Cerebrovascular Complications and Neurodevelopmental Sequelae of Neonatal Ecmo

NEUROLOGIC DISORDERS IN THE NEWBORN PART I

0095-5108/97 $0.00

+

.20

CEREBROVASCULAR COMPLICATIONS AND NEURODEVELOPMENTAL SEQUELAE OF NEONATAL ECMO Leonard J. Graziani, MD, Marcy Gringlas, PhD, and Stephen Baumgart, MD

Extracorporeal membrane oxygenation (ECMO) uses prolonged cardiopulmonary bypass to reduce mortality rates from about 80% to 20% in selected neonates with severe but reversible cardiorespiratory failure refractory to conventional ventilatory therapy. 3• 41 About 1 in 2000 live births results in a term or near-term neonate who develops neonatal cardiorespiratory failure that is sufficiently severe to be considered a candidate for ECMO intervention. The computer registry of the Extracorporeal Life Support Organization (ELSO) reported that 11,921 neonates since about 1985 have received ECMO intervention. 15 Until recently, venoarterial ECMO requiring cannulation of the right common carotid artery (RCCA) and the right jugular vein has been the predominant neonatal bypass procedure (Fig. 1). In 1990, venovenous ECMO using a single cannula with a double lumen inserted into the inferior vena cava via the jugular vein was reported to be effective in selected infants, thereby avoiding RCCA cannulation. 15 Venovenous ECMO is now often selected for neonates who have a jugular vein size adequate for the double-lumen cannula and who do not have severe circulatory failure requiring arterial pump support. Venoarterial ECMO is often begun when the neonate is at a relatively high risk for hypoxic-ischemic injury, may result in altered brain blood flow,7· 37- 39 • 61 and is associated with a high incidence of neonatal neuroimaging abnormalities.23· 28· 35 • 49 • 59 • 60 • 66• 67 In survivors of venoarterial ECMO, overt neurologic and audiologic sequelae occur in 10% to 20%,H~20· 22 • 64 • 66• 70 and another 20% to 30% with no evidence of a severe handicap at ages 1 to 3 years have cognitive and This work supported by NIH grants NS-27463 and HND21453.

From the Divisions of Child Neurology and Development (LJG) and Neonatology (SB), Departments of Pediatrics (LJG, MG, SB), and Neurology (LJG), Thomas Jefferson University Hospital, Jefferson Medical College of Thomas Jefferson University, Philadelphia, Pe1msylvania

CLINICS IN PERINATOLOGY VOLUME 24 ·NUMBER 3 ·SEPTEMBER 1997

655

656

GRAZIANI et al

@ Figure 1. Venoarterial ECMO circuit with cannulation of the right common carotid artery and right jugular vein. Shown are four major circuit components: (1) the venous and arterial cannulae positioned in the inferior vena cava and aortic arch, respectively (the bridge between the cannulae is clamped to permit blood to flow from the infant's vena cava, through the circuit, and return reoxygenated blood through the arterial cannula); (2) the servoregulated roller pump to maintain blood flow within the circuit; (3) the artificial membrane lung for exchange of gases between the infant's venous blood and the external source; and (4) the heat exchanger for rewarming blood prior to recirculation to the infant. Traps to exclude clots, emboli or gas bubbles, and physiologic or mechanical monitoring modules incorporated into the ECMO circuit are not shown. (From Baumgart S: Extracorporeal membrane oxygenation. Jn Spitzer AR [ed]: Intensive Care of the Fetus and Neonate. St. Louis, Mosby, 1996; with permission.)

visual-motor deficiencies at early school age that increase the risk for learning disabilities. 18• 26 Possible mechanisms of brain injury associated with the ECMO procedure include (1) cannulation of the common carotid artery, jugular vein, or both, resulting in cerebral blood-flow abnormalities and cerebrovascular damage; (2) the need for systemic heparinization, which increases the risk of brain hemorrhage; (3) air or thrombic embolization, which may occur during bypass; and (4) reperfusion injury to neuronal structures at the onset of bypass and following cerebral ischemia. 14• 19• 35• 42 Brain injury or developmental abnormalities before birth or the effects of ischemia and hypoxemia associated with severe cardiorespiratory failure prior to ECMO, however, are often difficult to separate from the potential complications of the bypass procedure itself. Notably, a high incidence of moderate to marked electroencephalographic (EEG) abnormalities has been observed in neonates prior to initiation of ECMO treatment/'· 28• 33• 57 reflecting the adverse effects of neonatal cardiorespiratory failure, prenatal abnormalities, or both on brain bioelectric activity. Also, in nonrandomized studies, term and near-term infants with severe respiratory failure treated with conventional mechanical ventilation had neurodevelopmental sequelae comparable with those in similarly ill neonates treated with ECM0. 50• 68

657

COMPLICATIONS AND SEQUELAE OF NEONATAL ECMO

In this article we explore the clinical and laboratory correlates that may be of predictive value or that may be causally related to the neurologic complications in the term and near-term neonate with severe cardiorespiratory failure who requires ECMO therapy. In addition, the incidence and possible pathogenesis of neurodevelopmental and auditory sequelae in survivors of neonatal ECMO are discussed. CLINICAL SIGNIFICANCE OF NEONATAL EEG AND BAEP STUDIES IN VENOARTERIAL ECMO-TREATED NEONATES

