Oxygenation management during veno-arterial ECMO support for cardiogenic shock: a multicentric retrospective cohort study

In this multicentric cohort study of cardiogenic shock supported by VA ECMO, results may be summarized as follow: (1) median value of F_SO_{2 mean (day 1−7)} was 70 [57; 79] %; (2) P_POSTO₂ (a surrogate of lower body PO₂) monitoring was infrequent and revealed significant hyperoxemia, and (3) higher F_SO_{2 mean (day 1−3)} and right radial P_aO_{2 mean (day 1−3)} were independently associated with in-ICU mortality.

Published data on F_SO₂ management are scarce. Most of studies report either only punctual F_SO₂ values [17, 18], or protocol for the initial F_SO₂ setting [1, 2, 5, 6, 19]. In a retrospective monocentric study on 54 patients, median F_SO₂ decreased from 80% [70–100] at baseline, to 70% [65–80] at 48 h [20]. In another study, mean F_SO₂ was around 80% between day 1 and day 10 of ECMO support [21]. In the study by Moussa et al., mean F_SO₂ ranged from 50% to 70% between baseline and day 2 [12]. Our data are in line with the previous studies, showing that F_SO₂ is usually set around 70%. Interestingly, we found that F_SO₂ varied inversely with ECMO center case volume. Because right radial P_aO₂ was also higher in low volume centers, we can hypothesize that low volume centers were more tolerant with hyperoxemia, leading to less down titration of F_SO₂ compared to most experienced centers. However, such observation needs to be interpreted cautiously because only 23% and 14% of patients were enrolled in low and high-volume centers. Then, this difference may reflect practices of a very few centers rather than a real case volume effect.

Using the threshold of 300 mmHg for the right radial P_aO₂, we found a quite low prevalence of severe hyperoxemia (12%). Our results are concordant with those of Jentzer et al. who found a 19.8% prevalence of severe hyperoxemia 24 h after ECMO start for cardiogenic shock [8]. This observation should probably be linked to an early F_SO₂ titration, as only 17% of patients of our cohort experienced at least one day with a F_SO₂ set at 100%.

Although P_POSTO₂ was a main objective of the study, we found that P_POSTO₂ monitoring was infrequent. It is important to distinguish P_POSTO₂ monitoring, i.e. PO₂ monitoring of the blood reinjected in the abdominal aorta which can be assimilated to the hepato-splanchnic PO₂ monitoring, and functional membrane assessment evaluated once daily by increasing F_SO₂ transiently at 100% to determine the gas transfer capacity of the membrane. Some clinicians may consider P_POSTO₂ monitoring useless, as right radial P_aO₂ may be sufficient to detect ECMO-induced hyperoxemia, and differential hypoxemia. In the subgroup of patients with available data, we found significant hyperoxemia with a median P_POSTO₂ value of 198 mmHg, and two third of values above the ELSO recommended target of 150 mmHg [14]. Previously, only one study on 45 patients has reported data regarding to P_POSTO₂. They found that median P_POSTO₂ decreased from 301 [215–386] mmHg at baseline, to 140 [78–220] mmHg at H48. In this study, only one third of P_POSTO₂ values were below 150 mmHg [20]. A possible reason for this tolerance with hyperoxemia might be the fear of unrecognized hypoxemia of the lower part of the body. Indeed, devices for continuous monitoring of P_POSTO₂ or postoxygenator oxygen saturation exist but are not widespread. Then down titration of F_SO₂ might theoretically result in unknown low P_POSTO₂, and hepato-splanchnic hypoxia [15]. One could also argue that for now, the safe P_POSTO₂ target is still unknown, as randomized trials are ongoing [22].

Although in the setting of ECPR, most of studies have reported an association between P_aO₂ and outcome [2, 4‐7, 9‐11, 23], results are not so clear for patients with cardiogenic shock supported by VA-ECMO. Based on the ELSO registry, Munshi et al. did not found any association between P_aO₂ 24 h after ECMO start and mortality in the subgroup of 775 patients with cardiogenic shock [7]. In a smaller cohort by Ross et al., mean P_aO₂ of the 72 first hours was also not associated with mortality [1]. However, more recently, Moussa et al. reported that early hyperoxemia was associated with mortality in a cohort of 430 patients. In this study, mean P_aO₂, absolute peak P_aO₂, and mean daily peak P_aO₂ during the 48 first hours were associated with 28-day mortality [12]. An analysis of 9959 patients in the ELSO registry found an association between hyperoxemia (P_aO₂ > 150 mmHg 24 h after ECMO start) and in-hospital mortality [8]. Our results are in line with these findings.

Beyond the already described link between P_aO₂ and outcome, we found that F_SO_{2 mean (day 1−3)} was independently associated with in-ICU mortality. Such finding is of interest because it may help to distinguish if hyperoxemia is a culprit or only a covariate [22]. Indeed, such link between hyperoxemia and prognosis might either be mediated by the proper harm of ECMO-associated hyperoxemia or be biased by the severity of cardiac failure. In the setting of peripheral VA-ECMO, the differential hypoxemia phenomenon results in heterogeneous PO₂ along the aorta, depending of the location of the mixing zone [24]. In the most severe cardiac failure, the mixing zone is in the aortic arch, close to the brachiocephalic trunk. Then, right radial P_aO₂ is mainly determined by F_SO_2, and its value is closed to the P_POSTO₂ value. In the absence of measurement of stroke volume or its surrogates (pulse pressure or end tidal CO₂ [25]), we cannot rule out that patients with higher right radial P_aO₂ were those with the most severe cardiac failure, having per se a higher mortality [26]. The fact that a higher F_SO_{2 mean (day 1−3)} was independently associated with a higher in-ICU mortality supports the hypothesis of the proper harm of ECMO-induced hyperoxemia, as F_SO₂ may not be impacted by cardiac failure severity. Indeed, higher F_SO_{2 mean (day 1−3)} may have resulted in higher P_POSTO₂, and potentially more reperfusion injury of the gut [27], liver, and kidneys. Interestingly, in the study of Moussa et al., the mean F_SO₂ was lower in survivors than in non-survivors [12]. In the study of Justus et al., the median F_SO₂ also tended to be lower in survivors than in non-survivors (72% versus 78%) [21].

Our study has strengths. First, the multicentric design allowed us to detect a signal of variability of oxygenation practices according to center’s case volume. Second, we collected data during seven days after ECMO implantation, which corresponds to almost the whole duration of VA-ECMO support. We think that it is more adapted to study the real impact of ECMO-induced hyperoxemia than focusing on the first 24 h of support.

Our study has also several limitations. First, while we hypothesized that the association between F_SO_{2 mean (day 1−3)} and in-ICU mortality was mediated by the P_POSTO₂ value, we were unable to confirm it because P_POSTO₂ monitoring was infrequent, even in high-volume centers. We also did not collect data on ECMO blood flow, which has been recently demonstrated to be a major determinant of right radial P_aO₂ [20]. Because the primary objective of the study was description of oxygenation practices, we also did not collect data allowing to evaluate the impact of ECMO-induced hyperoxemia on organ dysfunction. Second, our study was mostly conducted in France. Epidemiological data from another country may have led to different observation. Third, we did not find any association between centers case volume and in-ICU mortality. It is however commonly admitted that center’s case volume is associated with improved outcome [28]. These results might be explained by the limited sample size, the fact that most of patients were admitted in medium volume centers, and the short-term outcome (in-ICU mortality). Finally, due to the inclusion criteria (last 10 patients with cardiogenic shock supported VA ECMO for more than 24 h), data may reflect oxygenation practices over up to two years in low volume centers, compared to only 3 months in high volume centers.