Extracorporeal membrane oxygenation cannulation trends for pediatric respiratory failure and central nervous system injury

Journal of Pediatric Surgery (2012) 47, 68 75 www.elsevier.com/locate/jpedsurg Extracorporeal membrane oxygenation cannulation trends for pediatric respiratory failure and central nervous system injury Michael D. Rollins a,, Ania Hubbard b, Luke Zabrocki c, d, Douglas C. Barnhart a, Susan L. Bratton b a Division of Pediatric Surgery, Primary Children's Medical Center, University of Utah, Salt Lake City, UT 84113, USA b Division of Pediatric Critical Care Medicine, Primary Children's Medical Center, University of Utah, Salt Lake City, UT 84108, USA c Department of Pediatrics, Primary Children's Medical Center, University of Utah, Salt Lake City, UT 84113, USA d Naval Medical Center San Diego, San Diego, CA 92134, USA Received 26 September 2011; accepted 6 October 2011 Key words: Extracorporeal membrane oxygenation; Pediatric respiratory failure; Central nervous system injury; Venovenous extracorporeal life support; Venoarterial Abstract Background: Guidelines regarding arterial cannula site and cannula site-specific risks of central nervous system (CNS) injury for pediatric patients requiring extracorporeal membrane oxygenation (ECMO) support are lacking. We reviewed cannulation trends for pediatric respiratory failure and evaluated CNS complication rates by cannulation site and mode of support. Methods: The Extracorporeal Life Support Organization (ELSO) registry was queried for all pediatric respiratory failure patients b18 years treated from 1993-2007. The primary outcome was radiographic evidence of CNS injury. Results: Venoarterial (VA) support was used in 62% of 2617 ECMO runs. The carotid artery was used in 93% of VA patients. Femoral artery use increased in patients N5 years of age and N20 kg. Venovenous (VV) ECMO was used in N50% of children N10 years. No significant difference was identified in CNS injury between carotid and femoral cannulation in any age group but the femoral group was small (4.4%). VA support was independently associated with increased odds of CNS injury compared to VV cannulation (OR, 1.6). Conclusion: VA ECMO is the most common mode of support in pediatric respiratory failure patients. Although no significant difference in CNS injury was noted between carotid and femoral artery cannulation, the odds of injury were significantly higher than VV support. 2012 Elsevier Inc. All rights reserved. The views expressed in this article are those of the author(s) and do not necessarily reflect the official policy or position of the Department of the Navy, Department of Defense, or the US Government. Corresponding author. Division of Pediatric Surgery, Surgery Primary Children's Medical Center, Salt Lake City, Utah 84113, USA. Tel.: +1 801 662 2950; fax: +1 801 662 2980. E-mail addresses: michael.rollins@imail.org, michael.rollins@hsc.utah.edu (M.D. Rollins). Central nervous system hemorrhage and infarct are two of the most serious complications of extracorporeal life support (ECLS). The choice of cannulation site must be weighed carefully with an initial decision of venovenous (VV) versus venoarterial (VA) support, which is typically determined by the degree of cardiovascular failure. VV theoretically should pose decreased risk of embolic injury 0022-3468/$ see front matter 2012 Elsevier Inc. All rights reserved. doi:10.1016/j.jpedsurg.2011.10.017

Extracorporeal membrane oxygenation cannulation trends although a past report examining extracorporeal membrane oxygenation (ECMO) use in pediatric respiratory failure did not show this benefit [1]. If VA support is required, a decision is generally made between the carotid and femoral artery. In the newborn, the common carotid artery is typically used for VA ECMO owing to inadequate size of the femoral artery for the necessary pump flows and the risk of limb ischemia. Informal recommendations have been made (Bartlett; JPS Lecture, American Pediatric Surgical Association 41 st annual meeting 2010) to use the femoral artery as the preferred cannulation site in children requiring VA if they are N15 kg in order to reduce the risk of central nervous