This multicenter observational study was conducted between January 1, 2017 and December 31, 2022 in 5 Intensive Care Units (ICU) of Italian university-affiliated hospitals, overall accounting for a total of 70-ICU beds (i.e., Mater Domini Hospital (Catanzaro); Padua University Hospital; Verona University Hospital; Policlinico University Hospital (Bari) and Fondazione IRCSS Gerardo Hospital dei Tintori Hospital (Monza)). We included adult patients, over 18 years of age, who received V-V ECMO for respiratory support during the study period. The exclusion criteria were age < 18 years old, pregnancy, veno-arterial (V-A) or mixed ECMO-configuration (e.g., V-VA), incomplete records for the main outcomes (absence of 1-year mortality and/or microbiological surveillance), and survival < 24 hours after cannulation. The study was conducted in compliance with the Declaration of Helsinki and the approval for the investigation was granted by the local Ethics Committee "Comitato Etico Territoriale Regione Calabria" (approval n. 22 on September 27, 2023), which waived the need for informed consent due to the retrospective observational nature of the study. All patient data was anonymised and de-identified before analysis. This study followed the ‘Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement guidelines for observational cohort studies’ (additional-Table 1)19.
The decision to start V-V ECMO treatment was made by senior intensivists (PN, FL, EB, GF, LG, SG), according to the ELSO guidelines/recommendations1. All V-V ECMOs were placed exclusively in ICU and a femoro-jugular configuration was preferred. All centers kept the ECMO circuit, as much as possible, isolated (e.g. withdrawals or infuse medications were not recommended). Antimicrobial prophylaxis, at the time of cannulation, was uniformly not administrated1.
Routine microbiological surveillance was uniformly conducted in all centers: rectal swabs and respiratory tract samples were collected at ICU admission and, subsequently, every 48-72 hours. Blood and urine samples were collected on clinical suspicion, as well as other biological samples collected from skin, soft tissue, cannula insertion etc13,20. The routine protocol for infection control and prevention, shared by all enrolled ICUs, is available in additional-Methods 1.
To prevent the occurrence of MDR patterns, each participating center adopted an institutional antimicrobial stewardship program, which involved rapid molecular diagnostic tests, strict communication between ICUs and microbiology laboratories, and daily review of antibiotic regimens by a dedicated infectious disease specialist consultant with the aim to promote the selection of the optimal antimicrobial drug regimen21. The antimicrobial therapy was defined as ‘empiric’ when started before any microbiological evidence; or as ‘targeted’ strategy when started after microbiological evidence and according to in vitro susceptibility tests22.
Patients were divided into three groups according to the time of MDR GN bacteria detection: 1) 'precolonized MDR GN bacteria’ group, including those patients with a history of MDR GN isolation within 48 hours after ECMO cannulation; 2) ‘V-V ECMO-acquired MDR GN bacteria’ group, including patients with isolation of MDR GN bacteria after 48 hours from ECMO start and 48 hours after disconnection23,24; and 3) ‘non-MDR GN bacteria’ group, including patients never culturing MDR GN bacteria during V-V ECMO support.
Data Collection
The electronic health records were retrospectively examined and the following variables were collected: i) demographic and baseline characteristics before V-V ECMO placement, Charlson Comorbidity index, Sequential Organ Failure Assessment (SOFA) score at ICU admission and at cannulation, initiation of invasive mechanical ventilation (IMV), indications for V-V ECMO, interfacility transport (defined as transfer from any medical facility outside of our ECMO centers and without ECMO capabilities1,25), year of V-V ECMO connection (Table 1); ii) outcomes of interest (see full description below and in Table 2); iii) culture results during V-V ECMO support, type of isolated bacteria, site of isolation, resistance profiles, and data on antibiotic usage (Figure 1, Tables 3 and 4).
