Ann Thorac Surg 1995;59:975-980
© 1995 The Society of Thoracic Surgeons
Neonatal Extracorporeal Membrane Oxygenation Complicated by Sepsis
Dan M. Meyer, MD,
Michael E. Jessen, MD,
Robert C. Eberhart, PhD Extracorporeal Life Support Organization
Division of Thoracic and Cardiovascular Surgery, Department of Surgery, University of Texas Southwestern Medical Center, Dallas, Texas and the University of Michigan, Ann Arbor, Michigan
Accepted for publication January 9, 1995.
 |
Abstract
|
|---|
The onset of sepsis in neonates while on extracorporeal membrane oxygenation (ECMO) may portend adverse results. Nevertheless, ECMO has been used as a therapy in the management of septic conditions. This study assessed morbidity and mortality in neonates in whom septic complications developed while they were on ECMO. Of 5,123 neonates in the Extracorporeal Life Support Organization Registry undergoing ECMO for nonseptic indications, 217 patients had development of septic complications. A multivariate logistic regression analysis that considered 15 pre-ECMO criteria was performed to evaluate outcome. Mortality was higher in the septic group (35% versus 17%; p < 0.002) and ECMO duration averaged 85 hours longer (p < 0.001). Septic neonates had a greater frequency of complications including seizures, gastrointestinal bleeding, renal dysfunction, and metabolic problems (all p < 0.05). Transfusion requirements were doubled. Oxygenator thrombi and hemofilter malfunction occurred more often in septic patients (p < 0.03). New strategies to prevent sepsis and associated thrombotic and metabolic complications may be indicated. A critical reappraisal of continued aggressive support may be warranted when septic complications develop in neonates during ECMO.
|
Introduction
|
|---|
Since its introduction in 1975, extracorporeal membrane oxygenation (ECMO) has evolved into the standard of care for neonates with severe respiratory failure [1]. The use of this technology is widely accepted for infants with meconium aspiration syndrome [2], persistent pulmonary hypertension of the newborn [3], and congenital diaphragmatic hernia with persistent fetal circulation [4]. Extracorporeal membrane oxygenation has also been used in neonates with a primary diagnosis of sepsis, with acceptable results [5, 6]. However, when sepsis develops in patients already on ECMO, less information is available to assess the benefits of this therapy. Unfortunately, other standard therapeutic options in seriously ill neonates are largely ineffective in this setting. The development of sepsis during ECMO is infrequent and any individual center is unlikely to have significant experience with this problem. Therefore, data from the Extracorporeal Life Support Organization (ELSO) Registry was analyzed to evaluate critically the effectiveness of this technology in neonates in whom sepsis develops while they are on ECMO.
|
Material and Methods
|
|---|
Patients
This study consisted of 5,123 neonates (aged 14 days or less) placed on ECMO for nonseptic indications betweenNovember 1987 and June 1993 and entered in the ELSO Registry. The frequency of each indication is presented in Figure 1
. Two hundred seventeen patients (4.2%) in this series had development of sepsis while on ECMO and were identified as group 1. Group 2 included all patients on ECMO for nonseptic indications in whom sepsis did not develop (n = 4,906). One thousand sixty patients were placed on ECMO for a primary indication of sepsis and were not included in this analysis.
View larger version (54K):
[in this window]
[in a new window]
|
Fig 1. . Frequency distribution of 5,123 neonates placed on extracorporeal membrane oxygenation for nonseptic indications. (CDH = congenital diaphragmatic hernia; MAS = meconium aspiration syndrome; PPHN = persistent pulmonary hypertension of the newborn; RDS = respiratory distress syndrome.)
|
|
Outcome Variables
Outcome variables studied included: ``survival'' (at time of reporting), ``recovery'' (of lung function leading to discontinuation of ECMO), and duration of ECMO support (in hours). Forty-six specific ECMO complications were available for analysis. Additional variables used in the comparison were the volume of packed red blood cells, fresh frozen plasma, and platelets administered during the peri-ECMO period, and the use of drugs (vasoactive drugs, bicarbonate, and surfactant) in the ECMO patients. The 63 outcome variables are listed in Appendix 1.
