Is Piperacillin-Tazobactam an Appropriate Empirical Agent for Hospital-Acquired Sepsis and Community-Acquired Septic Shock of Unknown Origin in Australia?

Early appropriate empirical antibiotics are critical for reducing mortality in sepsis. For hospital-acquired sepsis of unknown origin in Australia, piperacillin-tazobactam (TZP) is recommended as an empirical therapy. Anecdotally, some institutions also use TZP for community-acquired septic shock. This narrative review aimed to scrutinise the appropriateness of TZP as an empirical agent for undifferentiated hospital-acquired sepsis and community-acquired septic shock. An online database (Medline) was searched for relevant studies in adults published in the last 10 years. Studies were included if they addressed separately reported clinical outcomes related to a relevant aspect of TZP therapy in sepsis. Of 290 search results, no studies directly addressed the study aim. This review therefore explores several themes that emerged from the contemporary literature, all of which must be considered to fully interrogate the appropriateness of TZP use in this context. This review reveals the paucity and low quality of evidence available for TZP use in sepsis of unclear origin, while demonstrating the urgent need and equipoise for an Australian audit of TZP use in patients with sepsis of unknown origin.


Introduction
Sepsis is a serious infectious syndrome caused by myriad pathogens and is a significant contributor to mortality, morbidity, and healthcare costs in Australia and worldwide. Sepsis is defined by The Third International Consensus Definitions for Sepsis and Septic Shock as 'life-threatening organ dysfunction caused by a dysregulated host response to infection' [1]. In clinical settings, sepsis can be rapidly identified in patients with suspected infection plus at least two criteria of the quick Sequential Organ Failure Assessment (qSOFA) tool: respiratory rate ≥22/min, altered mentation, or systolic blood pressure ≤100 mmHg [1]. Septic shock is diagnosed when vasopressors are required to maintain mean arterial pressure ≥65 mmHg and serum lactate is ≥2 mmol/L despite adequate volume resuscitation [1]. Every year, 11 million or almost one-fifth of all deaths worldwide are associated with sepsis [2]. Of those who survive, many live with permanent disability. In 2017, there were 104,912 hospitalisations with sepsis Australia-wide, of which 11% resulted in death in hospital, with an estimated cost per episode of AUD 39,300 [3,4]. In 2020, 69% of sepsis cases in Australia were due to gram-negative pathogens, with the most common being Escherichia coli [5]. Gram-positive pathogens, most commonly Staphylococcus aureus, accounted for the remaining 31% of sepsis cases [5].
Early and adequate empirical antibiotics are a key predictor of outcome in sepsis [6]. The choice of an appropriate empirical agent is based on a range of factors including hospital or community acquisition, septic focus, illness severity, allergy status, local antimicrobial resistance patterns, patient risk factors for antimicrobial resistance, and patient co-morbidities such as impaired renal function. If the source of infection is not immediately apparent, termed 'sepsis of unknown origin', empirical antibiotics must cover a wide range of possible pathogens. For hospital-acquired sepsis of unknown origin, the Australian Therapeutic Guidelines: Antibiotic Version 16 recommend empirical intravenous (IV) piperacillin-tazobactam (TZP) as first-line therapy for patients with no penicillin hypersensitivity [7]. Anecdotally, some institutions also prescribe TZP as empirical therapy for community-acquired septic shock, although this is not recommended by the Australian Therapeutic Guidelines.
