactly should be considered susceptible, how frequently that occurs, and whether, in light of those findings, there is a rationale path forward for TZP for ESBLs. What we definitely should not do is assume that the addition of a random concentration of tazobactam (4 mg/L) returning piperacillin to or below its susceptibility breakpoint in a test tube means that labeled doses of TZP will work for the management of infections due to beta-lactamase producing organisms. Finally, we should not assume that because TZP appears suboptimal for ESBL-producing infections, that this would automatically apply to other tazobactam based combinations which will differ in the stability of the parent beta-lactam to hydrolysis by the enzyme, the dose/infusion strategy of both the parent drug and tazobactam, and the ‘susceptibility’ breakpoint.

So that leads us to the other tazobactam based combination available, CT. CT comes with two labeled doses of 1.5 grams every 8 hours for intra-abdominal infections (IAI) and urinary tract infections (UTI) and 3 grams every 8 hours for ventilated nosocomial pneumonia. Notably, that equates to 500–1000 mg of tazobactam every 8 hours as a 60-minute infusion. In general, ceftolozane is more stable to hydrolysis by common ESBLs than piperacillin, and thus tends to have better activity and potency. In fact, at the CLSI breakpoint of 2/4 mg/L, CT demonstrated greater in vitro activity against a collection of 139 ESBL + E. coli (91% vs. 78%) and 107 ESBL + K. pneumoniae (71% vs. 52%) than TZP at a breakpoint of 16/4 mg/L [22].

The question arises however, given the PK/PD issues described with TZP, how does CT fare in this regard? VanScoy and colleagues assessed tazobactam pharmacodynamics in an in vitro PK/PD model similar to the previously described model for TZP [23]. Analogous to the findings for TZP, the authors demonstrated that time above a threshold tazobactam concentration was the PK/PD index that best correlated to restoration and that free tazobactam concentrations needed to be above the threshold concentration 66% and 77% of the time for bacteriostasis, and 1 log kill, respectively. The difference between CT and TZP was that whereas the threshold concentration for tazobactam when combined with piperacillin was the TZP MIC, the threshold concentration for tazobactam when combined with ceftolozane was 0.5 x the CT MIC [23]. Therefore, at the CT breakpoint of 2 mg/L, the threshold concentration needed would be 1 mg/L.

As both the breakpoint (2 mg/L), and threshold concentration (0.5 x MIC) are significantly lower, and the tazobactam dose (1000 mg every eight hours over 60 minutes with ‘high dose’) is higher with CT, in general there is a decreased concern from a PK/PD standpoint compared to TZP. However, the previously described data suggest challenges achieving target exposures at threshold values of 0.5–1 mg/L (which would equate to a CT MIC of 1–2 mg/L), suggesting some concern at the upper end of the susceptibility breakpoint and that perhaps the EUCAST breakpoint (1 mg/L) is more appropriate [24].

While no clinical data has specifically compared CT and a carbapenem for ESBL BSIs, the experience with CT for infections due to ESBL producing organisms from clinical trials has been encouraging. Popejoy and colleagues reported on patients receiving CT for infections due to ESBL producing Enterobacteriaceae from the UTI and IAI studies (at a dose of 1.5 grams every 8 hours.) Clinical cure was demonstrated in 76/78 (97%) of CT patients, which was comparable to the cure rates with meropenem in the cIAI study [25]. More recently, results from ASPECT-NP, the RCT comparing CT (3 grams every 8 hours as a 60-minute infusion) and meropenem in patients with ventilated nosocomial pneumonia were presented in poster format at a scientific conference. In a post-hoc analysis of patients with ceftazidime-resistant Enterobacteriaceae, mortality was demonstrated in 18/81 (22.2%) of patients receiving CT compared to 20/71 (27.8%) of patients receiving meropenem [26]. While these data are still preliminary they are encouraging for the role of CT in ESBL infections, including high innocula sources.

3. Expert opinion

So, is there a role for tazobactam based combinations in the management of invasive ESBL infections? Absolutely! The main concern is that we just do not currently know exactly what that role is. What we do know is that the dose of TZP appears too low, the need for prolonged or continuous infusions to improve tazobactam threshold exposures is necessary, and in the absence of likely both of these factors, the breakpoint is clearly too high. CT looks closer to appropriate, however limited data determining the adequacy of the dosing regimen currently exist and there are concerns near the CLSI breakpoint. Without robust PK/PD data to support a dosing regimen and then clinical validation of that dosing regimen, one would be foolish to routinely recommend one of these agents over the established carbapenems for invasive ESBL infections. Additionally, even if/when optimized, the ecological advantage of the tazobactam based combinations over a carbapenem needs to be thoroughly analyzed. If the driver of the desire to utilize these agents over carbapenems is to minimize the development of carbapenem-resistant organisms, surely the finding of increased infections due to carbapenem-resistant organisms after treatment with TZP compared to meropenem in MERINO should give us pause that our logic is possibly flawed. Speaking personally, until better data become available to further delineate the role of tazobactam based combinations for ESBL, both authors would respectfully ask you to treat us with a carbapenem if we present at your institution with an ESBL infection.

Declaration of interest

JM Pogue serves as a consultant for Merck & Co., Melinta, Achaogen, Shionogi, Tetraphase, and Nabriva. He only has received research grant funding from Merck & Co. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Additional information

Funding

This manuscript has not been funded.