27 ECMO circuits, used for advanced extracorporeal respiratory support in 20 COVID-19 ARDS patients hospitalized in our intensive care unit (ICU), were included in this observational non-interventional study. Each patient satisfies the inclusion and exclusion criteria shown in Table 1.
Table 1
Inclusion and exclusion criteria.
Inclusion criteria | Exclusion criteria |
• Age ≥ 18 years old • Acute respiratory failure, refractory to mechanical ventilation • Veno-venous ECMO settings with HLS Maquet© (PLS or Cardiohelp) | • Age < 18 years old • ECMO settings other than veno-venous or not exploiting HLS Maquet© |
Each circuit belonged to an HLS system (Maquet©, Getinge Group). Measurements started from day 0 of ECMO run start and/or ECMO circuit exchange. Every 48 hours MLDS and oxygenation and decarboxylation functions were measured, and laboratory tests, together with the presence of active anticoagulation treatment, were derived from our electronic medical records (Sunrise Clinical Manager, Allscripts Healthcare Solutions, Inc., Chicago, Illinois). Each event possibly related to VV ECMO failure (refractory hypoxemia, respiratory acidosis, active bleeding) was recorded in a time-set database, relating the membrane lung functional parameters with the clinical events responsible for the extracorporeal circuit exchange. Any septic derangement was recorded in terms of serum white blood cells and C-reactive protein daily kinetics.
Coagulative balance was recorded (partial thromboplastin time time, fibrinogen, D-dimer and unfractionated heparin infusion). Heparin infusion rate was targeted so that activated prothrombin time values were kept between 50 and 60 seconds.
Membrane lung dead space determination and decarboxylation function.
To determine membrane lung dead space, we used the Bohr-Enghoff equation:
$$\left(1\right)\frac{Vd}{Vt}=\frac{PaCO2-PECO2}{PaCO2}*100$$
where PaCO2 in arterial partial pressure, considering the membrane lung, coincides with carbon dioxide partial pressure in the blood approaching the outflow blood cannula (POUTCO2). This value is obtained by blood gas analyses (RAPIDPoint 500, Siemens). The equation in (1), in a VV ECMO setting, can be written as (2):
$$\left(2\right)\frac{VdML}{VtML}=\frac{PoutCO2-ETCO2}{PoutCO2}*100$$
where ETMLCO2 is the exhaled CO2 at the exhaust gas flow of the membrane lung.
Furthermore, measuring the ETMLCO2 as a percentage allowed us to quantify the exact membrane lung work in terms of decarboxylation (mlVCO2): multiplying ETMLCO2 as a percentage of the actual SGF (expressed in mL/min) as in (3) provides the exact number of milliliters per minute (mL/min) of carbon dioxide effectively removed from the blood by the extracorporeal system.
$$\left(3\right)mlVCO2=\%CO2xSGF\left(\frac{mL}{min}\right)$$
While MLDS is independent of a patient’s clinical setting and venous carbon dioxide content, it is clear that MLVCO2 directly depends on carbon dioxide venous content at the inflow side of the membrane lung and can be unrelated to effective membrane lung performance.
Determinants of membrane lung oxygenation performance
Through the determination of oxygen blood partial pressure, the oxygenated fraction of hemoglobin and its serum concentration (pO2, HbO2 and [Hb]) and the oxygenation function of VV ECMO in terms of oxygen blood transfer (MLVO2) were easily assessed, as shown in Eq. (4).
Since residual oxygen delivery is present in venous blood facing the inflow side of the membrane lung, HbO2 and pO2 referring to membrane lung action alone are considered as differentials between the outflow and inflow values (dHbO2, dPO2), which correspond to the difference in oxygen content between the outflow and the inflow tracts.
$$\left(4\right)mlVO2=mlBF*10*\left(1.39*\left[Hb\right]*dHbO2\left(\text{\%}\right)+0.003*dPO2\right)$$
To better assess membrane lung efficiency in oxygenation, we measured its intrinsic shunt (QS/ECBF)10 calculated with the equation in (5):
$$\left(5\right)\frac{Qs}{ECBF}=\frac{CoxyO2-CoutO2}{CoxyO2-CinlO2}*100$$
where CoxyO2 is the content of oxygen ideally in equilibrium with the gas in the hollow fibers, CinlO2 is the content of oxygen in the membrane lung inlet, and CoutO2 is the content of oxygen in the membrane lung outlet. Each value is expressed in milliliters per minute (mL/min).
CoxyO2 is defined by the gas equations in (6),(7), (8) and (9):
$$\left(6\right)PoxyO2=Patm*FiO2-\frac{pCO2}{RQ}$$
$$\left(6.1\right)PoxyO2=713mmHg*1-\frac{pCO2}{RQ}$$
where Patm = 760 mmHg – 47 mmHg (atmospheric water tension) and RQ is the respiratory quotient, which consists of the extracorporeal VCO2 to VO2 ratio.
$$\left(7\right)CoxyO2=10*\left(1.39*\left[Hb\right]+0.003*PoxyO2\right)$$
$$\left(8\right)CoutO2=10*\left(1.39*\left[Hb\right]*HbO2out+0.003*pO2out\right)$$
$$\left(9\right)CinlO2=10*\left(1.39*\left[Hb\right]*HbO2inl+0.003*pO2inl\right)$$
The same considerations made for MLDS and MLVCO2 with regard to membrane lung performance are also valid for the membrane lung intrinsic shunt (which is independent of inlet blood oxygen concentration) and MLVO2 (which depends on inlet blood oxygen concentration).
