- Email: [email protected]
Neuroimaging of brain injury in neonates treated with extracorporeal membrane oxygenation: Lessons learned from serial examinations
Neuroimaging of Brain Injury in Neonates Treated With Extracorporeal Membrane Oxygenation: Lessons Learned From Serial Examinations By Eric L. Lazar, Sara J. Abramson,
Samuel Weinstein,
and Charles J.H. Stolar
New York, New York l The head ultrasound (US) is used extensively at most extracorporeal membrane oxygenation (ECMO) centers to screen for intracranial pathology in the ECMO candidate. Daily head US examinations are obtained in patients on ECMO to detect the emergence of intracranial hemorrhage (ICH). The authors asked whether these serial studies could be correlated with more definitive diagnostic studies, such as computed tomography (CT) and magnetic resonance imaging (MM) scans, autopsy data, or the long-term neurodevelopmental status, to discern the predictive value of these daily examinations. Seventy-four consecutively treated neonates with reversible respiratory failure refractory to conventional support met institutional criteria for placement on ECMO. In addition to a pre-ECMO US, daily real-time portable head US images were evaluated for changes in echotexture, ventricular configuration, and extraaxial fluid. Follow-up CT and MRI scans were evaluated for the presence of hemorrhagic or ischemic lesions. Autopsy data were obtained from nonsurvivors. Survivors were examined by a neurodevelopmental specialist at regular intervals and classified as normal or delayed for chronological age. In this series of 74 patients, CT/MM scanning and autopsy data demonstrated structural injury in 19 patients; there were 16 ischemic infarctions and three hemorrhages. The incidence of hemorrhage in this series was considerably lower than that previously reported. Ten of the 19 patients had serial head US findings demonstrating a progression from focal increases in echotexture to diffuse effacement of cerebral architecture. In the remaining nine, serial head US examinations did not show injury. An additional 10 children had a clear delay in neurological development despite no evidence of anatomic injury on serial head US examinations or CT/MRI scanning. The mortality was 42% for neonates who had either an abnormal head US result or an abnormal result on the follow-up neuroimaging study. In the 43 patients without evidence of such injury, the mortality was 16%. Copyright o 1994 by W.B. Saunders Company INDEX WORDS: Extracorporeal membrane (ECMO); cerebral injury, neuroimaging.
oxygenation
From the Divisions of Pediatric Surgery and Radiology, College of Physicians and Surgeons, The Babies Hospital, Columbia Presbyterian Medical Center, New York, NY Presented at the 24th Annual Meeting of the American Pediatric Surgical Association, Hilton Head, South Carolina, May 15-18, 1993. Supported by The Charles Edison Fund and The Anya Foundation. Address reprint requests to Charles J.H. Stolar, MD, Division of Pediatric Surgery, College of Physicians and Surgeons, The Babies Hospital, Columbia Presbyterian Medical Center, 3959 Broadway. Room 212N, New York, NY10032. Copyright o 1994 by WB. Saunders Company 0022-3468/94/2902-0010$03.00/O 186
E
XTRACORPOREAL membrane oxygenation (ECMO) has successfully treated refractory respiratory failure in full-term infants since 1978.lJ However, the neurological consequences of this therapy remain of paramount concern. In the early ECMO experience, intracranial hemorrhage (ICH) occurred in up to 35% of treated infants, although this has decreased to 15% with time.3,4 Several technical features of ECMO may potentiate neurological injury: systemic heparinization, platelet depletion by the extracorporeal circuit, and right carotid artery ligation. In the patient with sepsis, these may be superimposed on an ongoing coagulopathy. In most ECMO centers, the head ultrasound (US) examination is relied on to identify hemorrhage before beginning ECMO, and is used daily in the course of ECMO to detect new or evolving hemorrhage.5 Although head US is a sensitive tool for major ICH, it may not detect many clinically significant injuries.” For example, ischemic infarction may not be detected early in its clinical course. The evolution of serial head US findings in the context of the neurological examination, other neuroimaging techniques, autopsy data, and long-term neurodevelopmental outcome is uncertain. We reviewed the serial head US examinations of neonatal ECMO patients to determine whether there was a relationship to the available follow-up instruments-CT/MRI scans and long-term neurodevelopmental assessment for survivors, and autopsy data for nonsurvivors. MATERIALS
AND METHODS
Seventy-four consecutively treated neonates with reversible respiratory failure refractory to conventional pharmacological and ventilatory support met institutional criteria for ECMO: 36 to 42 weeks’ gestation, birth weight of more than 2 kg, absence of named genetic syndromes or multiple congenital anomalies, and no grade II or greater ICH on initial head US. Access for bypass was venoarterial via the right jugular and carotid vessels, which were ligated.
ECMO Management During ECMO, infants were awake, lightly sedated as needed, and were not paralyzed. Anticoagulation was achieved via continuous heparin infusion and monitored hourly, keeping activated clotting times (ACT) elevated 1.5 to 1.8 times normal (180 to 220 seconds). Platelet counts ranged from 75,000 to 150,000 and were maintained with volume-reduced transfusions. Bypass flows were weaned from 100 to 120 ml/kg/mm to 15 to 30 mL/kg/min to maintain the umbilical arterial PO? at 50 to 80 mm Hg, the PCOZat JournalofPediatric Surgery, Vol 29, No
2 (February), 1994: pp 166-191
ECMO, NEUROIMAGING,
187
AND OUTCOME
35 to 55 mm Hg, and the mixed venous oxygen saturations (Smvoz) above 65%. ECMO support was discontinued when hemodynamic stability was achieved, no vasoactive drugs were required, and measured arterial PO?, Pco?. Smvoz were normal on minimal ventilator settings: 35% inspired oxygen, ventilatory rates of 10 to 20 breaths per minute, and peak inspiratory pressures of 15 to 25 mm Hg. Anticoagulation was terminated at the completion of bypass. No vascular repairs were performed.