Pre-Extracorporeal Membrane Oxygenation Factors
In addition to the development of sepsis, 15 variables were selected for inclusion in a model for prediction of outcome from ECMO. These variables included patient age at the time of ECMO, sex, birth weight, and history of cardiopulmonary arrest before ECMO. Also, the results of the most recent arterial blood gas analysis before ECMO, the ventilator parameters of the patient immediately before initiation of ECMO, and the most recent ``oxygenation index'' (calculated as [fractional concentration of oxygen {FiO2} x mean airway pressure x 100]/arterial oxygen tension) were used in the analysis. The type of ECMO was not included, as 89% underwent ECMO through a venoarterial technique. The pre-ECMO variables used are listed in Appendix 2.
Statistical Analysis
Data were analyzed using the SAS software package (SAS Institute, Cary, NC). A univariate analysis of all outcome variables was performed initially to compare results between groups 1 and 2. A 
2 analysis was used for discrete variables, with a p value less than 0.05 by two-tailed Fisher's exact test considered significant. For continuous variables, Student's t test with Welch's approximation for unequal variances was used to compare results between the two groups.
To account for interactions between pre-ECMO variables, additional tests were performed. For dichotomous outcome variables, a multivariate stepwise logistic regression analysis for each outcome event was performed. All variables listed in Appendix 2 were candidates for entry into the model. For each resulting model, the ``group'' variable was always included, whereas only those remaining variables that met the 0.05 significance level were entered. For each element in the model, a parameter estimate was calculated from which a p value and an odds ratio for the variable were derived.
For continuous outcome variables, a stepwise multiple regression analysis was performed. All pre-ECMO candidate variables were the same (Appendix 2). Parameter estimates for the group variable in the resulting models were calculated. From each model, a p value of less than 0.05 was considered significant for the group variable.
|
Results
|
|---|
Overall Results
For the entire group of 5,123 neonates, survival was documented as 82.1%. Lung recovery leading to separation from ECMO occurred in 87.0% of the total population. Although any individual ECMO complication occurred infrequently, at least one complication occurred in 3,873 patients (75.6%) and more than one complication was documented in 2,686 patients (52.4%). Complications occurred more frequently in group 1, with 58.5% of this group having more than three of the complications listed in Appendix 1. In contrast, only 20.3% of group 2 patients had more than three complications.
Univariate Analysis
A univariate analysis of each outcome variable was undertaken to compare the 217 neonates with the complication of sepsis while on ECMO (group 1) against the remaining 4,906 patients (group 2). Outcome variables that were found to be significantly different between groups are listed in Tables 1 and 2 along with their mean values or frequency of occurrence, as appropriatetab 1,2Au: have renumbered tables. Of note, mortality was significantly worse in group 1 (35%) when compared with group 2 (17%), p < 0.001. Recovery of lung function was also less likely (p < 0.001) by this analysis. Duration of ECMO averaged 85 hours longer for septic patients (p < 0.001). Patients with sepsis while undergoing ECMO were also more likely to sustain neurologic, hemorrhagic, renal, hypertensive, and metabolic (hyperglycemia, hypoglycemia, and hypernatremia) complications. Red blood cell transfusion requirements were significantly greater in the septic group as well.
View this table:
[in this window]
[in a new window]
|
Table 1. . Outcome Variables Found to Occur Significantly More Frequently (p < 0.05) in Patients Developing Sepsis as Determined by Univariate Analysis
|
|
View this table:
[in this window]
[in a new window]
|
Table 2. . Outcome Variables Found to Occur Significantly Less Frequently (p < 0.05) in Patients Developing Sepsis as Determined by Univariate Analysis
|
|
Multivariate Analysis
To account for potential interactions between pre-ECMO variables, a stepwise regression was performed for each outcome variable as described above. Outcome variables that were found to be significantly different between the two groups at p < 0.05 are listed in Tables 3 and 4tab 3,4. As expected, fewer outcomes were significantly different between groups by this analysis, but hemorrhagic, renal, metabolic, and mechanical complications remained important. The development of sepsis was associated with a greater need for red cell transfusion. Survival, lung recovery, and ECMO duration remained significantly different, as well. A total of 17 complications occurred more frequently in those neonates in whom sepsis developed on ECMO than in those for whom sepsis was not a complication.