We studied the prognostic significance of EEGs recorded serially at 2- to 4day intervals during the acute neonatal course of 119 term or near-term infants with severe respiratory failure treated by venoarterial ECM0. 22 A poor prognosis was defined as either early death (n = 27), an abnormally low developmental score (n = 14), or cerebral palsy (CP) (n = 14) at 12 to 45 months of age. The neonatal EEG criteria of Tharp and Laboyrie 63 were used to grade each abnormality as mild, moderate, or marked. Marked EEG abnormalities included the following patterns: isoelectric, burst suppression, generalized or focal suppression, and generalized or focal slowing. An electrographic seizure, usually beginning as low-amplitude activity in the 8- to 15-Hz range, slowing to rhythmic, monomorphic waveforms at 0.5 to 3.0 Hz, and then stopping abruptly, also was classified as a marked abnormality. The only EEG abnormalities that were significantly related to a poor prognosis were burst suppression (BS) and electrographic seizure (ES). The 30 infants with two or more recordings of BS or ES, when compared with the 58 neonates without such EEG abnormalities, had a significantly increased odds ratio for a poor prognosis (Table 1). The 31 infants with a single ES or BS recording, when compared with the 58 neonates without Table 1. RISK RATIOS FOR AN ADVERSE OUTCOME RELATIVE TO EEGs RECORDED BEFORE AND DURING NEONATAL VENOARTERIAL ECMO IN 119 INFANTS* EEG Recordings of BS or ESt

All infants Normal Death or DH§ OR (95% CL)ll Survivors Normal DH§ OR (95% CL)ll

None

One

2::Two

P:j:

41 17 1.0

15 16 2.6 (0.9-7.0)

8 22 6.6 (2.2-20.2)

< 0.0001

41 9 1.0

15 8 2.4 (0.7-8.6)

8 11 6.3 (1.7-23.9)

0.006

*All infants had at least two technically satisfactory EE Gs recorded alter admission to nursery and before termination of ECMO. tBurst suppression (BS) or electrographic seizure (ES). :f:Chi-squared test of significance for table. §Developmental handicap (DH) owing to cerebral palsy (n = 14), or any Bayley or Mullen Scale score more than 2 SD below normal at age 12 to 45 months. l!Odds ratios (OR); 95% confidence limits (95% CL) that include 1 are not statistically significant. Adapted from Graziani LJ, Streletz LJ, Baumgart S, et al: The predictive value of neonatal electroencephalograms before and during extracorporeal membrane oxygenation. J Pediatr 125:969-975, 1994; with permission.

658

GRAZIANI et al

such EEG abnormalities, did not have a significantly increased risk for a poor prognosis. In a univariate analysis, cardiopulmonary resuscitation (CPR) immediately before ECMO (n = 8) and the lowest systolic blood pressure before or during ECMO were significantly related to the occurrence of ES or BS recordings (Table 2). We have not found a significant predilection of ES or other EEG abnormalities for either the right or left cerebral hemisphere, and noted that most of our ECMO-treated neonates had moderate to marked EEG abnormalities either before (85%) or during (72%) ECM0. 22• 57 Despite this high incidence of EEG abnormalities, the tendency toward improvement in recordings of survivors following the initiation of ECMO was statistically significant, probably owing, in part, to improved oxygenation and perfusion of cerebral structures. Our results suggest that multiple risk factors associated with the need for ECMO and the acuteness of the cardiorespiratory failure are responsible for the EEG abnormalities. The prognosis and pathogenesis of EEG abnormalities other than ES or BS are uncertain in neonates treated with ECMO and require further study. Other researchers also have emphasized that EEG abnormalities reflect the acuteness and the severity of brain dysfunction in critically ill neonates, and that the EEG abnormalities evolve during the nursery course. 25• 31 • 32• 58 Serial recordings during the early treatment of respiratory failure, therefore, should increase the predictive value of EEG studies in ECMO-treated neonates. Our studies indicate that a single EEG recording obtained during the treatment of

Table 2. NEONATAL CLINICAL FEATURES IN 119 VENOARTERIAL ECMO-TREATED INFANTS WITH NONE, ONE, OR '°" TWO EEG RECORDINGS OF ELECTROGRAPHIC SEIZURE (ES) OR BURST SUPPRESSION (BS) EEG Recordings with BS or ES* Clinical Featurest Apgar scores 1 minute 5 minute Lowest Pa0 2 before ECMO§ Oxygenation indexll Age at start of ECMO (hr) Duration of ECMO (days) Lowest blood pressure~ CPR before ECMO (n[%])**

None (n=58)

One (n =31)

'°"Two (n=30)

Pt

5 (0-9) 8 (2-10) 28 (12-47)

5 (0-9) 8 (2-9) 31 (15-115)

5.5 (0-10) 7 (2-10) 28 (13-63)

NS NS NS

47 (9-208) 36 (3-310) 6.0 (2.6-30.7) 44 (18-66) 1 (2)

35 (3-260) 40 (5-316) 5.8 (2.3-27.5) 40.5 (24-80) 2 (7)

43.5 (5-250) 35 (6-185) 6.7 (3.3-37.0) 35 (26-67) 5 (17)