system (CNS) injury from use of the carotid artery. The risk of either CNS infarct or hemorrhage while on ECMO in neonates, pediatrics, and adults is reported by ELSO as 14% [2], 8%[3], and 11% to 15% [4,5], respectively. This study seeks to (1) describe the current cannulation modalities used in children with respiratory failure, (2) determine if children supported with VV ECMO have lower risk of CNS injury compared to VA support, and (3) evaluate CNS complication rates by arterial cannulation site. 1. Materials and methods ELSO collects case data from over 115 ECMO centers worldwide through voluntary reporting. Data include a primary indication for ECMO therapy designated as Pulmonary, Cardiac, or E-CPR, and detailed patient demographics and ECMO support data. The ELSO registry (Ann Arbor, Mich) was queried for all pediatric patients b18 years of age treated from January 1, 1993, to December 31, 2007, for a primary pulmonary indication (n = 3717). Patients designated by the treating center as neonate were not included. Exclusion criteria included children for whom cannulation site information was not available or a site was coded as other (n = 1474). Patients with complex arterial cannulation: aorta (n = 120), multiple groin (n = 1), and both carotid and femoral (n = 6) were also excluded leaving 2617 ECMO runs for analysis. Primary diagnosis ICD-9 codes as well as secondary diagnoses and Current Procedural Terminology codes were examined independently by two investigators (LZ, SB) who assigned a primary respiratory diagnosis [6]. In a similar fashion, diagnosis codes were also used to determine comorbid conditions. Patient complications during ECMO were grouped using complication codes created by the ELSO registry into the following categories and subcategories: (1) Brain injury [seizures (clinical or electroencephalogram evidence of seizures), radiological evidence of CNS injury (ultrasound or computerized tomography scan evidence of hemorrhage or infarction), and brain death]; (2) Renal injury: [dialysis use (hemodialysis or continuous arterio-venous hemodialysis)]; and (3) mechanical failure of the ECMO circuit (thrombus in the ECMO circuit, cannulation site or surgical site bleeding). ECMO cannulation was classified as VV or VA using the carotid or femoral artery based on the final cannulation site data fields. The primary outcome was radiographic evidence of CNS injury (infarct or hemorrhage). Any CNS complication was defined as seizures, radiographic evidence of stroke, or hemorrhage. Carotid and femoral artery sites as well as VV cannulation were compared for an association with brain injury. Categorical variables were analyzed using a Fischer exact test or Pearson χ 2 test, while continuous variables were analyzed using the Mann- Whitney U test or the Kruskal Wallis test when comparing more than 2 groups. Data are reported as frequency (n) with proportion (%), or median values with inter-quartile range (25th-75th percentile) unless specified otherwise. All statistical analysis was performed using SPSS 17.0 for Windows (SPSS, Chicago, Ill). Significance was determined as P b.05. A multivariable logistic regression model was developed to evaluate patient demographics, pre-ecmo clinical factors, and initial ECMO features associated with radiographic evidence of CNS injury. Variables that were statistically associated with radiographic evidence of CNS injury in the bivariate analysis were considered as candidate variables in the multivariate model using a forward stepwise selection procedure. Inclusion cutoff for the model was set as P.05 and exclusion as.10. Primary pulmonary diagnoses associated with CNS injury (acute respiratory distress syndrome [ARDS] from sepsis, bacterial pneumonia, and pertussis) were grouped as a high-risk diagnosis and were compared to all other primary pulmonary diagnoses as a group. Acidosis immediately before ECMO support was associated with brain injury. To better quantify the magnitude, we made categorical variables such that the lowest quartile, 25th quartile to median value were compared to greater ph values. Because CNS complications did not significantly differ between carotid and femoral artery cannulation site, they were combined to compare VA to VV in this analysis. 