Impact of cannula configuration on efficiency and mechanical ventilation intensity during venovenous ECMO

Abstract

Background

The efficiency of venovenous extracorporeal membrane oxygenation (VV ECMO) may be influenced by cannula configuration, potentially affecting oxygenation and the ability to apply ultra-protective ventilation in severe ARDS.

Materials and methods

This retrospective single-center study included all adult patients with severe ARDS supported with femoro-jugular VV ECMO between January 2013 and December 2022. Patients were divided into two groups according to cannula configuration: cross configuration (CC) and non-cross configuration (NCC). Ventilatory parameters, mechanical power, blood gases, and ECMO settings were analyzed at 1, 4, 12, and 24 h after ECMO initiation. A longitudinal analysis using generalized estimating equations adjusted for APACHE II score, COVID-19 ARDS, and baseline mechanical power was performed.

Results

Sixty-two patients were included (CC, n = 33; NCC, n = 29). Patients had a median age of 56 [50–63] years, 66% were male, and 58% had COVID-19 ARDS. Baseline respiratory parameters, ECMO settings, and cardiac output were comparable between groups. Compared with NCC, the CC group had significantly lower FiO₂ (21 [21–27] vs. 60 [40–100]%, p < 0.0001), and mechanical power (6.5 ± 2.4 vs. 12.3 ± 6.2 J/min, p < 0.0001) at 1 h, with sustained differences up to 24 h. In adjusted longitudinal analysis, the adjusted difference in mechanical power between CC and NCC was6.80 J/min (95% CI 3.82–9.78, p < 0.001) at 1 h and 6.69 J/min (95% CI 3.62–9.77, p < 0.001) at 24 h, with no significant interaction over time (p = 0.681). Sensitivity analyses including additional adjustment for prone positioning, baseline compliance, and time from intubation to ECMO initiation showed consistent results.

Conclusion

Cross cannula configuration is associated with reduced mechanical ventilation intensity during VV ECMO, independent of disease severity and ARDS etiology. These findings suggest improved ECMO efficiency and support further evaluation in prospective studies.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12890-026-04363-3.

Introduction

In severe acute respiratory distress syndrome (ARDS), patients may develop refractory hypoxemia despite optimal conventional management and require the support of venovenous extracorporeal membrane oxygenation (VV-ECMO) [1].

The femoro-jugular cannulation is the most commonly used configuration in VV ECMO. This technic depends on multiple factors to be effective and provide optimal systemic oxygenation. Among them, adequate extracorporeal blood flow (> 60% of cardiac output) [2] and minimal recirculation are critical [3]. Recirculation describes the phenomenon in which reinjected oxygenated blood is directly aspirated by the drainage cannula, bypassing the systemic circulation [4]. Therefore, this phenomenon can significantly decrease the efficiency of VV ECMO support. Recirculation has been shown to increase with ECMO flow. In a patient with no residual lung function and high cardiac output, recirculation could therefore be clinically relevant. Since recirculation may impair effective oxygen delivery, it may necessitate higher ventilatory settings, thereby increasing mechanical power and the risk of ventilator-induced lung injury.

A number of factors have been identified that influence the recirculation fraction, including the distance between the two cannulas in the traditional femoro-jugular configuration. Indeed, the closer the cannulas, the greater the recirculation [5]. To mitigate this phenomenon, it has been recommended to position the cannulas at a distance of 15 cm ⁵. However, in clinical practice, it is often difficult to maintain this distance consistently, especially during patient mobilization or prone positioning. During these maneuvers, recirculation may increase, thereby significantly reducing extracorporeal oxygenation efficiency. In such instances, urgent cannula repositioning may be required.

During the COVID-19 pandemic, ICU resource constraints highlighted the need for ECMO strategies that minimize manipulation and improve efficiency. In this context, we adopted for a different VV ECMO configuration with the objective of limiting recirculation and avoiding time-consuming cannula repositioning. The cross configuration, a distinct cannulation strategy for VV ECMO, where the tip of the draining femoral cannula is located in the superior vena cava and the tip of the reinjecting jugular cannula is located in the right atrium, has been described as feasible, safe and more effective regarding systemic oxygenation [6] (Figs. 1 and 2). However, comparative data between this configuration and the conventional femoro-jugular approach are scarce, particularly regarding recirculation and its impact on ventilatory management.

Fig. 1.

Schematic representation of the VV ECMO cannula and circuit in the cross-configuration. VV ECMO: Veno-venous extracorporeal membrane oxygenation

Fig. 2.

Schematic representation of the VV ECMO cannula in the SVC, RA and IVC in the cross-configuration. Blue arrows: Drainage blood flow, Red arrows: Reinjection blood flow. VV ECMO: Veno-venous extracorporeal membrane oxygenation, SVC: Superior vena cave, RA: Right atrium, IVC, Inferior vena cava, TV: Tricuspid valve

The aim of this retrospective study was to compare two VV ECMO cannula configurations. The first group included patients in whom the cross-configuration strategy was systematically applied. The second group was a historical cohort of VV ECMO patients with the traditional femoro-jugular configuration (non-crossed). The groups were compared in terms of oxygenation capacities, ventilatory parameters, blood gas analyses, and ECMO parameters. Beyond oxygenation efficiency, we sought to determine whether cannula configuration could influence mechanical ventilation intensity and mechanical power during the early phase of VV ECMO support. We hypothesized that a configuration associated with potentially related to reduced recirculation would allow a more effective implementation of ultra-protective ventilation strategies and that this association would persist over time.