NS NS NS 0.0007 0.03

'All infants had at least two technically satisfactory EE Gs recorded after admission to nursery and before completion of ECMO. Gestational age and birthweight did not differ significantly among the three groups: Chi-squared test tlisted as the median and (range), except for CPR. :!:Chi-squared or Kruskal-Wallis test of significance; NS, not significant. §Partial pressure of oxygen in mm Hg. Lowest value in postductal arterial blood prior to start of ECMO. Illas! oxygenation index before ECMO in 71 infants. ~Lowest systolic pressure in mm Hg, before or during ECMO. "Vigorous cardiopulmonary resuscitation exclusive of the immediate resuscitation at birth but prior to or during cannulation: endotracheal intubation, ventilation, cardiac compressions, and cardiac pressor medications. Percentage is calculated from the total number of infants in each column. Adapted from Graziani LJ, Streletz LJ, Baumgart S, et al: The predictive value of neonatal electroencephalograms before and during extracorporeal membrane oxygenation. J Pediatr 125:969-975, 1994; with permission.

COMPLICATIONS AND SEQUELAE OF NEONATAL ECMO

659

severe respiratory failure in neonates requiring ECMO therapy is of limited clinical usefulness and may not be of significant predictive value. Therefore, a single markedly abnormal EEG prior to cannulation does not necessarily constitute criteria sufficient to withhold ECMO. In our opinion, the clinical management of term and near-term infants with severe respiratory failure during the neonatal period can be facilitated by the appropriate use of technically satisfactory serial EEG recordings; however, the initiation or discontinuation of ECMO must be decided not only by the results of EEG tests but also in the context of complex clinical features, the wishes of informed guardians, and alternative treatment options. Desai et al, 11 in a study of 80 ECMO survivors from our hospital, reported the sensitivity and specificity of neonatal brainstem auditory-evoked potential (BAEP) recordings for predicting hearing loss and developmental delay in infancy and early childhood. BAEP recordings were performed before discharge in ECMO-treated neonates, and two specific abnormalities were analyzed: (1) elevated wave V threshold to auditory stimuli (clicks at 85-dB nHL) and (2) delayed central auditory conduction (prolonged 1-V interwave interval). Fortysix of the 80 ECMO survivors (57%) who were tested had normal results, and 34 infants (43%) tested abnormally, with elevated wave V threshold, prolonged I-V interwave interval, or both prior to discharge. Of the 12 children with hearing loss, seven (58%) had neonatal BAEP threshold tests that were normal. Also, there was no significant difference in the incidence of subsequent hearing loss in neonates who had abnormal BAEP thresholds (5/21 or 24%) compared with those with normal BAEP thresholds (7 /59 or 12%) at discharge. Therefore, the sensitivity of neonatal BAEP testing for predicting subsequent hearing loss was only 42%, and specificity for excluding subsequent hearing loss was 76%. In contrast, of 19 ECMO survivors with receptive language delay, 12 (63%) had abnormal neonatal BAEP recordings, and seven (37°/,,) had normal threshold and central auditory conduction tests (P = 0.04). Receptive language ability was delayed significantly during early childhood more often in infants with an abnormal neonatal BAEP, a difference that was not explained by the comorbidity of hearing loss. Therefore, severe neonatal cardiorespiratory failure, ECMO treatment, or both may adversely affect central auditory pathways and language development independently of damage to the peripheral auditory nerves or other brain structures. Because of the potential limitations associated with neonatal testing, BAEP recordings for detecting early hearing loss is probably most useful clinically at 3 to 4 months of age. Therefore, hearing evaluations are essential following discharge in ECMO-treated infants, regardless of the neonatal BAEP results. STRUCTURAL AND LATERALIZED CEREBRAL HEMISPHERIC ABNORMALITIES, ALTERED CEREBRAL BLOOD FLOW AND RECONSTRUCTION OF THE RCCA IN VENOARTERIAL ECMO-TREATED NEONATES

A variety of acute and chronic structural brain abnormalities, including ischemic injury, hemorrhages, infarctions, and atrophy, have been reported in CT scanning, MR imaging, and head ultrasound (HUS) studies of venoarterial ECMO-treated neonates. 23 ' 35 , 42 • 49 ' 55 • 60• 67 Lago et al3 5 reported that enlarged cerebrospinal fluid (CSF) spaces identified on MR imaging after venoarterial ECMO and permanent ligation of the RCCA were associated with lower scores on the Bayley Scales of Infant Development during the first year of life. Although the

