The prevalence of infections caused by difficult-to-treat resistant Gram-negative bacteria (DTR-GNB) is rapidly increasing, and along with constantly evolving epidemiology, represents a major challenge to the management of hospital-acquired infections (HAIs).¹ In this regard, DTR describes treatment-limiting resistance to all first-line agents, that is, all β-lactams, including carbapenems and previous generation β-lactamase inhibitor combinations (BLICs), and fluoroquinolones.² Arguably, the greatest threat from DTR-GNB comes in the form of carbapenem-resistant Enterobacterales (CRE) such as carbapenemase-producing Klebsiella pneumoniae.³

There has been an unprecedented proliferation of carbapenemase encoding genes, the evolutionary epidemiology and resistome dynamics of which were recently highlighted in several local studies.^4,5,6 Widespread transmission is reported, and is mediated by clonal, multiclonal and horizontal mechanisms.^4,5,6 Enhanced surveillance for CRE causing blood-stream infections (BSIs) in South Africa, recently demonstrated a significant increase in the proportion of cases especially from Gauteng and the Western Cape.⁴

Moreover, the proportion of CRE isolates, specifically K. pneumoniae, phenotypically non-susceptible to class 2 carbapenems (imipenem, meropenem and doripenem) ranges from 50% to 53%, while non-susceptibility to ertapenem is 86%, reflecting production of diverse carbapenemases, with a predominance of OXA-48, and reduced permeability (porin reduction or alteration). The dire situation with regard to treatment options is reflected by the fact that up to 22% and 13% of CRE isolates are also non-susceptible to tigecycline and colistin, respectively, and the overall in-hospital mortality regardless of therapy is 38%.⁴

Available options for the treatment of HAIs caused by CR-Pseudomonas aeruginosa (P. aeruginosa) are also scarce, and recent reports emphasising the spread of colistin non-susceptibility in environments with high volumes of colistin or polymyxin usage, are a major concern.³ South Africa is no exception, with CR among P. aeruginosa isolated from bacteraemic patients in the public and private sectors, ranging from 30% to 40%, while concurrent colistin resistance has emerged and susceptibility to this antibiotic of last resort has decreased year on year.^7,8 Similarly, agents to which DTR-Acinetobacter baumannii are susceptible are very few in number and there is a paucity of new antibiotics in the pipeline for these pathogens.

A major challenge is how best to optimise care for infections caused by DTR-GNB in the context of diminishing antibiotic options that may be less effective and associated with higher toxicity. There is a need for appropriate and effective initial therapy while promoting antibiotic stewardship (ABS) principles, emphasising the continued use of existing antibiotics to which the bacteria remain susceptible.¹ In this regard, it is evident that new molecules and novel BLICs with in vitro activity against DTR-GNB, only partially address the currently prevalent mechanisms of resistance (Table 1).^9,10 As such, effective therapy is dependent on accurate and prompt identification of the organism, and on phenotypic and genotypic antimicrobial susceptibility and mechanisms of resistance testing, respectively.

Ceftazidime-avibactam (CA) and ceftolozane-tazobactam (CT) are the first of the new generation BLICs registered in South Africa. The potential for uncontrolled use of these agents as a ‘one-size fits all’ treatment for DTR-GNB will prove catastrophic for their future preservation as viable therapeutic options. Ceftolozane-tazobactam and CA were previously prescribed in South Africa as section 21 drugs since 2018, and utilisation data indicates a sharp increase in the compound annual growth rate, specifically a 383% increase of CA consumption, over the past three years (Data on file). The utilisation is almost exclusively driven by the private sector, where CA is also the dominant agent of the two, with an overall 100% market share. The global temporary recall of CT at the beginning of 2021 by the manufacturer, as well as other stock issues relating to this new BLIC are some of the reasons for exclusive CA consumption.