2. Results Selected demographics of 2617 children with respiratory failure supported with ECMO from 1993 to 2007 are shown in Table 1. VA cannulation was used in 1632 patients (62%), and the most frequent arterial site was the carotid artery (93%). VV cannulation was utilized in the remaining 985 (38%). Median age and weight of patients supported with VA-ECMO using the carotid artery were significantly lower than for those cannulated for VA-ECMO using the femoral artery. Among VA arterial cannulation sites, femoral artery use was rare (3%) among children 1 to 5 years of age but increased to 12% in those aged 5.1 to 10 years, and to 36% in those older than 10 years. 69

70 M.D. Rollins et al. Table 1 Children supported with ECMO for pediatric respiratory failure by cannulation type Venoarterial Venovenous Carotid Femoral VV n = 1516 n = 116 n = 985 n (%) n (%) n (%) Gender Males 738 (49) 59 (51) 481 (49) Unknown 43 (3) 1 (1) 12 (1) Age # b1 year 834 (55) 9 (8) 405 (41) 1-5 years 426 (28) 15 (13) 240 (24) 5-10 years 123 (8) 17 (15) 89 (9) 10-15 years 92 (6) 42 (36) 138 (14) 15 years 41 (3) 33 (28) 113 (12) Age # (median [IQR]) months 9.6 (2.8, 34.8) 155.0 (86.5,188.6) 20.9 (4.5,25.5) Weight # (median [IQR]) kg 7.3 (3.8, 14.0) 43.5 (25.3, 61.0) 11.0 (4.8,35.0) Year ECMO # 1993-1997 436 (29) 15 (13) 171 (17) 1998-2002 500 (33) 49 (42) 365 (37) 2003-2007 580 (38) 52 (45) 449 (46) Venous catheters # 1 1403(93) 83 (72) 592 (60) 2 80 (5) 28 (24) 335 (34) 3 3 (0.2) 3 (3) 58 (6) Unclear (other) 30 (2) 2 (2) 0 Percutaneous arterial catheter # 68 (5) 33 (28) 0 Percutaneous venous catheter placement # 99 (7) 46 (39) 335 (34) Pre ECMO arrest # 209 (14) 24 (21) 75 (8) Blood ph before ECMO 7.27 (7.17, 7.38) 7.26 (7.16, 7.38) 7.30 (7.19, 7.40) In hospital death 738 (49) 68 (59) 308 (31) Carotid arterial vs VV, Pb.05. # Carotid vs femoral artery, Pb.05. The use of VV-ECMO significantly increased over the study period (27% in 1993-1997 vs. 42% in 2003-2007). Percutaneous catheter placement was rare with carotid cannulation (5%) compared to femoral cannulation (28%). Children with arterial cannulation were significantly more likely to have a pre-ecmo cardiac arrest, lower blood ph just before initiation of bypass, and higher hospital mortality rate compared to VV cannulation. Brain injury leading to brain death was significantly more common in the femoral VA-ECMO patients compared to carotid VA-ECMO; however, the risk was not adjusted for severity of illness, and the femoral VA-ECMO patients suffered significantly more pre-ecmo cardiac arrests (21% vs 14%). One hundred sixty-seven patients required conversion from VV-ECMO to VA-ECMO, and 50% survived to hospital discharge. Significantly more surgical site and cannula bleeding complications were reported in the femoral VA-ECMO group compared to both other groups, whereas functional problems with the cannula were more common in the VV- ECMO group compared to carotid VA-ECMO group (Table 2). No difference was noted in the need for renal support among the groups. The overall incidence of CNS hemorrhage or infarct was 9.6% (VA 11.8%, VV 6%). Among all cases, the rates of CNS hemorrhage was significantly greater for VA carotid compared to VA femoral (7% vs. 2%), while rates of CNS infarct were similar (6% vs. 5%) and the presence of either injury was not significantly different (11% vs.7 %). All CNS complications were significantly less for VV compared to VA carotid. CNS complications stratified by age and cannulation site are reported in Table 3. Carotid VA-ECMO was used more commonly than femoral VA-ECMO among all age groups. VV-ECMO was used in over 50% of children N10 years of age requiring ECMO support for respiratory failure. CNS infarcts were more common with carotid cannulation in children N10 years of age; however, this was the only group where a significant difference was found, while rates of CNS infarcts were significantly different for most age groups when comparing VA carotid to VV (except age 5-10 years). When any CNS complications were evaluated (infarct, hemorrhage, or seizure), VV was consistently associated with lower rates than VA carotid. CNS complications were also stratified by weight (0-5, 5.1-10, 10.1-20, N20 kg) and cannulation site (VA carotid, VA femoral, VV). The femoral artery was used infrequently in patients b20 kg, but in patients N20 kg, it was used in 28%