Table 1. Outcome for 74 ECMO-Treated
NormalHead US Group A/normal follow-up 43 Neonates
In addition to the pre-ECMO US, daily real-time portable head US images were obtained in both the coronal and sagittal planes. These were evaluated for changes in echotexture (increased, decreased, or normal), ventricular configuration (compressed, hydrocephalic, or normal), and extraaxial fluid (present or absent). Within the first week off ventilator support, CT scans were obtained for all infants and evaluated for the presence of hemorrhagic or ischemic lesions: MRI scans were obtained on an outpatient basis and evaluated similarly. Autopsy data were obtained for nonsurvivors. Survivors were examined by a neurodevelopmental specialist at 6 weeks, 6 months, and then annually until age 5. Based on these evaluations, the children were classified as normal or delayed for chronological age.
AbnormalHead US Group C/normal follow-up 10 Neonates
7 Nonneurological deaths 36 Well survivors Group B/abnormal follow-up
Neurological Evaluation
Neonates With Refractory
Respiratory Failure
Group D/abnormal follow-up
19 Neonates
12 Neonates
7 Deaths
6 Deaths
6 lschemic infarctions
4 lschemic infarctions
1 Hemorrhage
1 Hemorrhage
12 Survivors 4 Neuromotor delay 6 Neuromotor and cognitive delay 2 lschemic infarctions on MRI
1 Unknown 6 Survivors 4 lschemic infarctions
1 Hemorrhage 1 Unknown
(but normal neurodevelopment)
ECMO was excellent, with 83% survival, whereas the mortality with sepsis was 50%.
Statistical Analyses The x2 and Student’s t test methods were used to assess differences between infants who had normal and abnormal serial US findings. RESULTS
ECMO Outcome The 74 infants were placed on ECMO for diverse diagnoses (Fig 1): 34 congenital diaphragmatic hernia (CDH), 20 meconium aspiration syndrome (MAS), 10 sepsis, 8 persistent pulmonary hypertension of the newborn (PPHN), and 2 with other diagnoses. The mean gestational age was 39 weeks (range, 36 to 43 weeks). The average pre-ECMO alveolar-arterial gradient was 620 mm Hg for greater than 4 hours preceding ECMO, and the average umbilical artery Paz was 34 mm Hg (range, 7 to 44 mm Hg). In no child was hemorrhage detected on the pre-ECMO head US. The average length of bypass was 93.5 hours (range, 10 to 413 hours). The overall survival to discharge was 76%. The prognosis of CDH requiring
Fig 1. Entry diagnoses for 74 ECMO-treated neonates. CDH, congenital diaphragmatic hernia; MAS, meconium aspiration syndrome; PPHN, persistant pulmonary hypertension of the newborn.
Neurological Outcome Four distinct patient groups were defined by the results of serial head US findings (normal or abnormal) and the outcome of the follow-up instrumentsautopsy or CT/MRI scan as well as neurodevelopmental assessment (Table 1). Group A consisted of 43 patients whose neurological examinations were completely normal before, during, and after ECMO, without evidence of seizure, focal motor finding, or altered tone. The serial head US findings were normal as were the follow-up CT/MRI scans and the neurodevelopmental assessments. In this group of 43 neonates, there were 7 nonneurological deaths, with normal brain and brain-stem specimens at autopsy. Group B comprised 19 neonates who had abnormal follow-up CT/MRI results or abnormal neurodevelopmental assessments for chronological age despite normal pre- and intra-ECMO serial head US findings. Autopsy findings demonstrated six ischemic infarctions and one hemorrhage in the seven patients in this group who died; these lesions were not demonstrated on head US. Of the 12 survivors in this group, four are judged delayed in neuromotor development; six others have combined neuromotor and cognitive delay evident on examinations. The remaining two have focal loss of gray matter consistent with ischemic injury on MRI but are completely asymptomatic, with normal neurodevelopmental assessments for age to date. Group Chad normal follow-up imaging or neurodevelopmentai assessments despite abnormal serial head US findings, thus representing false-positive head US studies. There were no patients in this group.
Group D consisted of 12 patients who had abnormal serial head US findings. These studies showed generalized changes in echotexture, evolving from focal alterations in echodensity to complete effacement of cerebral architecture (Fig 2). The initial US results were normal in six patients; in the other six there were diffuse areas of increased ethos consistent with edema. The earliest specific abnormal finding was increased ethos localized to the region of the thalamus in five patients. Ultimately, the thalamus or perinuclear structures were involved in eight of the 12 patients. The earliest sonographic findings occurred on the left side in three cases, the right in five cases, and bilaterally in the remaining four cases. The six deaths in this group were attributed to ischemic infarction in four cases and hemorrhage in one; the remaining death was not autopsied. CT/MRI scans confirmed four ischemic infarctions and one hemorrhage in the surviving six, one survivor had no follow-up imaging. In this series of 74 ECMO-treated neonates, there were 16 ischemic infarctions and three hemorrhages in 19 patients. Autopsy or CT and MRI scans showed these brain injuries to be on the left in four cases, on the right in five cases, and bilaterally in 10 cases, indicating no lateralization of findings (u = .97). The hemorrhagic and ischemic lesions occurred equally in those with normal and abnormal head US studies (P = .OS). When stratified by the presence or absence of evidence for neurological morbidity, the overall mortality was greater (P = .OOl) for those with neurological morbidity (13 deaths in 31 cases) than for t hose without any evidence of neurological injury (seven deaths in 43 patients). DISCUSSION
The head US examination
is used extensively at
nnost ECMO centers to screen for intracranial pathol-
ogy such as hemorrhage in the ECMO candidate. At our institution, like many others, daily head US examinations are obtained in the patient on ECMO to detect the emergence of ICH. We asked whether these serial studies could be correlated with more definitive diagnostic studies, such as the CT/MRI scans, autopsy data, or the long-term neurodevelopmental status, to discern the predictive value of these daily examinations. In this series of 74 patients, CT/MRI scanning and autopsy data demonstrated structural injury in 19 patients. Ten of the 19 had serial head US findings demonstrating a progression from focal increases in echotexture to diffuse effacement of cerebral architecture. In the remaining nine, serial head US examinations did not show injury. Additionally, 10 children had a clear delay in neurological development despite no evidence of anatomic
Fig 2. Evolution of injury seen in serial head US. (A) Coronal view. There is focal change in echotexture (arrows) in the right temporal horn, imparting a “Swiss cheese” appearance. (B) Coronal view. Increased parenchymal edema can result in this pattern of serpigenous echodensity (arrows). (C) Sagittal view. There is complete effacement of cerebral architecture by diffuse increased echodensity.