660

GRAZIANI et al

pathogenesis of this post-ECMO MR imaging abnormality was not determined, the authors suggested that impaired absorption of CSF or cerebral congestion resulting from jugular vein ligation or diffuse brain injury was responsible for the enlarged spaces. Of the infants studied after venoarterial ECMO by these authors, MR images of the brain were classified as normal in 61 % and as disclosing major lesions in 23% and minor lesions in 16%. Head circumference growth was significantly less at 6 or 12 months of age in the seven survivors with major MR imaging lesions compared with those with normal or minor lesions on MR imaging. Von-Allmen et al67 reported that HUS evidence of cerebral edema or periventricular white matter injury prior to ECMO increased the risk of major abnormalities on post-ECMO neuroimaging studies. In HUS studies from our hospital,49 grade I subependymal hemorrhage was noted in 43 of 212 neonates before or during venoarterial ECMO, but during bypass progressed to grade III in only one and to grade IV in only two infants and stabilized or resolved in all others. Thus, a grade I subependymal hemorrhage is not necessarily a contraindication for ECMO. Finally, mild to moderate neuroimaging abnormalities, though commonly noted following venoarterial ECMO, are of very limited predictive value in an individual survivor, as noted by others. 59• 60 Though infrequent in ECMO survivors, CT scan or MR image evidence of a major structural injury with permanent loss of parenchymal brain tissue, such as cystic encephalomalacia owing to a large or multiple infarctions, is associated with a very high probability of a severe neurodevelopmental handicap. Neuroimaging and other studies have not disclosed consistent evidence of selective or isolated damage to the right cerebral hemisphere in neonates treated with venoarterial ECMO. In a retrospective review using CT scans, EEG, and neurologic assessments, Schumacher et al52 reported that 8 of 59 survivors of neonatal venoarterial ECMO had right cerebral abnormalities and suggested that carotid artery ligation was not without risk to the brain, particularly the right hemisphere. Andrews et aF reported right-sided weakness in a venovenous ECMO-treated infant and left-sided weakness in another infant whose right carotid artery was cannulated. Lott et al3 6 reported that EEG abnormalities were not significantly lateralized but did note a reduction in the amplitude of right, compared with left, hemispheric auditory and somatosensory-evoked potentials recorded from 10 survivors of neonatal ECMO. Campbell et aF reported an increased incidence of left versus right focal seizures in 35 ECMO-treated infants but concluded that ischemia resulting from carotid ligation was insufficient to result in lateralized neurologic abnormalities at 2 years of age. Statistically significant lateralized EEG or HUS abnormalities in neonatal infants during or after venoarterial ECMO have not been found in previous studies from our hospital. 23 • 24• 57 single photon emission computed tomography (SPECT) scans, however, disclosed decreased regional cerebral blood flow (CBF) of the right hemisphere in five of 13 of our venoarterial ECMO-treated neonates.44 We also noted a slight but significant decrease of peak systolic flow in the right, compared with the left, middle cerebral arteries in a color Doppler imaging (CDI) study of brain blood flow during neonatal venoarterial ECM0. 24 Nevertheless, the clinical implications and prognostic significance of most EEG, CBF, and neuroimaging studies in venoarterial ECMO-treated neonates indicating lateralized abnormalities are uncertain except when associated with evidence of irreversible and severe structural brain damage. The usual collateral pathway after cannulation of the right common carotid artery at the initiation of ECMO includes left-to-right blood flow through the anterior communicating artery and retrograde flow in the Al segment of the right anterior cerebral artery39 (Fig. 2). Retrograde flow in the right internal

COMPLICATIONS AND SEQUELAE OF NEONATAL ECMO

661

ECMO

Normal

Al

ICA

ICA

Figure 2. Altered blood flow in anterior circle of Willis depicted in the coronal plane. Normal flow (A), and during and after venoarterial ECMO (8). With ECMO, flow is antegrade up the left internal carotid artery (ICA) and A 1 segment of the anterior cerebral artery (ACA). Blood then flows from left to right across the anterior communicating artery (AComA) and retrograde down the right A 1 and ICA, but antegrade in the right middle cerebral artery (MCA). R, right; L, left; A2, distal ACA. (From Mitchell DG, Merton D, Desai H, et al: Neonatal brain: Color Doppler imaging: II. Altered flow patterns from extracorporeal membrane oxygenation. Radiology 167:307-310, 1988; with permission.)

carotid artery persisted in some infants following ECMO until after discharge from our nursery. 37 In another CDI study from our hospital, a retrograde flow in the right vertebral artery that shunted blood from the brain's posterior circulation to the right arm via the subclavian artery (i.e., a subclavian steal) was noted in 17 infants during venoarterial ECM0. 24 When we compared the 17 infants with a subclavian steal to 37 others without a steal, we were unable to detect statistically significant mortality rate differences between the two groups; neonatal EEG, HUS, or CT scan abnormalities; or early neurodevelopment in survivors. Basilar artery perfusion was maintained by increased left vertebral artery flow in infants with a right subclavian artery steal, perhaps because of autoregulation of CBF. Only one of the 17 infants with a subclavian steal showed signs of peripheral ischemia, including decreased skin temperature and increased cyanosis in the right upper extremity, compared with the left, and all signs of ischemia subsided immediately after removal of the cannulas. Except for significantly lower blood pressures in the right than in the left arm during ECMO (P < 0.005), subclavian steal was not accompanied by clinical signs of extremity or brainstem ischemia in any of the other 16 affected infants. After removal of the cannulas, flow in the right vertebral artery immediately returned to antegrade, and blood pressures in the upper extremities were symmetric in all surviving infants. We have detected a subclavian steal in only four of 83 subsequent infants treated with venoarterial ECMO at our hospital following the routine use of 12-F or smaller RCCA camrnlas. Our studies suggest that a subclavian steal during ECMO is not injurious to the developing brain and does not adversely affect peak systolic blood-flow velocities in intracranial vessels other than the right vertebral artery. Therefore, CDI of the vertebral arteries is not clinically necessary during venoarterial ECMO, particularly when a 12-F or smaller RCCA cannula is used. Although reconstruction of the RCCA after ECMO has proved feasible, 4• 9

