This was a single centre, retrospective cohort study. The study was approved as low risk by the Alfred Hospital Human Ethics Committee (application: 28/20), with a waiver of individual patient informed consent.
The setting was a 50-bed, quaternary referral ICU in Melbourne, Australia that has almost 3000 admissions to ICU per year. It provides state services for burns, trauma, and hyperbaric medicine, heart and lung transplant, ventricular assist devices (VAD), and has had an ECMO-program since 1990, providing over 100 ECMO runs annually.
Adult patients (> 16 years old) admitted to the intensive care unit (ICU) following in-hospital or out of hospital CA of primary cardiac origin from 1st January 2013 to 31st August 2018 were included in this study. Patients were excluded if they had a non-cardiac cause of their CA (e.g. drug overdose, trauma, hypoxia), they were treated with veno-venous (VV) ECMO, or if they died < 24 hours after admission (as oxygen levels are unlikely to have impact on mortality in this group). Patients were either managed with ECPR or CCPR.
Data management and collection
We collected data concerning patient demographics (age, weight, height, gender), illness severity scores (APACHE Score II and III / SOFA score), the reason for CA, and the initial rhythm. We further extracted the two important CPR time variables: No-flow-time (period of CA where no CPR is performed) and low-flow-time (period of CA where CPR is performed but ROSC has not yet been established). The type of arrest (out of hospital or in hospital CA) was also recorded. Laboratory results, hemodynamic and ventilation-parameters, the need for renal replacement therapy and ECMO-parameters (site of cannulation, size of cannulas, mode, blood flow, fresh gas flow, ECMO complications and duration) were assessed on admission and daily from day 1 to day 8.
All arterial blood gas (ABG) samples in the ECPR group were taken from the right arm as per the institutional ECMO protocol due to the risk of differential arterial gas tensions during VA ECMO. Either arm could be used in the CCPR group. These were taken as per routine protocol (1–2 hourly) or more frequently following any change of condition. We assessed the highest, the mean, and lowest PaO2-value over a timeframe of 8 consecutive days. Hyperoxia was defined as a PaO2-level of 101–299 mmHg. Extreme hyperoxia was defined as any recorded PaO2 -level ≥ 300 mmHg over the first 8-days, as per previous authors. [8, 12, 14, 28]. In case of non-availability of PaO2-values these were classified as missing data. Additional data were retrieved from electronic database records. No assumptions were made regarding missing data. All proportions were calculated as percentages of the patients with available data.
CA management was standardized as per local hospital and ambulance service guidelines.[29− 31] CPR was initially performed manually in all cases, but if prolonged > 10 minutes was switched to the Lucas® chest compression system (Physiocontrol, Redmond) to facilitate transportation. Patients with CA were treated via two different pathways: CCPR was initiated in all patients, while ECPR with implementation of VA ECMO was initiated in refractory CA when there were no contraindications. ECPR inclusion and exclusion criteria for CA patients have been previously published.[25, 32] In the ECPR group hemodynamic support was maintained with either the Cardiohelp® (HLS, Maquet) or Rotaflow® (PLS, Getinge Group). The initial typical settings of the ECMO were a fresh gas flow oxygen fraction (FsO2) of 1.0, 3l/min blood flow, 3l/min fresh gas (sweep) flow and ventilator settings of FiO2 0.5, positive end expiratory pressure (PEEP) 10 cmH2O and pressure control level above PEEP (PC) 10cmH2O. These were both adjusted to maintain mean arterial pressure of ≥ 65 mmHg, normalization of lactate, SpO2 in the right arm of between 94–100% and normocapnia. Withdrawal of care occurred when clinical, radiological and biochemical information were consistent with a poor outcome.
The primary outcome was all cause 30-day mortality. Secondary outcomes included hospital and ICU length of stay, days of invasive ventilation and complications of treatment (infection, CNS complications, hepatic injury, renal failure, multiorgan failure, vascular injury) in each sub-group (ECPR and CCPR).
Continuous variables were summarised using means and standard deviation (SD) or medians and interquartile ranges (IQR) according to data type and distribution. Categorical variables were summarised using frequency counts and percentages. Comparisons between groups were made using Student’s t-test or Mann-Whitney U test as appropriate for continuous variables and chi-square test with Yates correction for categorical variables. The relationship between oxygen levels over the eight-day period in the ECPR group and CCPR group were determined using linear mixed modelling fitting main effects for group, time, and an interaction between group and time. Results from the mixed effects models were reported as adjusted means and 95% CI. Univariate and multivariate analyses for 30-day mortality were performed using logistic regression with results reported as odds ratios (OR) and 95% confidence intervals (95% CI). Variables with a p < 0.05 on univariate analysis and those deemed to be clinically important (age, gender) were included in the multivariable logistic regression model. A two-sided p-value less than 0.05 was chosen to indicate statistical significance. Analyses were performed using SPSS Statistics version 25 (IBM 2017) or SAS version 9.4 (SAS Institute, Cary, NC, USA).