TZP is a broad-spectrum IV antibiotic composed of piperacillin, a ureidopenicillin with extended spectrum activity, and tazobactam, a beta-lactamase inhibitor [8]. Piperacillin exerts bactericidal action by inhibiting penicillin-binding proteins, thereby disrupting bacterial cell wall cross-linkage. Tazobactam prevents hydrolysis of piperacillin by forming a stable acyl-enzyme complex with bacterial beta-lactamases [8]. In 2017, TZP was the seventh most used antibiotic in Australia, making up 4.6% of total antibiotic usage. This figure is probably an underestimate, as TZP use was restricted in 2017 owing to a national shortage [9]. As a first-line empirical agent, TZP has established activity against many grampositive, gram-negative, and anaerobic bacteria [8]. To address gaps in bacterial coverage, the Australian Therapeutic Guidelines have recommended adding IV vancomycin to cover methicillin-resistant S. aureus in septic shock or line-related sepsis and substituting TZP for IV meropenem when there are risk factors for multi-drug-resistant (MDR) gram-negative bacteria [7].
Emerging resistance mechanisms in gram-negative bacteria raise concerns regarding the continued appropriateness of TZP as a first-line broad spectrum antibiotic. In Australia in 2020, 14.7% of E. coli and 10.0% of Klebsiella pneumoniae bloodstream isolates had extended-spectrum beta-lactamase (ESBL) phenotype and 8.3% of bloodstream isolates were 'ESCAPPM' AmpC producers (Enterobacter, Serratia, Citrobacter freundii, Acetinobacter/Aeromonas, Providencia, Proteus vulgaris, and Morganella species) [5]. In 2020, 11.3% of Pseudomonas aeruginosa isolates tested in Australia were resistant to antibiotics of at least two classes [5]. ESBLs derive from mutations in chromosomal or plasmid-borne beta-lactamase genes and confer resistance to third generation cephalosporins and aztreonam [10]. Although carbapenems are the treatment of choice for ESBL-producing Enterobacteriaceae (ESBL-PE), the alarming emergence of carbapenem resistance has led to renewed consideration of carbapenem-sparing agents such as TZP [7,11]. AmpC betalactamases confer resistance to many antibiotics including TZP and are impervious to beta-lactamase inhibitors, including tazobactam [12]. AmpC beta-lactamases are chromosomally encoded in ESCAPPM bacteria [12]. The three primary resistance mechanisms for AmpC beta-lactamases are inducible chromosomal expression, stably derepressed chromosomal expression and plasmid-mediated resistance [12]. Even if initially beta-lactam susceptible, exposure to beta-lactams may temporarily lead to resistance in ESCAPPM species via inducible AmpC beta-lactamase production [12]. TZP is thought to be a weak inducer, suggesting it may be an appropriate agent for ESCAPPM and other organisms with inducible AmpC beta-lactamase [12].
Important considerations in evaluating TZP appropriateness are toxicity and dosing. Nephrotoxicity and neurotoxicity are potential adverse events reported in the literature, but studies in patients with defined sepsis are limited [13,14]. Appropriate dosing regimens aim to optimise clinical outcomes in patients with adverse risk factors including immunosuppression, infections with a high bacterial inoculum (e.g., deep-seated abscess or collection) or with a high minimum inhibitory concentration (MIC) to TZP. Dose optimization may also be appropriate for patients with augmented renal clearance, as well as critically ill and obese patients. In Australia, TZP is conventionally infused over 30 min and serum drug measurements are not routinely recommended [7].
The aim of this review is to address the appropriateness of TZP as an empirical agent for hospital-acquired sepsis and community-acquired septic shock of unknown origin in Australia. Because of a gap in the literature for efficacy of empirical TZP in sepsis of unknown origin in Australia, this review will focus more broadly on clinical outcomes of empirical or definitive TZP for sepsis of any source in any country. Moreover, this review will not separate clinical outcomes related to hospital versus community-acquired sepsis and sepsis versus septic shock, as these terms are often variably and poorly defined in the literature and are rarely analysed as separate entities. Therefore, this review summarises the contemporary literature surrounding clinical outcomes related to empirical or definitive TZP efficacy, toxicity, and dosing considerations in sepsis of any source. Subsequently, these findings are expected to establish the basis of a study addressing the original research question, relating to hospital-acquired sepsis and community-acquired septic shock of unknown origin in the Australian setting.