662

GRAZIANI et al

neurologic sequelae in a randomized study of this procedure have not been evaluated. We evaluated the results of RCCA reconstruction in 47 infants treated with venoarterial ECMO at our hospital and compared their ultrasound and neuroanatomic imaging studies, EEGs, and developmental outcomes with those of 93 infants who had no reconstruction. 4 Right internal carotid and bilateral anterior and middle cerebral arterial blood-flow velocities were generally higher and were more symmetrically distributed in infants with a reconstructed RCCA compared with infants whose RCCA was permanently ligated (Fig. 3). EEG, neuroimaging, and neurodevelopmental studies did not disclose a significantly increased risk of abnormalities in infants after reconstruction (34 of 40 [85%] were normal at 6 to 30 months) compared with ECMO survivors whose RCCA was not repaired (52 of 74 [70%] were normal). RCCA reconstruction after ECMO, therefore, facilitated normal distribution of CBF through the circle of Willis and augmented both left and right middle cerebral artery blood flow immediately after decannulation. Serial studies using pulsed Doppler ultrasound with spectral analysis were performed in 70 infants and children (age range: neonatal to 4.5 years) who had RCCA reconstruction immediately following venoarterial ECMO at our hospital. 13 Following discharge, the RCCA remained patent or had a mild stenosis in 51 of the 70 survivors (73%). The other 19 survivors (27%) had either severe stenosis (n = 11) at the site of the repair or total occlusion of the artery (n = 8), but there was no correlation between the patency of the RCCA and early neurodevelopment. The consequences of either RCCA ligation or reconstruction after ECMO will, therefore, require longer-term evaluation before either course is recommended as routine. CLINICAL ANTECEDENTS OF NEUROLOGIC AND AUDIOLOGIC ABNORMALITIES IN SURVIVORS OF VENOARTERIAL AND VENOVENOUS NEONATAL ECMO

Recently, we reported the neurodevelopment and hearing evaluations in 181 survivors of neonatal venoarterial ECMO, treated at our hospital, who were 15 months of age or older when last examined. 20 The primary diagnosis, gestational age, birthweight, and outcome of the survivors are summarized in Table 3. Seventeen ECMO survivors (9%) had a spastic form of CP ranging from

Figure 3. Right and left internal carotid arteries enter the circle of Willis. Anterior, middle, and posterior communicating cerebral arterial collateral vessels distribute rostral flow. Relative differences in circulatory distribution between infants whose RCCAs were ligated and those whose RCCAs were reconstructed after ECMO are shown for selected cerebral arterial branches by histogram. In infants with a ligated RCCA, collateral retrograde flow through the right A 1 segment of the anterior cerebral artery demonstrated negative systolic flow velocity, and directionally normal antegrade flow (positive velocity) as RCCA blood flow was restored by the reconstruction procedure after decannulation. Moreover, both right and left middle cerebral artery flows were augmented (greater anterograde positive systolic blood flow velocity values) in infants with RCCA reconstruction compared with infants whose RCCA was ligated after ECMO. Vertical histogram axis: peak systolic velocity (in cm/second); negative values indicate retrograde flow. Dark arrow = antegrade; shaded arrow = retrograde. RCCA = right common carotid artery; ECMO = extracorporeal membrane oxygenation. (From Baumgart S, Graziani LJ, Streletz LJ, et al: Right common carotid artery reconstruction after extracorporeal membrane oxygenation: Vascular imaging, cerebral circulation, electroencephalographic, and neurodevelopmental correlates to recovery. J Pediatr 125:295-304, 1994; with permission.)

COMPLICATIONS AND SEQUELAE OF NEONATAL ECMO

663

Left A-i Anterior Cerebral

Right A-1 Anterior Cerebral 150

60

p<0.001

40

20

100

-20

50

-40

0 Reconstructed

Reconstructed

Ligated

Right Middle Cerebral Artery

Ligated

Left Middle Cerebral Artery

BO

80

60

60

40

40

20

20

Reconstructed

Reconstructed

Ligated

Right Proximal Internal Carotid

Left Distal Internal Carotid

60

120

50

100

40

80

30 60 20 40 10 20

Reconstructed

Ligated

Reconstructed

Figure 3. See legend on opposite page

Ligated

Ligated

664

GRAZIANI et al

Table 3. PRIMARY DIAGNOSIS, GESTATIONAL AGE, BIRTHWEIGHT, AND OUTCOME IN 181 SURVIVORS OF NEONATAL VENOARTERIAL (n= 152) OR VENOVENOUS (n=29) ECMO* Outcomes Normal Neonatal Factors Primary diagnosis MAS/PPHN Diaphragmatic hernia RDS/HMO Sepsis/pneumonia Gestational age, weeks (mean [SD]) Birthweight, kg: (mean [SD])

Suspect§

Abnormal[[

12-46 mot (n=96)

4.8-6 yr:t: (n=34)

4.8-6 yr (n = 17)

16 mo-8 yr (n=34)

63 12 8 13 39.4 (±2)

23 4 5 2 40.1 (±2)

9 0 7 1 39.6 (±2)

20 2 6 6 39.2 (±2)

3.31 (±0.6)

3.33 (±0.5)

3.20 (±0.5)

3.18 (±0.5)

