A multicenter retrospective cohort study was performed at three tertiary care transplant centers in the United States: The Hospital of the University of Pennsylvania (HUP) (776 beds), The Johns Hopkins Hospital (JHH) (1154 beds), and the University of Maryland Medical Center (UMMC) (767 beds).

The initial source population included all SOT recipients with an Enterobacterales (EB) BSI identified at HUP or UMMC between January 1, 2007, and June 30, 2018, and at JHH between January 1, 2005, and December 31, 2015 (with the varying time periods being due to differences in data availability). Because the microbiology laboratories at each institution process both inpatient and outpatient cultures, the cohort included any SOT recipient with an EB BSI regardless of the geographic location at which the BSI was identified, although all subjects were ultimately hospitalized. For those SOT recipients with multiple EB BSI during the study period, only the first episode was included.

For the determination of the impact of CRE BSI on clinical outcomes, the cohort was divided into exposed and unexposed subjects: Exposed subjects were those with a CRE BSI. CRE BSI was defined by any EB on blood culture that exhibited in vitro nonsusceptibility to any carbapenem, that is, minimum inhibitory concentration [MIC] ≥2 μg/mL for ertapenem or MIC ≥4 μg/mL for meropenem, doripenem, or imipenem before 2011, or MIC ≥1 μg/mL for ertapenem or MIC ≥2 μg/mL for meropenem, doripenem, or imipenem after 2011 per Clinical and Laboratory Standards Institute guidelines.²² EB isolates were not routinely screened for carbapenemase production. Unexposed subjects were those with a non-CRE EB BSI, defined as any EB on blood culture that exhibited in vitro susceptibility to all carbapenems. This control population was selected to focus on the impact of carbapenem resistance, rather than EB BSI, on outcomes.

For the determination of risk factors for CRE BSI, a case-control study design was employed. Case subjects were those with a CRE BSI, and control subjects were those with a non-CRE EB BSI (as defined above). All cases and controls were included in the analysis; no matching was performed.

This study was approved by the Institutional Review Board at each of the participating transplant centers (see Text S1, SDC, http://links.lww.com/TP/C498).

There were two primary outcomes: (1) all-cause mortality within 60 d after the EB BSI, and (2) new-onset graft failure within 60 d after the EB BSI. Graft failure was defined by retransplantation of the same organ, relisting for transplantation of the same organ, return to dialysis for kidney transplant recipients,²³ or a cardiac index ≤2 or ejection fraction ≤40% with need for inotropes or persistent mechanical support in heart transplant recipients.²⁴ For the analysis of new-onset graft failure, those SOT recipients with graft failure before the EB BSI onset were excluded from the cohort. We chose not to restrict graft failure or death to those that were thought to be directly attributable to the EB BSI episode because the relatedness of graft failure or death to a single infectious event can be difficult to ascertain retrospectively and can lead to misclassification of outcomes. Subjects were followed for 60 d because this is a timeframe in which subsequent graft failure or death may be attributable, at least in part, to the EB BSI. The follow-up period began immediately after the first positive blood culture was collected.

Data on SOT recipients were abstracted from the electronic medical records at each study site by a combination of electronic data extraction, with validation of variables, and manual chart review. Information was collected on demographics, comorbidities, medications, and details of the EB BSI episode (see Text S2 parts B1 and B2, SDC, http://links.lww.com/TP/C498 for details).

All EB isolates identified from study subjects were tested as part of routine care for susceptibility to antibiotics at each of the centers’ clinical microbiology laboratories. At HUP, the semiautomated Vitek 2 identification and susceptibility system (bioMerieux, Inc, Durham, NC) was utilized; at JHH, the BD Phoenix Automated System (BD Diagnostics, Sparks, MD) was used; and at UMMC, disk diffusion was used before 2010 and the Vitek 2 (bioMerieux, Inc, Durham, NC) was used after 2010.

