176 A . G . Beshish et al .: J Extra Corpor Technol 2024 , 56 , 174 – 184 admission ) and again at hospital discharge utilizing appropriate documentation . FSS score determination was blinded from hyperoxia status . Newborns who had never achieved a stable baseline function were assigned a score of 6 . This was operationalized by assigning a baseline FSS score of 6 to all admissions for infants 0 – 2 days old and to transfers from another facility for infants 3 – 6 days old as previously reported [ 27 – 30 ]. New morbidity was defined as an increase in the total score of 3 points , and unfavorable functional outcome was defined as an increase of 5 points [ 31 ].
Clinical management
In our center , in the period 2010 – 2020 , there were multiple different iterations in the layout of the ECLS circuit , and we used different manufacturers for different circuit components / parts . All circuits were blood-primed before the start of ECLS with packed red blood cells , 25 % albumin , sodium bicarbonate , calcium gluconate , and heparin . It is common practice for ABGs to be obtained at the discretion of the clinical team , commonly approximately 30 min after initial ECLS cannulation , and then hourly for the first several hours . Subsequently , blood gases are obtained every 3 – 6 h and shortly after an adjustment in ECLS support . Target gas exchange parameters are not dictated by protocol at our center . Goal PaO 2 ranges have no established normal and the variation we describe is derived from measurements occurring during clinical care . We target goal arterial oxygen saturations > 80 %, and to ensure adequacy of cardiac output , we target pre-membrane saturations > 50 % using Spectrum Medical Perfusion Monitor M3 and M4 , Fort Mill , SC , USA . Goal PaCO 2 was typically 35 – 45 mmHg , and goal pH was typically 7.35 – 7.45 . While on ECLS , our center ’ s approach was not to mechanically limit or clamp any source of pulmonary blood flow ( i . e ., m-BTT or Sano shunt ).
Statistical analysis
Variables were described using medians with interquartile ranges ( IQR ) or counts with percentages . Patient characteristics were compared between hyperoxia and non-hyperoxia groups . Comparisons were made using chi-square tests or Fisher ’ s exact tests for categorical variables and Wilcoxon rank-sum tests for continuous variables . Average PaO 2 while on CPB was calculated for individual patients . The correlation coefficient with a 95 % confidence interval ( 95 % CI ) was calculated . ROC curve and area under the curve ( AUC ) analyses were performed to identify the optimal cut-off values of mean PaO 2 for predicting mortality within 30 days after the operation . All variables were entered into binary logistic regression analyses with pre-identified hyperoxia status as the outcome . The associations between hyperoxia status and adverse outcomes ( i . e ., Stage II or III AKI , PPLOS , and mortality ) were assessed using binary logistics and multivariable logistics regression controlling for those explanatory variables with a p-value < 0.05 . Odds ratios ( OR ) and adjusted odds ratios ( aOR ) with 95 % CI were presented . The association between average PaO 2 on ECLS and duration of ECLS was further assessed using Pearson correlation . Overall FSS and subdomain FSS were reported as mean and standard
Figure 1 . Flow chart of neonates requiring extracorporeal life support post-Norwood operation stratified based on PaO 2 levels in the first 48-hours while on ECLS .
deviation ( SD ) and paired Student ’ s t-test was used to compare scores at admission and at discharge . All p-values < . 05 were considered significant ( two-tailed ). All analyses were performed using SAS version 9.4 ( SAS Institute , Cary , NC ) and R statistical software ( version 4.0.2 ; R Core Team , 2020 ).
Results
Patient demographics and characteristics for overall cohort During the period , there were 269 neonates with univentricular physiology who underwent Norwood operation . Of these , 65 / 269 ( 24 %) required ECLS support ( Figure 1 ). The median age at the time of surgery was 6 ( IQR 4 , 7 ) days , and the median weight was 3.2 ( IQR 2.8 , 3.5 ) kg . The most frequent cardiac diagnosis was hypoplastic left heart syndrome in 50 / 65 ( 77 %) of cases , and mitral atresia with aortic atresia was the most common variant in 26 / 50 ( 51 %). Patient demographics and clinical characteristics are presented in Table 1 . Of note , none of our patients underwent a concomitant AVV repair at the time of Norwood ’ s operation .
Cut-point analysis
Using ROC analysis , PaO 2 > 182 mmHg had the optimal discriminatory ability for operative mortality ( sensitivity of 68 %, and specificity of 70 %) and was therefore used to define hyperoxia in our exploratory analysis ( Figure 2 ). The AUC for average PaO 2 during CPB and subsequent mortality was 0.69 , ( 95 % CI : 0.58 – 0.81 ; p = 0.001 ). When using the PaO 2 > 182 mmHg threshold 34 / 65 were in the hyperoxia group . On univariable analyses , this designation of hyperoxia was associated with more Sano shunts / RV-PA conduits ( 82 % vs . 29 %, p < 0.001 ), had longer median CPB times ( 187 vs . 165 min , p = 0.023 ), had higher median VIS-scores in the first 24-hours and hours 24 – 48 , [( 25 vs . 20 , p = 0.027 ), and ( 30 vs . 23 , p = 0.017 ) respectively ], higher rates of central ECLScannulation ( 85 % vs . 47 %, p = 0.003 ), shorter median duration from CICU arrival to ECLS-cannulation ( 13.3 vs . 232.6 h , p = 0.003 ), higher serum lactate within 2-hours from ECLScanulation ( 14.65 vs . 5.8 , p = 0.01 ), and had higher flows in