Results of Extracororeal Membrane Oxygenation in Children With Sesis Dan M. Meyer, MD, Michael E. Jessen, MD, and the Extracororeal Life Suort Organization University of Texas Southwestern Medical Center, Dallas, Texas, and The University of Michigan, Ann Arbor, Michigan Background. Desite good results in neonates, extracororeal membrane oxygenation (ECMO) is less well acceted in ediatric atients. Older children frequently undergo ECMO for severe bacterial, viral, or asiration neumonia and many have coexisting systemic sesis. We reviewed data from a national registry to study the influence of sesis on survival from ECMO. Methods. Six hundred fifty-five atients (aged 2 weeks to 17 years) with resiratory failure treated with ECMO were divided into two grous by the resence (n 76) or absence (n 579) of sesis. Grous were comared by univariate analysis and by multivariate logistic regression that considered 10 additional re-ecmo variables (age, sex, weight, arterial blood gas results, ventilator arameters, and renal failure). Results. By univariate analysis, survival was lower in setic children (36.8% versus 51.6%; < 0.02). However, by multivariate analysis, sesis was not an indeendent survival redictor (odds ratio, 0.578; confidence interval, 0.288 1.162; 0.12). The ECMO comlications redicted by the resence of sesis included (1) seizures, (2) other neurologic comlications, and (3) infection at other sites (all < 0.05). Conclusions. Systemic sesis does not indeendently influence survival in ediatric ECMO. This theray should not be withheld solely because of sesis, although neurologic comlications may occur more frequently. (Ann Thorac Surg 1997;63:756 61) 1997 by The Society of Thoracic Surgeons Acceted for ublication Oct 22, 1996. Address rerint requests to Dr Meyer, Division of Thoracic and Cardiovascular Surgery, University of Texas Southwestern Medical Center at Dallas, 5323 Harry Hines Blvd, Dallas, TX 75235-8879. Since its introduction in 1975, extracororeal membrane oxygenation (ECMO) has evolved into a standard treatment for neonates with severe resiratory failure [1]. Much of the success of this modality was based on the transient nature of the neonatal disease rocesses for which ECMO was selected, such as ersistent ulmonary hyertension of the newborn [2] or ersistent fetal circulation [3]. However, the use of this technology in the ediatric oulation is less acceted. In ediatric atients, ECMO is often used in cases of severe bacterial, viral, or asiration neumonia. Many of these atients resent with systemic sesis and data suorting the use of ECMO in this setting is even more limited. Some investigators have suggested that the resence of systemic sesis has a negative effect on outcome [4], whereas others have claimed that sesis does not adversely affect results [5 7]. However, these studies restricted their analysis to the neonatal oulation (age, 14 days), an age grou in which ECMO is infrequently alied for setic conditions. Data are lacking for the ediatric oulation. This study used data from the Extracororeal Life Suort Organization (ELSO) registry to critically evaluate the effectiveness of ECMO in children aged 14 days to 18 years with resiratory failure when systemic sesis was a concurrent condition. The study tested the hyothesis that coexisting systemic sesis adversely affects (1) survival and (2) comlication rates in children with resiratory failure treated with ECMO. Material and Methods Patients This study consisted of 655 children (aged 14 days to 17 years) laced on ECMO for resiratory failure between November 1987 and June 1993 and entered in the ELSO registry. The age distribution and secific indications for ECMO are resented in Figures 1 and 2. Patients with diagnoses labeled Other include atients laced on ECMO for resiratory failure for an uncommon or uncertain diagnosis. Coexisting systemic sesis was resent in 76 atients (11.6%) in this series, and these atients were identified as grou 1. Grou 2 included all atients on ECMO without systemic sesis (n 579). Sesis, as defined by the ELSO registry, was the resence of athogenic microorganisms or their toxins in the blood or other tissues. It may be diagnosed clinically by symtomatic evidence of infection, or by laboratory studies. It may also be diagnosed by a documented ositive culture. The resence or absence of sesis was determined by the This article has been selected for the oen discussion forum on the STS Web site: htt://www.sts.org/annals 1997 by The Society of Thoracic Surgeons 0003-4975/97/$17.00 Published by Elsevier Science Inc PII S0003-4975(96)01272-6