View this table:
[in this window]
[in a new window]
|
Table 3. . Outcome Variables Found to Occur Significantly More Frequently (p < 0.05) in Patients Developing Sepsis as Determined by Multivariate Analysis
|
|
View this table:
[in this window]
[in a new window]
|
Table 4. . Outcome Variables Found to Occur Significantly Less Frequently (p < 0.05) in Patients Developing Sepsis as Determined by Multivariate Analysis
|
|
|
Comment
|
|---|
Previous studies examining the role of ECMO in neonatal sepsis have provided conflicting data. Hocker and colleagues [6], in a retrospective study, found ECMO to have a survival advantage in neonates with early onset group B sepsis who developed acidosis or hypotension, or both, even if only minimal respiratory support was necessary. They reported an 87% survival in this subset of patients treated with ECMO. McCune and associates [5], in a series of patients with septic shock also reported encouraging survival rates with ECMO. Olson and colleagues [7] found a 70% survival using ECMO in 10 neonates with sepsis, not significantly different from results with neonates placed on ECMO for other reasons. In contrast, a small study by Kornhauser and colleagues [8] found a significantly decreased survival in neonates placed on ECMO for group B streptococcal pneumonia when compared to outcome in patients with meconium aspiration syndrome [2]. Thus, some small series have supported the use of ECMO for a primary indication of sepsis. However, no sizeable report has focused on the efficacy of ECMO when sepsis intervenes in patients without underlying infection.
This study demonstrated that the development of sepsis during extracorporeal membrane oxygenation support was an independent predictor of worsening survival, with raw survival data indicating a twofold increase in mortality rates in patients in whom sepsis developed while they were on ECMO. After accounting for 15 additional variables, survival remained significantly worse in this group, with an odds ratio of 0.372. Therefore, the development of sepsis during ECMO was a clear negative predictor of survival. Although this finding in itself is not surprising, further analysis of the associated complications that occur in this group provides some insights into the mechanisms behind this poor outcome.
Lung recovery leading to discontinuation of ECMO was markedly worse in neonates in whom sepsis developed (odds ratio, 0.272). As pulmonary failure is the underlying process involved in all of these neonates, inability to achieve separation from ECMO support will clearly play a major role in the survival issue. McCune and colleagues [5] found lung recovery to be slower in septic patients, relating this finding to pulmonary parenchymal or vascular damage mediated through complement activation and release of white blood cell inflammatory factors.
The significantly prolonged duration of support required in the neonates in whom sepsis developed may contribute to the frequency of complications observed in this group. By multivariate analysis, development of sepsis independently predicted the need for an additional 79 hours of ECMO support. The complications and outcome variables that occurred more frequently in those in whom sepsis developed during ECMO (Tables 3, 4
) involved multiple organ systems. Hemorrhagic complications were a frequent problem, with gastrointestinal hemorrhage requiring transfusion, bleeding at surgical sites, and other hemorrhagic problems significantly more common by multivariate analysis. Of interest, intracranial bleeding defined by either ultrasound, computed tomography, or magnetic resonance imaging was not significantly different between the two groups. By multivariate analysis, birth weight and acidosis before ECMO were the strongest predictors of this problem and may have rendered other factors unimportant. Nevertheless, seizures were more common in the group in whom sepsis developed.
In association with the increased bleeding complications, evidence of thrombotic problems, such as clots in the circuit, hemofilter malfunction, and other mechanical problems, were identified more frequently in the septic group. Extracorporeal support is known to induce problems with coagulation [8] and cause qualitative and quantitative platelet abnormalities [9]. McCune and associates [5], who found ECMO to be a viable option for septic neonates not responding to conventional therapy, also described a significant increase in hemorrhagic complications. These investigators suggested that the increased hemorrhagic problems may be related to hematogenous toxins released during sepsis, resulting in a consumptive coagulopathy. Such a process could explain the frequent observations of oxygenator thrombi and clots in the circuit found in septic patients in the present study. Close monitoring of activated clotting times and alterations in heparin management might decrease hemorrhagic or thrombotic complications and impact survival rates in these patients.
An increased incidence of complications in other systems was noted in the septic group. An elevated serum creatinine level was more likely (odds ratio, 2.076; p < 0.04) and the frequency of hypertension requiring vasodilators was nearly tripled. These findings, in addition to the greater incidence of hypernatremia, may reflect renal damage imparted by the underlying infection, treatments designed to combat the infectious agent, or an interaction of these factors with extracorporeal circulatory support. Alterations in the host response to infection may play a role in the adverse outcomes observed in the septic group, but no exact mechanism has been devised. Similarly, alterations in blood glucose levels found more often in neonates in whom sepsis developed may be a sequela of the septic process or the physiologic response to this stress.