ECMO, NEUROIMAGING,
AND OUTCOME
injury on CT/MRI scanning. Serial head US findings were normal in these children as well. There were no false-positive serial head US studies in this series. Viewed in total, these data suggest that a positive head US indicates likely neurological injury, but normal serial head US studies do not exclude the possibility of significant ischemic injury to the brain. Hemorrhagic Injuly
Hemorrhagic lesions were uncommon in this series, occurring in only 4% of the patients. Sell et al reported a 52% incidence of ICH in 25 ECMOtreated infants.’ In a larger series, Gross et al reported 91 intracranial hemorrhagic complications in 212 ECMO-treated neonates (43%).8 However, 22 of the 91 complications were grade I ICH according to the pre-ECMO US. In most other contemporary series, the reported incidence of intracranial hemorrhage attributable to ECMO is 10% to 35%.5,6,9There are two possible explanations for the significantly decreased rate of hemorrhagic lesions noted in this series. Many older studies include neonates who are younger than 35 weeks’ gestation and are at a greater risk of ICH because of the still-present germinal matrix. Bowerman et al reported eight hemorrhages in a series of 28 patients (28%), but this group included infants as young as 27 weeks’ gestation.5 In recognition of this increased risk, most ECMO centers exclude infants younger than 35 weeks’ gestation. The control of activated clotting times (ACT) may also be an issue. Babcock et al reported a 26% rate of intracranial hemorrhage in 50 infants whose ACTS were elevated 2.5 times normal9 Similarly, in the study by Sell et al, ACTS were kept 2.5 times normal.’ At our institution, ACTS are kept closer to control (1.5 to 1.8 times normal) in part because a high proportion (46%) of our ECMO cases are postoperative diaphragmatic hernia repairs; keeping ACTS tightly controlled to prevent postsurgical bleeding may have had some benefit in minimizing ICH. Ischemic Injuly
The most common structural injury demonstrated in this series was ischemic infarction. The thalamus was the first localized structure involved in 40% of patients whose head US result was abnormal; its involvement was ultimately noted on the serial examination in 75% of the 12 patients. Voit et al reported on four asphyxiated infants who had increased echodensity appearing specifically in the thalamus.10 The follow-up CT scan in these infants showed three infarctions and one hemorrhage. As in our series, US detected the injury but was unable to distinguish between infarction and hemorrhage. Similarly, Paster-
189
nak et al reported on two infants who had thalamic injury undetected by US and CT but clearly evident with MRI.” It has been proposed that this pattern of early thalamic injury suggests a brief period of total asphyxia.r2 The clinical correlate to brief total asphyxia would be circulatory arrest. This is distinct from partial prolonged ischemia in which there is diffuse cortical injury and edema without specific localization. This pattern of injury would seem to fit the respiratory history of most ECMO patients. However, we were unable to confirm brief total asphyxia in the eight patients who had thalamic injury. Lateralization of Injury
Much concern has been generated over carotid artery ligation and its effects on intracranial pressure and cerebral blood flo~.r~s’~ Presumably, ligation of the right carotid artery would be associated with right-sided cerebral injuries. We were unable to demonstrate lateralization of neurological injury in this series. In the acute setting, carotid artery ligation is well tolerated. Our data confirm this clinical observation; there was no right-left preference in either the initial head US findings or the follow-up autopsy or CT/MRI scans. This parallels previous observations by a number of authorities.‘-’ Currently, many centers are placing infants on venovenous bypass using a single coaxia1 double-lumen catheter placed in the right atrium via the jugular vein. This procedure does not involve the carotid artery and may further vitiate concern over this technical aspect. Overall, mortality is significantly increased in infants who have either an abnormal head US result or evidence of neurological morbidity. von Allmen et al reached similar conclusions.‘5 This is significant in that while head US findings cannot specifically predict outcome for a given infant, the parents may be counseled as to the likelihood of significant morbidity or mortality in the presence of an abnormal ultrasound either pre-ECMO or early in the course of ECMO. We speculate that pre-ECMO morbidity is largely responsible for neurological injury in ECMO patients. These infants have arterial PO? of less than 35 mm Hg for several hours preceding bypass, are acidotic with considerable oxygen debt, and have an estimated mortality of 80% unless treated with ECMO. We speculate that critically ill infants with a similar degree of prolonged hypoxemia but not treated with ECMO would have far worse neurological injury. However, survivorship in this cohort is small. As the complications of ECMO decrease with advances in anticoagulation and bypass technique, the margin of safety grows. This increase in the benefit/risk ratio would permit infants to be assigned to treatment with ECMO sooner, before prolonged hypoxemic damage occurs.