Material and method

Study population

This retrospective study included adult patients (≥ 18 years) presenting with severe ARDS refractory to conventional management and supported with VV ECMO at Geneva University Hospitals (Switzerland) between January 2013 and December 2022. All patients had a femoro-jugular cannulation. Severe ARDS was defined according to the Berlin definition [7]. VV ECMO candidacy was determined according to institutional protocols based on EOLIA criteria, including severe hypoxemia despite optimized conventional management (PaO₂/FiO₂ < 50 mmHg for > 3 h or < 80 mmHg for > 6 h), or uncompensated hypercapnia with acidosis (pH < 7.25), despite lung-protective ventilation and adjunctive therapies when appropriate, including prone positioning and inhaled nitric oxide [8].

Approval for this study was obtained from the Cantonal Research Ethics Commission of the Republic and Canton of Geneva (2024 − 01266).

The patients were divided into 2 groups according to the configuration used: cross-configuration (CC) group and non-cross configuration (NCC) group. In the CC group, the tip of the drainage cannula was positioned in the superior vena cava and the tip of the reinjection cannula in the right atrium. In the NCC group, the tip of the drainage cannula was positioned in the inferior vena cava, at the level of the supra-hepatic veins and the tip of the reinjection cannula in the right atrium or the lower part of the superior vena cava. All VV ECMO implantations were performed at the patient’s bedside at the ICU, under vascular and transesophageal ultrasound guidance, using a percutaneous technique according to the Seldinger method. Cannulas sizes were chosen according to patient’s anatomical characteristics and their cardiac output, evaluated before implantation. The FmO₂ (fraction of oxygen on the ECMO membrane) was initially set to 100% and then adjusted to the arterial oxygen saturation (PaO₂) and the sweep gas flow was titrated according to the arterial partial pressure of CO₂ (PaCO₂). At the end of the procedure, the position of the cannulas was checked by transesophageal echocardiography and chest X-ray. An ultra-lung-protective ventilation strategy was applied to all our VV ECMO patients, regardless of ARDS etiology, in line with the EOLIA protocol and ELSO recommendations [8]. Specifically, we maintained tidal volume below 4 mL/kg, plateau pressure (Pplat) below 24 cmH₂O, set PEEP above 10 cmH₂O, kept the driving pressure below 14 cmH₂O, and targeted a respiratory rate of less than 15 breaths/min. Heparin was used for anticoagulation in all patients. Anticoagulation therapy was monitored with specific anti-Xa assays.

Patients meeting the following criteria were excluded from analysis: (1) ECMO duration for less than 24 h, (2) death within 24 h after the start of ECMO, (3) bridge to VAV or VA-ECMO.

Data source, collection, and outcomes

Data were extracted from computerized patient records which included demographics, severity scores (APACHE II and SAPS II), ARDS etiology, time from intubation to ECMO initiation, inhaled nitric oxide use (iNO), prone positioning use, hemodynamic status and ventilation parameters before ECMO initiation. Ventilation parameters, blood gas analyses and ECMO parameters at 1, 4, 12 and 24 h after ECMO initiation were analyzed for descriptive comparisons. Longitudinal modeling focused on 1 and 24 h, while additional sensitivity analyses included all available early time points (1, 4, 12, and 24 h). Mechanical power (MP) was calculated using the equation derives from a mathematical simplification of the original mechanical power formula [9]. Baseline respiratory system compliance was calculated as tidal volume divided by driving pressure, using ventilatory parameters recorded before ECMO initiation [10]. In addition, the following variables were extracted: Cardiac output (CO) measurements, assessed either invasively via Swan-Ganz catheter or by cardiac echocardiography at ECMO implantation, variation in ScvO₂ from baseline to 24 h following ECMO initiation duration of mechanical ventilation and ECMO, prone positioning use, incidence of tracheotomy insertion, ICU-length of stay, 90-day mortality and ECMO-related complications, including: (1) blood infection defined by the presence of positive blood cultures after the start of ECMO [11]; (2) cannula-related bleeding at the puncture site and/or requiring a blood transfusion; (3) deep vein thrombosis (DVT) induced by ECMO cannulas and confirmed by angiology examination; (4) death by accidental decannulation.

For longitudinal analyses, mechanical power measurements at 1 and 24 h were treated as repeated observations within the same patient to evaluate the persistence of the effect of cannula configuration over time.

Statistical analysis

Continuous variables were presented based on their distribution, with normally distributed data expressed as mean ± standard deviation (SD) and non-normally distributed data as median with interquartile range [IQR], while categorical variables as absolute and relative frequencies expressed as percentage (%). Normality was assessed using the D’Agostino–Pearson omnibus test. The two groups were compared with the student’s t test or the Mann-Whitney U test according to the distribution of the data for the continuous variables and with a chi-square test or a Fisher’s exact test for the categorical variables.