The objectives of the article’s recommended approach to the management of MDR- and DTR-GNB (i.e. mechanism-based inhibition therapy) are to:

The mechanisms by which GNB develop and express resistance, may interfere with several facets of ABS and impact on clinical treatment pathways, including the choice of an empiric antibiotic regimen, the potential for de-escalation, and the management of clinical failure because of the emergence of resistance on therapy.^11,12 Antibiotic resistance in GNB results from the expression of antibiotic-inactivating enzymes and non-enzymatic mechanisms, both of which may be intrinsically expressed by a given species (chromosomal genes) or acquired by a subset of strains as a consequence of two distinct, albeit not mutually exclusive, genetic events³:

In non-fermenting GNB such as P. aeruginosa, MDR and DTR may emerge following sequential chromosomal mutations, which may lead to the overproduction of intrinsic β-lactamases such as AmpC and hyperexpression of efflux pumps, target modifications and permeability alterations.^3,7 P. aeruginosa also has the ability to acquire mobile genetic elements encoding for resistance determinants, including carbapenemases such as VIM, which is particularly prevalent in SA.^7,13,14,15 The spontaneous mutation rate for expression of resistance may occur as frequently as 1 in 10⁶ to 1 in 10⁷ wild type strains. This process may be accelerated by the use of antibiotics with anti-pseudomonal activity, particularly if therapy is prolonged.¹⁶

With this in mind, ABS should preferably involve the application of certain fundamental tenets when selecting antibiotics for treatment of DTR-GNB, chief among which are the following:

The available evidence regarding the use of currently available BLICs, as well as an emerging body of data on novel non-β-lactam agents in the antibiotic development pipeline, require a paradigm shift with regard to how, we recommend the use of both currently available and new antibiotics.

Translating microbiological data relating to β-lactamase genealogy in South Africa into clinical practice, and enablement of this information at the bedside remain a formidable challenge. β-lactamases such as OXA-48 (and its variants) and NDM-1, either alone, or simultaneously present on plasmids, are the predominant carbapenemases among Enterobacterales with very few, or no K. pneumoniae carbapenemases (KPC), or other carbapenemases evident.^4,5,6 However, recent unpublished data suggest a steady increase in KPC in Gauteng province. In P. aeruginosa-BSIs in South Africa,⁷ metallo-β-lactamases (MBL) such as VIM and the Ambler class C AmpC enzymes predominate, and in addition, MDR efflux pumps are expressed in the majority of the non-susceptible isolates investigated.

Therefore, recommendations for antibiotic use should be strategic and should rely, in the case of new BLICs, on categorisation by ability to inhibit specific β-lactamases, when directed therapy is required (Figure 1). Hence, in order to minimise antibiotic selective pressure, a ‘Best Practice’ guide for existing and new antibiotic options in the South African context (Table 2, Table 3 and Table 4), which includes dose and administration strategies for critically ill patients, is proposed (Table 5).^17,18

Some β-lactamase inhibitors inhibit ESBLs. Although hyperproduction of β-lactamases or additional resistance mechanisms may hamper the activity of these compounds, the previous generation BLICs such as piperacillin-tazobactam, remain active in vitro against a considerable proportion of ESBL-producing GNB. However, the use of non-carbapenem β-lactams for the treatment of ESBL infections has yielded conflicting and controversial results.^19,20,21

Until the results of the MERINO trial (piperacillin-tazobactam versus meropenem for the treatment of BSIs caused by ceftriaxone-resistant Escherichia coli or K. pneumoniae) were published,^20,21 based on previous comparative studies and reviews that reported the potential role of BLICs in treatment of serious ESBL-infections, piperacillin-tazobactam appeared to be a reasonable option for:^22,23