Materials and Methods
The online database Medline (Ovid) was searched on 13 August 2021 using MeSH terms and key words for 'sepsis' and 'TZP' and publication date from 2011 to 2021. Search history is listed in full in Appendix A.
Studies were excluded on title and abstract and full text screening for the following reasons: • Written in a language other than English; • Meta-analysis or study protocol; • Full text unavailable online; • The study exclusively concerned a paediatric population; • Outcomes relating to either TZP therapy or sepsis were not reported separately from other outcomes; • Study participants did not have known or suspected sepsis, as defined by positive blood culture, clinical diagnosis of sepsis, bacteraemia, or bloodstream infection; • The study did not address clinical outcomes related to a relevant aspect of TZP therapy; efficacy, resistance, toxicity, or dosing regimens.
Studies were included in the review if a major aim of the study was to investigate clinical outcomes of TZP in sepsis in adults. In order to specifically address the rapidly emerging contemporary evidence on the clinical efficacy of TZP in ESBL and AmpC beta lactamase-producing gram-negative pathogens, an additional search was performed using the key words 'piperacillin-tazobactam' and 'ESBL' or 'AmpC' on 26 April 2022.
The latest Antimicrobial Use and Resistance in Australia (AURA) and sepsis outcome program reports were retrieved from safetyandquality.gov.au (accessed on 14 August 2021) and included to assess the relevance of international studies in the Australian context. Additional studies referenced only in the introduction were retrieved from the reference lists of included articles.

Results
No studies identified from the database search directly addressed sepsis of unknown origin or the appropriateness of TZP monotherapy in sepsis of any source. One study addressed the appropriateness of TZP/vancomycin combination therapy in sepsis [15] but was excluded as the appropriateness of TZP alone was not reported separately. The following studies were included in this review: 24 addressing clinical efficacy of TZP (Table 1), 4 reporting toxicity (Table 2), and 5 exploring pharmacokinetic and dosing considerations (Table 3). Additional case reports are included in Tables S1 and S2.   Table 3. Studies reporting on clinical outcomes related to TZP pharmacokinetic and dosing considerations in sepsis. An additional case report [44] can be found in Table S2.

Discussion
This review addresses the appropriateness of TZP as an empirical agent for sepsis of unknown origin from three angles; clinical efficacy against known pathogens, reported toxicities in septic patients, and optimised dosing.

Efficacy against ESBL-Producing Enterobacteriaceae
In 2019, more than 1 in 10 Australian E. coli and K. pneumoniae blood isolates harboured ESBLs [5], but evidence is mixed regarding the utility of TZP as a carbapenem-sparing agent for these pathogens. Following the results of a recent randomised controlled trial (RCT), the Australian Therapeutic Guidelines recommended substituting TZP for meropenem where risk factors for ESBL producers are present [7]. However, several retrospective studies report conflicting results, with most supporting TZP as a useful alternative to carbapenems.
Harris et al. conducted a noninferiority RCT in nine countries, including Australia, that concluded that definitive TZP, compared with meropenem, was associated with increased 30-day mortality [32]. However, this conclusion may not apply to empirical TZP and septic shock, as only definitive therapy was investigated, and most study participants had low illness severity. Furthermore, the study was limited by lack of blinding and was underpowered to analyse subgroups where definitive TZP may have been noninferior to meropenem. A post-hoc analysis found that Harris et al. may have overestimated mortality in the TZP group by inclusion of isolates susceptible to TZP by automated testing methods that were found to be TZP non-susceptible post hoc by broth microdilution susceptibility testing [32,33]. When TZP non-susceptible strains were excluded, there was no longer any significant difference in mortality between TZP and meropenem treatment arms [33]. Isolates co-harbouring ESBL and narrow spectrum OXA-type beta lactamases were associated with increased TZP MIC and elevated mortality [33]. Therefore, rapid identification of beta lactamase type alongside susceptibility testing may guide the selection of appropriate definitive antibiotics. However, a larger study is needed to explore the association between mortality and other beta lactamase types in isolates with ESBL phenotype. In the absence of further data, meropenem should continue to be the first-line therapy for patients with sepsis with confirmed or suspected ESBL Enterobacteriaceae.
Out of eight retrospective observational studies directly comparing TZP and carbapenems in ESBL-PE sepsis, only two found an association between empiric TZP use and mortality [26,28]. One study in patients with haematological malignancy found that although empiric TZP was not linked to increased mortality, persistent bacteraemia and prolonged fever were more common than in the carbapenem group [29]. The two largest retrospective studies, with over 200 participants each, reported conflicting results. Tamma et al. concluded that the adjusted 14-day mortality risk was 1.92 times higher for patients receiving empiric TZP vs. carbapenem, while Ko et al. concluded no difference in 30-day mortality [28,35]. However, it is difficult to compare the two studies, as they differ in several important aspects: geographical location, sample size of the TZP group, and infection focus. Moreover, unlike Ko et al., Tamma et al. was a single-centre study, and did not exclude patients receiving the study antibiotic for less than 48 h, possibly obscuring the effect of the empirical therapy. Note also that the lower bound of the 95% confidence interval for the finding of 14-day mortality risk was 1.07, which although statistically significant, may be clinically irrelevant [28].
In two retrospective studies, lower TZP MIC was associated with reduced mortality in TZP-treated ESBL-PE sepsis than higher MIC, even within the Clinical and Laboratory Standards Institute (CLSI) susceptible range of ≤16 mg/L [30,37]. Most dramatically, Tsai et al. concluded that 30-day mortality was significantly lower for highly susceptible isolates with TZP MIC ≤ 0.5 mg/L compared to still susceptible isolates with MIC ≥ 1 mg/L, although these findings only apply to ESBL-producing Proteus mirabilis [37]. This supports the assertion that TZP should be used cautiously in suspected ESBL-PE sepsis. In contrast, another retrospective study observed no difference in 30-day mortality between patients with low (≤4 mg/L) and higher (8-16 mg/L) MIC in both ESBL and non-ESBL-PE sepsis [22]. How-ever, all three studies had small overall or subgroup sample sizes, limiting the ability to adjust for confounding variables and perform meaningful MIC subgroup analyses.
Apart from the general limitations of retrospective studies discussed later, most studies discussed so far have only included TZP-susceptible infections. Therefore, the available evidence should be applied cautiously to routine clinical practice, where empirical therapy is implemented without knowing the susceptibility profile of the causative organism.