To evaluate whether the effect of cannula configuration persisted over time, a longitudinal analysis was performed using generalized estimating equations (GEE) with a Gaussian distribution and an exchangeable correlation structure to account for within-patient repeated measurements. The model included cannula configuration, time (1 h vs. 24 h), interaction term (configuration × time), APACHE II score, COVID-19 ARDS, and baseline mechanical power. Adjusted contrasts were calculated to estimate the effect of cannula configuration at each time point. Sensitivity analyses were performed by including all available early time points (1, 4, 12, and 24 h) and by additionally adjusting for prone positioning on ECMO, baseline respiratory system compliance, and time from intubation to ECMO initiation.

The working correlation structure was specified as exchangeable, assuming constant correlation between repeated measurements within each patient. Robust standard errors were computed to account for potential model misspecification. Model coefficients are reported as adjusted β estimates with 95% confidence intervals.

A two-tailed p value of less than 0.05 indicated statistical significance.

Univariable comparisons were performed using GraphPad Prism (GraphPad Software, San Diego, CA, USA). Multivariable and longitudinal analyses were performed using Python (statsmodels package, version 0.14.2).

Results

Between 2013 and 2022, 64 patients underwent VV ECMO. After exclusion of two patients (death within 48 h, n = 1; conversion to VAV-ECMO, n = 1), 62 patients were included in the analysis. The main characteristics of this population are provided in Table 1. Briefly, their mean age was 55 ± 11 years old, 41 patients (66%) were male, and the mean BMI was 28 ± 5. In the overall cohort, 36 (58%) had COVID-19 associated ARDS. Among the CC group, 97% presented COVID-19 associated ARDS, while Influenza was the most frequent cause of ARDS in the NCC group. SAPS 2 and APACHE II score were respectively 54 ± 18 and 26 ± 9. The time from intubation to ECMO initiation did not differ significantly between groups (2.0 [0.8–4.8] vs. 1.0 [0.2–4.1] days, p = 0.114). Before ECMO implantation, 36 (58%) patients had undergone prone positioning and 22 (35%) received iNO.

Table 1.

General charateristics

	All patients (62)	Cross (n = 33)	Non-Cross (n = 29)	p
Gender (Male), n (%)	41 (66)	22 (67)	19 (66)	0.924
Age, years	56 [50–63]	57 [52–63]	54 [44–62]	0.293
BMI	27 [24–31]	28 [26–32]	26 [23–31]	0.070
Diagnosis				< 0.001
Bacterial pneumonia	3 (5)	0	3 (10)
Influenzae	10 (16)	0	10 (34)
CMV/PJP	1 (2)	1 (3)	0
SARS-CoV-2	36 (58)	32 (97)	9 (31)
PE	1 (2)	0	1 (3)
Fibrosis	3 (5)	0	3 (10)
Pulmonary edema	3 (5)	0	3 (10)
SAPS II	54 ± 18	53 ± 17	56 ± 18	0.513
APACHE 2	28 [20–33]	25 [18–30]	30 [21–37]	0.010
Parameters before ECMO
Intubation to ECMO, days	1.01 [0.7–4.7]	2.0 [0.8–4.8]	1.0 [0.0-4.1]	0.114
PP, n (%)	36 (58)	27 (82)	9 (31)	< 0.001
iNO, n (%)	22 (35)	9 (27)	13 (45)	0.149
FiO2, %	100 [100–100]	100 [100–100]	100 [95–100]	0.530
PEEP, cmH2O	11 [10–13]	12 [10–13]	10 [8–14]	0.223
Vt, ml	420 [360–480]	430 [390–480]	360 [340–470]	0.014
RR, /min	25 [22–28]	24 [20–28]	26 [25–30]	0.026
MV, l/min	10.4 ± 2	6.8 ± 3.4	10 ± 2	0.336
Pplat, cmH2O	29 ± 4	29 ± 3	29 ± 4	0.732
Ppeak, cmH2O	36 ± 6	36 ± 6	36 ± 6	0.643
Baseline ΔP, cmH₂O	18 [16–20]	17 [16–19]	19 [15–22]	0.435
Baseline compliance, mL/cmH₂O	22.8 [17.5–28.9]	25.0 [21.5–29.3]	18.9 [16.3–25.3]	0.022
pH	7.29 ± 0.12	7.34 ± 0.09	7.23 ± 0.13	0.180
PaO2, kPa	7.6 [7.0-8.3]	7.2 [7.0-8.2]	8.0 [7.1–8.8]	0.202
PaCO2, kPa	8.3 ± 1.8	8.1 ± 1.7	8.64 ± 1.9	0.177
SaO2, %	89 [88–91]	89 [88–91]	89 [87–92]	0.507
PaO2/FiO2, %	8.3 ± 1.9	8 ± 1.6	8.7 ± 2.1	0.138
ScvO2, %	65 [60–71]	70 [60–75]	65 [59–68]	0.013
Cardiac Output, L/min	5.7 ± 1.4	5.5 ± 1.4	5.6 ± 1.5	0.949
Bicarbonates, mmol/l	28.2 ± 7.6	30.3 ± 5.4	25.7 ± 2.4	0.016
Lactate, mmol/l	1.5 [1.2–2.3]	1.5 [1.1-2]	1.5 [1.1–2.9]	0.326
Hb, g/l	113 ± 20	114 ± 20	111 ± 20	0.460
Norepinephrine, µg/kg/min	0.09 [0.02–0.19]	0.05 [0.01–0.16]	0.13 [0.05–0.37]	0.016
Dobutamine, nb of patients (%)	4 (6)	0 (0)	4 (14)	0.027