Ann Thorac Surg MEYER ET AL 1997;63:756 61 RESULTS OF ECMO IN CHILDREN WITH SEPSIS 757 Fig 1. Age distribution of 655 ediatric atients laced on extracororeal membrane oxygenation. Both nonsetic or setic grous had similar frequency distributions over each age range. reorting center. No atients laced on ECMO for cardiac suort were included. Outcome s Outcome variables studied included survival (at time of reorting) and 34 ECMO comlications. These variables comrised hemorrhagic, neurologic, renal, cardiovascular, ulmonary, setic, and metabolic comlications. The 35 outcome variables are listed in Aendix 1. The ELSO definition of survival was successful searation from ECMO. However, most the individual centers, including The University of Michigan, required the atient to be discharged from the hosital to be classified as a survivor. Factors Before Extracororeal Membrane Oxygenation In addition to the resence of systemic sesis, 10 variables were used in the evaluation as redictors of outcome. These variables included atient age at the time of ECMO, sex, weight, the results of the most recent arterial Fig 2. Primary diagnosis for 655 ediatric atients in nonsetic and setic children treated with extracororeal membrane oxygenation. (ASPIR asiration neumonia; BPNEU bacterial neumonia; IPH/PNCYS intraulmonary hemorrhage/neumocystis; OTHER resiratory failure of uncommon or uncertain causes; VPNEU viral neumonia.) blood gas analysis before ECMO, the ventilator arameters of the atient immediately before initiation of ECMO, and the resence of renal failure before ECMO. The mode of ECMO was not included as 83% underwent ECMO through a venoarterial technique. The variables used before ECMO are listed in Aendix 2. Statistical Analysis Data were analyzed using commercially available statistical software (SAS Institute, Cary, NC). A univariate analysis of all outcome variables was erformed initially to comare results between grous 1 and 2, with a value of less than 0.05 by two-tailed Fisher s exact test considered significant. To account for interactions between variables before ECMO, additional tests were erformed. For each outcome variable, a multivariate stewise logistic regression analysis was erformed. All variables listed in Aendix 2 were candidates for entry into the model. For each resulting model, the grou variable (ie, the resence or absence of systemic sesis) was always included, whereas only those remaining variables that met the 0.05 significance level were entered. For each element in the model, a arameter estimate was calculated from which a value and an odds ratio for the variable were derived. Results Overall Results For the entire grou of 655 children, survival was 49.9%. When examined by age intervals, survival was 55.0% for children less than 24 months (n 407), 47.4% for children 24 to 72 months (n 133), 38.2% for those 72 to 156 months (n 76), and 28.2% for those more than 156 months (n 39). This trend of reduced survival with advancing age was statistically significant ( 0.05). Univariate Analysis A univariate analysis of atient survival was undertaken to comare the 76 children with sesis while on ECMO (grou 1) with the remaining 579 atients (grou 2). Survival was significantly better in grou 2 (51.6%) than in grou 1 (36.8%) ( 0.02 by Fisher s two-tailed exact test). Multivariate Analysis To account for otential interactions between variables before ECMO, a stewise regression was erformed for each variable as described reviously. Survival was not significantly different between grous when the variables before ECMO were accounted for by this analysis (odds ratio, 0.578; confidence interval, 0.288 1.162; 0.12). Only older age, lower H, and lower arterial oxygen tension in arterial blood gas levels before ECMO were redictors of increased mortality. The effect of sesis was tested by this multivariate analysis for each of the age ranges defined. In no subset was sesis a significant redictor of survival (all 0.10). Because the resence of before ECMO variables affected the analysis of survival, only the results of the multivariate analysis are reorted for ECMO comlications.