190
LAZAR ET AL
REFERENCES 1. Bartlett RH. Andrews AF. Toomasian JM, et al: Extracorporeal membrane oxygenation for newborn respiratory failure: Fortyfive cases. Surgery 92:425-433, 1982 2. Neonatal Extracorporeal Life Support Organization Registry: January 1993 Report, Ann Arbor, MI. 3. 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 84:957-963, 1989 4. Andrews AF, Roloff DN, Bartlett RH: Use of extracorporeal membrane oxygenation in persistent pulmonary hypertension of the newborn. Clin Perinatol 11:729-735,1984 5. Bowerman RA, Zwischenberger JB, Andrews AF, et al: Cranial sonography of the infant treated with extracorporeal membrane oxygenation. AJR 145:161-166,1985 6. Taylor GA, Fitz CR, Miller MK, et al: Intracranial abnormalities in infants treated with extracorporeal membrane oxygenation: Imaging with US and CT. Radiology 165:675-678, 1987 7. Sell LL, Cullen ML, Whittlesey GC, et al: Hemorrhagic complications during extracorporeal membrane oxygenation: Prevention and treatment. J Pediatr Surg 21:1087-1091, 1986 8. Gross GW, Baumgart S, Antunes MJ: Intracranial complications associated with ECMO in neonates-Patterns and frequency. Presented at the Children’s National Medical Center 9th Annual ECMO Symposium, Keystone, CO, March 1993
9. Babcock DS, Han BK, Weiss RG. et al: Brain abnormalities in infants on extracorporeal membrane oxygenation: Sonographic and CT findings. AJR 153:571-576, 1989 10. Voit T, Lemburg P, Neuen E, et al: Damage of thalamus and basal ganglia in asphyxiated full-term neonates. Neuropediatrics 18:176-181, 1987 11. Pasternak JF, Predey TA, Mikhael MA: Neonatal asphyxia: Vulnerability of basal ganglia, thalamus, and brainstem. Pediatr Neural 7:147-149, 1991 12. Myers RE: Two patterns of perinatal brain damage and the conditions of occurrence. Am J Obstet Gynecol 112:246-276, 1977 13. Stolar CJH, Reyes C: Extracorporeal membrane oxygenation causes significant changes in intracranial pressure and carotid artery blood flow in newborn lambs. J Pediatr Surg 23:1163-1168, 1988 14. Short BL, Walker LK, Traystman RJ: ECMO worsens the alteration of cerebral autoregulation caused by hypoxia. Presented at the Children’s National Medical Center 9th Annual ECMO Symposium, Keystone, CO, March 1993 15. von Allmen D, Babcock D, Matsumoto J, et al: The predictive value of head ultrasound in the ECMO candidate. J Pediatr Surg 27:36-39, 1992
Discussion J.M. Wilson (Boston, MA): The authors have pre-
sented their series of 74 carefully analyzed patients. The aim of their study was to determine whether ultrasound, which is the industry standard for preECMO and intra-ECMO evaluation, was a sensitive tool for predicting neurological outcome. Their conclusion was emphatically no. Follow-up is clearly essential. This is a point that Dr Lund and I hope to underscore when he presents our follow-up data with diaphragmatic hernia patients tomorrow. The lack of lateralization of the lesions continues to allay my fears about the dangers of carotid artery ligation. Nevertheless, conflicting reports exist, and personally I continue to reconstruct the carotid when feasible. The authors have also speculated, and I believe correctly, that neurological injury results from pre-ECMO hypoxia rather than from bypass itself. Their data, taken together with the fact that ECMO has become safer (much of ECMO can be done via the venovenous route now) would argue for earlier ECMO in order to avoid these neurological complications. Unfortunately, with ECMO available as a bail-out, many centers are becoming more aggressive at pushing conventional and high-frequency ventilation to their limits so that in fact today’s ECMO patients are often far sicker than those of only a few years ago. It wiil be interesting to see if the neurologi-
cal outcome of these patients is better or worse than that of predecessors. I will restrict my questions to one remarkable number. Hidden quietly in this manuscript are the best survival statistics for congenital diaphragmatic hernia (CDH) since Dr Gross reported an 89% survival in newborns. In the present study, 83% of CDH patients treated with ECMO survived. This is a remarkable statistic that is clearly leagues ahead of anybody else?. I would like to ask a few questions in that regard. First, is this a selected series that excludes any patients, and if so on what basis? Second, how have you defined survival? Is it decannulation from ECMO or discharge home? Finally, to what do you attribute your success? What is it that you are doing right that the rest of us can do? E.L. Lam- (response): I would like to thank Dr. Wilson for his prepared remarks and the consideration he has given this paper. He has asked about the the high survival rate for patients with the diagnosis of diaphragmatic hernia at our institution, and whether or not there is any selection bias in this sample. Under optimal ventilatory management, infants who do not have a preductal Po2 of near 100 mm Hg or better at some point prior to hernia repair are not offered ECMO. There have been eight such infants since 1983; if we include these children and assume that they would have died despite ECMO, the 83%
ECMO, NEUROIMAGING,
AND OUTCOME
survival we experienced would drop to about 79%. I guess that, strictly speaking, since we exclude some patients from analysis, there is selection bias, but the survival drops only marginally if we include these eight children. We have defined survival in this series as discharge home, not just to decannulation. I would speculate
191
the increased survival reflects less late mortality from chronic lung disease, which may be attributable to ventilatory strategies that include high-pressure hyperventilation. Because we use lower pressures to ventilate unparalyzed, spontaneously breathing infants, I believe we avoid crippling long-term pulmonary pathology.