Efficacy against AmpC Beta-Lactamase Producers
A pilot RCT comparing meropenem and TZP for definitive treatment of bloodstream infections caused by ESCAPPM species known to possess inducible chromosomal AmpC beta-lactamases demonstrated no difference in clinical outcomes [21]. Five AmpC genotypes were identified with CMY being the most prevalent, although the degree of AmpC induction was not measured, and the sample size was too small for genotype subgroup analysis of TZP efficacy [21]. Four retrospective observational studies have found TZP to be generally as effective as cefepime or carbapenems for treating ESCAPPM bloodstream infections [16][17][18][19]. One retrospective cohort study did observe lower rates of early treatment response for patients treated with empiric TZP compared with carbapenems, but the daily dose of TZP given was relatively low (13.5 g) [20]. In all five retrospective studies, no distinction was made between chromosomal inducible, derepressed and plasmid mediated AmpC production. Notably, only one study performed AmpC genotyping and found that AmpC was detected in >95% of Enterobacter and Serratia species but in only 19% of Citrobacter species [17]. The remaining three studies either performed no AmpC screening or low specificity phenotypic screening, with cefoxitin-resistant isolates assumed to be AmpC producers [16,18,19]. However, McKamey et al. did find lower rates of clinical and microbiological resolution with TZP than cefepime only for isolates with both cefoxitin and third generation cephalosporin resistance, which the authors conjecture may indicate derepressed or plasmid-mediated AmpC production [18]. Given the heterogeneity in AmpC expression among ESCAPPM species, the relationship between TZP efficacy and AmpC type and level of expression is unclear and requires further research.

Efficacy against Pseudomonas aeruginosa
TZP appears to be a generally effective agent for sepsis caused by TZP-susceptible P. aeruginosa, although this is based on low-quality evidence. Only 5.5% of P. aeruginosa blood isolates were TZP-resistant in Australia in 2020 [5], but there are concerns that infections with MICs below the CLSI breakpoint may have poorer outcomes [50]. Reassuringly, in a case-control study of patients receiving empirical TZP, there was no difference in 30-day mortality between MICs in the CLSI low (≤16 mg/L) and intermediate (32-64 mg/L) ranges [23]. Moreover, in P. aeruginosa sepsis susceptible to TZP by disc testing, there was no significant difference in 30-day mortality or clinical/microbiological response between definitive TZP and other anti-pseudomonal agents, including carbapenems and ceftazidime [24]. As with many pathogens, TZP-resistance may develop within a septic episode, rendering initially successful empirical therapy ineffective. In one case report of P. aeruginosa sepsis associated with liver abscesses, after initial resolution of fever with empirical TZP, the patient developed a new liver abscess harbouring TZP-resistant P. aeruginosa [51].