Outcomes
The primary outcome was assessing the rate of MDR GN bacteria in a homogeneous cohort of V-V ECMOs. The MDR GN pathogens (i.e., ESBL, AmpC, CRE, DTR profiles, and CRAB), were classified according to the Center for Disease Control definition (https://www.cdc.gov/infectioncontrol/index.html)18and in vitro susceptibility tests (https://www.eucast.org/clinical_breakpoints). Infection was defined as ≥ 104 CFU/ml and clinical signs of sepsis or septic shock, according to the Sepsis-3 criteria26; while colonization occurred in case of less than 104 CFU/ml and in absence of clinical signs of infection23,24,26. The definitions of ventilator-associated pneumonia (VAP), community-acquired pneumonia (CAP), bloodstream infection (BSI), and urinary tract infection (UTI) are provided in additional-Methods 1.
Other outcomes of interest included: i) 1-year mortality; ii) annual hospital V-V ECMO (defined as the specific number of patients treated with V-V ECMO per year25); iii) weaning success (defined as extubation and absence of invasive ventilatory support 48 hours following extubation) and weaning failure (defined as failure of the first spontaneous breathing trial, and/or reintubation or resumption of ventilatory support within 48 hours after extubation and/or death within 48 h following extubation27); iv) ventilation free days (VFD) (reference: 28 days)28; v) overall V-V ECMO duration; vi) need of renal replacement therapy (RRT) after V-V ECMO start, and vii) ICU length of stay (Table 2).
Statistical analysis
Continuous variables are presented as medians and interquartile ranges [IQR] or as mean and standard deviation (SD); while categorical variables are presented as numbers (percentages). Baseline patients’ characteristics and outcome variables were compared between two or three pre-defined subpopulations, as follows: 1) ‘precolonized MDR GN bacteria’ group, 2) ‘V-V ECMO-acquired MDR GN bacteria’ group, and 3) ‘non-MDR GN bacteria’ group. The sample size could not be calculated due to the explorative design of our investigation and the scarcity of data regarding the prevalence of precolonizations in patients eligible to V-V ECMO support. The t-test, Mann-Whitney test, ANOVA or Kruskal Wallis test were properly used to compare continuous variables and adjusted by Benjamini and Hochberg method. Chi-square and Fisher’s exact tests were used for comparing categorical variables.
Regarding 1-year mortality, the Kaplan Meier curves were provided only as graphical support. For investigating the risk of mortality, the unadjusted (HR) and adjusted hazard ratio (aHR), 95% confidence intervals [CI], were calculated using Cox-proportional hazards models (additional-Tables 2-4). Cox-proportional hazards models assume that the hazard ratio is constant over time, therefore the test for proportional-hazard assumption was verified for each covariate included in the univariable model. The time-dependent variable started from V-V ECMO connection for patients without MDR GN pathogens and in case of previous colonizations; while, for patients acquiring MDR GN bacteria after V-V ECMO connection, the time-dependent variable started from the first MDR GN bacteria isolation (to avoid immortal time bias). All variables described in Tables 1, 3 and 4,with a significance p-value < 0.10, were included in an univariable Cox-proportional hazards model investigating 1-year mortality (*) (additional-Table 4). Then, as shown in additional-Tables 2 and 3, the multivariable adjustment was provided according to significant confounders (p-value < 0.05) identified through the univariable Cox-proportional hazards model, as mentioned above (*).
For investigating the impact of the annual hospital V-V ECMO volume25 on the occurrence of MDR GN bacteria acquisition after V-V ECMO connection (precolonized patients were excluded from this analysis), a multivariable logistic regression was applied, after adjustment for potential confounders (p-value < 0.05) identified through an univariable logistic regression model exclusively focused on the risk of MDR GN bacteria isolation after V-V ECMO start (additional-Table 5). The unadjusted (OR) and adjusted odds ratio (aOR), 95% CI were calculated, and all tests were two-sided and p-values < 0.05 were considered statistically significant. The analyses were performed using R (version 4.0.3, R foundation for Statistical Computing, Vienna, Austria).