Samuel Weinstein,
and Charles J.H. Stolar
New York, New York l The head ultrasound (US) is used extensively at most extracorporeal membrane oxygenation (ECMO) centers to screen for intracranial pathology in the ECMO candidate. Daily head US examinations are obtained in patients on ECMO to detect the emergence of intracranial hemorrhage (ICH). The authors asked whether these serial studies could be correlated with more definitive diagnostic studies, such as computed tomography (CT) and magnetic resonance imaging (MM) scans, autopsy data, or the long-term neurodevelopmental status, to discern the predictive value of these daily examinations. Seventy-four consecutively treated neonates with reversible respiratory failure refractory to conventional support met institutional criteria for placement on ECMO. In addition to a pre-ECMO US, daily real-time portable head US images were evaluated for changes in echotexture, ventricular configuration, and extraaxial fluid. Follow-up CT and MRI scans were evaluated for the presence of hemorrhagic or ischemic lesions. Autopsy data were obtained from nonsurvivors. Survivors were examined by a neurodevelopmental specialist at regular intervals and classified as normal or delayed for chronological age. In this series of 74 patients, CT/MM scanning and autopsy data demonstrated structural injury in 19 patients; there were 16 ischemic infarctions and three hemorrhages. The incidence of hemorrhage in this series was considerably lower than that previously reported. Ten of the 19 patients had serial head US findings demonstrating a progression from focal increases in echotexture to diffuse effacement of cerebral architecture. In the remaining nine, serial head US examinations did not show injury. An additional 10 children had a clear delay in neurological development despite no evidence of anatomic injury on serial head US examinations or CT/MRI scanning. The mortality was 42% for neonates who had either an abnormal head US result or an abnormal result on the follow-up neuroimaging study. In the 43 patients without evidence of such injury, the mortality was 16%. Copyright o 1994 by W.B. Saunders Company INDEX WORDS: Extracorporeal membrane (ECMO); cerebral injury, neuroimaging.
oxygenation
From the Divisions of Pediatric Surgery and Radiology, College of Physicians and Surgeons, The Babies Hospital, Columbia Presbyterian Medical Center, New York, NY Presented at the 24th Annual Meeting of the American Pediatric Surgical Association, Hilton Head, South Carolina, May 15-18, 1993. Supported by The Charles Edison Fund and The Anya Foundation. Address reprint requests to Charles J.H. Stolar, MD, Division of Pediatric Surgery, College of Physicians and Surgeons, The Babies Hospital, Columbia Presbyterian Medical Center, 3959 Broadway. Room 212N, New York, NY10032. Copyright o 1994 by WB. Saunders Company 0022-3468/94/2902-0010$03.00/O 186
E
XTRACORPOREAL membrane oxygenation (ECMO) has successfully treated refractory respiratory failure in full-term infants since 1978.lJ However, the neurological consequences of this therapy remain of paramount concern. In the early ECMO experience, intracranial hemorrhage (ICH) occurred in up to 35% of treated infants, although this has decreased to 15% with time.3,4 Several technical features of ECMO may potentiate neurological injury: systemic heparinization, platelet depletion by the extracorporeal circuit, and right carotid artery ligation. In the patient with sepsis, these may be superimposed on an ongoing coagulopathy. In most ECMO centers, the head ultrasound (US) examination is relied on to identify hemorrhage before beginning ECMO, and is used daily in the course of ECMO to detect new or evolving hemorrhage.5 Although head US is a sensitive tool for major ICH, it may not detect many clinically significant injuries.” For example, ischemic infarction may not be detected early in its clinical course. The evolution of serial head US findings in the context of the neurological examination, other neuroimaging techniques, autopsy data, and long-term neurodevelopmental outcome is uncertain. We reviewed the serial head US examinations of neonatal ECMO patients to determine whether there was a relationship to the available follow-up instruments-CT/MRI scans and long-term neurodevelopmental assessment for survivors, and autopsy data for nonsurvivors. MATERIALS
AND METHODS
Seventy-four consecutively treated neonates with reversible respiratory failure refractory to conventional pharmacological and ventilatory support met institutional criteria for ECMO: 36 to 42 weeks’ gestation, birth weight of more than 2 kg, absence of named genetic syndromes or multiple congenital anomalies, and no grade II or greater ICH on initial head US. Access for bypass was venoarterial via the right jugular and carotid vessels, which were ligated.
ECMO Management During ECMO, infants were awake, lightly sedated as needed, and were not paralyzed. Anticoagulation was achieved via continuous heparin infusion and monitored hourly, keeping activated clotting times (ACT) elevated 1.5 to 1.8 times normal (180 to 220 seconds). Platelet counts ranged from 75,000 to 150,000 and were maintained with volume-reduced transfusions. Bypass flows were weaned from 100 to 120 ml/kg/mm to 15 to 30 mL/kg/min to maintain the umbilical arterial PO? at 50 to 80 mm Hg, the PCOZat JournalofPediatric Surgery, Vol 29, No
2 (February), 1994: pp 166-191
ECMO, NEUROIMAGING,
187
AND OUTCOME
35 to 55 mm Hg, and the mixed venous oxygen saturations (Smvoz) above 65%. ECMO support was discontinued when hemodynamic stability was achieved, no vasoactive drugs were required, and measured arterial PO?, Pco?. Smvoz were normal on minimal ventilator settings: 35% inspired oxygen, ventilatory rates of 10 to 20 breaths per minute, and peak inspiratory pressures of 15 to 25 mm Hg. Anticoagulation was terminated at the completion of bypass. No vascular repairs were performed.
