Identifying Effective Durations of Antibiotic Therapy for the Treatment of Carbapenem-resistant Enterobacterales Bloodstream Infections: A Multicenter Observational Study

Abstract

At least 3 clinical trials indicate that approximately 7 days of antibiotic therapy are sufficient for the treatment of gram-negative bloodstream infections (GN-BSI) [1–3]. It is unknown, however, if durations need to be extended for the treatment of GN-BSI caused by pathogens with more drug-resistant phenotypes—such as those exhibiting carbapenem resistance. Among the 1356 patients enrolled in 3 trials focusing on the optimal duration of therapy for GN-BSI, only 3 were infected with carbapenem-resistant Enterobacterales (CRE)—limiting the applicability of trial findings to patients with CRE BSI [1–3]. As limited antibiotic options are available for the treatment of CRE BSI, identifying the shortest yet most effective treatment duration for patients with CRE BSI is of paramount importance to further limit the emergence of resistance [4]. Our objective was to determine if relatively short durations of active antibiotic therapy (ie, 7–10 days) are as effective as prolonged durations of antibiotic therapy for CRE BSI.

METHODS

Study Cohort

This is a retrospective, observational study of unique consecutive hospitalized patients ≥18 years of age with monomicrobial CRE BSI admitted between 1 January 2019 and 31 December 2019 at any of 24 hospitals in the United States [5]. All 24 hospitals used an identical REDCap database with an accompanying data dictionary. Data were entered by infectious diseases trained physicians, pharmacists, or trainees closely supervised by infectious diseases specialists. Regular virtual meetings were held to review consistent data collection practices. Extensive data cleaning occurred to address outlier values, missing data, and other inconsistencies prior to analysis. Data collection occurred after IRB approval was granted, with waivers of informed consent.

Eligibility Criteria

Patients had to be infected with Enterobacterales isolates resistant to at least 1 carbapenem antibiotic to meet eligibility criteria, using Clinical and Laboratory Standards Institute interpretive criteria [6]. Bacteria generally exhibiting intrinsic resistance to imipenem (eg, Proteus species, Morganella species, Providencia species), required resistance to at least 1 carbapenem other than imipenem for study inclusion [6]. Patients meeting any of the following conditions were excluded: (1) BSI complicated by osteoarticular infection, endocarditis, endovascular infection, or central nervous system infection; (2) receipt of ≥72 hours of empiric antibiotic therapy without in vitro activity against the infecting organism; (3) receipt of agents other than novel beta-lactams (eg, ceftazidime/avibactam) for isolates resistant to anti-pseudomonal carbapenems (ie, meropenem or imipenem-cilastatin); (4) receipt of monotherapy with aminoglycosides, fluoroquinolones, polymyxins, or tetracycline-derivatives; (5) receipt of <7 days of antibiotic therapy; (6) receipt of ≥21 days of antibiotics as it was anticipated that these patients may have a complex foci of infection; or (7) patients for whom an antibiotic course was prematurely terminated due to withdrawal of care. Patients had to be alive at least 1 day after completing antibiotic therapy for inclusion.

Exposures and Outcomes

The primary exposure was days of active antibiotics, defined as calendar days in which the patient received an in vitro active antibiotic that is considered appropriate for the treatment of a CRE BSI [4]. Short course and prolonged course therapy were defined as 7–10 days and 14–21 days of active antibiotics, respectively. A dichotomous exposure was selected to increase clinical applicability. The primary outcome was a composite of all-cause mortality or a recurrent BSI with the same bacterial species, both within 30 days of completing antibiotic therapy.

Data Collection

Detailed information was collected on all patients including the following: demographics, pre-existing medical conditions, severity of illness, source of BSI and source control interventions, microbiological data, antibiotic treatment, and clinical outcomes. Adequate source control was defined as drainage of any potentially infected fluid collections or removal of all catheters and hardware at the time of antibiotic completion. Twenty-two of the 24 hospitals reported routinely testing CRE BSI for the presence of carbapenemase genes (Xpert® Carba-R [n = 6], the Verigene® Gram-Negative Blood Culture Nucleic Acid Test [n = 6], BioFire® FilmArray® Blood Culture Identification Panel [n = 5], or the ePlex® panel [n = 5]).

Statistical Analysis

Inverse probability of treatment weighting (IPTW) was performed on baseline variables to account for selection bias that may have impacted the duration of antibiotic treatment selected [7]. The following covariates were included in the generation of propensity scores: age >65 years, male sex, body mass index (BMI) ≥30, severe immunocompromise, intensive care unit (ICU) admission, Pitt bacteremia score ≥4, Charlson comorbidity index score ≥5, source of bacteremia, inadequate source control, a carbapenemase-producing isolate, mono-ertapenem resistance, and use of a novel beta-lactam.

