Physiologic Effects of Extracorporeal Membrane Oxygenation in Patients with Severe Acute Respiratory Distress Syndrome

Abstract

Rationale

Blood flow rate affects mixed venous oxygenation (Sv_O2) during venovenous extracorporeal membrane oxygenation (ECMO), with possible effects on the pulmonary circulation and the right heart function.

Objectives

To describe the physiologic effects of different levels of Sv_O2 obtained by changing ECMO blood flow in patients with severe acute respiratory distress syndrome receiving ECMO and controlled mechanical ventilation.

Methods

Low (Sv_O2 target, 70–75%), intermediate (Sv_O2 target, 75–80%), and high (Sv_O2 target, >80%) ECMO blood flows were applied for 30 minutes in random order in 20 patients. Mechanical ventilation settings were left unchanged. The hemodynamic and pulmonary effects were assessed with pulmonary artery catheter and electrical impedance tomography.

Measurements and Main Results

Cardiac output decreased from low to intermediate and to high blood flow/Sv_O2 (9.2 [6.2–10.9] vs. 8.3 [5.9–9.8] vs. 7.9 [6.5–9.1] L/min; P = 0.014), as well as mean pulmonary artery pressure (34 ± 6 vs. 31 ± 6 vs. 30 ± 5 mm Hg; P < 0.001) and right ventricular stroke work index (14.2 ± 4.4 vs. 12.2 ± 3.6 vs. 11.4 ± 3.2 g × m/beat/m²; P = 0.002). Cardiac output was inversely correlated with mixed venous and arterial Po₂ values (R² = 0.257; P = 0.031; and R² = 0.324; P = 0.05). Pulmonary artery pressure was correlated with decreasing mixed venous Po₂ (R² = 0.29; P < 0.001) and with increasing cardiac output (R² = 0.378; P < 0.007). Measures of $\dot{\mathrm{V}}$/$\dot{\mathrm{Q}}$ mismatch did not differ between the three steps.

Conclusions

In patients with severe acute respiratory distress syndrome, increased ECMO blood flow rate resulting in higher Sv_O2 decreases pulmonary artery pressure, cardiac output, and right heart workload.

At a Glance Commentary

Scientific Knowledge on the Subject

Physiological effects of extracorporeal membrane oxygenation blood flow rates have not been studied. Change in mixed venous saturation (Sv_O2) could impact pulmonary hemodynamics and $\dot{\mathrm{V}}$/$\dot{\mathrm{Q}}$ matching.

What This Study Adds to the Field

Increasing extracorporeal membrane oxygenation blood flow rate and the patient’s Sv_O2 decreases cardiac output and pulmonary artery pressure. Pulmonary artery compliance and right heart workload improve, too. Complex physiological interactions drive such changes, with the aim of minimizing cardiac output while maintaining adequate systemic oxygen delivery. $\dot{\mathrm{V}}$/$\dot{\mathrm{Q}}$ mismatch is not affected by higher Sv_O2 in the range explored by this study (70–90%).

Life-threatening impairment of gas exchange, risk of ventilation-induced lung injury (VILI) (1), and pulmonary arterial hypertension leading to right ventricular failure (2) represent major clinical challenges for the treatment of patients with severe acute respiratory distress syndrome (ARDS). Venovenous extracorporeal membrane oxygenation (ECMO) is currently applied in patients with severe ARDS and refractory hypoxemia, despite optimization of protective ventilation (3). Venovenous ECMO is highly effective in obtaining adequate gas exchange and decreasing ventilatory pressures (3), whereas it does not provide direct cardiovascular support for the heart.

Lately, clinical evidence suggested that the start of venovenous ECMO could unload the right ventricle by attenuating pulmonary hypertension (4–6). Normalization of blood gases and reduced ventilation likely play a major role upon ECMO initiation (7), but higher saturation of mixed venous blood produced by ECMO blood flow returning to the right heart could also impact the pulmonary circulation (8). Early experimental studies applied ECMO to manipulate mixed venous oxygen saturation (Sv_O2) and showed reversal of hypoxic pulmonary vasoconstriction (HPV), albeit with variable effects on pulmonary arterial pressures (PAPs) and intrapulmonary shunting (9–11). Theoretically, reversal of HPV because of increased Sv_O2 could decrease PAP, possibly at the expenses of a higher intrapulmonary shunt (12).

Among clinical ECMO settings, blood flow rate is the main determinant of Sv_O2 (13), but there is a large variation in values reported by observational studies and clinical trials (3, 13). No previous study has formally tested the relationships between ECMO blood flow, the resulting Sv_O2, and cardiopulmonary physiology. If the modulation of Sv_O2 could induce physiologically relevant protective effects on the right heart and the pulmonary circulation, then titration of ECMO blood flow in patients with severe ARDS might aim at optimization of the right heart workload, providing an additional benefit beyond the reduction of Vt and the achievement of viable gas exchange (although changes in arterial O₂ content could have indirect cardiac effects).