*No significant differences in primary diagnosis, gestational age, or birthweight among the four groups: chi-squared test tPresumed normal because of young age. No definite evidence of MR, CP, or SNHL :j:WPPSl-R full scale IQ ;o- 85; normal neurologic examination, and normal hearing thresholds for both ears at school age. §WPPSl-R full scale IQ between 71 and 84, without CP and SNHL at school age. llCP (n = 17); SNHL without CP or MR (n = 12); MR without CP or SNHL (n =5). MAS/PPHN: Meconium aspiration syndrome/persistent pulmonary hypertension of the newborn. RDS/HMD: Respiratory distress syndrorne/hyaline membrane disease. Adapted from Graziani LJ, Baumgart S, Desai S, et al: Clinical antecedents of neurologic and audiologic abnormalities in survivors of neonatal ECMO. J Child Neurol 1997, in press; with permission.

a mild hemiparesis or diparesis to severe quadriparesis diagnosed by age 15 months and confirmed by subsequent examinations. Of the 17 children with CP, 10 (including seven with spastic quadriparesis) also had mental retardation. Twelve others (7%) without CP or mental retardation had sensorineural hearing loss (SNHL) in one (n = 6) or both (n = 6) ears diagnosed by age 6 months to 4.5 years and confirmed by repeat testing. Ninety-six survivors (53%) were classified as presumably normal because their early assessments disclosed no definite CP, SNHL, or mental retardation, and because they were not yet ageeligible for Wechsler Preschool and Primary Scale of Intelligence-Revised (WPPSI-R) IQ tests. 69 Five survivors (3%) without CP or SNHL had full-scale WPPSI-R IQ scores below 71 and in the mental retardation range. Seventeen survivors (9%) without CP or SNHL had borderline full-scale WPPSI-R IQ scores between 71 and 84 at school age and were, therefore, classified as suspect. The remaining 34 survivors (19%) were classified as normal at school age because they had bilateral normal hearing thresholds, full-scale WPPSI-R IQ scores above 84, and normal neurologic examinations. Neonatal clinical variables in the 34 normal school-age survivors were then compared with the 17 with CP, to the 12 with hearing loss, and to the 22 without CP or SNHL who had borderline or abnormally low WPPSI-R IQ scores. A univariate analysis disclosed the potential significance of a low systolic blood pressure before or during ECMO, CPR before to ECMO (exclusive of resuscitation immediately at birth), and a high oxygen index (OI) before ECMO in relationship to the development of CP (Table 4). Because of venoarterial ECMO

COMPLICATIONS AND SEQUELAE OF NEONATAL ECMO

665

Table 4. CEREBRAL PALSY (n = 17) COMPARED WITH NORMAL SCHOOL-AGE OUTCOME (n=34) IN 51 VENOARTERIAL ECMO SURVIVORS: NEONATAL CLINICAL FEATURES Normal*

Cerebral Palsyt

Clinical Features

(11=34)

(11=17)

P:j:

Lowest pH Pre-ECMO Lowest Pa02 Pre-ECMO Lowest PaC02 Pre-ECMO Oxygenation index§ Lowest systolic blood pressurell CPR before ECMO (n[%])

7.16±0.19 29.9±8.5 22.9±9.6 63.7±46.9 46.2±9.5 0 (0)

7.19±0.23 31.6±23.5 18.5±7.7 92.7±64.5 40.9±11.3 8 (47)

NS NS NS <0.02 <0.01 <0.0001

Data represent the mean ± SD, except for CPR (number[%]). Gestational age, birthweight, Apgar scores, age at start of ECMO, duration of ECMO, and highest neonatal indirect bilirubin level did not differ significantly between the two groups. 'Normal neurologic examination, full-scale WPPSl-R IQ scores above 84, and normal bilateral hearing at age 4.8 to 6.0 years. tSpastic form of CP at age 15 months to 8 years, confirmed by repeated examinations. :j:Chi-squared or Kruskal-Wallis tests of significance. NS, not significant. §Last oxygenation index before ECMO. Normal, n = 15; cerebral palsy, n = 6. llln mm Hg, before or during ECMO. Pa0 2 , partial pressure of oxygen; PaC0 2 , partial pressure of carbon dioxide (lowest value (mm Hg) in postductal arterial blood before ECMO). CPR, vigorous cardiopulmonary resuscitation prior to ECMO exclusive of the immediate resuscitation at birth (see Table 2). Percentage calculated from the total number of survivors in each column. Adapted from Graziani LJ, Baumgart S, Desai S, et al: Clinical antecedents of neurologic and audiologic abnormalities in survivors of neonatal ECMO. J Child Neurol 1997, in press; with permission.

pump effects that result in a narrow pulse pressure, we elected to analyze systolic blood pressure measurements exclusively. A univariate odds ratio disclosed that survivors with a systolic blood pressure before or during ECMO of 38 mm Hg or less had a significantly increased relative risk of CP compared with survivors whose systolic blood pressures remained above 38 mm Hg (Table 5). The relationship between duration of maximally low systolic blood pressures and ischemic brain injury could not be analyzed in this study. Brain blood flow, however, is dependent upon intracranial perfusion pressure, critical closing pressure of brain blood vessels, and vascular resistance, which, in turn, are influenced by numerous variables including intracranial pressure, vasomotor tone, and autoregulation in response to changes in blood pressure or blood carbon dioxide and oxygen tensions 43 • 46• 47 Also, impaired autoregulation of CBF has been noted following asphyxia in term infants 10• 48 and during ECMO in animal studies. 54 Thus, prior to or during ECMO, prolonged hypotension with intact autoregulation, normal vasomotor tone in brain blood vessels, and normal intracranial pressure may not be as injurious as a brief but comparably low blood pressure in infants who have impaired cerebrovascular autoregulation and increased intracranial pressure with decreased vasomotor tone and a low critical closing pressure. In our studies, maximally low Pa0 2 levels before ECMO were not significantly related to an increased risk of CP,2° suggesting that hypoxemia in the absence of cerebral ischemia does not exert a permanent injurious effect on the neonatal brain, a concept previously discussed by Vannucci.65 We have not, as yet, been able to analyze for possible additive effects of hypotension, hypocarbia, and hypoxia in our studies of ECMO treated neonates. Also, except for six other ECMO survivors with evidence of a congenital devel-