Efficacy against Less Common Pathogens
Several case reports suggest that TZP is an appropriate antibiotic for less common pathogens involved in anaerobic and aerobic gram-negative sepsis, although this finding must be balanced against the known limitations of case reports including publication bias [52][53][54][55][56]. As an exception, TZP should be used cautiously if sepsis due to the anaerobic gram-positive bacillus Eggerthella lenta is suspected. A retrospective cohort study of E. lenta sepsis found high MICs to TZP and an independent association with 30-day mortality and Intensive Care Unit (ICU) stay in 43 patients treated with empirical TZP [38]. However, these findings are confounded by 40% of the bloodstream infections being polymicrobial. Attempts to adjust for this and other confounders by logistic regression were restricted by the small sample size. Additional case reports of less common pathogens identified in this study are available in Supplementary Table S1.

TZP-Related Toxicity in Sepsis
In judging the appropriateness of any empirical antibiotic, it is important to balance toxicity against benefit. There are few published reports of TZP-related toxicity in septic patients when used as monotherapy, although this review only included reports in patients with clinically or microbiologically diagnosed sepsis. TZP appears appropriate for patients at risk of renal impairment, as a large retrospective cohort study of patients with gramnegative bacteraemia found no association of TZP receipt or duration of therapy with nephrotoxicity [43]. Patients with baseline serum creatinine of >310 µmol/L were excluded, limiting the application of these findings in patients with severe pre-existing chronic kidney disease.
Neurotoxicity at high concentrations of penicillins, including piperacillin, is a known adverse effect and has been described in case reports [13], but the evidence for neurotoxic effects in patients with known sepsis is very limited. In a single-centre retrospective study of ICU patients with severe sepsis, increasing TZP trough concentrations were weakly correlated with worsening neurological status based on changes in the Glasgow Coma Scale [42]. As there are many potential causes of worsening neurological status in critically unwell patients, better controlled studies are needed to explore this correlation, preferably using more nuanced neurological examination to measure neurotoxicity.
TZP-induced thrombocytopaenia was reported in two case studies [40,41] and is an uncommon differential diagnosis to be considered in a septic patient who develops thrombocytopaenia following initiation of empirical TZP.

Therapeutic Drug Monitoring (TDM) and Dosing Considerations for TZP in Septic Patients
TZP may only be an appropriate empirical antibiotic for sepsis of unknown origin if dosed to maintain optimal therapeutic concentrations for sepsis due to organisms with high MIC. In P. aeruginosa bacteraemia, maintaining ≥60% of the dose interval above the MIC is a significant predictor of in-hospital survival [46]. Despite concerns that augmented renal clearance in sepsis would reduce the efficacy of TZP as a renally cleared drug, a prospective observational study found no difference in rate of therapeutic failure for patients with creatinine clearance of >170 mL/min [45]. However, in both these studies, steady state TZP concentrations were derived from simulated pharmacokinetic models, either based on patient characteristics such as renal function [46] or serum samples with high intra-and inter-patient variability [45]. It is therefore unclear how dose adjustments could be made based on these studies, and TDM may be required to tailor the most effective dose for each septic patient.
Extended infusion regimens, where each dose is generally infused over 4 hours instead of 30 min, may improve TZP serum concentrations without the need for individualised TDM. An RCT in a Hong Kong ICU demonstrated significantly reduced 14-day mortality in bacteraemic patients receiving extended vs. standard intermittent TZP infusion [48]. Conversely, two larger retrospective cohort studies found no overall difference in mortality rates between extended and standard infusion regimens [47,49]. Interestingly, however, Cutro et al. found that mortality rates were lower with extended infusions for patients with urinary or intra-abdominal septic foci [47]. This suggests that when septic focus is unknown extended infusion regimens may be preferred. Note that comparisons are hindered by variation in TZP doses between studies, though the doses used are generally similar to the Australian Therapeutic Guideline recommendation of 4 + 0.5 g 6-hourly [7].