We designed the present physiologic study to investigate, in patients with severe ARDS, the effects of ECMO blood flow targeted to three levels of Sv_O2. The aim was to assess the impact of ECMO blood flow and Sv_O2 on the pulmonary circulation, right ventricular workload, and intrapulmonary $\dot{\mathrm{V}}$/$\dot{\mathrm{Q}}$ matching.

Methods

This was a prospective physiologic randomized crossover study of patients with severe ARDS who were undergoing venovenous ECMO for clinical reasons between June 2021 and June 2022. The study was conducted in the general ICUs of Maggiore Policlinico (Milan, Italy) and San Gerardo (Monza, Italy) hospitals. The study protocol was approved by the ethical committees of the two participating centers (reference numbers 968_2020 and 3896, respectively), and informed consent requirements were met according to local regulations.

We enrolled intubated, deeply sedated, paralyzed patients undergoing ECMO for severe ARDS. All patients were receiving controlled mechanical ventilation and had a pulmonary artery catheter in place. Exclusion criteria were age <18 years, pregnancy, contraindication to the use of electrical impedance tomography (EIT; e.g., pacemaker or implantable cardioverter defibrillator), hemodynamic instability, oxygen saturation as measured by pulse oximetry <80% despite ECMO fraction of delivered oxygen in the sweep gas flow (Fd_O2) and Fi_O2 set at 100%, moribund status, and refusal by the attending physician. Patients with severe ARDS receiving ECMO were screened daily and enrolled when they fulfilled all the inclusion criteria and no exclusion criteria were present.

At enrollment, we collected the patients’ main demographics and clinical data. Clinical ventilation and ECMO settings and gas exchanges were recorded. After enrollment, ventilation and ECMO were adjusted to the following standardized settings:

• •Ventilation: pressure control mode with clinical positive end-expiratory pressure (PEEP), inspiratory pressure 12 cm H₂O above PEEP, respiratory rate 10–12 breaths per minute, inspiration/expiration ratio 1:2, clinical Fi_O2
• •ECMO: sweep gas flow set to obtain arterial pH between 7.35 and 7.45 at clinical blood flow, FdO₂ equal to clinical Fi_O2

All the standardized ventilation and ECMO settings were left unchanged throughout all the study steps. EIT monitoring (PulmoVista; Dräger) was applied before starting the protocol.

The study protocol consisted of three randomized crossover steps (each step lasted 30 min), corresponding to three levels of ECMO blood flow rate manually adjusted to target three ranges of Sv_O2:

• •Low ECMO blood flow, set to obtain Sv_O2 of 70–75%
• •Intermediate ECMO blood flow, set to obtain Sv_O2 of 75–80%
• •High ECMO blood flow, set to obtain Sv_O2 >80%

Invasive systemic and PAP, heart rate (HR), and oxygen saturation as measured by pulse oximetry were continuously monitored. Additional details about the study protocol are reported in the online supplement.

During the last minutes of each study step, we collected:

• •ECMO blood flow rate
• •Hemodynamics, including HR, cardiac output (CO; thermodilution technique), cardiac index (CI = CO/body surface area), and central pressures (central venous pressure [CVP]; pulmonary arterial occlusion pressure; systolic, diastolic, and mean PAP: PAPs, PAPd, and PAPm) measured with the pulmonary artery catheter
• •Arterial and mixed venous blood gases
• •Recordings of EIT waveforms with $\dot{\mathrm{V}}$/$\dot{\mathrm{Q}}$ data. Perfusion was assessed during an inspiratory breathhold plus bolus injection of 5% saline (10 ml) in the central line, as previously described (14).

Data from arterial and mixed venous blood gases were used to calculate the intrapulmonary shunt. Oxygen exchange through the natural lung ($\dot{\mathrm{V}}$o₂ NL) was calculated as previously reported (15).

The impact of ECMO blood flow rate on pulmonary circulation was evaluated also by calculation of pulmonary vascular resistance [PVR = (PAPm − pulmonary arterial occlusion pressure)/CO]. Pulmonary arterial compliance (PA compliance) was calculated as stroke volume (SV) divided by PA pulse pressure [PA compliance = (CO/HR)/(PAPs − PAPd)] (16). Right heart workload was quantified through the calculation of the right ventricular stroke work index [RVSWI = (CI/HR) × (PAPm − CVP) × 0.0136] (17).