Patients in the short-course group were weighted by the inverse of the propensity score; patients in the prolonged-course group were weighted by the inverse of 1 minus the propensity score. Individuals in each group were upweighted or downweighted to ensure similarity across baseline variables, except for duration of therapy. Standardized mean differences were calculated to ensure similarity between groups for each included variable. Baseline data were compared using the Fisher exact and Pearson χ² test for categorical data and the Wilcoxon rank sum test for continuous variables. In the IPTW analysis, odds ratios (ORs) and 95% confidence intervals (CIs) for 30-day mortality and recurrent BSI were calculated using weighted regression, adjusting for variables with a standardized mean difference >10%. Statistical analysis was completed using R version 4.2.0 (R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

In total, 183 patients met eligibility criteria; 66 (36%) received a short-course of active antibiotics (median: 9; interquartile range [IQR]: 8–10 days), and 117 (64%) received a prolonged-course of active antibiotics (median: 14; IQR: 14–15 days). The most common CRE were Klebsiella pneumoniae (96 [52%]) and Enterobacter cloacae complex (66 [36%]). Of the 183 isolates, 64 (35%) had a carbapenemase gene; all belonged to the bla_KPC family. Thirty-six isolates (20%) were ertapenem-resistant but retained susceptibility to either meropenem or imipenem-cilastatin. Common treatment regimens included ceftazidime-avibactam (75, 41%); meropenem-vaborbactam (19, 11%); and high-dose extended-infusion meropenem with or without an aminoglycoside, fluoroquinolone, or polymyxin (89, 49%). The addition of an aminoglycoside, fluoroquinolone, or polymyxin to a beta-lactam/beta-lactam inhibitor was administered to 15% of patients. The proportion of patients who received either ceftazidime-avibactam or meropenem-vaborbactam was equally distributed between the two groups (52% vs 51%), even before application of IPTW. No patients received imipenem-cilastatin-relebactam or cefiderocol. All patients with a bla_KPC identified received a beta-lactam/beta-lactamase inhibitor.

Baseline characteristics are displayed in Table 1. Short course therapy was more likely in patients ≥65 years (52 [79%] vs 71 [61%], P = .01) and less likely in patients with incomplete source control (5 [8%] vs 18 [15%], P = .038). The 2 groups were well balanced after IPTW.

In the IPTW cohort, 13.6% and 11.9% of patients in the short course and prolonged-course groups met criteria for the primary composite outcome (OR 1.21 [95% CI: .55–2.31]). There were no differences with regards to all-cause mortality (3.4% vs 4.6%) or recurrent bacteremia (6.1% vs 5.7%) within 30 days between the short-course and the prolonged-course groups.

DISCUSSION

We investigated the impact of short- versus prolonged courses of active antibiotic therapy on clinical outcomes of patients with CRE BSI in a cohort of 24 hospitals. We found no difference in 30-day mortality or recurrent BSI between patients receiving 7–10 days versus 14–21 days of antibiotic therapy in an IPTW analysis. To increase generalizability of our findings to populations most at risk of CRE BSI, we included patients regardless of underlying medical conditions (eg, solid organ transplant, neutropenia) or severity of illness.

It is often assumed that infections caused by organisms with more resistant phenotypes require more aggressive treatment, including longer durations of therapy, given these patients may have significant comorbidities and immunocompromise [8]. Clinical trials investigating the optimal duration of therapy for GN-BSI suggest 7 days are sufficient, however, < 1% of trial participants were infected with CRE [1–3]. Zhou and colleagues determined that durations of therapy <10 days were independently associated with increased 30-day mortality in a cohort of patients with CRE BSI [9]. This study included people who died while receiving antibiotic therapy, potentially resulting in shorter durations artificially appearing worse as a disproportionate number of people who die of gram-negative sepsis pass away within the first few days of culture collection [10, 11]. Additionally, several non-beta lactam treatment regimens were included (eg, colistin monotherapy).

Studies evaluating optimal durations of antibiotics for patients with highly drug-resistant infections present several challenges. First, patients often receive several days of ineffective antibiotic therapy as empiric regimens generally do not provide adequate coverage for CRE. To reduce the impact of this practice, we limited eligibility to patients receiving active antibiotics ≤72 hours and to those who received ≥7 consecutive days of active antibiotics, as determined by consensus from two infectious diseases specialists. Second, there is the possibility of immortal time bias. To circumvent this issue, we required that patients had to survive until at least 1 day after antibiotics were discontinued to be eligible for inclusion so that they had the “opportunity” to be evaluated for treatment failure. For example, if a patient received 7 days of therapy, (s)he had to be alive on day 8 to ensure that the duration of therapy was not truncated due to death. Third, there is a risk for confounding by indication where more severely ill or medically complex patients receive prolonged treatment durations than their healthier counterparts. IPTW was utilized to reduce confounding by indication but we acknowledge that this does not eliminate all selection bias that still may have impacted treatment durations selected.

Our findings suggest that 7–10 days of antibiotic therapy may be sufficient for patients with CRE BSI. As approximately 90% of patients in our cohort had adequate source control interventions, durations may need to be extended for patients with persistent sources of infection. Moreover, commonly employed empiric antibiotic regimens may not be active against CRE and the duration of 7–10 days represents days of active antibiotic therapy.

Notes

Financial support. This work is funded by the Prevention Epicenters Program of the Centers for Disease Control and Prevention (grant numbers 6 U54CK000617–01-02 and 5 U54CK000617-02-00).

References

Author notes