Data were analyzed according to distribution: continuous variables are presented as mean ± SD or median [IQR], and were compared using Student’s t-test or Mann-Whitney U test; categorical variables are presented as counts (%) and were compared using chi-square or Fisher’s exact test, as appropriate. Bold values indicate statistically significant associations or between-group differences (p < 0.05)

BMI Body Mass Index, CMV Cyto-Megalo Virus, PJP Pneumocystis jirovecii pneumonia, PE Pulmonary Embolism, ΔP Driving pressure, PP Prone Positioning, iNO Inhaled Nitrid Oxyde, FiO2 Inspired Oxygen Fraction, PEEP Positive End-Expiratory Pressure, Vt Tidal Volume, RR Respiratory Rate, MV Minute Ventilation, Pplat Plateau Pressure, Pmax Peak Pressure, PaO2 Arterial Oxygen partial Pressure, Arterial Carbone Dioxide partial Pressure, Hb Hemoglobin

Although the APACHE II score was significantly higher in the NCC group (median 30 [IQR 21–37] vs. 25 [IQR 18–30], p < 0.001), reflecting greater overall severity, baseline arterial oxygenation, PaCO₂, pH, and driving pressure were comparable between groups. Baseline respiratory system compliance, however, was significantly higher in the CC group than in the NCC group (25.0 [21.5–29.3] vs. 18.9 [16.3–25.3] mL/cmH₂O, p = 0.022). SOFA scores and lactate levels were also similar, suggesting comparable overall organ dysfunction and tissue hypoperfusion despite differences in APACHE II score. (Table 1).

There were no documented differences in the incidence of cardiogenic shock or myocarditis. Cardiac output (CO) measurements, assessed either invasively via Swan-Ganz catheter or by cardiac echocardiography at ECMO implantation, did not differ between groups (5.5 ± 1.4 vs. 5.6 ± 1.5 L/min, p = 0.949). ECMO flow, membrane settings, and cannula size were also comparable (Table 1). However, the cross femoro-jugular configuration (CC) was associated with reduced ScvO₂ variation during the first 24 h of support (1 [0–2]% vs. 8 [0–10]%, p < 0.001). Moreover, after one hour of ECMO support, patients in the CC group required significantly lower FiO₂ (21 [21–27] vs. 60 [40–100]%, p < 0.001), RR (11 ± 2 vs. 18 ± 5 /min, p < 0.001), MV (3 ± 1 vs. 5.5 ± 2.6 L/min, p < 0.001), Pplat (23 [22–24] vs. 25 [23–28] cmH₂O, p = 0.001) and MP (6.5 ± 2.4 vs. 12.3 ± 6.2 J/min, p < 0.001). These benefits were sustained at 4, 12 and 24 h in univariable analyses (Tables 2, 3 and 4, and 5).

Table 2.

Ventilatory, blood gas analyses and ECMO parameters at 1 h on ECMO

	Cross (n = 33)	Non-Cross (n = 29)	p
FiO2, %	21 [21–27]	60 [40–100]	< 0.001
PEEP, cmH2O	11 ± 2	10 ± 4	0.095
Vt, mL	265 ± 83	304 ± 115	0.134
RR, /min	11 ± 2	18 ± 5	< 0.001
MV, L/min	3 ± 1	5.5 ± 2.6	< 0.001
Pplat, cmH2O	23 [22–24]	25 [23–28]	0.001
Mechanical Power, J/min	6.5 ± 2.4	12.3 ± 6.2	< 0.001
pH	7.49 ± 0.08	7.38 ± 0.13	< 0.001
PaO2, kPa	9.5 [8.1–15.2]	9.6 [8.1–12.0}	0.587
PaCO2, kPa	4.6 [4.2–5.5]	5.0 [4.6–6.6]	0.014
SaO2, %	96 [93–98]	96 [91–98]	0.485
ECMO Blood Flow, L/min	4.8 ± 0.6	4.5 ± 0.8	0.062
ECMO Sweep Gas Flow, L/min	4 [3–5]	3.5 [2.3-4]	0.271
FmO2, %	100 [100–100]	100 [100–100]	0.359

Data were analyzed according to distribution: continuous variables are presented as mean ± SD or median [IQR], and were compared using Student’s t-test or Mann-Whitney U test; categorical variables are presented as counts (%) and were compared using chi-square or Fisher’s exact test, as appropriate. Bold values indicate statistically significant associations or between-group differences (p < 0.05)

FiO₂ Inspired Oxygen Fraction, PEEP Positive End-Expiratory Pressure, Vt Tidal Volume, RR Respiratory Rate, MV Minute Ventilation, P_plat Plateau Pressure, PaO₂ Arterial Oxygen partial Pressure, PaCO₂ Arterial Carbone Dioxide Partial Pressure, SaO₂ Arterial Saturation in Oxygen, FmO₂ ECMO Membrane Oxygen Fraction

Table 3.