EIT-based distribution of ventilation (V) and perfusion (Q) and $\dot{\mathrm{V}}$/$\dot{\mathrm{Q}}$ mismatch were assessed offline, as in previous studies (14). The following parameters were quantified:

• •Regional lung $\dot{\mathrm{V}}$/$\dot{\mathrm{Q}}$ in three horizontal gravitational regions of interest of the same size (ventral, middle, dorsal)
• •The percentage of unmatched units (only ventilated units + only perfused units)

Additional details about the study measurements can be found in the online supplement.

Statistics

Normal distribution of the study variables was assessed with the Shapiro-Wilk test. Data are expressed as mean ± standard deviation or median [interquartile range] and graphed as bars or boxplots, as appropriate. Comparison between physiologic variables (means or medians) measured at the end of the three steps was performed by one-way repeated measures ANOVA or Friedman repeated measures ANOVA on ranks, as appropriate. Post hoc pairwise multiple comparisons were performed using the Holm-Sidak or Tukey multiple comparison test, respectively.

Then, the data from the three study steps were pooled together to explore physiological interactions with a more continuous approach. Linear correlations between variables were assessed using linear mixed-effect models for repeated measures with patient as a random effect or Spearman’s test for nonrepeated measures, as appropriate. For nonlinear correlations, the Akaike information criterion was applied to identify the best-fit curve to describe correlations between physiological variables (e.g., between PAP and CO). Correlations prone to mathematical coupling (e.g., between CO and PA compliance or RVSWI) were not tested. Statistical analysis was conducted with JMP Pro version 15 software (SAS Institute Inc.).

Results

Study Population

Twenty patients were enrolled. Patients’ characteristics are reported in Table 1. The etiology of ARDS was infectious for all patients: 70% had COVID-19, and 30% had bacterial pneumonia. Right before ECMO initiation, patients’ Pa_O2/Fi_O2 was 76 ± 18 mm Hg, Pa_CO2 63 ± 11 mm Hg, and arterial pH 7.32 ± 0.07.

Table 1.

Main Characteristics of the Study Population

Characteristics	Data
Age, yr	51 ± 9
Male sex, n (%)	11 (55)
ARDS etiology, n (%)
COVID-19	14 (70)
Bacterial pneumonia	6 (30)
SOFA score	6 ± 3
Days of intubation before ECMO start	3 [0.5–5]
ECMO configuration, n (%)
Femorojugular	12 (60)
Femorofemoral	8 (40)
Femoral drainage cannula size, Fr	25 ± 1
Days of ECMO at enrollment	4 [2–6]
Gas exchange before ECMO start
PaO2/FiO2 (mm Hg)	75 ± 18
PaO2 (mm Hg)	63 ± 11
PaCO2 (mm Hg)	62 ± 11
Arterial pH	7.31 ± 0.07
Total days on ECMO	17 ± 10
ICU length of stay, d	30 ± 14
Hospital mortality, %	35

Definition of abbreviations: ARDS = acute respiratory distress syndrome; ECMO = extracorporeal membrane oxygenation; SOFA = Sequential Organ Failure Assessment.

Patients were started on ECMO after 3 [0.5–5] days of intubation, and the study was performed on ECMO Day 4 [2–6]. The total number of days on ECMO was 17 ± 10, with in-hospital mortality of 35%.

Study Protocol

During the three study steps, all ventilation and ECMO settings remained unchanged, apart from ECMO blood flow rate. PEEP was 15 [12–16] cm H₂O, Vt was 4.1 ± 1.0 ml/kg predicted body weight, respiratory rate was 12 ± 0.5 breaths/min, plateau pressure was 27 [25–29] cm H₂O, Fi_O2 and Fd_O2 were 66 ± 16%, and sweep ECMO gas flow was 4.5 [3.8–6.5] L/min.

The blood flow rates at the low, intermediate, and high steps were 1.51 [1.16—1.94] versus 2.44 [2.03–2.93] versus 3.43 [3.01–3.74] L/min (P < 0.001), corresponding to progressively more negative but acceptable ECMO drainage pressures and well-separated levels of Sv_O2 (Table 2 and Figure E1 in the online supplement). All patients tolerated the three randomized study steps without any safety issue.

Table 2.

Extracorporeal Membrane Oxygenation Settings and Hemodynamics for the Three Study Steps

