Study type
This study used an observational, prospective monocentric design in the Surgical Intensive Care and Circulatory Support Unit of the University Hospital of Dijon, France between September 2015 and February 2017. As the samples were realized systematically following the standard protocols of care for patients benefiting from VA ECMO, no oral or written consent was required. The study was submitted to the Ethics Committee of the French Society of Anaesthesia and Critical Care and was authorized and registered under the number IRB00010252018179. In order to utilize the patients’ information, and in compliance with the law on personal data protection (Loi informatique et libertés, 6 January 1978; modified in 2004), the study was submitted to the National Commission for Data Protection (CNIL); it gained authorization and was registered under the number 1855426 v 0.
Patients and VA ECMO protocol
All patients over 18 years of age who were treated using peripheral VA ECMO for refractory cardiogenic shock, refractory cardiac arrest (RCA), post-cardiotomy cardiac failure, and/or post-cardiac arrest syndrome were included in the study. Exclusion criteria included patients with a central or pulmonary artery VA ECMO, and patients who already benefited from VA ECMO prior to the present episode. In the study institution, peripheral VA ECMO support initiation is standardized and follows the usual standard of care, as previously described [18]. Patients were initially intubated and sedated with continuous infusion of propofol. VA ECMO cannulas were systematically positioned using transoesophageal echocardiography guidance. Venous cannula was positioned in the right atrium in order to drain the venous return form superior and inferior vena cava. Arterial cannula guidewire was visualized in the descending thoracic aorta prior to cannulation.
For patients who were stabilized under VA ECMO and who did not have residual organ hypoperfusion, sedation and mechanical ventilation weaning were initiated as soon as possible. Vasopressors were introduced when needed to maintain a minimal mean arterial pressure of 60 mmHg. Left ventricular unloading was set up in the absence of arterial pulsatility. Lower leg with arterial cannula was monitored twice daily by the nurse in charge, with Doppler examination.
VA ECMO weaning was initiated when left ventricular ejection fraction > 20%, subaortic velocity time integral > 10 cm, right ventricular tricuspid annular systolic excursion > 17 mm, right ventricular end diastolic basal diameter < 35 mm, and arterial lactate level < 2 mmol.L-1 with a VA ECMO pump speed ≤ 1500 RPM.
Study protocol
Baseline (H0) was set when the desired VA ECMO flow rate was reached. Blood samples were drawn from arterial (after the oxygenator) and venous (before oxygenator) VA ECMO cannulas after purging 5 mL of blood to perform blood gas analyses and to measure lactate concentration on the VA ECMO at H0, H6, and H24. Arterial blood samples were drawn form VA ECMO arterial line due to variation of arterial catheter position (radial or femoral). The syringes were pneumatically sent to the laboratory (ABL800, Radiometer, Copenhagen, Denmark). For each series of samples, pH, CO2 partial pressure (PCO2), oxygen partial pressure (PO2), oxygen saturation (SaO2), bicarbonates, and lactate concentration were collected. The venous saturation of the VA ECMO was considered as the venous saturation (SvO2) of the patient. The PCO2 gap was calculated as the difference between the venous partial pressure in CO2 (PvCO2) and the arterial partial pressure in CO2 (PaCO2):
PCO2 gap (mmHg) = PvCO2 - PaCO2 (1)
Arterial oxygen content (DaO2), venous oxygen content (DvO2), and arteriovenous difference in oxygen content (Da–vO2) were calculated using the following equations:
DaO2 = SaO2 x Hb x 1.34 + 0.0031 x PaO2 (2)
DvO2 = SvO2 x Hb x 1.34 + 0.0031 x PvO2 (3)
Da-vO2 = DaO2 - DvO2 (4)
The PCO2/Da–vO2 ratio was calculated as follows:
PCO2 gap/Da–vO2 (6)
For each patient, demographic data, SOFA score, SAPS II score, hemodynamic and ventilatory parameters (heart rate, systolic, diastolic and mean arterial pressure, tidal volume, positive expiratory pressure, and respiratory rate), vasopressor, inotropic doses, and VA ECMO parameters at each time point (flow rate, sweep gas flow, inspired fraction of O2 [FIO2]) were collected. Duration of VA ECMO, duration of intensive care unit (ICU) stays, and 28-day mortality were collected as outcomes.
Definition
Early mortality was defined as any death occurring under VA ECMO or in the 72 hours following VA ECMO weaning secondary to multiple organ failure. This led to two groups of patients: patients who died under VA ECMO or in the 72 hours following VA ECMO weaning (i.e., early death), and patients who survived VA ECMO weaning and beyond 72 hours (i.e., survival).
Study endpoints
The primary endpoint was the ability to use the PCO2 gap to determine early mortality. The secondary end-points were the PCO2 gap/Da–vO2 ratio, SvO2, the SOFA score, and the IGS II score. Overall mortality was also evaluated at 28 days.
Statistical analysis
The quantitative data are presented as medians and interquartile ranges; the qualitative data are presented as numbers and percentages. Appropriate parametric or non-parametric tests were also performed. Normality was assessed using the Shapiro–Wilk test. Bonferroni correction was applied to interpret the p-values of repeated measures. Kaplan–Meier curves were drawn for censored data, and log rank tests were carried out using the reported cut-off values of 6 mmHg for the PCO2 gap [13] and 1.4 for the PCO2 gap/Da–vO2 ratio [16]. Retrospectively, the power to assess the PCO2 gap difference observed between the two groups at H6 was 77%. Correlation was assessed using Spearman’s method. ROC curves were drawn to represent the ability of the PCO2 gap, the PCO2/Da–vO2 ratio, and lactatemia to discriminate early death. AUC and optimal cut-off were determined using Youden’s method. Statistical analysis was performed with R Studio Version 1.0.143 (© 2009–2016 R Studio, Inc. from R version 3.5.0 Patched; 2018-05-03 r74699).