Ventilatory, blood gas analyses and ECMO parameters at 4 h on ECMO

	Cross (n = 33)	Non-Cross (n = 29)	p
FiO2, %	21 [21–23]	51 [40–80]	< 0.001
PEEP, cmH2O	11 ± 2	10 ± 4	0.069
Vt, mL	254 ± 85	289 ± 106	0.151
RR, /min	11 ± 2	18 ± 6	< 0.001
MV, L/min	2.8 [1.9–3.8]	4.3 [3.0-7.4]	< 0.001
Pplat, cmH2O	23 [22–24]	25 [23–27]	< 0.001
Mechanical Power, J/min	6.1 ± 2.4	12.3 ± 7.4	< 0.001
pH	7.46 ± 0.06	7.39 ± 0.12	0.002
PaO2, kPa	9.3 [8.4–10.5]	9.2 [8.2–11.1]	0.886
PaCO2, kPa	5.1 ± 0.8	5.56 ± 1.3	0.081
SaO2, %	94 ± 3	94 ± 2	0.372
ECMO Blood Flow, L/min	4.7 ± 0.6	4.5 ± 0.9	0.266
ECMO Sweep Gas Flow, L/min	4 [3–5]	3.5 [3-4.3]	0.469
FmO2, %	100 [100–100]	100 [100–100]	0.359

Data were analyzed according to distribution: continuous variables are presented as mean ± SD or median [IQR], and were compared using Student’s t-test or Mann-Whitney U test; categorical variables are presented as counts (%) and were compared using chi-square or Fisher’s exact test, as appropriate. Bold values indicate statistically significant associations or between-group differences (p < 0.05)

FiO₂ Inspired Oxygen Fraction, PEEP Positive End-Expiratory Pressure, Vt Tidal Volume, RR Respiratory Rate, MV Minute Ventilation, P_plat Plateau Pressure, PaO₂ Arterial Oxygen partial Pressure, PaCO₂ Arterial Carbone Dioxide Partial Pressure, SaO₂ Arterial Saturation in Oxygen, FmO₂ ECMO Membrane Oxygen Fraction

Table 4.

Ventilatory, blood gas analyses and ECMO parameters at 12 h on ECMO

	Cross (n = 33)	Non-Cross (n = 29)	p
FiO2, %	21 [21–21]	40 [33–52]	< 0.001
PEEP, cmH2O	12 [10–12]	10 [7–12]	0.120
Vt, mL	254 ± 87	272 ± 106	0.480
RR, /min	10 [10–12]	15 [14–21]	< 0.001
MV, L/min	2.8 [2.1–3.7]	4.2 [3.1–5.9]	< 0.001
Pplat, cmH2O	23 [22–24]	24 [22–26]	0.009
Mechanical Power, J/min	6.0 [4.4–8.2]	9.4 [6.3–12.3]	< 0.001
pH	7.45 ± 0.05	7.40 ± 0.08	0.003
PaO2, kPa	9.4 [8.5–10.6]	9.2 [8.3–11.3]	0.997
PaCO2, kPa	5.2 [4.8–5.5]	5.1 [4.7–5.8]	0.847
SaO2, %	95 [93–97]	96 [93–98]	0.517
ECMO Blood Flow, L/min	4.7 ± 0.7	4.3 ± 0.9	0.091
ECMO Sweep Gas Flow, L/min	4 [3–5]	4 [3–5]	0.582
FmO2, %	100 [100–100]	100 [100–100]	0.359

Data were analyzed according to distribution: continuous variables are presented as mean ± SD or median [IQR], and were compared using Student’s t-test or Mann-Whitney U test; categorical variables are presented as counts (%) and were compared using chi-square or Fisher’s exact test, as appropriate. Bold values indicate statistically significant associations or between-group differences (p < 0.05)

FiO₂ Inspired Oxygen Fraction, PEEP Positive End-Expiratory Pressure, Vt Tidal Volume, RR Respiratory Rate, MV Minute Ventilation, P_plat Plateau Pressure, PaO₂ Arterial Oxygen partial Pressure, PaCO₂ Arterial Carbone Dioxide Partial Pressure, SaO₂ Arterial Saturation in Oxygen, FmO₂ ECMO Membrane Oxygen Fraction

Table 5.