666

GRAZIANI et al

Table 5. RELATIVE RISKS FOR CEREBRAL PALSY ASSOCIATED WITH RANGE OF LOWEST SYSTOLIC BLOOD PRESSURES BEFORE OR DURING NEONATAL VENOARTERIAL ECMO IN 51 SURVIVORS Lowest Systolic Blood Pressure: Range*

246 mm Hg (n =20) Cerebral palsy No (n=34)t Yes (n=17):j: Odds ratio 95% CL§

16

4 1

39-46 mm Hg (n = 17) 12

5 1.7 0.4-7.6

:s38 mm Hg (n=14)

6 8 5.3 1.2-24.5*

*Lowest systolic blood pressure (mm Hg) recorded before or during ECMO. t34 survivors of neonatal ECMO who at school age (4.8 to 6.0 years) had a full-scale WPPSl-R IQ >85, normal neurologic examinations, and normal bilateral hearing thresholds. :j:17 survivors of neonatal ECMO who at ages that ranged from 15 months to 8 years had a spastic form of CP on serial neurological examinations. §95% confidence limits; limits that do not include 1 are statistically significant (P<0.05). Adapted from Graziani LJ, Baumgart S, Desai S, et al: Clinical antecedents of neurologic and audiologic abnormalities in survivors of neonatal ECMO. J Child Neurol 1997, in press; with permission.

opmental brain disorder who were not included in our final data analyses, prenatal variables including placental, gestational, and obstetric factors that may be causally related to the neurologic and audiologic sequelae in our studies have not been completely evaluated. Apgar scores, gestational age, and birthweight, however, did not differ when the normal and abnormal survivors were compared, suggesting that fetal maturation and those perinatal disorders that may affect the condition of the neonate immediately at birth were not closely associated with the observed neurologic and audiologic sequelae of neonatal ECMO treatment. Finally, the neonatal variables analyzed were not significantly related to borderline or abnormally low WPPSI-R IQ scores in survivors of ECMO who did not have CP or SNHL. Another univariate analysis of our data disclosed the potential significance of hypocarbia prior to ECMO and the postnatal age at the start of ECMO in relationship to the development of SNHL in survivors (Table 6). A univariate odds ratio disclosed that survivors with a PaC0 2 of 14 mm Hg or less before ECMO had a significantly increased relative risk of SNHL compared with survivors whose PaC0 2 levels remained above 14 mm Hg (Table 7). HendricksMunoz and Walton29 reported that hyperventilation and the duration of ventilation were significantly related to the occurrence of SNHL in infants with persistent fetal circulation who were not treated with ECMO. Severe hypocarbia also has been noted to significantly increase the risk of cerebral abnormalities in preterm infants. 21 The injurious effect of hypocarbia on neural tissue presumably is due to constriction of the brain vasculature, resulting in ischemia,45 though in our studies we were unable to demonstrate a significant association between maximally low PaC02 levels prior to neonatal ECMO and neurodevelopment other than SNHL. 20 Del Toro et al1° noted, however, that cerebrovascular responses of the immature brain during hypocapnia are not consistent. Glass et al1 8 and Vaucher et al,66 in recent reports of survivors from two other large ECMO centers, noted an incidence of audiologic and severe neurologic sequelae similar to that observed in our studies. In general, severe neurodevelopmental abnormalities identified during early childhood in ECMO survivors

COMPLICATIONS AND SEQUELAE OF NEONATAL ECMO

667

Table 6. SENSORINEURAL HEARING LOSS (n = 12) COMPARED WITH NORMAL OUTCOME AT SCHOOL AGE (n=34) IN 46 VENOARTERIAL ECMO SURVIVORS Clinical Features

Normal* (n=34)

Hearing losst (n = 12)

P:t:

Lowest pH Pre-ECMO Lowest Pa0 2 Pre-ECMO Lowest PaC02 Pre-ECMO Oxygenation index§ Age at start of ECMO (hr) Duration of ECMO (hr) Lowest systolic blood pressurell Highest bilirubin level'll