Continuous variables were summarized using median and interquartile range, and categorical variables were summarized using proportions. Continuous variables were compared using the Wilcoxon rank-sum test (because distributions were non-normal), and categorical variables were compared using the χ² or Fisher exact test.

For the determination of the association between CRE BSI and risk of death, we performed a survival analysis. Time zero was defined as the date of the first blood culture that grew an EB organism, and the time at risk was measured in days. The day on which the SOT recipient first met criteria for the outcome (ie, date of death) following the EB BSI was the failure date, and subjects were censored after 60 d of follow-up. For the unadjusted analyses, a Kaplan Meier curve was plotted to assess the time to the primary outcome (death), stratified by exposure status (CRE versus non-CRE EB BSI), and a log-rank test was performed. For the adjusted analysis, a mixed-effects multivariable frailty model using the Weibull distribution was developed, with a random effect for study site. Bivariate regression was used to examine the relationship between the primary exposure, as well as other baseline factors, and the outcome. Variables from bivariate analyses with P < 0.20 were considered for inclusion in the final multivariable model. Variables were retained in the final multivariable model if they were confounders of the primary association (defined by a 15% or more change in the effect estimate of the association between the primary exposure and outcome) or had a P of <0.05 in the multivariable model. Hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated to evaluate the strength of any associations.

For the determination of the association between CRE BSI and risk of new-onset graft failure, we again performed a survival analysis, but used multivariable competing risks regression, with death modeled as a competing risk. Study site was incorporated as a fixed effect. Measurement of time at risk, duration of follow-up, and variable selection for the multivariable model were performed in the same manner as described above. For this analysis, subhazard ratios (SHRs) and 95% CIs were calculated.

For the determination of risk factors for CRE BSI, we performed mixed-effects multivariable logistic regression analyses. Bivariate mixed-effects logistic regression was used to examine the relationship between each potential risk factor and CRE BSI. Variables from bivariate analyses with P < 0.20 were considered for inclusion in the final multivariable model. Manual forward selection was performed to build the multivariable model. Variables were retained in the final model if they were significantly associated with the outcome (P < 0.05). Odds ratios (ORs) and 95% CIs were calculated to evaluate the strength of any association.

We performed several secondary analyses, including: (1) Stratified analyses by organ transplant type. (2) An analysis of outcomes after further stratifying the exposure into three groups: CRE BSI, extended-spectrum beta-lactamase (ESBL)–producing EB BSI (with ESBL status determined by either confirmatory testing using the double disk method with both cefotaxime and ceftazidime,²⁵ the ESBL ETEST [bioMerieux, Durham, NC], or a ceftriaxone MIC of ≥8 μg/mL²⁶), and susceptible EB BSI. (3) An analysis of outcomes after stratifying the CRE exposure into two mutually exclusive groups: CRE BSI nonsusceptible to all carbapenems and CRE BSI nonsusceptible to ertapenem only.

A mediation analysis was performed in which we evaluated whether time to effective antibiotic therapy (measured in days) was the primary driver of the association between CRE BSI and all-cause mortality and/or new-onset graft failure. Effective antibiotic therapy is defined in Text S2 part B3, SDC, http://links.lww.com/TP/C498.

An exploratory analysis was performed in which we identified the risk factors for death among those with CRE BSI. In this analysis, only those with a CRE BSI were included in the cohort. Cases were those who died within 60 d after the CRE BSI, and controls were those who did not.

The same statistical approach was used for all secondary analyses as is described for the primary analyses. All analyses were performed with STATA/SE 15.1 (StataCorp, College Station, TX).