Ventilatory, blood gas analyses and ECMO parameters at 24 h on ECMO

	Cross (n = 33)	Non-Cross (n = 29)	p
FiO2, %	21 [21–21]	40 [30–55]	< 0.001
PEEP, cmH2O	12 [10–13]	10 [6–12]	0.073
Vt, mL	251 ± 89	277 ± 119	0.332
RR, /min	10 [10–12]	15 [14–22]	< 0.001
MV, L/min	2.8 [2.0-3.5]	4.8 [2.9–6.7]	< 0.001
Pplat, cmH2O	23 [22–24]	24 [23–27]	0.004
Mechanical Power, J/min	6.0 [4.4–7.6]	10.4 [5.6–14.5]	< 0.001
pH	7.46 ± 0.05	7.41 ± 0.07	0.002
PaO2, kPa	9.3 [8.7–11.0]	8.9 [8.4–10.3]	0.220
PaCO2, kPa	5.2 [4.7–5.7]	5.2 [4.6–5.9]	0.703
SaO2, %	94 [93–96]	94 [93–96]	0.637
ScvO2, %	70 ± 8	72 ± 6.2	0.307
ScvO2 variation, %	1.0 [0.0-2.2]	8.0 [0.0–10.0]	< 0.001
ECMO Blood Flow, L/min	4.8 ± 0.7	4.4 ± 0.9	0.075
ECMO Sweep Gas Flow, L/min	4.5 ± 1.7	4 ± 1.5	0.195
FmO2, %	100 [100–100]	100 [100–100]	0.359

Data were analyzed according to distribution: continuous variables are presented as mean ± SD or median [IQR], and were compared using Student’s t-test or Mann-Whitney U test; categorical variables are presented as counts (%) and were compared using chi-square or Fisher’s exact test, as appropriate

FiO₂ Inspired Oxygen Fraction, PEEP Positive End-Expiratory Pressure, Vt Tidal Volume, RR Respiratory Rate, MV Minute Ventilation, P_plat Plateau Pressure, PaO₂ Arterial Oxygen partial Pressure, PaCO₂ Arterial Carbone Dioxide Partial Pressure, SaO₂ Arterial Saturation in Oxygen, ScvO₂ Central venous oxygen saturation, ΔScvO₂ Central venous oxygen saturation difference between pre-ECMO and after 24 h on ECMO, FmO₂ ECMO Membrane Oxygen Fraction

In the longitudinal analysis using generalized estimating equations adjusted for APACHE II score, COVID-19 ARDS, baseline mechanical power, and accounting for within-patient correlation, cross configuration remained independently associated with lower mechanical power over time. The adjusted difference between CC and NCC was 6.80 J/min (95% CI 3.82–9.78, p < 0.001) at 1 h and 6.69 J/min (95% CI 3.62–9.77, p < 0.001) at 24 h (Table 6). The interaction term between configuration and time was not statistically significant (p = 0.681), indicating that the magnitude of the effect did not significantly change between 1 and 24 h. COVID-19 ARDS was independently associated with lower mechanical power (β = −6.76, 95% CI − 9.36 to − 4.16, p < 0.001), whereas APACHE II score and baseline mechanical power were not significantly associated with mechanical power in the adjusted model.

Table 6.

Longitudinal association between cannula configuration and mechanical power at 1 and 24 h. Model: GEE (Gaussian), exchangeable correlation structure, adjusted for APACHE II, COVID-19 ARDS and baseline MP (n = 62)

Variable / Contrast	Adjusted β (95% CI)	SE	p
Cross vs. Non-cross at 1 h	6.80 (3.82 to 9.78)	1.52	< 0.001
Cross vs. Non-cross at 24 h	6.69 (3.62 to 9.77)	1.57	< 0.001
Cross × Time interaction	−0.11 (− 0.61 to 0.40)	0.26	0.681
APACHE II score	−0.12 (− 0.28 to 0.03)	0.08	0.119
COVID-19 ARDS	−6.76 (− 9.36 to − 4.16)	1.33	< 0.001
Baseline Mechanical Power	0.10 (− 0.05 to 0.25)	0.08	0.200

β coefficients represent adjusted mean differences in mechanical power (J/min)

The interaction term (Cross × Time) tests whether the effect of cannula configuration changes between 1 and 24 h. Robust standard errors were used

Additional sensitivity analyses including all available time points (1, 4, 12, and 24 h) and further adjustment for prone positioning on ECMO, baseline respiratory system compliance, and time from ICU admission to ECMO initiation are reported in Supplementary Table S1

ARDS Acute respiratory distress syndrome, CI Confidence interval, COVID-19 Coronavirus disease 2019, ECMO Extracorporeal membrane oxygenation, GEE Generalized estimating equations, MP Mechanical power, SE Standard error

Sensitivity analyses including all available early time points (1, 4, 12, and 24 h) showed consistent results. In the fully adjusted sensitivity model additionally including prone positioning on ECMO, baseline respiratory system compliance, and time from intubation to ECMO initiation, mechanical power remained significantly higher in the NCC group at 1 h (adjusted NCC–CC difference 3.38 J/min, 95% CI 0.58–6.19, p = 0.018), 4 h (3.87 J/min, 95% CI 1.25–6.49, p = 0.004), and 24 h (2.73 J/min, 95% CI 0.17–5.28, p = 0.037), with a similar trend at 12 h (1.83 J/min, 95% CI − 0.41 to 4.07, p = 0.110) (Supplementary Table S1).

Prone position was used more frequently in the CC group (14 (42%) vs. 3 (10%), p = 0.005). ECMO duration, mechanical ventilation duration, and ICU length of stay were significantly longer in the CC group (25 [13–36] vs. 7 [5–12] days, p < 0.001; 34 [20–55] vs. 12 [7–30] days, p < 0.001; and 34 [20–56] vs. 15 [8–33] days, p < 0.001, respectively). ECMO-related complications and 90-day mortality did not differ between groups (Table 7).