Table 1. Outcome for 74 ECMO-Treated
NormalHead US Group A/normal follow-up 43 Neonates
In addition to the pre-ECMO US, daily real-time portable head US images were obtained in both the coronal and sagittal planes. These were evaluated for changes in echotexture (increased, decreased, or normal), ventricular configuration (compressed, hydrocephalic, or normal), and extraaxial fluid (present or absent). Within the first week off ventilator support, CT scans were obtained for all infants and evaluated for the presence of hemorrhagic or ischemic lesions: MRI scans were obtained on an outpatient basis and evaluated similarly. Autopsy data were obtained for nonsurvivors. Survivors were examined by a neurodevelopmental specialist at 6 weeks, 6 months, and then annually until age 5. Based on these evaluations, the children were classified as normal or delayed for chronological age.
AbnormalHead US Group C/normal follow-up 10 Neonates
7 Nonneurological deaths 36 Well survivors Group B/abnormal follow-up
Neurological Evaluation
Neonates With Refractory
Respiratory Failure
Group D/abnormal follow-up
19 Neonates
12 Neonates
7 Deaths
6 Deaths
6 lschemic infarctions
4 lschemic infarctions
1 Hemorrhage
1 Hemorrhage
12 Survivors 4 Neuromotor delay 6 Neuromotor and cognitive delay 2 lschemic infarctions on MRI
1 Unknown 6 Survivors 4 lschemic infarctions
1 Hemorrhage 1 Unknown
(but normal neurodevelopment)
ECMO was excellent, with 83% survival, whereas the mortality with sepsis was 50%.
Statistical Analyses The x2 and Student’s t test methods were used to assess differences between infants who had normal and abnormal serial US findings. RESULTS
ECMO Outcome The 74 infants were placed on ECMO for diverse diagnoses (Fig 1): 34 congenital diaphragmatic hernia (CDH), 20 meconium aspiration syndrome (MAS), 10 sepsis, 8 persistent pulmonary hypertension of the newborn (PPHN), and 2 with other diagnoses. The mean gestational age was 39 weeks (range, 36 to 43 weeks). The average pre-ECMO alveolar-arterial gradient was 620 mm Hg for greater than 4 hours preceding ECMO, and the average umbilical artery Paz was 34 mm Hg (range, 7 to 44 mm Hg). In no child was hemorrhage detected on the pre-ECMO head US. The average length of bypass was 93.5 hours (range, 10 to 413 hours). The overall survival to discharge was 76%. The prognosis of CDH requiring
Fig 1. Entry diagnoses for 74 ECMO-treated neonates. CDH, congenital diaphragmatic hernia; MAS, meconium aspiration syndrome; PPHN, persistant pulmonary hypertension of the newborn.
Neurological Outcome Four distinct patient groups were defined by the results of serial head US findings (normal or abnormal) and the outcome of the follow-up instrumentsautopsy or CT/MRI scan as well as neurodevelopmental assessment (Table 1). Group A consisted of 43 patients whose neurological examinations were completely normal before, during, and after ECMO, without evidence of seizure, focal motor finding, or altered tone. The serial head US findings were normal as were the follow-up CT/MRI scans and the neurodevelopmental assessments. In this group of 43 neonates, there were 7 nonneurological deaths, with normal brain and brain-stem specimens at autopsy. Group B comprised 19 neonates who had abnormal follow-up CT/MRI results or abnormal neurodevelopmental assessments for chronological age despite normal pre- and intra-ECMO serial head US findings. Autopsy findings demonstrated six ischemic infarctions and one hemorrhage in the seven patients in this group who died; these lesions were not demonstrated on head US. Of the 12 survivors in this group, four are judged delayed in neuromotor development; six others have combined neuromotor and cognitive delay evident on examinations. The remaining two have focal loss of gray matter consistent with ischemic injury on MRI but are completely asymptomatic, with normal neurodevelopmental assessments for age to date. Group Chad normal follow-up imaging or neurodevelopmentai assessments despite abnormal serial head US findings, thus representing false-positive head US studies. There were no patients in this group.
Group D consisted of 12 patients who had abnormal serial head US findings. These studies showed generalized changes in echotexture, evolving from focal alterations in echodensity to complete effacement of cerebral architecture (Fig 2). The initial US results were normal in six patients; in the other six there were diffuse areas of increased ethos consistent with edema. The earliest specific abnormal finding was increased ethos localized to the region of the thalamus in five patients. Ultimately, the thalamus or perinuclear structures were involved in eight of the 12 patients. The earliest sonographic findings occurred on the left side in three cases, the right in five cases, and bilaterally in the remaining four cases. The six deaths in this group were attributed to ischemic infarction in four cases and hemorrhage in one; the remaining death was not autopsied. CT/MRI scans confirmed four ischemic infarctions and one hemorrhage in the surviving six, one survivor had no follow-up imaging. In this series of 74 ECMO-treated neonates, there were 16 ischemic infarctions and three hemorrhages in 19 patients. Autopsy or CT and MRI scans showed these brain injuries to be on the left in four cases, on the right in five cases, and bilaterally in 10 cases, indicating no lateralization of findings (u = .97). The hemorrhagic and ischemic lesions occurred equally in those with normal and abnormal head US studies (P = .OS). When stratified by the presence or absence of evidence for neurological morbidity, the overall mortality was greater (P = .OOl) for those with neurological morbidity (13 deaths in 31 cases) than for t hose without any evidence of neurological injury (seven deaths in 43 patients). DISCUSSION
The head US examination
is used extensively at
nnost ECMO centers to screen for intracranial pathol-
ogy such as hemorrhage in the ECMO candidate. At our institution, like many others, daily head US examinations are obtained in the patient on ECMO to detect the emergence of ICH. We asked whether these serial studies could be correlated with more definitive diagnostic studies, such as the CT/MRI scans, autopsy data, or the long-term neurodevelopmental status, to discern the predictive value of these daily examinations. In this series of 74 patients, CT/MRI scanning and autopsy data demonstrated structural injury in 19 patients. Ten of the 19 had serial head US findings demonstrating a progression from focal increases in echotexture to diffuse effacement of cerebral architecture. In the remaining nine, serial head US examinations did not show injury. Additionally, 10 children had a clear delay in neurological development despite no evidence of anatomic
Fig 2. Evolution of injury seen in serial head US. (A) Coronal view. There is focal change in echotexture (arrows) in the right temporal horn, imparting a “Swiss cheese” appearance. (B) Coronal view. Increased parenchymal edema can result in this pattern of serpigenous echodensity (arrows). (C) Sagittal view. There is complete effacement of cerebral architecture by diffuse increased echodensity.