758 MEYER ET AL Ann Thorac Surg RESULTS OF ECMO IN CHILDREN WITH SEPSIS 1997;63:756 61 Table 1. Outcome s Found to Occur Significantly More Frequently a in Children With Coexisting Systemic Sesis as Determined by Multivariate Analysis The ECMO comlications that were found to be significantly different between the two grous at 0.05 are listed in Table 1. Not surrisingly, culture-roven infections at locations other than the site of rimary infection were more common in the setic grou. Neurologic sequelae (seizures and other neurologic comlications) were also more frequent in grou 1. The duration of ECMO did not differ significantly between grous (276 192 hours versus 248 195 hours for the nonsetic and setic atients, resectively, mean standard deviation). These outcomes were also examined for each of the four age grous defined reviously. The variables before ECMO used in the model remained the same. The ECMO comlications affected by the resence of sesis are resented for children aged 2 weeks to 24 months in Table 2, those aged 24 to 72 months in Table 3, and those aged 72 to 106 months in Table 4. In children more than 156 months of age, sesis had no significant effect on any of the 34 reortable ECMO comlications. For these analyses, any comlication influenced by sesis at the 0.10 level are reorted. Table 2. Outcome s Found to Occur Significantly More Frequently a in Children Less Than 24 Months of Age With Coexisting Systemic Sesis as Determined by Multivariate Analysis Gastrointestinal 1.5724 4.818 1.101 21.080 0.037 hemorrhage Seizures 1.5450 4.688 1.373 16.001 0.014 Other neurologic comlications 2.0493 7.763 1.142 52.760 0.036 a 0.05. Seizures 0.9846 2.677 1.033 6.935 0.0426 Other neurologic 1.4864 4.421 1.062 18.399 0.0410 comlications Culture-roven infections 0.7489 2.115 1.011 4.424 0.0467 a 0.05. Table 3. Outcome s With Trends Toward Occurring More Frequently in Children Aged 24 to 72 Months With Coexisting Systemic Sesis as Determined by Multivariate Analysis Dialysis or hemofiltration Cardioulmonary resuscitation required Cardiac arrhythmias 1.2197 3.386 0.788 14.555 0.101 1.7198 5.583 1.107 28.160 0.037 1.5950 4.929 0.762 31.881 0.094 Comment Mortality Results In this study, sesis was associated with resiratory failure in 76 of the 655 atients (11.6%). Although comarison by univariate analysis showed a lower survival in the setic children (36.8% versus 51.6%; 0.02), this was not confirmed in the multivariate model. Comonents of the model including older age, lower arterial H, and lower arterial oxygen tension before ECMO were negative redictors of survival. Overall survival results with ECMO for the management of resiratory failure in the ediatric oulation have not matched the excellent statistics seen in neonates. Survival rates more than 90% have been recorded for neonatal resiratory failure attributable to meconium asiration syndrome, and survival for all indications in the neonatal oulation averages 80% [8]. In the ediatric grou, the use of ECMO for resiratory suort has resulted in survival rates aroaching only 60%. The best results in this grou of atients have been those with asiration neumonia (64%), and the lowest survival rate has been noted for ediatric atients with neumocystis (38%) [9]. Part of the reason for this difference in survival between the neonatal and ediatric atient grous may relate to differences in the underlying disease conditions. In the neonate, resiratory failure is often the result of hysiologic erturbations such as ersistent ulmonary hyertension of the newborn or ersistent fetal circulation. These conditions are frequently reversible with the time rovided by ECMO suort. In the older atient, where resiratory failure is most commonly secondary to viral or bacterial neumonia, ulmonary dysfunction is less likely to be a fully reversible rocess. Furthermore, survival rates with ECMO rogressively decline with advancing age, even within the ediatric oulation. It is Table 4. Outcome s With Trends Toward Occurring More Frequently in Children Aged 72 to 156 Months With Coexisting Systemic Sesis as Determined by Multivariate Analysis Cardiac 2.1282 8.400 0.961 73.439 0.054 arrhythmias Seizures 2.1844 8.885 0.679 116.218 0.096 Culture-roven infections 2.1282 8.400 0.691 73.439 0.054