In the MERINO trial, however, 30-day mortality was 12.3% in patients randomised to piperacillin-tazobactam versus 3.7% in those randomised to meropenem.^20,21 Nevertheless, the results of the MERINO study have been shown to have numerous biases. Henderson et al.²⁴ reviewed the association of piperacillin-tazobactam and meropenem MICs and β-lactam resistance genes with mortality in the MERINO trial. According to reference broth microdilution, almost 20% of isolates were incorrectly categorised as susceptible, highlighting issues with piperacillin-tazobactam susceptibility testing itself. A high combined prevalence of OXA-1 β-lactamase and ESBLs was also demonstrated. Differences in outcome may also have related to the varying piperacillin-tazobactam dosing and administration schedules used.^22,25

Therefore, piperacillin-tazobactam may be a carbapenem-sparing agent for BSIs as directed therapy, if the criteria are met, and dose is optimised and response to treatment is closely monitored (Table 5). Regarding the new BLICs, given its limited utility in South Africa because of the high prevalence of metallo-β lactamases, specifically VIM in P. aeruginosa, CT should be investigated as a carbapenem-sparing strategy for serious ESBL infections.

Regarding carbapenems, ertapenem as a class 1 carbapenem with no, or limited, activity against non-fermentative GNB including Pseudomonas, remains particularly suitable for ESBL-GNB infections.²⁶ This includes BSI, as recently highlighted in a multinational retrospective cohort study (INCREMENT) in which comparable outcomes, for empiric and targeted therapy of mono-microbial BSI because of ESBL-producing pathogens, were seen with ertapenem and the broader spectrum carbapenems.²⁷ The advantage of recommending ertapenem as a treatment option in this setting is that the drug does not appear to increase antibiotic resistance in P. aeruginosa.^28,29,30

For patients with BSI because of ESBL-producing Enterobacterales, with severe infection, imipenem or meropenem instead of ertapenem has been recommended.³¹ However, this remains contentious with recent studies not demonstrating differences in outcomes, even in patients with septic shock.³² The distinct pharmacokinetic differences between ertapenem and the other carbapenems may account for conflicting findings and, therefore, choice of carbapenem in critically ill patients should best be decided by multi-disciplinary teams or institutional guidelines.

AmpC resistance can be classified into three categories³³:

It is important to note, that infections that pose a management conundrum, relate to the emergence of clinically relevant AmpC expression during antibiotic treatment, which has most frequently been described for Enterobacter cloacae, Klebsiella aerogenes (formerly Enterobacter aerogenes) and Citrobacter freundii.^33,34 In this regard, few studies have provided reliable insights into effective management approaches for infection caused by such AmpC-producers. Nonetheless, it may be prudent to avoid third generation cephalosporins for the treatment of these organisms posing the greatest risk of AmpC induction. This has best been described in the context of especially, E. cloacae infections, even if susceptible in vitro.

In contrast, piperacillin-tazobactam and cefepime are potential directed therapeutic options in the treatment of susceptible AmpC producing bacteria. Both piperacillin and tazobactam are weak inducers of Amp-C β-lactamases.^33,34 Cefepime has the advantage of being a weak inducer while withstanding hydrolysis by AmpC β-lactamases because of the formation of a stable acyl enzyme complex. As such the available clinical evidence does not support the use of alternative agents over piperacillin-tazobactam or cefepime, nor that resistance should be inferred (reporting an agent as resistant on the antibiogram) based on the Amp-C phenotype if piperacillin-tazobactam or cefepime is susceptible in vitro.^34,35 Based on a systematic review and meta-analysis comparing the effects of different antibiotics on mortality in patients with BSIs caused by Enterobacterales-producing chromosomal AmpC β-lactamases, no strong evidence exists to suggest that piperacillin-tazobactam, cefepime or quinolones are inferior to carbapenems (meropenem or imipenem), provided that they are susceptible.³⁶ Notably, there are no comparative data evaluating outcomes of ertapenem treatment for infections with AmpC-producing Enterobacterales.