Table 7.

Outcomes

	Cross (n = 33)	Non-Cross (n = 29)	p
Drainage cannula, French	26 [25-29]	25 [25-25]	0.081
Reinjection cannula, French	20 [19–21]	19 [19–20]	0.149
Time on ECMO, days	25 [13–36]	7 [5–12]	< 0.001
Time on mechanical ventilation, days	34 [20–55]	12 [7–30]	< 0.001
ICU length of stay, days	34 [20–56]	15 [8–33]	< 0.001
Tracheotomy, n	14 (42)	7 (24)	0.157
Prone positioning on ECMO, n	14 (42)	3 (10)	0.005
ECMO-related complications
Bleeding, n	6 (18)	3 (10)	0.490
Infection, n	3 (9)	2 (7)	1.000
Venous thrombosis, n	6 (18)	4 (14)	0.740
Decannulation, n	1 (3)	0	1.000
Death, n	16 (48)	12 (41)	0.610

Data were analyzed according to distribution: continuous variables are presented as mean ± SD or median [IQR], and were compared using Student’s t-test or Mann-Whitney U test; categorical variables are presented as counts (%) and were compared using chi-square or Fisher’s exact test, as appropriate

Discussion

This retrospective study compared two femoro-jugular VV ECMO configurations and found that the cross configuration significantly reduced mechanical ventilation intensity and FiO₂ requirements compared to the traditional configuration, despite comparable ECMO blood flow, membrane oxygenation settings, and cardiac output.

One of the primary objectives of VV ECMO support in ARDS is to enable “lung rest”, minimizing alveolar strain and stress during mechanical ventilation to the greatest extent. The concept of mechanical power, the energy transferred from the ventilator to the respiratory system per unit time, has gained considerable attention [9] and is associated with mortality in ARDS [12, 13]. Although optimal mechanical ventilation settings on ECMO remain unknown, international guidelines recommend reducing mechanical power through an ultra-protective ventilation strategy (plateau pressure ≤ 24 cmH₂O; driving pressure ≤ 14 cmH₂O, PEEP > 10 cmH₂O, RR 4–15/min), aimed at improving lung recovery [9, 12–15]. Of note, increasing respiratory rate to improve gas exchange during VV ECMO, as observed in both the LIFEGUARDS study and the EOLIA trial, may be counterproductive, as it increases mechanical power, and may contribute to ventilator-induced lung injury despite extracorporeal support [8, 16].

Achieving this requires VV ECMO to efficiently meet systemic oxygenation demands. Our findings show that the cross configuration reduced FiO₂ and all components of mechanical power (respiratory rate, tidal volume, and plateau pressure), consistent with consistent with improved ECMO efficiency, potentially related to reduced recirculation. Importantly, baseline respiratory parameters, including severity of hypoxemia, hypercapnia, and respiratory compliance, were comparable between groups. In addition, an ultra-protective ventilation strategy based on EOLIA protocols was consistently applied in the entire cohort. ECMO blood flow, membrane settings, cardiac output, and the ratio of ECMO flow to cardiac output were also similar, reducing potential confounding factors. To support our inference, ScvO₂ variation, an indirect marker of recirculation [17], was greater in the non-cross group.

The longitudinal analysis further strengthened these findings. In the generalized estimating equation model adjusted for APACHE II score, COVID-19 ARDS, and baseline mechanical power, the cross configuration remained independently associated with lower mechanical power at both 1 and 24 h. The adjusted difference between CC and NCC was 6.80 J/min at 1 h and 6.69 J/min at 24 h, with no significant interaction between configuration and time. The magnitude of the effect remained stable over the first 24 h, suggesting a sustained association rather than a transient post-implantation effect. Sensitivity analyses including prone positioning on ECMO, baseline respiratory system compliance, and time from intubation to ECMO initiation showed that the association between cannula configuration and mechanical power remained significant at 1, 4, and 24 h.

Interestingly, COVID-19 ARDS was independently associated with lower mechanical power in the adjusted model. This may reflect evolving ventilatory practices during the pandemic, with broader implementation of ultra-protective ventilation strategies. The effect of cannula configuration remained significant after adjustment for COVID-19 status, indicating that the observed benefit cannot be solely attributed to differences in ARDS etiology or temporal practice changes.

Recirculation in VV ECMO is a multifactorial phenomenon influenced by cannula position, cannula configuration, ECMO flow, and patient cardiac output [18]. The cross configuration may reduce recirculation by improving drainage of desaturated blood and directing reinfused oxygenated blood more effectively toward the tricuspid valve, minimizing venous mixing. This contrasts with the traditional femoro-jugular configuration, where mispositioning of the drainage cannula, too low in the IVC or too high at the RA, can result in suboptimal drainage or increased recirculation, respectively [5, 19, 20].