Variables	ECMO Blood Flow Target			P Value
	Low SvO2	Intermediate SvO2	High SvO2
ECMO settings
Blood flow rate, L/min	1.51 [1.16–1.94]*†	2.44 [2.03–2.93]‡	3.43 [3.01–3.75]	<0.001
BF/CO, %	20 ± 8*†	32 ± 9‡	44 ± 9	<0.001
ECMO drainage pressure, mm Hg	15 [4–25]	−5 [−12, 12]	−27 [−36, 9]	<0.001
BF/C2, ml/min/Fr2	2.56 [1.93–3.11]*†	3.90 [3.47–4.77]‡	5.49 [4.82–6.35]	<0.001
Hemodynamics
CO. L/min	9.2 [6.2–10.9]*†	8.3 [5.9–9.8]	7.9 [6.5–9.1]	0.014
CI, L/min/m2	4.7 [3.4–5.3]*†	4.1 [3.2–4.6]‡	3.9 [3.5–4.4]	0.014
Heart rate, beats/min	95 ± 18	87 ± 19	87 ± 20	0.186
Stroke volume, ml	94 ± 25	87 ± 4‡	87 ± 8‡	<0.01
PAPs, mm Hg	50 ± 12*†	46 ± 12‡	41 ± 8	<0.001
PAPm, mm Hg	34 ± 6*†	31 ± 6	30 ± 5	<0.001
PAPd, mm Hg	23 ± 5*	22 ± 5	21 ± 6	0.014
CVP, mm Hg	11 [8–14]	10 [9–13]	10 [8–12]	0.044
PAOP, mm Hg	13 [10–16]	13 [10–14]	12 [10–14]	0.062
PAPm – PAOP, mm Hg	21 ± 2	19 ± 2	18 ± 2	<0.001
Diastolic pulmonary gradient, mm Hg	10 [7–12]	8.5 [6–10]	8 [6–11]	0.132
Pulmonary artery RC time constant, ms	659 [582–815]	665 [575–858]	747 [564–955]	0.705
PVR, dyn · s · cm-5	180 [154–232]	170 [152–240]	159 [138–232]	0.064
PAC, ml/mm Hg	3.8 ± 1.5	4.0 ± 1.5	4.6 ± 1.9	<0.001
DO2, ml/min	1,110 ± 335	1,027 ± 266	1,041 ± 217	0.044

Definition of abbreviations: BF/CO = ratio between blood flow and cardiac output; BF/C² = ratio between blood flow and square of drainage cannula size; CI = cardiac index; DO₂ = oxygen delivery; PAOP = pulmonary artery occlusion pressure; PAPm = mean pulmonary artery pressure; PAPd-PAOP = diastolic pulmonary gradient; Pulmonary artery RC time constant = PA compliance × PVR; PAC = PA compliance; PVR = pulmonary vascular resistance.

Data are expressed as mean ± SD or as median [IQR], as appropriate. Comparisons were performed by ANOVA for repeated measures followed by the Holm-Sidak multiple comparisons test for normally distributed values (P value reported in the graph) and with Friedman ANOVA on ranks test for repeated measures followed by Tukey’s multiple comparison test for nonnormally distributed values.

^*P < 0.05, and P < 0.001 low versus high.

^† P < 0.05, and P < 0.001 low versus intermediate.

^‡ P < 0.05, and P < 0.001 intermediate versus high.

Blood Gases

Mixed venous oxygen tension (Pv_O2) increased at higher ECMO blood flow rates (42 ± 3 vs. 47 ± 3 vs. 53 ± 4 mm Hg for low, intermediate, and high blood flow, respectively; P < 0.001), whereas the $\dot{\mathrm{V}}$o₂ NL progressively decreased from the low to intermediate to high blood flow steps (199 [148–253] vs. 140 [124–195] vs. 107 [94–141] ml/min, respectively; P < 0.001). As expected, Pv_CO2 and Pa_CO2 also decreased at higher ECMO blood flow rates, albeit slightly, and pH values remained within the normal range (Table 3).

Table 3.

Blood Gases and Electrical Impedance Tomography Data on Ventilation, Perfusion, and $\dot{\mathrm{V}}$/$\dot{\mathrm{Q}}$ Mismatch for the Three Study Steps

Variables	ECMO Blood Flow Target			P Value
	Low SvO2	Intermediate SvO2	High SvO2
Blood gases
SvO2, %	73.9 ± 2.8*†	79.4 ± 2.7‡	86.7 ± 3.5	<0.001
PvO2, mm Hg	42 ± 3*†	47 ± 3‡	53 ± 4	<0.001
PvCO2, mm Hg	53 ± 5	50 ± 5	48 ± 5	<0.001
PaO2	63 [57–70]*†	70 [62–81]‡	83 [72–99]	<0.001
PaCO2, mm Hg	52 ± 6*†	49 ± 5‡	47 ± 6	<0.001
Intrapulmonary shunt, %	51.5 [41.9–64.6]	50.9 [40.2–60.2]‡	52.6 [45.3–64.9]	0.638
Arterial pH	7.39 ± 0.05	7.41 ± 0.05	7.42 ± 0.05	<0.001
EIT ventilation and perfusion, %
Ventral ventilation	28 [21–34]	29 [24–36]	29 [24–35]	0.511
Middle ventilation	60 [43–69]	56 [43–69]	58 [44–68]	0.443
Dorsal ventilation	11 [7–25]	16 [6–25]	15 [5–24]	0.511
Ventral perfusion	21 [15–29]	21 [15–28]	21 [15–27]	0.520
Middle perfusion	67 [53–69]	65 [56–67]	64 [54–70]	0.336
Dorsal perfusion	15 [10–20]	16 [13–22]	15 [13–19]	0.115
EIT $\dot{\mathrm{V}}$/$\dot{\mathrm{Q}}$ mismatch, %
Only ventilated units	10.7 [4.7–14.5]*	6.7 [3.8–10.9]	10.9 [5.5–19.6]	0.016
Only perfused units	16.7 [7.9–22.5]	17.3 [12.2–22.5]	16.6 [12.8–21.2]	0.212
Unmatched units	25.2 [10.3–27.7]	23.1 [19.6–31.8]	27.8 [24.7–31.3]	0.086