Extracorporeal membrane oxygenation cannulation trends Table 2 ECMO Complications by cannulation type VA VA VA venovenous Carotid Femoral Venous N = 1516 N = 116 N = 985 Insertion/surgical n (%) n (%) n (%) Surgical site bleeding # 162 (11) 31 (27) 98 (10) Cannulation site 209 (14) 32 (28) 185 (19) bleeding # Hemorrhage DIC 96 (6) 11 (10) 35 (4) Cannula problems 214 (14) 19 (16) 174 (18) Mechanical Clots in bladder # 140 (9) 4 (3) 73 (7) Hemolysis 165 (11) 15 (8) 62 (6) Renal Support Hemofiltration 42 (3) 5 (4) 31 (3) Dialysis 83 (6) 11 (10) 38 (4) Central Nervous System Seizures 136 (9) 6 (5) 36 (4) Central nervous 98 (7) 2 (2) 41 (4) system bleed # Central nervous 87 (6) 6 (5) 18 (2) system infarct Central nervous system 171(11) 8 (7) 54 (6) bleed or infarct Brain death # 92 (6) 13 (11) 33 (3) Carotid arterial vs VV, P b.05. # Carotid vs femoral artery, P b.05. (91/320). No consistent difference was identified in CNS injury between carotid and femoral VA-ECMO among the weight groups. Any CNS complication in patients N 20 kg was 10% among femoral artery cannulation group compared with 17% among carotid VA-ECMO group (P =.52). Table 4 includes demographic and clinical factors of children supported by ECMO for respiratory failure whose infarct or CNS hemorrhage was documented by radiographic imaging compared to those without CNS infarct or hemorrhage. Lower age and weight, as well as more recent treatment (2003-2007) were significantly associated with these CNS injuries. Primary pulmonary diagnosis of bacterial pneumonia, ARDS from sepsis and pertussis, as well as comorbid conditions such as liver insufficiency and renal failure were associated with higher rates of CNS injury. Likewise more acidotic pre-ecmo blood gas values and carotid VA-ECMO were associated with increased rates of CNS injury. Hospital mortality was 70% among those with CNS infarct or hemorrhage compared to 40% among those with no CNS injury. Table 5 shows a multivariable logistic regression model for the odds of CNS infarct or hemorrhage. VA support was independently associated with increased odds of CNS injury compared to VV cannulation (OR, 1.6; 95% confidence interval, 1.1-2.3) after adjustment for age in years, the presence of liver insufficiency, renal failure, or cancer before institution of ECMO, a primary pulmonary diagnosis of Table 3 Central nervous system complications stratified by cannulation site and age b1 y 1-5 y 5-10 y N10 y Patients (n) VA carotid 834 426 123 133 VA femoral 9 15 17 75 VV 405 240 89 251 Seizures n, (%) VA carotid 103 (12) 17 (4) 8 (7) 8 (6) VA femoral 0 0 2 (12) 4 (5) VV 23 (6) 4 (2) 3 (3) 6 (2) CNS hemorrhage n,(%) VA carotid 58 (6) 27 (6) 6 (5) 7 (5) VA femoral 0 0 0 2 (3) VV 17 (4) 14 (7) 3 (3) 7 (3) CNS infarct, n (%) VA carotid 42 (5) 29 (7) 4 (3) 12 (9) # VA femoral 1 (11) 2 (13) 2 (12) 1 (1) VV 6 (1) 7 (3) 0 5 (2) CNS injury VA carotid 92 (11) 53(12) 9 (7) 17(13) VA femoral 1 (11) 2 (13) 2 (12) 3 (4) VV 23 (6) 18 (8) 3 (3) 10 (4) Any CNS complication, n (%) VA carotid 174 (20) 67 (16) 17 (19) 23 (17) VA femoral 1 (11) 2 (13) 3 (18) 7 (9) VV 41 (10) 20 (8) 6 (7) 15 (6) CNS injury indicates hemorrhage or infarct; any CNS complication, seizures or infarct or hemorrhage. VA carotid vs VV, P b.05. # VA carotid vs VA femoral, P b.05. bacterial pneumonia, pertussis or ARDS from sepsis, pre- ECMO ph and presence of pre ECMO infection. 3. Discussion Decision on ECLS support type (VA vs. VV) and site of the cannulation remains one of the critical issues during the process of placing children on extracorporeal life support for failure. Choice of VA vs. VV support should be predominantly directed by degree of hemodynamic instability, but it also depends on patient age, size, and institutional experience. While VV support has many benefits for children in respiratory failure, VA remains the predominant mode of support for these patients across the ECLS centers. Our review of the ELSO trends for vascular cannulation in pediatric patients with respiratory failure demonstrates significantly increased odds of CNS injury with VA support even when severity of disease is controlled. While similar findings have been demonstrated in neonates with congenital diaphragmatic hernia [7,8], there has been lack of clinical evidence to confirm this in pediatric patients with respiratory failure [1]. This difference may be related to a smaller patient population supported with VV ECMO (22%) in the study by 71

