Cefoxitin versus carbapenems as definitive treatment for extended-spectrum β-lactamase-producing Klebsiella pneumoniae bacteremia in intensive care unit: a propensity-matched retrospective analysis

Since the 1990s, the emergence and diffusion of extended-spectrum beta-lactamase-producing Enterobacterales (ESBL-E) has been a global concern [1]. In 2021, the rate of invasive ESBL-E isolates has reached 7.5% in French healthcare settings [2]. ESBL-E infections treatment is becoming challenging due to high frequency of co-resistance mechanisms to other antibiotics classes, such as fluoroquinolones and aminoglycosides; therefore, therapeutic options are often very limited. Carbapenems have long been considered to be the reference treatment for infections caused by ESBL-E [3]. Unfortunately, their large prescription constitutes a strong selective pressure, which has led to the emergence and rapid worldwide spread of carbapenemase-producing Enterobacterales [4].

One possible alternative to carbapenems could be cefoxitin, a cephamycin developed in the 1970s, resistant to ESBL hydrolysis and highly active against ESBL-E in vitro [5, 6]. Currently, cefoxitin is the standard treatment for perioperative surgical prophylaxis, and is proposed as second-line antibiotic therapy for acute ESBL-Escherichia coli (ESBL-EC) pyelonephritis in adults [7]. Indeed, cefoxitin and other cephamycins have been shown to be effective in several small cases series of mild infections, such as urinary tract infections and bacteremia without signs of severity [8‐11]. In a multicenter retrospective study, Senard et al. showed no significant difference in clinical and microbiological success between cefoxitin and carbapenems as the definitive treatment of ESBL-EC urinary tract infections in men [12].

ESBL-Klebsiella pneumoniae (ESBL-KP) infections are also common nosocomial infection, but clinical data on their treatment with cefoxitin are scarce [13‐15]. In Guadeloupe University Hospital, the observed incidence rate of nosocomial ESBL-E infections is one of the highest in France, with a large predominance of Klebsiella pneumoniae [16, 17]. Since 2015, to limit the use of carbapenems, antibiotic guidelines in our intensive care unit (ICU) proposed cefoxitin rather than carbapenems as definitive treatment in cefoxitin-susceptible ESBL-E systemic infections, whenever clinical condition and microbiological findings are compatible. Until now this therapeutic strategy had not been formally evaluated. The aim of our study was thus to compare the clinical efficacy of cefoxitin as definitive antibiotic therapy for ICU patients with ESBL-KP bacteremia, versus standard carbapenem therapy. Selection of cefoxitin- and carbapenem-resistant Enterobacterales in both groups was also analyzed.

To the best of our knowledge, this original study is the first to compare the efficacy of cefoxitin versus carbapenems as definitive treatment in ESBL-Klebsiella pneumoniae bacteremia in ICU. The study’s main finding, confirmed by PS-matched analysis, shows that there was no statistical difference in 30-day clinical success when cefoxitin was used as definitive antibiotic therapy for ESBL-KP bacteremia, compared with carbapenems. The second interesting finding is that the use of cefoxitin instead of carbapenems reduced the selection of carbapenem-resistant Pseudomonas aeruginosa in these critically ill patients, often at risk of new nosocomial infection.