Study design and population
The present study was submitted to an ethics committee and is identified as IRRB.41.21.
Over a period of time from September 2021 to November 2021, 13 HLS circuit systems of 20 patients admitted to our COVID-ICU were included in the study. Our dataset consisted of information about the type of ECMO circuit, and measures of membrane lung were collected and registered every 2 days. Blood flow (BF), Sweep Gas Flow (SGF), and membrane lung pressure drop (Pd) were measured and recorded.
After an oxygenator sight maneuver (SGF augmented to 12 L/min for 30 seconds) to allow condensate to escape from hollow fibers, we set BF and SGF respectively at 3 and 4.5 liters per minute (L/min), so as to homologate the SGF to BF ratio to a value of 1.5, which is crucial for maintaining the same extracorporeal perfusion to ventilation ratio.18 To avoid any possible shift in the patient’s oxidative metabolism (body VO2, VCO2), every patient received adequate sedation and paralysis (bispectral index monitoring was exploited to reach values between 40 and 60). Peripheral oxygen saturation (SpO2) and invasive arterial blood pressure (ABP) was continuously monitored.
When a mechanical steady state was reached and extracorporeal BF could be maintained without developing suction phenomena, we connected an end-tidal CO2 (ETCO2) detector (Vamos plus, Draeger) to the membrane lung gas exhaust. This was obtained with the help of an endotracheal tube linked to a heat and moisture exchanger (HME), which in turn was connected to the ETCO2 detector (Fig. 1). At this point, ETCO2 was measured directly from the membrane lung air outflow and shown on the detector monitor. We performed blood gas analysis (BGA) from blood entering the membrane lung inflow tract and BGA from blood exiting the membrane lung outflow. On PLS systems, we measured blood pressure before and after passage through the membrane lung, to determine the pressure drop (Pd) relative to the extracorporeal blood flow (BF), which is a commonly used parameter of membrane lung vitality. For Cardiohelp systems, Pd is shown on the monitor. For each circuit, a dataset was registered.
Membrane lung settings at the time of observation were recorded (BF, SGF, Pd, rpm), as well as oxygen and carbon dioxide partial pressures, and oxygen-bound hemoglobin (HbO2) from inflow and outflow BGAs. The electronic spreadsheet was built to automatically determine membrane lung carbon dioxide removal and oxygen transfer (respectively MLVCO2 and MLVO2) as shown in equations (3) and (4). As shown in equations (2) and (5), membrane lung dead space and shunt were also determined. For every determination, circuit age was considered (day of treatment with the current circuit).
When the clinical conditions suggested circuit derangement (for example, an unsafe increase in membrane lung pressure drop indicative of impending ECMO failure, or evidence of a discrepancy between MLVO2 and the patient’s oxygen/cardiac output requirements, or evidence of bleeding), we performed the circuit’s last measurements and set the resulting extracorporeal dead space value to ECMO failure.
Two different groups were identified: in group 1 ECMO circuits undergoing circuit exchange due to mechanical failure (n = 10), in group 2 the ECMO circuits that did not request a circuit exchange (n = 17).
The following parameters were collected every 2 days and before circuit exchange due to mechanical failure:
-
Membrane lung pressure drop (Pd)
-
Membrane lung revolutions-per-minute (rpm)
-
Partial pressure of oxygen in the ECMO outflow tract (PaO2post)
-
Oxygen-bound haemoglobin in the ECMO inflow tract (HbO2pre)
-
Blood haemoglobin concentration ([Hb])
-
Membrane lung dead space (MLDS) as described in (2), such as carbon dioxide partial pressure respectively in the ECMO inflow and outflow tract (pCO2pre; pCO2post) and membrane lung END tidal CO2 (ETMLCO2).
-
Membrane lung shunt (Qs/ECBF) as described in (5).
-
Membrane lung oxygen and carbon dioxide transfer (MLVO2; MLVCO2) as in (3) and (4).
The comparison between group 1 and group 2 was made on the last observation before the circuit exchange or the withraw of the extracorporeal treatment (respectively in Group 1 and Group 2).
Data analysis
Data were collected retrospectively. Continuous and categorical variables were expressed as mean, median with standard deviation and as frequency with percentage, respectively. To compare continuous variables, Wilcoxon tests or T test were used when appropriate. Chi-Square tests were used to compare categorical variables among groups. To explore the differences between the two groups, logistic regression models were used in order to assess OR. The accuracy of membrane lung dead space to discriminate between the two groups was evaluated using Receiver Operating Characteristic (ROC) curve analysis and Youden Index was used to identify a threshold of the variable.
Levels of significance were set with p < 0.05. Statistical analyses were performed using SAS software version 9.4.