Data from the ELSO Registry are incomplete regarding the type of organism involved or to the exact site of sepsis when it occurs in these patients. The timing of septic events during extracorporeal support also is not provided. As well, information on antibiotic use in these patients is not available. Whether the complications that occur may be avoided or reduced in severity by increasing the spectrum of the antibiotic prophylaxis or by early identification and intervention remains speculative. However, the findings in this study suggest potential strategies to combat unfavorable outcomes in patients in whom sepsis develops. First, a regimen of prophylaxis against infectious problems should be considered. Although most centers routinely cover patients on ECMO on antibiotics, more frequent surveillance cultures may be warranted to enable initiation of appropriate therapy in a timely manner. When sepsis occurs, early aggressive treatment of the primary source is warranted. Second, close monitoring of coagulation parameters may be beneficial in these patients. Further research into the underlying hematologic and inflammatory mediator responses that are induced by both ECMO and sepsis may yield other techniques to combat these problems. Strategies capable of reducing both hemorrhagic and thrombotic complications would appear to be particularly useful in this subset of patients. An example of such a strategy could include the use of heparin-bonded circuits, which has been shown to inhibit the initial formation of C5a [10].
This retrospective study contains several important limitations. First, not all patients undergoing ECMO are captured by the Registry, and a bias from centers that do report cannot be excluded. Second, there may be an underreporting of the incidence of septic complications, as only 217 of 5,123 neonates in this study (4.2%) had documented septic complications. Some septic problems may have escaped diagnosis, especially more minor episodes. Data are incomplete as to the infectious organism involved and the management scheme used to treat the septic condition. These factors may influence complications in other organ systems. Finally, no long-term follow-up data are available, and the ultimate fate of surviving patients remains unknown.
As with any therapy, one must evaluate continually the benefits of the therapy as conditions in the patient's clinical status change. The development of sepsis in patients on ECMO for respiratory failure clearly alert the physician of an increased likelihood of adverse outcome. Although ECMO is very effective in neonatal respiratory failure, results are significantly worse when sepsis develops. Patients in whom sepsis develops while they are on ECMO spend an average of 3 additional days on extracorporeal support and require multiple additional therapeutic interventions to treat other complications. Survival rates decline and neurologic, hemorrhagic, and metabolic complications occur more frequently. Mechanical complications (oxygenator failure and hemofilter malfunction) are frequent sequelae of new onset sepsis. Although cost data are not available directly, undoubtedly the expenses are magnified in this group.
At present, our analysis cannot identify a definite subset of patients in whom mortality rates approach those of conventional therapy. Therefore, it is difficult to exclude any particular patient from continued therapy. However, these data suggest that efforts must be taken to prevent sepsis during ECMO. Moreover, a heightened awareness of potential complications is required when sepsis develops. A periodic critical reappraisal of continuing aggressive support may be warranted when neonates develop septic complications. Further work will be required to evaluate predictors of outcome in this group and to design appropriate new strategies for preventing and managing the complications that may arise.
|
Appendix 1. Outcome Variables Used in Analysis
|
|---|
- Survival
- Recovery of lung function leading to discontinuation of ECMO
- Duration of ECMO support
- Oxygenator failure
- Raceway rupture
- Other tubing rupture
- Pump malfunction
- Heat exchanger malfunction
- Clots in circuit
- Cannula placement or reposition problems
- Restrictive sutures
- Hemofilter malfunction
- Kinking in cannula
- Other mechanical problem
- Creatinine level more than 1.5 mg/dL
- Dialysis or hemofiltration
- Other renal problem
- Hemorrhagic Complications
- Gastrointestinal hemorrhage requiring transfusion
- Bleeding at cannula site