72 M.D. Rollins et al. Table 4 Factors associated with central nervous system infarct or hemorrhage CNS Injury No CNS injury n = 233 n = 2384 n (%) n (%) Age in days 366 (81, 1093) 415 (106, 2188) Weight (kg) 7.9 (4.0,15.0) 9.5 (4.2, 21.1) Male gender 105 (45) 1173 (49) Year ECMO 1993-1997 34 (15) 588 (25) 1998-2002 85 (36) 829 (35) 2003-2007 114 (49) 967 (41) Primary pulmonary diagnosis RSV 43 (19) 406 (17) Bacterial pneumonia 49 (21) 381 (16) Other viral pneumonia 17 (7) 240 (10) Sepsis ARDS 23 (10) 157 (7) Aspiration pneumonia 7 (3) 125 (5) Trauma ARDS 9 (4) 104 (4) Pertussis 15 (6) 60 (3) Pulmonary hemorrhage 2 (1) 59 (3) Other ARDS 2 (1) 26 (1) Other 66 (28) 826 (34) Confounding conditions prior to ECMO Liver insufficiency 12 (5) 37 (2) Renal failure 46 (20) 239 (10) Chronic lung disease 19 (8) 229 (10) One ventricle CHD 5 (2) 33 (1) Two ventricle CHD 12 (5) 176 (7) Immunodeficiency 11 (5) 42 (2) Cancer 14 (6) 62 (3) Pre ECMO 44 (19) 264 (11) cardiac arrest Pre ECMO infection 131 (56) 955 (40) ph prior to ECMO 7.23 (7.11, 7.33) 7.29 (7.18, 7.39) paco 2 prior to ECMO 64 (46, 87) 55 (43, 75) Cannulation type Carotid artery 171 (73) 1345 (56) Femoral artery 8 (3) 108 (5) Veno venous support 54 (23) 931 (39) Percutaneous cannulation Venous 25 (11) 454 (19) Arterial 11 (5) 90 (4) ECMO complications Hemorrhage cannulation 40 (17) 217 (9) Clots in bladder 21 (9) 196 (8) DIC 20 (9) 122 (5) Hemolysis 25 (9) 217 (9) Cannula problems 40 (17) 364 (16) Hours on ECMO 176 (88, 326) 201 (107, 356) Died in hospital 163 (70) 951 (40) P b.05. Table 5 Multivariable model for odds of central nervous system infarct or hemorrhage Factors Odds ratio 95% Confidence interval Age in years 0.96 0.93-0.98 ph pre ECMO 6.4-7.18 2.1 1.5-2.8 7.19-7.29 1.5 1.1-2.1 N7.29 1 Reference group Pre ECMO infection 1.7 1.3-2.2 Pre ECMO non pulmonary conditions Cancer 2.3 1.3-4.3 Renal failure 1.7 1.3-2.4 Liver insufficiency 2.0 1.0-4.5 High risk pulmonary diagnosis 1.4 1.1-2.0 Venoarterial support 1.6 1.1-2.3 Venovenous support 1 Reference group Zahraa et al [1] compared to this study (38%) or may be related to recent changes in the ECMO circuit or differences in patient management. The use of the femoral artery was compared to carotid cannulation to determine differences in the odds of CNS injury during extracorporeal support. There was increasing use of the femoral artery with increasing patient age and size, but no statistically significant difference in CNS injury was identified. This finding may be explained by the small number of patients in the registry supported using the femoral artery. We believe that given the limited number of children who underwent femoral cannulation, this study should not be considered to refute the hypothesis that femoral cannulation may have a lower risk of CNS injury than carotid cannulation. More surgical site bleeding occurred in the femoral cannulation group compared to the carotid group. This could be explained by the different surgical techniques used for arterial cannula placement. More cannula problems occurred in the VV-ECMO group overall; however, owing to the limits of the ELSO registry, it is not possible to identify a specific problem in question. The use of femoral VA-ECMO or VV-ECMO requires a greater commitment to cannula management as 27% and 40% respectively required more than one venous cannula compared to 5% of the carotid group. However, there is a clear benefit to using these modes as hospital mortality was 70% among those with CNS infarct or hemorrhage compared to 40% among those with no CNS injury. The overall rate of CNS complications including stroke or hemorrhage was 9.6%, which is lower than previous reports from the ELSO registry [9]. Patients with conditions associated with coagulopathy such as bacterial pneumonia and sepsis had higher rates of CNS infarct or hemorrhage. Pre-ECMO severity of illness evidenced by acidosis, hypercarbia and cardiac arrest were also associated with central nervous system injury. However, it is unknown how many patients had CNS injury before ECLS initiation.

