“High blood flow” versus “low blood flow” ECCO₂R: how to classify—author’s reply

We sincerely thank Wang et al. [1] for their correspondence and thoughtful consideration of our article published in Critical Care [2]. We appreciate their acknowledgment of the valuable insights into lung protection provided by our work. Our study demonstrated that blood flow plays a predominant role in determining the feasibility of ultra-low tidal volume ventilation (≤ 3 mL/kg predicted body weight). We fully agree on the need for better standardization criteria for defining high- and low-blood-flow ECCO₂R devices in the future [3]. As noted by the authors, an internationally validated definition of these terms is still lacking [4].

Distinguishing high- and low-blood-flow devices

As underlined by the authors of the correspondence there is some threshold variability in defining high- and low-blood-flow devices. In the Supernova study [5] devices were categorized as follows: lower extraction (Hemolung Respiratory Assist System, blood flow between 300 and 500 mL/mn) and higher extraction (iLA activve, Novalung, and Cardiohelp, Gettinge, blood flow between 800 and 1000 mL/mn). We adopted the same definition for low-blood-flow devices in our study. Regarding high-blood-flow devices, the Supernova study specified: “Blood flow with the iLA activve (Novalung) and Cardiohelp HLS 5.0 (Getinge) can range between 0.5 and 4.5 L/min but was limited by study protocol to 800–1000 mL/min.” Similarly, in our study protocol, we set limits for high blood flow rates (albeit slightly higher than in the Supernova study). This ensured no overlap between low- and high-blood-flow rates, avoiding the ambiguities of a “gray area” where classification might become problematic. Indeed, devices using almost similar blood flow could have been more difficult to classify.

The authors of the correspondence mention the risk of misclassification of devices. Interestingly, had we adjusted the cut-off between high and low blood flow to 800 mL/mn our results would have remained unchanged. Indeed, under real-life conditions, blood flow rates in our study were either above 1000 mL/min or below 500 mL/min. Consequently, whether the threshold was set at 800 or 1000 mL/min would not have affected our findings. While several cut-off values could have been chosen, in the absence of a consensus, we adopted this approach. Recent clinical and experimental studies have highlighted that the combination of blood flow and sweep gas flow is the key determinant of CO₂ removal efficiency [6, 7]. With sweep gas flow reaching 10 to 15 L/min (standardize to 10 L/min in our study), blood flow remains the primary modifiable factor for enhanced CO₂ removal, as we demonstrated.

In Vitro vs. In Vivo membrane performance

Wang et al. [1] suggest prioritizing membrane performance. However, our study was not designed to assess the in vitro capabilities of different membranes and/or devices. Instead, it aimed to analyze the factors contributing to the failure of ultraprotective ventilation in a clinical, rather than an experimental, setting. Indeed, in vitro performance does not always translate to in vivo efficacy, particularly in critically ill patients with multiple organ failures. Even a membrane with optimal purification capacity may be less efficient if blood flow is low or subject to fluctuations. Theoretical blood flow rates should not be confused with actual blood flow rates measured under real conditions. The latter depend on multiple factors, including anticoagulation, hemodynamic status, cannula size, treatment interruptions, and circuit or membrane thrombosis. An observation applicable to all types of extracorporeal support: kidney [8], liver [9, 10], etc. Ultimately, optimal in vitro membrane performance does not necessarily equate to clinical efficacy in achieving ultra-low tidal volume ventilation.

The authors [1] also suggest that some low-blood-flow membranes may surpass the efficiency of high-blood-flow membranes. However, we did not find any solid published data to substantiate this hypothesis. In our study, the membrane surface areas were significantly different between the two groups. Membrane in the high blood flow group were systematically larger (1.3 m² vs. 0.32–0.8 m²). It is highly unlikely that a smaller membrane combined with lower blood flow would result in superior CO₂ extraction capacity, as previously confirmed in a secondary analysis of the Supernova study [11, 12].

Device heterogeneity

The authors also raise concerns about device heterogeneity when using a high- and low-blood-flow classification [1]. As they as they rightly point out high-blood-flow devices may combine very different devices (pumpless arteriovenous devices and veno-venous devices). However, we did not include pumpless arteriovenous devices in our study. In our work, only one device was used in the high-blood-flow category. That said, we acknowledge that some heterogeneity exists among low-blood-flow devices.

Improving future trials and clinical practice

We agree with the authors that CO₂ extraction rate could serve as a more precise classification method for low- and high-CO₂ extraction devices. However, actual CO₂ extraction should be measured rather than potential extraction. Normalizing this data based on membrane surface area is relevant only if the objective is membrane performance assessment. Keeping in mind that, regardless of membrane efficiency, blood flow (and sweep gas flow) will always be the primary determinants of the actual clinical efficacy of ECCO₂R therapy. We also support other suggestions made by the authors, including the potential benefits of device-specific subgroups. However, this would require larger patients cohorts in each subgroup to achieve statistical significance. Finally, integrating dynamic performance assessments into future trial designs would further enhance our understanding and clinical application of ECCO₂R therapy.