Limitations of This Review
This review demonstrates the lack of studies directly addressing the research question. Therefore, the scope of the review was broadened to identify general themes that might be used to help frame the original research question. Nonetheless, some caveats apply to the proposed future use of these review findings in addressing the research question. Sepsis as an inclusion criterion was defined broadly as a clinical diagnosis of known or suspected sepsis, bacteraemia, or bloodstream infection or positive blood culture. Although a broad definition was needed to encompass all relevant research, patients receiving empirical TZP for sepsis of unknown origin will fulfil the clinical but not necessarily the microbiological definitions of sepsis used. Confining search results to patients with known infectious syndromes also limits knowledge of the effects of empirical TZP, particularly toxicity, on patients who have non-infectious causes for a sepsis-like presentation.
Allowing for the lack of studies concerning sepsis of unknown origin, this review focused more broadly on sepsis of any source. A future review could be used to analyse only studies with a septic focus classified as 'unknown' or 'primary'. However, many studies do not stratify the analyses of TZP efficacy by infection focus, and typically, the sample sizes of patients with an unknown infection focus are small. Furthermore, such a limited analysis does not allow for recognition that a septic focus is often identified after commencement of empirical antibiotics.
This review only included studies with a major focus on clinical outcomes related to TZP therapy and may thus exclude studies with a non-clinical focus or where clinical outcomes are a minor aspect but still contribute valuable information, for example, in vitro susceptibility data for important sepsis-causing pathogens and pharmacokinetics in certain patient populations. As this review included studies investigating either empirical or definitive TZP in sepsis, the interpretation of its findings is limited when TZP is used in the critical early period before the causative organism and its antibiotic susceptibilities are known.
Apart from four RCTs and one prospective observational study, all studies discussed in this review are retrospective and observational. Unlike RCTs, retrospective observational studies are only designed to determine association, not causation, and are limited by the presence of uncontrolled confounding variables that could influence clinical outcome independently of variables of interest such as TZP treatment. Most retrospective studies reviewed here attempted to adjust for known confounders with multivariate analysis and case-control matching. However, both these techniques do not adjust for undocumented confounders and multivariate analysis has limited power with small sample sizes. Moreover, retrospective studies are more vulnerable to some biases than RCTs, such as selection, performance, and measurement biases.
The contemporary literature centres on the role of TZP in MDR sepsis compared to broader spectrum agents such as carbapenems. This focus not only biases studies towards TZP therapeutic failure, but it also fails to recognise the potential misuse of TZP in non-MDR sepsis, where, in Australia, a narrower spectrum of antimicrobials may be more appropriate. Further research is required to determine the appropriateness of TZP compared to narrower spectrum options in Australia, where sepsis of unknown origin may be caused by non-MDR organisms.

Conclusions
In conclusion, the appropriateness of TZP as an empirical agent for sepsis of unknown origin in Australia is unclear. Further research is required, particularly where antimicrobial stewardship efforts could substitute TZP for narrower spectrum agents. Despite some limitations, the highest quality evidence rejects TZP as a carbapenem-sparing agent in gram-negative ESBL producers. Limited evidence suggests that TZP may be an acceptable treatment for AmpC producers, but the impact of AmpC expression levels and genotype on TZP appropriateness is so far unknown and requires further investigation.
Only low-quality studies have investigated the appropriateness of TZP in susceptible P. aeruginosa and anaerobic sepsis. High TZP serum concentrations have been weakly linked to neurotoxicity in ICU patients, but better controlled studies are needed to determine the clinical impact of this link. So far, there are insufficient data to support the use of extended infusion regimens of TZP in sepsis, as is reflected in the Australian guidelines. Due to the difficulties of conducting large RCTs of empirical antibiotics, the majority of studies included in this review were retrospective with small sample sizes. Although the emerging evidence is compelling, this review highlights an urgent need to prospectively evaluate the appropriateness of antibiotic interventions in the Australian setting; until then, there remains insufficient equipoise to recommend a change to current clinical practice.