A total of 897 SOT recipients developed an EB BSI during the study period, of which 70 (8%) cases were due to CRE. Among those with a CRE BSI (Table 1), the median age was 56 y old (interquartile range [IQR], 48–63), 24 (34%) were women, 40 (57%) had received a liver transplant, 13 (19%) a kidney, 12 (17%) a lung, 4 (6%) a heart, and 1 (1%) a pancreas transplant. The remainder of the cohort has been described previously.¹²

Among those with a CRE BSI (Table 2), the most common identifiable sources were biliary (19, 27%), genitourinary (13, 19%), lower respiratory tract (13, 19%), and intra-abdominal (12, 17%). The most common CRE organisms were Klebsiella (37, 53%) and Enterobacter (25, 36%) species. The median duration of bacteremia was 1 d (IQR, 1–2). The microbiology of the remainder of the cohort has been described previously.¹²

Among those SOT recipients with a CRE BSI, 33 (47%) died within 60 d of the BSI, compared to 102 (12%) among those with a non-CRE EB BSI (Figure 1A, log-rank P < 0.001). On multivariable analysis (Table 3A), CRE BSI was associated with a significantly increased hazard of death within 60 d (adjusted hazard ratio [aHR] 2.85; 95% CI, 1.68-4.84; P < 0.001), after adjusting for age, organ type, prior CRE and EB colonization/infection, prior exposure to polymyxins, graft failure before the EB BSI, and recent allograft rejection treated with corticosteroids. Notably, the year of transplant and year of EB BSI were not confounders or significantly associated with the outcome.

There were 663 SOT recipients with a functioning graft at the time of their EB BSI, of which 44 (7%) had a CRE BSI. Among those with a CRE BSI, 15 (34%) developed new-onset graft failure within 60 d of the EB BSI, compared to 72 (12%) in the non-CRE BSI group (Figure 1B, log-rank P < 0.001). On multivariable analysis, in which death was modeled as a competing risk (Table 3B), CRE BSI was associated with a significantly increased hazard of new-onset graft failure within 60 d (aSHR, 2.42; 95% CI, 1.33-4.39; P = 0.004), after adjusting for organ type, study site, prior antibiotic exposures, prior EB colonization/infection, and recent allograft rejection.

On multivariable analysis, the independent risk factors for CRE BSI (compared to non-CRE EB BSI) included (Table 4) prior CRE colonization/infection (adjusted odds ratio [aOR], 9.86; 95% CI, 4.88-19.93; P < 0.001), liver transplantation (aOR, 2.64; 95% CI, 1.23-5.65; P = 0.012), lung transplantation (aOR, 3.76; 95% CI, 1.40-10.09; P = 0.009), exposure to a third-generation cephalosporin in the prior 6 mo (aOR, 2.21; 95% CI, 1.17-4.17; P = 0.015), and exposure to a carbapenem in the prior 6 mo (aOR, 2.80; 95% CI, 1.54-5.10; P = 0.001). Conversely, there was a significantly decreased odds of CRE BSI in SOT recipients on mycophenolate immunosuppression (aOR, 0.47; 95% CI, 0.25-0.89; P = 0.021), and a reduction in the odds of CRE BSI with increasing time posttransplant (aOR, 0.99; 95% CI, 0.99-1.00; P = 0.051 per day forward posttransplant).

Among liver transplant recipients, there was not a significant association between CRE BSI and the hazard of all-cause mortality (aHR, 1.60; 95% CI, 0.83-3.05; P = 0.158) or new-onset graft failure (aHR, 0.70; 95% CI, 0.25-1.99; P = 0.506) (Table S1, SDC, http://links.lww.com/TP/C498). In examining risk factors for CRE BSI in liver recipients, there was a significantly decreased odds of CRE BSI among those who received their allograft from a living donor (aOR, 0.18; 95% CI, 0.05-0.72; P = 0.016) (Table S2, SDC, http://links.lww.com/TP/C498).

Among kidney, lung, and pancreas transplant recipients (Table S3A, SDC, http://links.lww.com/TP/C498), CRE BSI was associated with a significantly increased hazard of all-cause mortality on bivariate analysis. Among all non–liver transplant recipients (Table S3B, SDC, http://links.lww.com/TP/C498), CRE BSI was associated with a significant increase in the hazard of new-onset graft failure on bivariate analysis.