In our work, cefoxitin appears to be effective in severe infections compared with carbapenems. As expected in ICU, the overall clinical success rate was low (57% in the cefoxitin group), but not lower than in the carbapenems group (53%) and slightly higher than results recently published in a study of ESBL-E infections (including 25 Klebsiella pneumoniae isolates) treated with cefoxitin in critical care which reported a therapeutic success rate of 37% [25]. Although patient severity appeared similar in both studies, our recurrence rate seemed lower (11% vs. 27%) as was the rate of antibiotic change (17% vs. 32%). This may be due to the higher number of pneumonias in their study, which was an independent risk factor for clinical failure in ours. However, 30-day mortality does not seem to be different in both studies (29%). Similar to our findings, the mortality rate between patients treated with flomoxef or cefmetazole, another cephamycins and those treated with carbapenem for ESBL-E bacteremia was not significantly different in previous studies [10, 13]. In contrast to our results, two studies comparing flomoxef with carbapenems showed excess mortality in patients treated with cephamycin [26, 27]. The first, conducted in a hemodialysis unit, focused on ESBL-KP bacteremia related to fistula, graft or catheter infections [26], while the second involved a large sample (n = 389) of ESBL-EC and ESBL-KP bloodstream infections of various origins and from any hospital ward [27]. The Infectious Diseases Society of America (IDSA) thus does not suggest the use of cephamycins for the treatment of ESBL-E infections until more clinical outcomes data are available [28].

In our study, unlike cefoxitin, no change of treatment was made in the carbapenem group. First of all, carbapenems are the reference, broad-spectrum, last-line treatment. In the absence of rapid clinical improvement in a severe patient or when the source of infection is not controlled, cefoxitin might be changed early as it is not the reference treatment, while carbapenem should only be changed in the event of proven microbiological failure. Another reason is that cefoxitin was systematically switched when co-infections with cefoxitin-resistant bacteria occurred, whereas carbapenem was maintained when there was co-infection with carbapenem-resistant bacteria, and another antibiotic was added.

Many studies of bloodstream infections have shown that the source of infection is a major prognostic factor in clinical success and mortality [29‐31]. Interestingly, 100% of patients with catheter-related bacteremia for whom thrombophlebitis was ruled out (n = 20) were cured on cefoxitin. Given the total absence of studies on this specific population in the literature, it would be interesting to carry out further studies to confirm these promising results. In contrast, it is difficult to conclude on the efficacy of cefoxitin in bacteremia secondary to pneumonia and intra-abdominal infections, in view of the extreme severity of our patients in the study.

Cefoxitin as definitive treatment for ESBL-KP bacteremia remains controversial. Indeed, this is explained by the description of clinical failures caused by the in vivo acquisition of cefoxitin-resistant mutant K. pneumoniae [15]. In 1989, the first treatment failure case of nosocomial pneumonia with cefoxitin (after initial treatment with imipenem-cilastatin) due to the selection of cefoxitin-resistant K. pneumoniae was reported [14]. Since then, the mechanism of resistance has been identified as the loss of an outer membrane porin called OmpK36, which provides a channel that allows a wide range of antibiotics to penetrate the Klebsiella pneumoniae cell wall [32, 33]. In our study, there was no greater selection of cefoxitin-resistant K. pneumoniae in patients treated with cefoxitin than with carbapenems within 30 days following the beginning of the treatment. Interestingly, a slower growth of the bacteria was observed when they lost the OmpK36 porin probably due to the fitness cost, a decrease in the virulence of the strain and an increased susceptibility to neutrophil phagocytosis [34].

There are some limitations with our study. First, this was a non-randomized retrospective observational study with the inherent shortcomings of these studies. At baseline, our two treatment groups were not comparable on several criteria, with more severe patients overall in the carbapenem group, probably due to an indication bias. Nevertheless, the variables with potential impact on outcome were searched and a PS-based matched analysis was performed to control for potential bias on antibiotic selection. It should be emphasized that considering SMD values, some differences existed between the two groups after PS matching but with a SMD value < 0.5. Second, although our study contains, to our knowledge, the largest published number of patients with ESBL-KP bacteremia treated with cefoxitin in ICU, the sample size was still small. Third, although there were no significant differences in primary and secondary endpoints, using a PS-based matched analysis, we acknowledge that the 95% confidence intervals for propensity-matched outcomes are very wide and include the possibility of harm. Fourth, cefoxitin was mainly administered on a discontinuous basis, whereas several recent studies confirm that continuous administration of large doses of cefoxitin appeared necessary to achieve the recommended beta-lactam PK/PD target in critically ill patients [6, 25, 35].