132 A. G. Beshish et al.: J Extra Corpor Technol 2025, 57, 129 – 136
Subset analysis of patients with a cardiac diagnosis who required ECLS post-cardiotomy
There were a total of 196 patients with a cardiac diagnosis who required VA-ECLS. Of those, 155 were post-cardiotomy and are shown in Supplemental Table 1. We stratified the post-cardiotomy patients using the PaO 2 cut point that was used above( PaO 2 233 mmHg) and found that patients in the hyperoxia group had a longer duration of ECLS run( 149 vs 88 h, p = 0.0004), and had a lower rates of expected recovery( 66.1 vs 78.1, p = 0.027). The full comparison between the cohorts is shown in Supplemental Table 2. Discussion Figure 1. Receiver operating characteristic( ROC) curve identifying the optimal discriminatory cut point for mortality was 233 mmHg( sensitivity 36 %, specificity 76 %).
initial ECLS flow in the first 2 h [ 128 mL / kg / min( IQR 97, 131) vs 109 mL / kg / min( IQR 100, 121), p = 0.018 ]. The most common indication to require ECLS( based on ELSO registry criteria) was cardiac in both groups( 60.3 % and 44.1 %, respectively). Patients in the hyperoxia group had a lower rate of expected recovery and ability to wean off ECLS( 54.4 % vs 73 %, p = 0.049). Compared to the non-hyperoxia group, we found that mortality was higher in the hyperoxia group( 54.4 % vs 40.45, p = 0.050)( Table 1, Fig. 2).
Outcomes analysis
In the univariable analysis, we found that the hyperoxia group had increased odds of mortality( OR 1.76, 95 % CI [ 0.995 – 3.12 ], p = 0.052). In the multivariable analysis when controlling for age group( neonates vs pediatrics) and indication for ECLS, hyperoxia continued to demonstrate higher odds of mortality( aOR 2.02, 95 % CI [ 1.03 – 3.97 ], p = 0.03) and higher odds of developing stage II or III AKI( aOR 2.12 95 % CI [ 0.82 – 5.50 ], p = 0.12)( Table 2), but this did not reach statistical significance. The association of average PaO 2 and ECLS duration is graphically demonstrated in Figure 3( correlation coefficient �0.11, 95 % CI [ �0.33 – 0.03 ], p = 0.093).
Functional Status Scale( FSS) of survivors and development of new morbidity and unfavorable outcomes
Among the non-hyperoxia group, 21 of the 96 survivors( 21.9 %) developed new morbidity along with 9 out of 31 survivors( 29.0 %) of the hyperoxia group. Unfavorable functional outcome( defined as an increase in FSS 5) occurred in 8 / 96 patients in the non-hyperoxia survivors( 8.3 %), and 2 / 31( 6.5 %) in the non-hyperoxia group. We failed to identify an association between designation as“ hyperoxia” and new morbidity or unfavorable outcomes( Table 3).
In this single-center retrospective report from a highvolume ECLS center, we describe an overall mortality rate of patients supported on VA-ECLS of 44.5 %. Using a ROC curve, ameanPaO 2 of 233 mmHg in the first 48 h of ECLS was determined to have the optimal discriminatory ability for mortality( sensitivity 36 % and specificity 76 %). Of the VA-ECLS runs, 68( 29.7 %) had mean PaO 2 > 233 mmHg and were categorized as the hyperoxia group. Patients in the hyperoxia group tended to be older( median age 4.6 vs 1.5 months, p = 0.019), to have a cardiac indication for VA-ECLS( 60 % vs 44 %, p = 0.0004), and had lower rates of expected recovery to come off ECLS( 54 % vs 73 %, p = 0.049). Multi-variable logistic regression controlling for confounding variables demonstrated a suggestive association, though it failed to reach statistical significance, between hyperoxia and mortality in the unadjusted analysis( OR 1.76, 95 % CI [ 0.995 – 3.12 ], p = 0.052). While hyperoxia during ECLS did not directly lead to death, we postulate that it may contribute to new morbidities, such as acute kidney injury, that would later lead to complications and mortality.
In other critical illness settings, an association between excessive oxygen delivery and poor clinical outcomes has been reported. In patients requiring ECLS post-cardiac arrest, hyperoxia( as defined by a mean PaO 2 > 193 mmHg) was associated with 30-day mortality and the need for dialysis [ 4, 15, 25 ]. In a large multicenter cohort study of adult patients admitted to the ICU after resuscitation from cardiac arrest( CA), Kilgannon et al. showed an association between hyperoxia and risk of in-hospital death consistent with a dose-dependent relationship [ 15 ]. In a prospective disease-specific CA database, Elmer and colleagues found that exposure to severe hyperoxia was independently associated with inpatient mortality [ 25 ]. A retrospective study by Al-Kawaz and colleagues demonstrated poor neurological outcomes in patients exposed to mild, moderate, and severe hyperoxia in the first 24 h of ECLS. Additional duration of severe hyperoxia was independently associated with in-hospital mortality [ 13 ]. Several reports of neonates with asphyxia have demonstrated an association between hyperoxia and an increased risk of brain injury and mortality [ 4, 26, 27 ]. Conversely, Raman et al, in a single-center study and systematic review of a heterogeneous cohort of critically ill patients, did not demonstrate an association between hyperoxia at the time of admission and mortality [ 28 ].