Extracorporeal membrane oxygenation cannulation trends Our study has limitations that should be considered. All databases are limited to the information recorded. For complication rates related to femoral arterial cannulation, there are no data available to assess rate of limb ischemia and the use of reperfusion catheters. Catheter information is not recorded uniformly and required searching text fields; furthermore, there were missing data regarding catheter placement precluding analysis of about 25% of potential patients. CNS injury was limited to clinically obtained data, and it is unknown if CNS injury preceded the initiation of ECMO in older patients in whom cranial ultrasound is not possible. Furthermore, while Lidegran et al [10] reported a 31% incidence of CNS injury in neonatal, pediatric, and adult patients on ECMO with frequent use of CT scanning, this is not standard, and the methods of detecting CNS injury either while a patient is on ECMO or after ECMO vary. Due to the lack of uniform management standards for ECMO patients among centers, we were not able to assess the effects of different treatment such as anticoagulation strategies and carotid artery reconstruction techniques on CNS complications. Prior reports of CNS infarct or hemorrhage have largely been single center experiences and may be related to the mode of ECMO support preferred in the specific institution [11-14]. 4. Conclusion The VV mode of support is associated with lower odds of cerebral infarct or hemorrhage, and should be considered as the preferred mode for patients with respiratory failure requiring ECLS. Based on our review of the ELSO registry, VV-ECMO is being used with increasing frequency in this patient population yet appears to be underutilized. The current data would support consideration of femoral artery VA-ECMO in patients N5 years old or N20 kg. References [1] Zahraa JN, Moler FW, Annich GM, et al. Venovenous versus venoarterial extracorporeal life support for pediatric respiratory failure: are there differences in survival and acute complications? Crit Care Med 2000;28:521-5. [2] VanMeurs KP, Hintz SR, Sheehan AM. ECMO for neonatal respiratory failure. In: Van Meurs K, Lally KP, Peek G, Zwischenberger JB, editors. ECMO Extracorporeal Cardiopulmonary Support in Critical Care. 3rd ed. Ann Arbor (Mich): Extracorporeal Life Support Organization; 2005. p. 273-95. [3] Frenckner B, Palmer P. Management of pediatric respiratory failure on ECLS. In: Van Meurs K, Lally KP, Peek G, Zwischenberger JB, editors. ECMO Extracorporeal Cardiopulmonary Support in Critical Care. 3rd ed. Ann Arbor (Mich): Extracorporeal Life Support Organization; 2005. p. 363-81. [4] Brogan TV, Thiagarajan RR, Rycus PT, et al. Extracorporeal membrane oxygenation in adults with severe respiratory failure: a multi-center database. Intensive Care Med 2009;35:2105-14. [5] Bartlett RH. Management of ECLS in adult respiratory failure. In: Van Meurs K, Lally KP, Peek G, Zwischenberger JB, editors. ECMO Extracorporeal Cardiopulmonary Support in Critical Care. 3rd ed. Ann Arbor (Mich): Extracorporeal Life Support Organization; 2005. p. 403-16. [6] Zabrocki LA, Brogan TV, Statler KD, et al. Extracorporeal membrane oxygenation for pediatric respiratory failure: Survival and predictors of mortality. Crit Care Med 2011;39:364-70. [7] Dimmitt RA, Moss RL, Rhine WD, et al. Venoarterial versus venovenous extracorporeal membrane oxygenation in congenital diaphragmatic hernia: the Extracorporeal Life Support Organization Registry, 1990-1999. J Pediatr Surg 2001;36:1199-204. [8] Guner YS, Khemani RG, Qureshi F, et al. Outcome analysis of neonates with congenital diaphragmatic hernia treated with venovenous vs venoarterial extracorporeal membrane oxygenation. J Pediatr Surg 2009;44:1691-701. [9] Cengiz