Definition of abbreviations: EIT = electrical impedance tomography; Pv_CO2 = mixed venous carbon dioxide tension; Pv_O2 = mixed venous oxygen tension; Sv_O2 = mixed venous blood oxygen saturation.

Data are expressed ad mean ± SD or as median [IQR], as appropriate. Comparisons were performed by ANOVA for repeated measures followed by the Holm-Sidak multiple comparisons test for normally distributed values (P value reported in the graph) and with Friedman ANOVA on ranks test for repeated measures followed by Tukey’s multiple comparison test for nonnormally distributed values.

^*P < 0.05, and P < 0.001 low versus high.

^† P < 0.05, and P < 0.001 low versus intermediate.

^‡ P < 0.05, and P < 0.001 intermediate versus high.

Pulmonary Circulation and Right Heart Hemodynamics

Systolic (50 ± 11 vs. 46 ± 12 vs. 41 ± 8 mm Hg; P < 0.001), diastolic (23 ± 5 vs. 22 ± 5 vs. 21 ± 6 mm Hg; P = 0.01), and mean (34 ± 6 vs. 31 ± 6 vs. 30 ± 5 mm Hg; P < 0.001) PAPs decreased from the low to intermediate to high blood flow steps (Figures 1A–1C). The decrease in pulmonary pressures was associated with a progressive parallel improvement of PA compliance (3.8 ± 1.5 vs. 4.0 ± 1.5 vs. 4.6 ± 1.9 ml/mm Hg for low, intermediate, and high steps, respectively; P < 0.001) (Figure 2A). The decrease in PVR also showed a trend between study steps (180 [154–232] vs. 170 [152–240] vs. 159 [138–232] dyn · s · cm⁻⁵ for low, intermediate, and high steps, respectively; P = 0.064).

Figure 1.

(A–C) Pulmonary arterial pressures (PAPs) at the three study steps. Systolic (A), mean (B), and diastolic (C) PAPs decrease from low to high extracorporeal membrane oxygenation (ECMO) blood flow and mixed venous saturation (Sv_O2). Data are expressed as scatterplots with bars and error bars (mean and standard deviation). Comparisons were performed by ANOVA for repeated measures for normally distributed values (P value reported in the graph) followed by the Holm-Sidak multiple comparisons test (*P < 0.05; low vs. high; ^§P < 0.05; low vs intermediate; ^P < 0.05; intermediate vs high). PAPs/m/d = pulmonary arterial pressure systolic/mean/diastolic.

Figure 2.

(A–C) Pulmonary arterial compliance (PA compliance), cardiac output, and right ventricular stroke work index at the three study steps. PA compliance increases, and cardiac output and right ventricular stroke work index (RVSWI) decrease, from low to high ECMO blood flow. Data are expressed as scatterplots with bars and error bars (mean and standard deviation). Comparisons were performed by ANOVA for repeated measures for normally distributed values (P value reported in the graph) followed by Holm-Sidak multiple comparisons test (*P < 0.05; low vs. high; ^§P < 0.05; low vs. intermediate; ^P < 0.05; intermediate vs. high).

In terms of right heart workload, CO decreased from low to intermediate to high blood flow steps (9.2 [6.2–10.9] vs. 8.3 [5.9–9.8] vs. 7.9 [6.5–9.1] L/min; P = 0.01) (Figure 2B), as did CVP, albeit slightly (11 [8–1] vs. 10 [9–13] vs. 10 [8–12] mm Hg; P = 0.044). The RVSWI also progressively decreased with higher blood flows (14.2 ± 4.4 vs. 12.2 ± 3.6 vs. 11.4 ± 3.2 g × m/beat/m² for low, intermediate, and high steps, respectively; P = 0.002) (Figure 2C). Hemodynamic variables during the three study steps are reported in Table 2.