Ann Thorac Surg MEYER ET AL 1997;63:756 61 RESULTS OF ECMO IN CHILDREN WITH SEPSIS 759 unknown whether this trend is exlained by changing indications (including the infectious agents resonsible for neumonia) or by age-related differences in ulmonary suscetibility to injury or in roensity to develo adult resiratory distress syndrome or multisystem organ failure [10]. Other researchers have examined redictors of survival from ECMO in the ediatric oulation. O Rourke and associates [11] reviewed ELSO data from 1982 to 1991 for 285 atients aged 14 days to 17 years and found an overall survival rate of 47%. The most common indications for this theray were viral neumonia and adult resiratory distress syndrome. Predictors of decreased survival were examined by univariate techniques and found to include higher mean airway ressure, lower arterial H, and higher eak insiratory ressure. However, multivariate analysis was not alied and the resence or absence of sesis was not considered in this study. A retrosective, multivariate analysis of survival in 220 ediatric atients conducted by Moler and colleagues [10] found younger age, shorter duration of re-ecmo mechanical ventilation, lower eak insiratory ressure, lower alveolar arterial oxygen tension difference, and more recent ECMO exerience as redictors of imroved survival. Again, sesis was not included as a candidate variable in this model. Information on ECMO for ediatric atients with systemic sesis is available only from limited case studies. Farmer and colleagues [12] studied 8 atients where ECMO was used as salvage from surgical emergencies. In 3 ediatric atients where sesis was resent before ECMO, 2 survivors were reorted. Ehren and associates [13] reorted favorable results using ECMO in 12 atients treated rimarily for asiration or neumonia of varied causes. Nine atients were weaned from ECMO and 8 survived long term. However, the incidence of systemic sesis was not stated. Beca and Butt [14] reorted the use of ECMO in 9 children with culture-roven refractory setic shock. A survival rate of 56% suggested that setic shock should not be a contraindication to ECMO. Most other studies examining ECMO in setic atients have been restricted to neonates. Kornhauser and associates [4] reorted no survivors treated with ECMO for neonatal Grou B stretococcal neumonia, whereas Hocker and colleagues [7] observed an 87% survival rate in hyotensive neonates with this same disease rocess. Recently, we examined sesis as a redictor of survival in neonates using a multivariate analysis of ELSO data [5]. Neonates with sesis laced on ECMO achieved a survival benefit equal to those atients without sesis, although neurologic, renal, and metabolic comlications were more frequent. Nevertheless, alying neonatal results to the ediatric oulation may not be valid. Several studies have theorized that extracororeal suort might be deleterious in the resence of sesis where an inflammatory resonse has been initiated. Prolonged circulatory suort has been shown to cause elevated neutrohil CD11b levels [15, 16], decreased neutrohil counts [17], elevated neutrohil lactoferrin and elastase levels [15, 17], and increases in lasma C3a, C5b-9, and interleukin-8 [16, 18]. These changes suggest that activation of host inflammatory resonses may be accelerated by ECMO and the body s ability to combat infection may be altered. In contrast, DePuydt and colleagues [19] demonstrated that neutrohil hagocytosis and killing were not significantly affected by 5 days of ECMO suort. Thus, the overall balance of effects on the host resonse to infection is incomletely defined. In view of the similar survival found in the setic and nonsetic grous in this study, any adverse effect attributable to inflammatory factors does not aear to affect mortality. However, these factors may lay a role in some of the comlications arising in setic children. ECMO Comlications Several comlications were more common in the setic grou. These included the develoment of seizures ( 0.05), other neurologic comlications ( 0.05), and culture-roven infection at other sites ( 0.05). The redominance of neurologic roblems is likely not related to intracranial hemorrhage or infarction, as these events are coded searately on the ELSO data forms. However, some of these neurologic roblems may relate to hyoerfusion or cerebral microemboli or microhemorrhages that are not detectable by standard imaging techniques. Trends toward increased comlication rates in other organ systems (gastrointestinal hemorrhage, renal failure, cardiac arrhythmias) began to aear in analysis of some of the older age grous. Younger