Global prevalence of cefiderocol non-susceptibility in Enterobacterales, Pseudomonas aeruginosa, Acinetobacter baumannii and Stenotrophomonas maltophilia: a systematic review and meta-analysis

Our latest research was just published in the prestigious journal Clinical Microbiology and Infection (European Society of Clinical Microbiology and Infectious Diseases), September 02, 2023.

Global prevalence of cefiderocol non-susceptibility in Enterobacterales, Pseudomonas aeruginosa, Acinetobacter baumannii and Stenotrophomonas maltophilia: a systematic review and meta-analysis

by

Stamatis Karakonstantis, Maria Rousaki, Loukia Vassilopoulou, and Evangelos I. Kritsotakis

Access the full paper via:

• Journal's website
• ScienceDirect
• PubMed
• White Rose repository
• Elsevier Share Link – Free access until November 18, 2023.

Abstract

Background

Cefiderocol is a last resort option for carbapenem-resistant (CR) Gram-negative bacteria, especially metallo-β-lactamase (MBL)-producing Pseudomonas aeruginosa and CR Acinetobacter baumannii. Monitoring global levels of cefiderocol non-susceptibility (CFDC-NS) is important.

Objectives

To systematically collate and examine studies investigating in-vitro CFDC-NS and estimate the global prevalence of CFDC-NS against major Gram-negative pathogens.

Data sources

PubMed and Scopus, up to May 2023.

Study eligibility criteria

Eligible were studies reporting CFDC-NS in Enterobacterales, P. aeruginosa, A. baumannii, or Stenotrophomonas maltophilia clinical isolates.

Methods

Two independent reviewers extracted study data and assessed risk of bias on the population, setting and measurement (susceptibility testing) domains. Binomial-Normal mixed-effects models were applied to estimate CFDC-NS prevalence by species, co-resistance phenotype and breakpoint definition (EUCAST, CLSI, FDA). Sources of heterogeneity were investigated by subgroup and meta-regression analyses.

Results

In all, 78 studies reporting 82,035 clinical isolates were analysed (87% published between 2020 and 2023). CFDC-NS prevalence (EUCAST breakpoints) was low overall, but varied by species [S. maltophilia 0.4% (95%CI 0.2-0.7%), Enterobacterales 3.0% (95%CI 1.5-6.0%), P. aeruginosa 1.4% (95%CI 0.5-4.0%)] and was highest for A. baumannii (8.8%, 95%CI 4.9-15.2%). CFDC-NS was much higher in CR Enterobacterales (12.4%, 95%CI 7.3-20.0%) and CR A. baumannii (13.2%, 95%CI 7.8-21.5%), but relatively low for CR P. aeruginosa (3.5%, 95%CI 1.6-7.8%). CFDC-NS was exceedingly high in NDM-producing Enterobacterales (38.8%, 95%CI 22.6-58.0%), NDM-producing A. baumannii (44.7%, 95%CI 34.5-55.4%), and ceftazidime/avibactam-resistant Enterobacterales (36.6%, 95%CI 22.7-53.1%). CFDC-NS varied considerably with breakpoint definition, predominantly among CR bacteria. Additional sources of heterogeneity were single-centre investigations and geographical regions.

Conclusions

CFDC-NS prevalence is low overall, but alarmingly high for specific CR phenotypes circulating in some institutions or regions. Continuous surveillance and updating of global CFDC-NS estimates are imperative while cefiderocol is increasingly introduced into clinical practice. The need to harmonize EUCAST and CLSI breakpoints was evident.

Key words

carbapenem-resistant, Cefiderocol, drug resistance, global epidemiology, gram-negative bacteria, prevalence.

Meta-analysis methods

The primary outcome for meta-analysis was CFDC-NS prevalence, stratified by combinations of microorganism species, the definition of breakpoints, and underlying co-resistance phenotype. Population-averaged proportions were estimated using a random intercept logistic regression model with maximum likelihood estimation [20]. The model assumed a Binomial distribution for the observed number of CFDC-NS isolates in each study and a Normal distribution for the random population-level effects after the logit transformation. This approach correctly handles proportions near 0% or 100% (when variance tends towards zero and classic inverse variance weights are problematic) [21], and maintains the confidence limits of pooled proportions within the zero to one range. The resulting CI contains highly probable values for the population-averaged (pooled) prevalence proportion of CFDC-NS.

Higgin–Thompson's I-squared statistic was used as a summary index of the amount of variability of CFDC-NS proportions across studies that cannot be attributed to sampling error. Because I-squared is usually high and may not be discriminative for prevalence data [22], we additionally reported between-study variance (tau-squared) with a respective 95% prediction interval (PI). The PI describes the range CFDC-NS proportions that can be expected in new studies or settings [23]. We used summary forest plots to present pooled estimates with heterogeneity statistics for each microorganism and each different combination of breakpoint threshold with co-resistance phenotype. In addition, we constructed forest plots for the study-specific data used in each meta-analysis to illustrate the distributions of CFDC-NS proportions across the studies along with 95% CIs calculated by Wilson's score method (presented in the Supplement).

To examine potential sources of variation in CFDC-NS prevalence, we conducted a series of univariate meta-regressions with the Binomial-Normal mixed-effects model. We applied this to CR Enterobacterales, P. aeruginosa, A. baumannii, and K. pneumoniae based on the CLSI breakpoints, as this was the most frequent combination of breakpoints with co-resistance phenotype with a minimum of 15 studies available for each meta-regression. ORs with respective 95% CΙ were calculated to summarize the strength and direction of associations between study-level covariates and CFDC-NS prevalence. A covariate-specific R-squared statistic was calculated as the portion of between-study variance that was reduced after the inclusion of that covariate in a null model. Moreover, for each covariate level, we calculated pooled estimates of CFDC-NS prevalence based on subgroup analysis. Candidate covariates were decided a priori in our study protocol. The following study-level variables were examined: year when data collection ended, geographical region (WHO classification), multinational setting (yes/no), multicentre setting (yes/no), and risk of bias (classified as either low or moderate-to-high). Despite our protocol plans, we were unable to examine substantive characteristics related to clinical settings (such as type and level of care, age of the patients, clinical specialty, and type of infection), which were largely unreported in the studies. The number of retrieved studies was inadequate to justify multivariable meta-regression.

All analyses were carried out in STATA (Version 17; Statcorp, College Station, TX, USA).

Figure. Summary forest plot of cefiderocol non-susceptibility against Enterobacterales

S is the number of independent sets of data in the analysis. n / N is the ratio of the cumulative number of isolates that were non-susceptible to cefiderocol (CFDC-NS) over the total number of isolates, according to the respective definition of breakpoints and resistance phenotype. The centre of the diamond is the population-averaged (pooled estimate) prevalence of CFDC-NS isolates. The length of the diamond indicates the 95% confidence interval (CI) for the pooled estimate. The horizontal thick lines extending from the diamond represent the 95% prediction interval (PI) for the prevalence of CFDCN-S isolates in new studies. 

CAR, carbapenem; MBL, metallo-β-lactamase; NDM, New Delhi metallo-β-lactamase; CZA, ceftazidime/avibactam; CTA, ceftolozane/tazobactam