$\dot{\mathbf{V}}$/$\dot{\mathbf{Q}}$ Mismatch

Changes in pulmonary hemodynamics and right heart workload were not associated with worsening $\dot{\mathrm{V}}$/$\dot{\mathrm{Q}}$ mismatch. Calculated intrapulmonary shunt did not differ for the three study steps (51.5 [41.9–64.6]% vs. 50.9 [40.2–60.2]% vs. 52.6 [45.3–64.9]% for low, intermediate, and high steps, respectively; P = 0.638), and more advanced regional EIT data showed similar stable values for the distribution of $\dot{\mathrm{V}}$/$\dot{\mathrm{Q}}$ and for the percentage of unmatched units (Table 3).

Physiological Interactions Underlying Main Cardiopulmonary Effects

When data from the three study steps were pooled together, PAP was inversely correlated with Pv_O2 (R² = 0.29; P < 0.001 for PAPs; R² = 0.16; P = 0.012 for PAPm) and directly with CO (nonlinear quadratic R² = 0.378; P < 0.007 for PAPs; R² = 0.287; P = 0.011 for PAPm) (Figures 3A and 3B). CO was inversely correlated with Pv_O2 and Pa_O2 (R² = 0.257; P = 0.031; and R² = 0.324; P = 0.05, respectively) (Figures 4A and 4B) and $\dot{\mathrm{V}}$o₂ NL showed an inverse correlation with Pv_O2, too (R² = 0.33; P < 0.0001) (Figure 4C). Right ventricular workload (RVSWI) decreased at higher Pv_O2 values (R² = 0.21; P = 0.004) (Figure 4D). PA compliance was directly correlated with Pv_O2 (R² = 0.198; P = 0.029) (Figure E1). The correlations between cardiopulmonary hemodynamic variables and Sv_O2, Pv_CO2, and ECMO blood flow rates were much weaker or inexistent (Tables E1–E4).

Figure 3.

(A and B) Relationships between systolic pulmonary artery pressure (PAPs) and mixed venous oxygen tension (Pv_O2) and between PAPs and cardiac output (CO). PAPs in relation to Pv_O2 (A). Values of Pv_O2 are grouped into quintiles, and the PAPs is presented as mean ± standard error. The regression line is computed on individual data points using linear mixed-effect models for repeated measures with patient as a random effect (R² and P values are reported in the graph). PAPs in relation to CO (B). Values of CO are grouped into quintiles, and the PAPs is presented as mean ± standard error. The best-fit curve demonstrated a nonlinear (quadratic) relationship between PAPs and CO (Akaike information criterion weight, 0.92; R² and P values are reported in the graph).

Figure 4.

(A–D) Correlations between CO and oxygenation (A and B). Correlations between mixed venous oxygen tension (Pv_O2) and $\dot{\mathrm{V}}$o₂ NL (C) and between Pv_O2 and RVSWI (D). CO is presented in relation to Pv_O2 (A) and Pa_O2 (B). Values of the x-axis variable (Pv_O2 and Pa_O2) are grouped into quintiles, and the CO is presented as mean ± standard error. Regression lines are computed on individual data points using linear mixed-effect models for repeated measures with patient as a random effect (R² and P values are reported in the graph). $\dot{\mathrm{V}}$o₂ NL and RVSWI are presented in relation to Pv_O2 (C and D, respectively). Values of Pv_O2 are grouped into quintiles, and the y-axis variable is presented as mean ± standard error. Regression line is computed on individual data points using linear mixed-effect models for repeated measures with patient as random effect (R² and P values reported in the graph). CO = cardiac output; RVSWI = right ventricular stroke work index; $\dot{\mathrm{V}}$o₂ NL = oxygen exchange through the natural lung.

Variables Associated with Decrease of the Right Heart Work

The decrease in RVSWI from low to high blood flow steps was directly correlated with PAPs (Figure 5A) and PAPm (R² = 0.39; P = 0.003; and R² = 0.33; P = 0.009, respectively), CO (R² = 0.66; P < 0.0001) (Figure 5B), and HR (R² = 0.33; P = 0.009) (Figure 5C) measured at the low blood flow step, likely meaning that patients with more severe pulmonary hypertension and higher sympathetic stimulation at the lowest ECMO blood flow may obtain greater unloading of the right heart at higher flow rates.

Figure 5.

(A–C) Correlations between the improvement in right ventricular stroke work index (RVSWI) between low and high blood flow and PAPs, CO, and heart rate measured at low blood flow and mixed venous oxygenation (Sv_O2) step. The decrease in RVSWI from low to high extracorporeal membrane oxygenation (ECMO) blood flow is presented in relation to PAPs (A), CO (B), and heart rate (C) at low ECMO blood flow step. Regression lines with intervals of confidence (R² and P values for Spearman correlation are reported in the graphs). CO = cardiac output; PAPs = systolic pulmonary arterial pressure.