ediatric atients, who comrised the majority of this study, may be less suscetible to these roblems, and their rofile of comlications more closely aroximated those seen in neonates. Study Limitations Several imortant limitations of this study should be noted. The study is retrosective and sans a considerable time eriod during which both ECMO and conventional ventilatory techniques have evolved. For examle, 83% of the children in this study were treated with venoarterial ECMO, whereas currently venovenous techniques have become more oular in hemodynamically stable atients. Multivariate analysis was alied to control for the effects of other associated variables on survival. Nevertheless, these statistical techniques are limited and will not account for factors not included in the model, which may also affect survival. The analysis is further limited by the size of the study grou, articularly when older subgrous are studied indeendently. Here, tye II statistical errors (where actual differences are not found) remain a substantial risk. Moreover, the ELSO registry is a voluntary data base, and bias toward centers that are major contributors cannot be excluded. The ediatric ECMO oulation is very heterogeneous in age distribution, indication for ECMO, and secific organisms identified (in cases of neumonia). It remains ossible that sesis may influence outcome in some atient subsets. The definition of sesis itself is roblematic in this analysis. According to the ELSO registry, no
760 MEYER ET AL Ann Thorac Surg RESULTS OF ECMO IN CHILDREN WITH SEPSIS 1997;63:756 61 formal documentation of sesis, such as ositive blood cultures, is required for a atient to be designated as setic. Many of the atients were classified as setic based urely on clinical data, as the designation of sesis was at the discretion of the reorting institution. Moreover, identification of the actual target organism was rarely available. Also, the associated treatments for sesis (antibiotics and steroids) are not rovided in the data base, and the effects of these treatments cannot be evaluated. Finally, no long-term follow-u data are available from the registry and conclusions must be restricted to early outcomes only. Clearly, although ECMO is very effective in neonatal resiratory failure, results are significantly worse in the ediatric oulation. Examination of the raw survival data suggests that children with sesis have an even more dismal outlook. However, our analysis has demonstrated that the resence of systemic sesis does not indeendently redict a lower survival from ECMO in children when factors such as age, arterial blood gas results before ECMO, and ventilator arameters before ECMO are considered. On the basis of these data, ECMO should not be withheld from children solely on the basis of sesis. This study demonstrates the value of large data registries in the analysis of comlex questions where many variables need to be considered. Pooled data from multile centers may allow adequate numbers to better examine issues that may not be less well studied by any individual center. We are grateful to Thomas DeLosh and Charles J. H. Stolar, MD, of the Extracororeal Life Suort Organization for roviding access to the ELSO Registry data base, and to Donald McIntire, PhD, for assisting in the statistical analysis. References 1. Bartlett RH, Gazzaniga AB, Jefferies R, et al. Extracororeal membrane oxygenation (ECMO) cardioulmonary suort in infancy. Trans Am Soc Artif Intern Organs 1976;22:80 8. 2. O Rourke PP, Cone RK, Vacanti JP, et al. Extracororeal membrane oxygenation and conventional medical theray in neonates with ersistent ulmonary hyertension of the newborn: a rosective randomized study. Pediatrics 1989; 84:957 63. 3. Langham MR Jr., Krummel TM, Bartlett RH, et al. Mortality with extracororeal membrane oxygenation following reair of congenital diahragmatic hernia in 93 infants. J Pediatr Surg 1987;22:1150 4. 4. Kornhauser MS, Gilbert PL, Desai HJ, Branca PA. The efficacy of extracororeal membrane oxygenation (ECMO) in meconium asiration syndrome and grou B stretococcal neumonia. Pediatr Res 1988;23:414A. 5. Meyer DM, Jessen ME, Extracororeal Life Suort Organization. Results of extracororeal membrane oxygenation in neonates with sesis. The Extracororeal Life Suort Organization exerience. J Thorac Cardiovasc Surg 1995;109: 419 27. 6. McCune S, Short BL, Miller MK, Lotze A, Anderson KD. Extracororeal membrane oxygenation theray in neonates with setic shock. J Pediatr Surg 1990;25:479 82. 