ECMO, NEUROIMAGING,
AND OUTCOME
injury on CT/MRI scanning. Serial head US findings were normal in these children as well. There were no false-positive serial head US studies in this series. Viewed in total, these data suggest that a positive head US indicates likely neurological injury, but normal serial head US studies do not exclude the possibility of significant ischemic injury to the brain. Hemorrhagic Injuly
Hemorrhagic lesions were uncommon in this series, occurring in only 4% of the patients. Sell et al reported a 52% incidence of ICH in 25 ECMOtreated infants.’ In a larger series, Gross et al reported 91 intracranial hemorrhagic complications in 212 ECMO-treated neonates (43%).8 However, 22 of the 91 complications were grade I ICH according to the pre-ECMO US. In most other contemporary series, the reported incidence of intracranial hemorrhage attributable to ECMO is 10% to 35%.5,6,9There are two possible explanations for the significantly decreased rate of hemorrhagic lesions noted in this series. Many older studies include neonates who are younger than 35 weeks’ gestation and are at a greater risk of ICH because of the still-present germinal matrix. Bowerman et al reported eight hemorrhages in a series of 28 patients (28%), but this group included infants as young as 27 weeks’ gestation.5 In recognition of this increased risk, most ECMO centers exclude infants younger than 35 weeks’ gestation. The control of activated clotting times (ACT) may also be an issue. Babcock et al reported a 26% rate of intracranial hemorrhage in 50 infants whose ACTS were elevated 2.5 times normal9 Similarly, in the study by Sell et al, ACTS were kept 2.5 times normal.’ At our institution, ACTS are kept closer to control (1.5 to 1.8 times normal) in part because a high proportion (46%) of our ECMO cases are postoperative diaphragmatic hernia repairs; keeping ACTS tightly controlled to prevent postsurgical bleeding may have had some benefit in minimizing ICH. Ischemic Injuly
The most common structural injury demonstrated in this series was ischemic infarction. The thalamus was the first localized structure involved in 40% of patients whose head US result was abnormal; its involvement was ultimately noted on the serial examination in 75% of the 12 patients. Voit et al reported on four asphyxiated infants who had increased echodensity appearing specifically in the thalamus.10 The follow-up CT scan in these infants showed three infarctions and one hemorrhage. As in our series, US detected the injury but was unable to distinguish between infarction and hemorrhage. Similarly, Paster-
189
nak et al reported on two infants who had thalamic injury undetected by US and CT but clearly evident with MRI.” It has been proposed that this pattern of early thalamic injury suggests a brief period of total asphyxia.r2 The clinical correlate to brief total asphyxia would be circulatory arrest. This is distinct from partial prolonged ischemia in which there is diffuse cortical injury and edema without specific localization. This pattern of injury would seem to fit the respiratory history of most ECMO patients. However, we were unable to confirm brief total asphyxia in the eight patients who had thalamic injury. Lateralization of Injury
Much concern has been generated over carotid artery ligation and its effects on intracranial pressure and cerebral blood flo~.r~s’~ Presumably, ligation of the right carotid artery would be associated with right-sided cerebral injuries. We were unable to demonstrate lateralization of neurological injury in this series. In the acute setting, carotid artery ligation is well tolerated. Our data confirm this clinical observation; there was no right-left preference in either the initial head US findings or the follow-up autopsy or CT/MRI scans. This parallels previous observations by a number of authorities.‘-’ Currently, many centers are placing infants on venovenous bypass using a single coaxia1 double-lumen catheter placed in the right atrium via the jugular vein. This procedure does not involve the carotid artery and may further vitiate concern over this technical aspect. Overall, mortality is significantly increased in infants who have either an abnormal head US result or evidence of neurological morbidity. von Allmen et al reached similar conclusions.‘5 This is significant in that while head US findings cannot specifically predict outcome for a given infant, the parents may be counseled as to the likelihood of significant morbidity or mortality in the presence of an abnormal ultrasound either pre-ECMO or early in the course of ECMO. We speculate that pre-ECMO morbidity is largely responsible for neurological injury in ECMO patients. These infants have arterial PO? of less than 35 mm Hg for several hours preceding bypass, are acidotic with considerable oxygen debt, and have an estimated mortality of 80% unless treated with ECMO. We speculate that critically ill infants with a similar degree of prolonged hypoxemia but not treated with ECMO would have far worse neurological injury. However, survivorship in this cohort is small. As the complications of ECMO decrease with advances in anticoagulation and bypass technique, the margin of safety grows. This increase in the benefit/risk ratio would permit infants to be assigned to treatment with ECMO sooner, before prolonged hypoxemic damage occurs.