Discussion

This study describes the physiologic effects induced by different ECMO blood flows and Sv_O2 across a spectrum of clinically acceptable values in patients with severe ARDS. The study’s main findings can be summarized as follows: Higher Sv_O2 achieved by increasing ECMO blood flow resulted in a reduction of CO and PAP. The main hemodynamic changes were also associated with improved PA compliance and reduced work of the right heart at higher Sv_O2 and ECMO blood flow. The changes in pulmonary circulation were not associated with evident and measurable changes in the distribution of ventilation and lung perfusion, nor were they associated with worsening of $\dot{\mathrm{V}}$/$\dot{\mathrm{Q}}$ mismatch. Finally, the protective effect of higher blood flow and Sv_O2 on the right heart was more pronounced in patients with higher PA pressure, CO, and HR at the lowest blood flow.

Early experimental studies investigated the cardiopulmonary effects of isolated changes of Sv_O2 at constant CO, with variable results depending on the target oxygenation values (Pv_O2 and/or Sv_O2) and on the lung condition. In healthy dogs, high Pv_O2 (>100 mm Hg) abolished the HPV in atelectatic lung regions, and this effect was associated with a dramatic decrease in PVR and an increase of intrapulmonary shunting (roughly from 20% to 50%) in comparison with lower Pv_O2 (∼30 mm Hg) (9). However, smaller changes in Pv_O2 (from ∼30 to ∼50 mm Hg) did not induce any of these effects. Changes in Sv_O2 from ∼30% to ∼60% caused minimal variations in intrapulmonary shunting (3–4%) with no changes in PVR in dogs with oleic acid–induced pulmonary edema (11).

Although providing a physiological basis for hypothesizing the effects of higher Sv_O2 on the pulmonary circulation, these experimental data did not allow us to precisely predict the cardiopulmonary effects of different levels of ECMO blood flow in patients with severe ARDS for three main reasons. First, modulation of ECMO blood flow could target values of Sv_O2 within a more limited range, both for technical limits and for patient safety (18). Second, in patients with severe ARDS, even if ECMO allows control of Sv_O2, modifying the amount of oxygen delivered by ECMO triggers complex hemodynamic responses that possibly include variations in CO (19). Third, the heterogeneity of ARDS pathophysiology (e.g., percentage of nonventilated lung, degree of activation of HPV, presence of pulmonary vascular occlusions) could lead to variable responses between patients.

Few clinical observations reported the cardiopulmonary effects of ECMO initiation in patients with ARDS (5, 6, 20). A decrease in PAP was observed within a few hours after ECMO initiation (5). However, the contribution of increased Sv_O2 could not be separated from the concomitant correction of hypercapnic acidosis (21) and severe hypoxemia and from the reduction of ventilation. It is reasonable to speculate that all these factors could contribute to the unloading of the right ventricle, which has been documented by echocardiographic examination in patients with severe ARDS during the first days of ECMO (6).

Our study investigated the physiological effects of different ECMO blood flow targeting varying Sv_O2 levels without changing ventilation. We explored the range of Sv_O2 that can be “safely” applied in the acute phase of severe ARDS on ECMO (i.e., 70–90%): Lower values could induce severe arterial desaturation, whereas higher values could be difficult to achieve for reasons that limit the increase of blood flow rate (smaller cannula size and relative or absolute hypovolemia) and/or of Sv_O2 (high CO, recirculation) (18).

The most physiologically and clinically relevant result of our study is probably the decrease in PA pressure and in CO at higher ECMO blood flow (Figure E2). Correlations between physiological variables were investigated to explore the interactions between all the physiological effects induced by higher ECMO blood flow. The correlation between higher Pv_O2/Pa_O2 and lower CO confirms that the goal of hemodynamic regulation is lowest cardiac workload to maintain an adequate $\dot{\mathrm{V}}$o₂/oxygen delivery relationship (22). Increasing Sv_O2 by increasing the ECMO flow rate decreased the need for native CO to maintain oxygen delivery. Moreover, the decrease in CO likely contributed to the decrease in PAP. However, a direct effect of Pv_O2 on pulmonary circulation cannot be excluded, and it is strongly suggested by the relationship between Pv_O2 and PAP and PA compliance. Relatively small changes in Pv_O2 (e.g., 40 to 55 mm Hg; see Figure 5) have a significant impact on pulmonary vascular function, even during protective ventilation. However, if the improvement in pulmonary vascular mechanics were the only effect of higher ECMO blood flow, we might have rather found stable PAP with increased CO or lower PAP with stable CO (15). Of note, decreased CO in our patients was not associated with higher CVP, suggesting changes in systemic venous resistance or in systemic vascular capacitance dampening excessive venous return.