- Bleeding at other surgical site
- Hemolysis
- Other hemorrhagic problem
- Cardiopulmonary Complications
- Pulmonary hemorrhage
- Cardiopulmonary resuscitation required
- Myocardial stunning
- Cardiac arrhythmias
- Symptomatic patent ductus arteriosus
- Hypertension requiring vasodilator therapy
- Other cardiopulmonary problem
- Pneumothorax requiring treatment
- Other pulmonary problem
- Brain death
- Seizure
- Infarction or hemorrhage by head ultrasound
- Infarction or hemorrhage by computed tomography or mag netic resonance imaging
- Other neurologic problem
- Serum potassium level less than 2.5 mEq/L
- Serum potassium level more than 6.0 mEq/L
- Serum sodium level less than 125 mEq/L
- Serum sodium level more than 150 mEq/L
- Serum calcium level less than 6 mg/dL
- Serum calcium level more than 12 mg/dL
- Blood glucose level less than 40 mg/dL
- Blood glucose level more than 240 mg/dL
- Other blood sugar problems
- pH less than 7.20
- pH more than 7.60
- Hyperbilirubinemia
- Other metabolic problems
- Volume of packed red blood cells transfused
- Volume of fresh frozen plasma transfused
- Volume of platelets administered
- Dopamine administered
- Dobutamine administered
- Tolazoline administered
- Nitroprusside administered
- Isoproterenol administered
- Prostaglandin E administered
- Epinephrine administered
- Nitroglycerine administered
- Other vasoactive drug administered
- Surfactant administered
- Bicarbonate administered
|
Appendix 2. Pre-Extracorporeal Membrane Oxygenation Variables Used in Evaluation
|
|---|
- 1. Patient age at time of ECMO
- 2. Patient sex
- 3. Birth weight
- 4. Cardiopulmonary arrest before ECMO
- Most recent arterial blood gas prior to ECMO
- 5. pH
- 6. Carbon dioxide tension
- 7. Oxygen tension
- 8. Bicarbonate
- 9. Oxygen saturation
- Most recent ventilator settings before ECMO
- 10. Ventilator rate
- 11. FiO2Au: OK as on p 5
- 12. Peak inspiratory pressure
- 13. Positive end expiratory pressure level
- 14. Mean airway pressure
- 15. Oxygenation index (FiO2 x mean airway pressure x 100/oxygen tension)
|
Acknowledgments
|
|---|
We are grateful to Thomas DeLosh and Charles J. H. Stolar, MD, of the Extracorporeal Life Support Organization for providing access to the ELSO Registry database, and to Donald McIntire, PhD, for reviewing the statistical analysis.
|
Footnotes
|
|---|
Presented in part at the Tenth Annual Children's National Medical Center ECMO Symposium, Keystone, CO, Feb 27-Mar 3, 1994.
Address reprint requests to Dr Meyer, Department of Thoracic and Cardiovascular Surgery, The Cleveland Clinic Foundation, 9500 Euclid Ave, Cleveland, OH 44195-5066.
|
References
|
|---|
- Bartlett RH, Gazzaniga AB, Jefferies R, et al. Extracorporeal membrane oxygenation (ECMO) cardiopulmonary support in infancy. Trans Am Soc Artif Intern Organs 1976;22:80–8.[Medline]
- Bartlett RH, Gazzaniga AB, Toomasian J, et al. Extracorporeal membrane oxygenation (ECMO) in neonatal respiratory failure. Ann Surg 1986;204:236–45.[Medline]
- O'Rourke PP, Crone RK, Vacanti JP, et al. Extracorporeal membrane oxygenation and conventional medical therapy in neonates with persistent pulmonary hypertension of the newborn: a prospective randomized study. Pediatrics 1989;84:957–63.[Abstract/Free Full Text]
- Langham MR Jr, Krummel TM, Bartlett RH, et al. Mortality with extracorporeal membrane oxygenation following repair of congenital diaphragmatic hernia in 93 infants. J Pediatr Surg 1987;22:1150–4.[Medline]
- McCune S, Short BL, Miller MK, et al. Extracorporeal membrane oxygenation therapy in neonates with septic shock. J Pediatr Surg 1990;25:479–82.[Medline]
- Hocker JR, Simpson PM, Rabalais GP, et al. Extracorporeal membrane oxygenation and early-onset Group B Streptococcal sepsis. Pediatrics 1992;89:1–4.[Abstract/Free Full Text]
- Olson GL, Couch CL, McDonald JV. ECMO experience with septic newborns. Sixth Annual Children's Hospital National Medical Center ECMO Symposium. Breckenridge, CO, February 1990. A17.
- Kornhauser MS, Gilbert PL, Desai HJ, et al. The efficacy of extracorporeal membrane oxygenation (ECMO) in meconium aspiration syndrome and Group B Streptococcal pneumonia. Pediatr Res 1988;23:414A.
- Robinson TM, Kickler TS, Walker LK, et al. Effect of extracorporeal membrane oxygenation on platelets in newborns. Crit Care Med 1993;21:1029–34.[Medline]
- Mollnes TE, Videm V, Gotze O, et al. Formation of C5a during cardiopulmonary bypass: inhibition by precoating with heparin. Ann Thorac Surg 1991;52:92–7.[Abstract]
Top
Abstract
Introduction