7. Hocker JR, Simson PM, Rabalais GP, Stewart DL, Cook LN. Extracororeal membrane oxygenation and early-onset grou B stretococcal sesis. Pediatrics 1992;89:1 4. 8. Toomasian JM, Snedecor SM, Cornell RG, Cilley RE, Bartlett RH. National exerience with extracororeal membrane oxygenation for newborn resiratory failure. ASAIO Trans 1988;34:140 7. 9. ECMO registry of the Extracororeal Life Suort Organization (ELSO). International summary. Ann Arbor, MI: University of Michigan Medical Center, January 1996. 10. Moler FW, Palmisano J, Custer JR. Extracororeal life suort for ediatric resiratory failure: redictors of survival from 220 atients. Crit Care Med 1993;21:1604 11. 11. O Rourke PP, Stolar CJH, Zwischenberger JB, Snedecor SM, Bartlett RH. Extracororeal membrane oxygenation: suort for overwhelming ulmonary failure in the ediatric oulation. Collective exerience from the Extracororeal Life Suort Organization. J Pediatr Surg 1993;28:523 9. 12. Farmer DL, Cullen ML, Philiart AI, Rector FE, Klein MD. Extracororeal membrane oxygenation as salvage in ediatric surgical emergencies. J Pediatr Surg 1995;30:345 8. 13. Ehren H, Palmer K, Eriksson M, Frenckner B. Pediatric ECMO for ulmonary suort: exerience from 12 cases. Acta Paediatr 1995;84:442 6. 14. Beca J, Butt W. Extracororeal membrane oxygenation for refractory setic shock in children. Pediatrics 1994;93:726 9. 15. Fortenberry J, Bhardwaj V, Bland L, Cornish D, Niemer P, Wright J. Effects of neonatal extracororeal membrane oxygenation (ECMO) on neutrohil activation and cytokine levels. Crit Care Med 1994;22:A211. 16. Moat NE, Rebuck N, Shore DF, et al. Humoral and cellular activation in a simulated extracororeal circuit. Ann Thorac Surg 1993;56:1509 14. 17. Wachtfogel YT, Kucich U, Greenlate J, et al. Human neutrohil granulation during extracororeal circulation. Blood 1987;69:324 30. 18. Salama A, Hugo F, Heinrich D, et al. Deosition of terminal C5b-9 comlement comlexes on erythrocytes and leukocytes during cardioulmonary byass. N Engl J Med 1988; 318:408 14. 19. DePuydt LE, Schuit KE, Smith SD. Effect of extracororeal membrane oxygenation on neutrohil function in neonates. Crit Care Med 1993;21:1324 7. Aendix 1. Outcome s Used in Analysis General outcome Survival Hemorrhagic comlications Intracranial infarct by comuted tomograhic scan Intracranial hemorrhage by comuted tomograhic scan Significant gastrointestinal hemorrhage Significant surgical site bleeding Other hemorrhagic roblem Neurologic comlications Brain death Probable or definite seizure Excessive jitteriness Other neurologic roblem Renal comlications Creatinine level more than 1.5 or less than 3.0 mg/dl Creatinine level more than 3.0 mg/dl Dialysis or hemofiltration Other renal roblem Cardiovascular comlications Cardioulmonary resuscitation required Cardiac arrhythmias Inotroes on ECMO Hyertension (systolic blood ressure 150 mm Hg; diastolic blood ressure 90 mm Hg) Other cardiovascular roblem
Ann Thorac Surg MEYER ET AL 1997;63:756 61 RESULTS OF ECMO IN CHILDREN WITH SEPSIS 761 Pulmonary comlications Pneumothorax requiring chest tube Other ulmonary roblem Setic comlications Culture-roven infection White blood cell count less than 1,500/ L Other infectious roblem Metabolic comlications Serum otassium level less than 2.5 meq/l Serum otassium level more than 7.5 meq/l Serum sodium level less than 120 meq/l Serum sodium level more than 160 meq/l Serum calcium level less than 6 meq/l Serum calcium level more than 14 meq/l Blood glucose level less than 20 mg/dl Blood glucose level more than 350 mg/dl H less than 7.05 H more than 7.75 Other metabolic roblem Aendix 2. s Before ECMO Used in Evaluation General Patient age at time of ECMO Patient sex Weight Most recent arterial blood gas before ECMO H Partial ressure of carbon dioxide Partial ressure of oxygen Most recent ventilator settings before ECMO Ventilator rate Fractional concentration of oxygen Other factors Systolic blood ressure Renal failure before ECMO Imortant Announcement In addition to ostal services and FAX, The Annals of Thoracic Surgery Editorial Office may now be reached by e-mail. The address is: ats@wudos2.wustl.edu The Annals of Thoracic Surgery Web site address is: htt://www.sts.org/annals/ 1997 by The Society of Thoracic Surgeons Ann Thorac Surg 1997;63:761 0003-4975/97/$17.00 Published by Elsevier Science Inc PII S0003-4975(97)00035-3