Bonacchi et al. first described the cross configuration to improve oxygenation during VV ECMO support in ARDS and reported improved oxygenation efficiency and reduced mechanical ventilation intensity [6]. Although their original description relied on a modified reinjection cannula, we used standard commercially available cannulas. Our findings suggest that some of the physiological benefits associated with the cross configuration may be reproducible in routine clinical practice without the need for device modification.

The presence of two large-bore cannulas in the SVC may raise concerns about thrombosis risk. However, systematic post-decannulation ultrasonography and CT scans did not show an increased incidence of thrombosis. Despite more frequent prone positioning in the CC group, complication rates, including accidental decannulation and cannula repositioning, were similar between groups.

Patients in the cross-configuration group had longer ECMO duration, mechanical ventilation duration, and ICU length of stay. This finding should be interpreted in light of the higher proportion of COVID-19 ARDS in this group. COVID-19 ARDS may differ from other ARDS etiologies in terms of lung mechanics, inflammatory profile, duration of respiratory failure, and response to ventilatory strategies [21]. Moreover, COVID-19-related ARDS requiring ECMO has been associated with longer ECMO runs compared with influenza- or non-COVID-related ARDS [22]. Therefore, the longer support duration and ICU stay observed in the cross-configuration group may partly reflect differences in ARDS etiology rather than the ECMO configuration itself.

This study has several limitations. First, it is a retrospective, single-center study, which may inherently be subject to residual confounding factors unrelated to cannula configuration. Second, it must be acknowledged that recirculation was not directly measured. Although ScvO₂ variation may provide indirect information on recirculation, it remains a surrogate marker influenced by sampling site, venous mixing, cannula position, and systemic oxygen extraction. Therefore, a reduction in recirculation could only be indirectly inferred and should be interpreted cautiously [23]. Third, although the sample size was modest, the longitudinal model yielded consistent effect estimates across time points. Nevertheless, the study may have been underpowered to detect subtle interaction effects or smaller between-group differences in secondary outcomes. Fourth, there was a marked imbalance in ARDS etiology between groups, which may have influenced lung mechanics, ventilatory management, and ECMO duration. Although COVID-19 status was included in the multivariable longitudinal model, residual confounding related to temporal changes in clinical practice cannot be excluded. In fact, several retrospective studies have reported similar findings in COVID-19 ARDS patients supported with VV ECMO [22, 24, 25].

Conclusions

In ARDS patients supported with femoro-jugular VV ECMO, the cross configuration was safe, feasible, efficient, and associated with significant and sustained reductions in mechanical ventilation intensity. In longitudinal analyses, this association persisted during the first 24 h of support and remained significant after adjustment for disease severity and ARDS etiology. Although recirculation was not directly measured, our findings are consistent with improved ECMO efficiency, potentially related to reduced recirculation. Prospective studies with direct quantification of recirculation are warranted to confirm these observations.

Abbreviations

APACHE II

Acute Physiology and Chronic Health Evaluation II

ARDS

Acute respiratory distress syndrome

anti-Xa

Anti-factor Xa

BMI

Body mass index

CC

Cross configuration

CI

Confidence interval

CO

Cardiac output

COVID-19

Coronavirus disease 2019

CT

Computed tomography

DVT

Deep vein thrombosis

ECMO

Extracorporeal membrane oxygenation

ELSO

Extracorporeal Life Support Organization

EOLIA

Extracorporeal Membrane Oxygenation for Severe Acute Respiratory Distress Syndrome

FiO₂

Fraction of inspired oxygen

FmO₂

Fraction of oxygen on the ECMO membrane

GEE

Generalized estimating equations

ICU

Intensive care unit

iNO

Inhaled nitric oxide

IQR

Interquartile range

IVC

Inferior vena cava

MP

Mechanical power

MV

Minute ventilation

NCC

Non-cross configuration

NO

Nitric oxide

PaCO₂

Arterial partial pressure of carbon dioxide

PaO₂

Arterial partial pressure of oxygen

PEEP

Positive end-expiratory pressure

Pplat

Plateau pressure

RA

Right atrium

RR

Respiratory rate

SAPS II

Simplified Acute Physiology Score II

ScvO₂

Central venous oxygen saturation

SD

Standard deviation

SOFA

Sequential Organ Failure Assessment

SVC

Superior vena cava

TV

Tricuspid valve

VA-ECMO

Veno-arterial extracorporeal membrane oxygenation

VAV-ECMO

Veno-arterio-venous extracorporeal membrane oxygenation

VV-ECMO

Veno-venous extracorporeal membrane oxygenation

Authors’ contributions

BA, CS, KB and RG conceived and designed the study. CS and RG collected the data, BA, CS and RG performed the analysis. BA, CS, KB and RG wrote the first draft. CL and CB made substantial contributions to revise the manuscript and provided intellectual input. All authors approved the final version of the manuscript.

Funding

None.

Data availability

The datasets generated and/or analyzed during the current study are not publicly available due to institutional and national data protection regulations but are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the Declaration of Helsinki. The study protocol was approved by the local ethics committee (BASEC No. 2024 − 01266). Due to the retrospective observational design, the requirement for informed consent was waived in accordance with Swiss legislation and institutional policies.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Conflict of interest

None of the authors has any conflict of interest to disclose.