P, Seidel K, Rycus P, et al. Central nervous system complications during pediatric extracorporeal life support: incidence and risk factors. Crit Care Med 2005;33:2817-24. [10] Lidegran M, Palmer K, Jorulf H, et al. CT in the evaluation of patients on ECMO due to acute respiratory failure. Pediatr Radiol 2002;32: 567-74. [11] Alsoufi B, Al-Radi OO, Nazer RI, et al. Survival outcomes after rescue extracorporeal cardiopulmonary resuscitation in pediatric patients with refractory cardiac arrest. J Thorac Cardiovasc Surg 2007;134:952-9. [12] Buesing KA, Kilian AK, Schaible T, et al. Extracorporeal membrane oxygenation in infants with congenital diaphragmatic hernia: followup MRI evaluating carotid artery reocclusion and neurologic outcome. AJR Am J Roentgenol 2007;188:1636-42. [13] Lamers LJ, Rowland DG, Seguin JH, et al. The effect of common origin of the carotid arteries in neurologic outcome after neonatal ECMO. J Pediatr Surg 2004;39:532-6. [14] Mehta NM, Turner D, Walsh B, et al. Factors associated with survival in pediatric extracorporeal membrane oxygenation a single-center experience. J Pediatr Surg 2010;45:1995-2003. Discussion Unidentified discussant: This is a very interesting and important subject, and it is one of the worst complications we can have on ECMO. When you looked at the data, was there a way to look at the mode of how anticoagulation was administered and controlled, what kind of ACTs people used, and did they use antithrombin-3, etc.? Barnhart (response): No, that is not in the ELSO registry so I am sure it is quite variable across centers and across times as well. Charles Stolar, MD (New York, NY): Thank you very much for that nice registry report. A couple of comments and I guess a question. The comment is you are looking at data over a very long period of time during which the non-ecmo care has undergone dramatic evolution, so it is really hard to correct for that, number one. Number two, I think we all would agree that the patients that end up on venoarterial ECMO, however you cannulate them, are sicker. Their pump does not work, so they are generally sicker and they 73

74 M.D. Rollins et al. are more likely to have a complication. I think even correcting for that I think both of these very nice presentations have shown that. I would suggest that the central nervous system injury you are seeing is unrelated to tying off the carotid artery. I would suggest that it is more related to the loss of autoregulation of cerebral blood flow. We know that we do this in babies all the time. We always are concerned about the baby's first stroke when we tie off the carotid but nothing happens. The CNS injuries are either posterior fossa bleeds or if there is a loss of the A1 segment in the circle of Willis, and the same thing is noted in the older kids. If they have a central nervous system injury, it is usually not an infarct or hemorrhage. It is usually ischemic encephalopathy, and it is not necessarily related to the side. I would encourage people not to be intimidated by using the carotid artery. I believe it is largely misplaced anxiety. The question I have for you, and you may or may not have these data, but you say that the central nervous system complications are a consequence of ligating the carotid. If you compare the incidence of CNS complications after carotid ligation versus the incidence of extremity tissue loss from an arterial cannulation of the leg, which is more prevalent? Barnhart (response): Thank you for both the question and the comments. First of all for clarity, I would not argue that the injuries are from ligation of the carotid. I share your interpretation as well as the possibility that probably many of them are embolic and the fact that on venovenous ECMO you have the benefit of having the lungs between the ECMO circuit and the brain. This question came up as we wrestled with which patient should go on venovenous ECMO. One of the things we discovered as we talked to various centers many centers would say they do venovenous ECMO and as you look it is always simpler for the surgeon and the intensivist to put the child on VA ECMO and be done with it. There is a greater hassle factor to everyone