Piperacillin-tazobactam (at a MIC ≤ 8 mg/L) and cefepime (at a MIC ≤ 2 mg/L) offer a carbapenem-sparing opportunity as directed therapy of susceptible Amp-C organisms, particularly for those Enterobacterales that have a low risk for clinically significant over-expression of AmpC (< 5%) such as S. marcescens, M. morganii and Providencia spp. In E. cloacae and K. aerogenes infections, de-escalation to piperacillin-tazobactam or cefepime may be considered once the MIC is known.

Therefore, it is reasonable to base treatment decisions for AmpC producers on:

Regarding the new BLICs, clinical data regarding the efficacy of CT and CA in the treatment of infections caused by AmpC- producing isolates are limited.³⁷ Ceftolozane-tazobactam and CA have similar spectra of antibacterial activity, but with some important differences. The key microbiologic difference is that avibactam inhibits carbapenemases, particularly KPC and OXA-48, while tazobactam does not. Notably, although tazobactam does not inhibit Amp-C β-lactamases, recent data have demonstrated that CT is active in vitro against 99.7% and 94.7% of isolates with moderately and strongly up-regulated efflux mechanisms respectively, and also against 96.6% of bacteria with de-repressed AmpC, confirming earlier data that ceftolozane may overcome the two most prevalent mechanisms of resistance (up-regulated efflux and de-repressed AmpC) in P. aeruginosa.³⁸ While both CT and CA have been shown to be microbiologically and clinically effective in ESBL, AmpC and P. aeruginosa infections,^39,40,41 from a stewardship point of view, the specific efficacy of CT against MDR SPICE bacteria, upon failure of preferred agents, is important (Table 3).

In a global cohort of CR P. aeruginosa, including isolates from one centre in Cape Town, South Africa, CA was the most active agent with 72% susceptibility as per CLSI breakpoints compared with 63% for CT.¹⁴ In the cohort, 87% were genotypically positive for a carbapenemase with the most common being the MBLs; VIM most frequently followed by class A GES, which explains the relatively lower CT activity. In this regard, comparative CT antibiograms for CR-P. aeruginosa in South Africa should be performed, to ascertain geographic differences in non-susceptibility and inform potential use. In the study, 46% of isolates remained susceptible to both ceftazidime and cefepime.¹⁴ Notably, 90% and 86% of carbapenemase-negative CR isolates remained susceptible to CA and CT, respectively.¹⁴

Soon after registration in the United States of America, several studies reported the in vivo emergence of CT resistance in MDR-P. aeruginosa, associated with de novo mutations, rather than by acquisition of resistance from nosocomial isolates.^42,43 Further studies are, therefore, necessary to optimise and validate administration techniques designed to minimise emergence of resistance, such as infusion strategies derived from PK/PD data, and to verify the mechanism by which resistance occurs in vivo.^42,44

Overall, the data as presented seem to support the preferred stratified and strategic use of CT in the management of serious infections with AmpC producers in the absence of MBLs as directed therapy particularly for DTR-P. aeruginosa (Figure 1, Table 3).

The clinical activity of CA against MDR-Enterobacterales and P. aeruginosa isolates, pooled from the various adult Phase III clinical trials (complicated intra-abdominal infection [cIAI] [RECLAIM 1 and 2; RECLAIM 3], complicated urinary tract infection [cUTI] [RECAPTURE 1 and 2], hospital-acquired pneumonia [HAP], including ventilator-associated pneumonia [VAP] [REPROVE] and cIAI or cUTI caused by ceftazidime-non-susceptible GNB [REPRISE]) demonstrated, that CA has similar clinical efficacy to predominantly carbapenem comparators against MDR Enterobacterales and P. aeruginosa.⁴⁵

Ceftazidime-avibactam may improve outcomes among patients with CRE infections and appear to be superior to conventional therapies, including colistin, against BSI caused by KPC with clinical cure in the majority of DTR infections that had failed previous therapy.^{46,47,48,49,50} Qualitative evidence of successful use of CA for treatment of hospitalised patients with GNB infections with limited treatment options, based on clinical and microbiological outcomes and mortality, including evidence of effectiveness against CRE and MDR-P. aeruginosa, were recently reviewed.⁵¹