After stratifying the exposure into 3 groups based on resistance to carbapenems and extended-spectrum cephalosporins (namely, CRE, ESBL-EB, and susceptible EB BSI) (Figure S1, SDC, http://links.lww.com/TP/C498), an unadjusted analysis showed that CRE BSI was associated with a significant increase in the hazard of all-cause mortality (HR, 5.46; 95% CI, 3.56-8.36; P < 0.001), whereas ESBL-EB BSI was not (HR, 1.22; 95% CI, 0.81-1.84; P = 0.348).

After stratifying the CRE exposure into 2 groups, namely, CRE nonsusceptible to all carbapenems and CRE nonsusceptible to ertapenem only (Figure S2, SDC, http://links.lww.com/TP/C498), an unadjusted analysis showed that both CRE types were associated with a significantly increased hazard of all-cause mortality (CRE BSI nonsusceptible to all carbapenems HR, 6.64; 95% CI, 4.32-10.18; P < 0.001; CRE nonsusceptible to only ertapenem HR, 2.48; 95% CI, 1.09-5.65; P = 0.031).

Among those with a CRE BSI, 27 (39%) received effective antibiotic therapy within 24 h of their first positive blood culture, compared to 597 (72%) among those with a non-CRE EB BSI (χ² P < 0.001) (Table S4, SDC, http://links.lww.com/TP/C498). Among those with a CRE BSI, the median time to effective antibiotics was 3 d (IQR, 0–90) compared to 1 d (IQR, 0–2) among those with a non-CRE EB BSI (Wilcoxon rank-sum P < 0.001). However, among those with a CRE BSI, the median time to effective antibiotics was reduced to 1 d (IQR, 0–4) for those with prior CRE colonization/infection. Furthermore, among those with a CRE BSI, the median time to effective antibiotics was 2 d (IQR, 0–8) in liver recipients, compared to 4 d (IQR, 1–90) in non–liver recipients (Wilcoxon rank-sum P = 0.058). There was no significant multicollinearity observed between prior CRE colonization/infection, liver transplantation, prior polymyxin exposure, and time to effective antibiotics (variance inflation factor values, 1.00–1.29; tolerance values, 0.78–1.14).

When time to effective therapy was incorporated into the analysis of CRE BSI and its impact on all-cause mortality and new-onset graft failure (Table S5, SDC, http://links.lww.com/TP/C498), we found that (1) with increasing time to effective antibiotics, there was a significant increase in the hazard of all-cause mortality (aHR, 1.01; 95% CI, 1.00-1.01; P = 0.034 per day without effective antibiotics) and new-onset graft failure (aSHR, 1.01; 95% CI, 1.00-1.02; P = 0.009); (2) there remained a significant association between CRE BSI and all-cause mortality (aHR, 2.60; 95% CI, 1.54-4.41; P < 0.001) and new-onset graft failure (aSHR, 2.03; 95% CI, 1.09-3.78; P = 0.025).

In the subgroup with a CRE BSI (N = 70), there were 33 (47%) who died within 60 d of the BSI. The independent risk factors for death within 60 d included (Table S6, SDC, http://links.lww.com/TP/C498): polymyxin exposure in the prior 6 mo (aOR, 6.44; 95% CI, 1.01-40.93; P = 0.049), Pitt bacteremia score (aOR, 1.36; 95% CI, 1.00-1.84; P = 0.053 per additional point), and graft failure before the CRE BSI (aOR, 7.66; 95% CI, 1.16-50.33; P = 0.034). Conversely, a genitourinary source of the CRE BSI (eg, urinary tract infection) was associated with a reduced odds of death (aOR, 0.05; 95% CI, 0.002-0.93; P = 0.045). Time to effective antibiotics was not significantly associated with the odds of death (aOR, 1.00; 95% CI, 0.98-1.02; P = 0.966).