to go on venovenous. Our motivation in going into this data was not to necessarily parse out scientifically all these variables, but I think it was to provide some data to incentivize us to continue to work hard to keep a kid on venovenous ECMO rather than going to the easy way by going on to venoarterial ECMO. It is interesting that if you go back and look in the literature this seems intuitive, but the only literature that is out there in the pediatric pulmonary failure patient is a paper in 2000 that showed that there was no difference. Even though the observation is intuitive, what we took away from this is that it is worth getting up in the middle of the night one more time to try to adjust the venovenous cannula to spare the baby the potential risk of neurologic injury by whatever mechanism it might occur. Richard Ranne, MD (Lubbock, TX): The VA system of using femoral cannulation mainly is based on adult data of limb ischemia and limb loss. Did you find any evidence of that in pediatric patients? And, there are strategies with decreased collaterals in adults to try to alleviate that. Has anyone been using that in children? Barnhart (response): I can answer from our personal and institutional experience. Unfortunately, there are no data in the ELSO registry about limb complications, so that is not only unanswered but at the moment unanswerable. I agree with you, I think that when you start down a path where you use the femoral artery, again like venovenous ECMO it is going to become more complicated and you may need to put in distal perfusion cannulas and you need to be prepared to reconstruct the femoral artery. Whether or not that is worth it I think is an unanswered question. We went into this to try to figure out whether or not there was going to be enough of an advantage of femoral cannulation to merit those extra efforts and I am not certain of that. I think we did convince ourselves that there is enough data that it is worth the struggle to try to stay on VV if you can support the child, but in terms of whether it is worth distal reperfusion cannulas and femoral reconstruction, I do not know the answer to that. Ronald Hirschl, MD (Ann Arbor, MI): Nice presentations and of course we are big proponents of venovenous. One of the issues may be the availability of cannulas and clearly we did not have a double-lumen cannula for anything over about 6 kg until you got somewhere about 15 to 20 kg and even then I think using two cannulation sites is not something people want to do. When you look at venovenous cannulation you see that it increases as you get older and in some ways that is not surprising. One of my questions is that the ELSO registry is a cumulative registry in terms of its data. Did you see any trends specifically over the last few years as venovenous has become more popular if you will or implemented at various sites and also as we have started to have availability of the Avalon double-lumen cannula of varying sizes so that we could utilize VV with one cannulation site? Barnhart (response): This series ended before the Avalon cannula became available, so I cannot answer that question. I know individually it certainly encouraged us to have a single-site, double-lumen cannula be our first modality and I think it has become our default. In spite of the fact that there was not availability of the new double lumen catheter, in this series, we did see that venovenous was becoming more popular, which I think is probably because as more centers become more confident with ECMO in general, they recognize that once you are on ECMO you can continue to make manipulations and adjustments and it is not a one-time decision of whether

Extracorporeal membrane oxygenation cannulation trends you cannulate VA or VV. What we have done as a result of this data and our experience is to try to come up with some pretty specific criteria that you need to meet to go on VA ECMO and otherwise you have to go through VV ECMO before you get to cannulate the carotid artery. We attempt to stabilize the infants using VV ECMO knowing that oxygen is a great inotrope and babies on VV ECMO look better than they did before they were put on ECMO. 75