190
LAZAR ET AL
REFERENCES 1. Bartlett RH. Andrews AF. Toomasian JM, et al: Extracorporeal membrane oxygenation for newborn respiratory failure: Fortyfive cases. Surgery 92:425-433, 1982 2. Neonatal Extracorporeal Life Support Organization Registry: January 1993 Report, Ann Arbor, MI. 3. 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 84:957-963, 1989 4. Andrews AF, Roloff DN, Bartlett RH: Use of extracorporeal membrane oxygenation in persistent pulmonary hypertension of the newborn. Clin Perinatol 11:729-735,1984 5. Bowerman RA, Zwischenberger JB, Andrews AF, et al: Cranial sonography of the infant treated with extracorporeal membrane oxygenation. AJR 145:161-166,1985 6. Taylor GA, Fitz CR, Miller MK, et al: Intracranial abnormalities in infants treated with extracorporeal membrane oxygenation: Imaging with US and CT. Radiology 165:675-678, 1987 7. Sell LL, Cullen ML, Whittlesey GC, et al: Hemorrhagic complications during extracorporeal membrane oxygenation: Prevention and treatment. J Pediatr Surg 21:1087-1091, 1986 8. Gross GW, Baumgart S, Antunes MJ: Intracranial complications associated with ECMO in neonates-Patterns and frequency. Presented at the Children’s National Medical Center 9th Annual ECMO Symposium, Keystone, CO, March 1993
9. Babcock DS, Han BK, Weiss RG. et al: Brain abnormalities in infants on extracorporeal membrane oxygenation: Sonographic and CT findings. AJR 153:571-576, 1989 10. Voit T, Lemburg P, Neuen E, et al: Damage of thalamus and basal ganglia in asphyxiated full-term neonates. Neuropediatrics 18:176-181, 1987 11. Pasternak JF, Predey TA, Mikhael MA: Neonatal asphyxia: Vulnerability of basal ganglia, thalamus, and brainstem. Pediatr Neural 7:147-149, 1991 12. Myers RE: Two patterns of perinatal brain damage and the conditions of occurrence. Am J Obstet Gynecol 112:246-276, 1977 13. Stolar CJH, Reyes C: Extracorporeal membrane oxygenation causes significant changes in intracranial pressure and carotid artery blood flow in newborn lambs. J Pediatr Surg 23:1163-1168, 1988 14. Short BL, Walker LK, Traystman RJ: ECMO worsens the alteration of cerebral autoregulation caused by hypoxia. Presented at the Children’s National Medical Center 9th Annual ECMO Symposium, Keystone, CO, March 1993 15. von Allmen D, Babcock D, Matsumoto J, et al: The predictive value of head ultrasound in the ECMO candidate. J Pediatr Surg 27:36-39, 1992
Discussion J.M. Wilson (Boston, MA): The authors have pre-
sented their series of 74 carefully analyzed patients. The aim of their study was to determine whether ultrasound, which is the industry standard for preECMO and intra-ECMO evaluation, was a sensitive tool for predicting neurological outcome. Their conclusion was emphatically no. Follow-up is clearly essential. This is a point that Dr Lund and I hope to underscore when he presents our follow-up data with diaphragmatic hernia patients tomorrow. The lack of lateralization of the lesions continues to allay my fears about the dangers of carotid artery ligation. Nevertheless, conflicting reports exist, and personally I continue to reconstruct the carotid when feasible. The authors have also speculated, and I believe correctly, that neurological injury results from pre-ECMO hypoxia rather than from bypass itself. Their data, taken together with the fact that ECMO has become safer (much of ECMO can be done via the venovenous route now) would argue for earlier ECMO in order to avoid these neurological complications. Unfortunately, with ECMO available as a bail-out, many centers are becoming more aggressive at pushing conventional and high-frequency ventilation to their limits so that in fact today’s ECMO patients are often far sicker than those of only a few years ago. It wiil be interesting to see if the neurologi-
cal outcome of these patients is better or worse than that of predecessors. I will restrict my questions to one remarkable number. Hidden quietly in this manuscript are the best survival statistics for congenital diaphragmatic hernia (CDH) since Dr Gross reported an 89% survival in newborns. In the present study, 83% of CDH patients treated with ECMO survived. This is a remarkable statistic that is clearly leagues ahead of anybody else?. I would like to ask a few questions in that regard. First, is this a selected series that excludes any patients, and if so on what basis? Second, how have you defined survival? Is it decannulation from ECMO or discharge home? Finally, to what do you attribute your success? What is it that you are doing right that the rest of us can do? E.L. Lam- (response): I would like to thank Dr. Wilson for his prepared remarks and the consideration he has given this paper. He has asked about the the high survival rate for patients with the diagnosis of diaphragmatic hernia at our institution, and whether or not there is any selection bias in this sample. Under optimal ventilatory management, infants who do not have a preductal Po2 of near 100 mm Hg or better at some point prior to hernia repair are not offered ECMO. There have been eight such infants since 1983; if we include these children and assume that they would have died despite ECMO, the 83%
ECMO, NEUROIMAGING,
AND OUTCOME
survival we experienced would drop to about 79%. I guess that, strictly speaking, since we exclude some patients from analysis, there is selection bias, but the survival drops only marginally if we include these eight children. We have defined survival in this series as discharge home, not just to decannulation. I would speculate
191
the increased survival reflects less late mortality from chronic lung disease, which may be attributable to ventilatory strategies that include high-pressure hyperventilation. Because we use lower pressures to ventilate unparalyzed, spontaneously breathing infants, I believe we avoid crippling long-term pulmonary pathology.