Interestingly, in our study, the decrease in PAP at higher ECMO blood flow was associated with a slight decrease of PVR and more pronounced improvement of PA compliance. The pulmonary vasomotor tone consists not only of a resistive component but also a capacitive component, which is measured by the PA compliance (23). It is known that PA compliance is a sensitive index of pulmonary vascular dysfunction, especially when PVR is not markedly increased (16), as it was in our patients. However, given the dynamic nature of this measure (PA compliance is measured while blood flows into the arterial vasculature) and the short observation period (30 min), we cannot exclude that the increase in PA compliance could have been subtended by reduced vascular resistance or might just reflect the decrease of CO.

Not only do the hemodynamic effects of higher ECMO blood flow all contribute to the reduction in right heart workload, but each one could also have a lung-protective effect per se. In fact, higher pulmonary vascular pressures and/or flows have been shown to contribute to VILI in experimental models (24–26). Our study shows that ECMO could potentially impact the “vascular side” of VILI through the decrease in arterial pressures, the reduction in pulmonary blood flow (CO), and the improved distensibility of the pulmonary circulation (PA compliance).

The present study also shows that, within the explored range of Pv_O2/Sv_O2, changes in ECMO blood flow do not have an impact on $\dot{\mathrm{V}}$/$\dot{\mathrm{Q}}$ mismatch, as assessed both by calculation of intrapulmonary shunt and by the percentage of unmatched units measured by EIT. These data confirm earlier experiments which showed that the changes in intrapulmonary shunting are minimal for comparable small variations of Pv_O2/Sv_O2. Similarly to a recent study (12), we also confirmed that these variations do not affect the EIT-based distribution of perfusion. Finally, in line with previous findings that the pulmonary vasculature directly reacts to Pa_O2 level (which could vary between patients even for similar Sv_O2 values) (27), the cardiopulmonary physiological changes induced by different ECMO blood flow correlated better with Pv_O2 than with Sv_O2 levels and were more relevant between the low and intermediate steps.

There are currently no recommended indications for setting ECMO blood flow rate in patients with severe ARDS. Both low and high ECMO blood flow rates could provide acceptable arterial oxygenation while significantly decreasing the ventilatory load (15, 28). In clinical practice, the risks of high blood flow rate include complications due to the use of large cannulas, increased fluid balance to facilitate venous drainage, hemolysis, hemorrhage, and possibly prolonged ECMO duration (29). On the other hand, lower blood flow could be associated with less effective lung rest and metabolic activation decreasing ECMO efficiency (15). A recent analysis of data from the Extracorporeal Life Support Organization Registry shows that a higher ratio of blood flow rate to square cannula size is associated with improved survival in patients with severe ARDS receiving ECMO (30). Our results indicate that the degree of unloading of the right ventricle correlates with the degree of pulmonary arterial hypertension and with higher CO at lower blood flow, in line with the data of early physiologic studies (9). Alternatively, a less invasive bedside measure (tachycardia at low ECMO blood flow) could identify more “activated” patients likely to benefit from an increased blood flow rate.

Limitations

Our study has several limitations. The duration of each study step was relatively short, allowing the assessment of the acute physiological changes but not of the long-term effects of different levels of ECMO support. Hemodynamic instability was an exclusion criterion for safety reasons, and repeating the study in patients with shock and severe right heart dysfunction could be even more interesting. We chose to target Sv_O2 to titrate ECMO blood flow at each study step; although this allows a meaningful analysis of the physiological effects, Sv_O2 is not available in many patients receiving ECMO, and surrogate targets for clinical blood flow titration remain to be assessed. Moreover, the achievement of high Sv_O2 could be more challenging in patients in more severe condition. Measures of CO by PA catheter during ECMO might be influenced by recirculation of the saline bolus but have already been reported in previous studies (31). Moreover, in the presence of partial suctioning of the bolus by the ECMO system, CO could be overestimated. Given a higher risk for bolus recirculation at higher ECMO blood flow, this limitation could have led to an underestimation of the results of this study (i.e., lower CO at higher ECMO blood flow). The study population mostly included patients with COVID-19 ARDS (32), and our data are limited to “classic” venovenous ECMO configuration with double-site cannulation (no double-lumen single cannula). Finally, even if performed after a few days on ECMO, patients’ hypoxemia was already improving, as indicated by the ECMO blood flow/CO ratios (Table 2), and this could limit the generalizability of the study results.

Conclusions

The present study explored the hemodynamic and pulmonary vascular modifications induced by three different levels of ECMO blood flow rate, titrated to three “clinical” ranges of Sv_O2 values. Higher ECMO blood flow rate with a resultant higher Sv_O2 decreased the need for CO to maintain the same O₂ delivery. The decrease in CO consequently decreased PAP, improving the PA compliance and the unloading of the right heart without worsening $\dot{\mathrm{V}}$/$\dot{\mathrm{Q}}$ mismatch. Our results suggest that personalized titration of ECMO blood flow may decrease the risk of right heart failure associated with severe ARDS and mechanical ventilation.