Ceftazidime-avibactam also has more favourable PK characteristics than colistin and has been shown to be better tolerated and several positive cost-effectiveness analyses have also been reported for BSIs, HAP and VAP.^52,53 Outcomes evaluated included quality-adjusted life-years, healthcare costs and incremental cost-effectiveness ratios. These studies found that CA was cost-effective for CRE-BSIs and -pneumonia based on accepted willingness to-pay-standards in both the United States (US) and Italy.^52,53 Similar conclusions were reached for low- to middle-income countries for CA versus colistin for CRE-BSIs and pneumonia in Peru.⁵⁴

Particularly relevant to South Africa, recent reports have focused on the role of CA in OXA-48-producing bacteria.^55,56 Avibactam is a diazabicyclooctane (DBO), which is a non β-lactam β-lactamase inhibitor that may have the ability to inhibit OXA-48 β-lactamases by forming a stable covalent complex. Although combination therapy is thought to improve the likelihood of clinical cure and survival in severe CPE infections, successful outcomes have been seen in approximately 70% of patients with infections caused by OXA-48-producing Enterobacterales treated with CA monotherapy.^55,56 A recent review confirmed no significant differences in mortality in the treatment of CRE infections with CA-combination therapy compared to CA-monotherapy.⁵⁷ Ceftazidime-avibactam thus shows promising results as monotherapy, and of the BLICs, is currently the preferred agent for the treatment of patients with severe infections because of OXA-48-producing Enterobacterales.⁵⁸

Several studies have demonstrated in vivo development of resistance following relatively short courses (10–19 days) of therapy for KPC-BSIs.^59,60 For CA, resistance mostly occurs in K. pneumoniae ST258 harbouring D179Y substitutions in the KPC enzyme, and proliferation of these KPC variants threatens its future use.^61,62,63 Notably, reports on the rise of resistance to CA are alarming, not only in K. pneumoniae, but recently among other common Enterobacterales species. Of major concern, in the South African private sector CA non-susceptibility in CR-K. pneumoniae ranges between 10% and 20% and increases to 38% in those isolates causing BSIs where concurrent OXA-48, NDM, or VIM carbapenemases are present (Data on file).

In line with the proposed conceptual approach to BLIC therapy (Figure 1) and ‘Best practices’ for use of existing and new antibiotic options (Table 4), CA is recommended as directed therapy for severe OXA-48 producing infections, on bacteriological or clinical failure of preferred agents.

The introduction of CT and CA into hospital formularies should be considered with great care, preferably with pre-planned stewardship-guided interventions emphasising their positioning as therapeutic options (Box 1). The focus on appropriate stewardship practices is vital to maximise the efficacy and longevity of all new agents that enter clinical practice. Preferably, no hospital should be allowed to implement the use of these drugs without the oversight of an infectious disease specialist, an experienced intensivist, microbiologist and pharmacists ABS teams. These new agents are generally not suitable for empiric use and as such is discouraged. Other clinical management challenges relate to uncertainty as to whether the new BLICs should be used as mono- or combination therapy.

The emergence of antibiotic resistance after short courses of therapy with CT and CA highlights the importance of establishing strict criteria for the use of these drugs and the continuing need for new antibiotics. Moreover, it substantiates the fact that these antibiotics at the current time, are the archetypal ‘antibiotics of last resort’. It further emphasises the ongoing challenges of treating infections caused by DTR-GNB species, and the substantial threat that resistance poses to other novel BLICs and future drug development. ‘An antibiotic steward knows how to use an antibiotic, a good antibiotic steward knows when to use an antibiotic, and a great antibiotic steward knows when not to use an antibiotic’ (Adapted from a quote regarding surgeons⁶⁵).