Optimal oxygen target levels for critically ill patients have been discussed for decades and very low but also high PaO₂-levels were both associated with higher mortality [1, 2]. In patients with ARDS and refractory hypoxemic respiratory failure, venovenous Extracorporeal Membrane Oxygenation (VV-ECMO) can be implemented as a rescue therapy and blood flow of the VV-ECMO offers some control over the PaO₂-levels. While severe hyperoxemia in patients treated with venoarterial ECMO was found associated with increased mortality and adverse events, data on patients with ARDS and VV-ECMO treatment are scarce [3].

In n = 443 patients with ARDS (all non-COVID19) admitted to a tertiary ARDS referral center and treated with VV-ECMO, 88.328 arterial blood gas measurements taken during therapy with VV-ECMO were available and time-weighted averages of PaO₂ and FiO₂ over VV-ECMO therapy time were calculated [4]. Mean PaO₂ differed significantly between survivors and non-survivors (95.70 [87.46,109.20] vs. 84.27 [75.36, 96.99], p < 0.001) as did mean FiO₂ (0.54 [0.45–0.62] vs. 0.66 [0.54–0.81], p < 0.001). Simple logistic regression analysis was performed for mortality and mean PaO₂ and showed a significant association for mean PaO₂ and ICU mortality (per mmHg OR 0.99 [95%CI 0.98-1.00], p = 0.002). Considering a possible non-linear relation between mortality and PaO₂, a restricted cubic spline (RCS) for mean PaO₂ was utilized for a simple logistic regression model.

RCS analysis showed a U-shaped association of mortality and mean PaO₂ (per mmHg PaO₂: PaO₂ 69.70-90.24mmHg: OR 0.92 [95%CI 0.89–0.94], p < 0.001, PaO₂ 90.24-123.40mmHg: OR 1.09 [1.06–1.13], p < 0.001, Fig. 1A) with both, low and high PaO₂-levels associated with an increased risk of death. The U-shaped relationship between mean PaO₂ and mortality was similarly evident when only the initial period of ECMO therapy (up to 72 h) was analyzed to address the potential reverse causality of PaO₂-levels (Additional file 1).