7.16±0.19 29.9±8.5 22.9±9.6 63.7±46.9 59.4±47.1 130.4±60.6 46.2±9.5 7.0+2.6

7.19±0.24 28.0±9.2 13.6±5.3 31.7±14.3 126.6±92.8 193.2±106.9 43.3±5 6.9±2.9

NS NS :50.001 <0.02 :50.05 NS NS NS

Data represent the mean ± SD. Gestational age, birthweight, Apgar scores, and the need for CPR did not differ significantly between the two groups. 'Normal neurologic examination, full-scale WPPSl-R IQ scores above 84, and normal bilateral hearing at age 4.8 to 6.0 years. tAt age 15 months to 6.3 years. :j:Kruskal-Wallis test. NS, not significant. §Last oxygenation index before ECMO. Normal, n = 15; hearing loss, n = 6. l[ln mm Hg, lowest measurement before or during ECMO (see text). ~In mg/dl, neonatal maximal concentration of indirect reacting bilirubin in serum. Pa0 2 , partial pressure of oxygen; PaC0 2 , partial pressure of carbon dioxide (lowest value (mm Hg) in postductal arterial blood before start of ECMO). Adapted from Graziani LJ, Baumgart S, Desai S, et al: Clinical antecedents of neurologic and audiologic abnormalities in survivors of neonatal ECMO. J Child Neurol 1997, in press; with permission.

are predictive of later developmental status. 18• 42 CP, mental retardation, and SNHL noted during the first 2 years of life have persisted as permanent handicaps in our studies to date. 20 Less severe cognitive, language, or hearing deficits may not be detected until early school age, however, even when ECMO survi-

Table 7. RELATIVE RISKS FOR SNHL ASSOCIATED WITH RANGE OF LOWEST PaC0 2 PRIOR TO NEONATAL VENOARTERIAL ECMO IN 46 SURVIVORS lowest PaC0 2 : Range*

SNHL No (n=34)t Yes (n=12):j: Odds Ratio 95% CL§

2=20 mm Hg (n=27)

15-29 mm Hg (n=9)

:s14 mm Hg (n = 10)

25 2 1

6 3 6.2 0.85-46.0

3 7 29.1 4.0-210.0

'Partial pressure of CO, (mm Hg) in postductal arterial blood: range of lowest value prior to start of ECMO. t34 survivors of neonatal ECMO who at early school age (4.8 to 6.0 years) had a WPPSl-R fullscale IQ <-85, normal neurologic examinations, and normal bilateral hearing thresholds. :J:12 survivors of neonatal ECMO without CP or MR who at ages that ranged from 15 months to 6.3 years had SNHL of at least 35 dB in one (n = 6) or both (n = 6) ears. §95% confidence limits; limits that do not include 1 are statistically significant (P<.05) Adapted from Graziani LJ, Baumgart S, Desai S, et al: Clinical antecedents of neurologic and audiologic abnormalities in survivors of neonatal ECMO. J Child Neurol 1997, in press; with permission.

668

GRAZIANI et al

vors are examined periodically during infancy and early childhood, as discussed in the next section. Kornhauser et al34 studied the relationship between bronchopulmonary dysplasia (BPD) and neurodevelopmental outcome following ECMO treatment of 64 term or near-term infants from our hospital. BPD was defined as a supplemental oxygen or mechanical ventilation requirement with appropriate chest radiographic abnormalities at age 28 days and older. Neonates with primary pulmonary hypoplasia or congenital diaphragmatic hernia (CDH) were excluded from the analyses. BPD occurred in 17 of the 64 (27%) survivors. Those with BPD were more likely to have had RDS as a primary diagnosis than were those without BPD (9/17 versus 6/47, P < 0.01). ECMO initiation was significantly later (127 ± 66 versus 53 ± 39 hr, P < 0.001), and the duration of ECMO treatment was longer (192 ± 68 versus 119 ± 53 hr, P < 0.001) in those with BPD than in those without. EEG and brain-imaging abnormalities did not differ between the two groups. Scores on the Bayley Scales of Infant Development (BSID) 5 at less than 30 months of age were significantly lower in BPD compared to non-BPD survivors (MDI 76 ± 21 versus 103 ± 17, P < 0.001; and PDI 74 ± 21 versus 97 ± 18, P < 0.001). Also, 3 of 4 Mullen Scales of Early Learning40 subtest scores in survivors older than 30 months were lower in those with BPD than in those without it (P < 0.001). At early school age, 11 of the 17 BPD patients (65%) had suspect or mild neurologic disabilities and three (18%) had severe disabilities compared with 16 of 47 non-BPD patients (34%) who had suspect or mild disabilities and two (4%) who had severe disabilities (P < 0.01). Also noted were an older age at initiation of ECMO and a longer duration of bypass as antecedents of neurodevelopmental disability (P < 0.05). We concluded that the occurrence of BPD following ECMO was associated with an increased risk of an adverse neurodevelopmental outcome. Predisposing factors included a delayed initiation and a prolonged duration of ECMO therapy and a primary diagnosis of respiratory distress syndrome. Venovenous bypass has been initiated in about 43% of the ECMO-treated neonates at our hospital since 1992. Of 55 neonates initially placed on venovenous ECMO at our center between 1992 and 1996, four (7%) were converted to venoarterial bypass because of clinical deterioration; 49 of the 51 (96%) who remained on venovenous ECMO survived. In preliminary nonrandomized studies of our patient population, Wiswell et aF2 reported that the incidence of neonatal neuroimaging abnormalities and BSID scores at age 1 year were similar in venovenous compared with venoarterial ECMO-treated neonates. We have not as yet noted any severely handicapped survivors of venovenous bypass therapy. Further studies are necessary, however, to determine whether brain injury is less likely to be associated with venovenous than with venoarterial ECMO in comparably ill neonates. NEUROLOGIC OUTCOME IN RELATIONSHIP TO CAUSE OF NEONATAL CARDIORESPIRATORY FAILURE

There are few reports of the relationship between the primary respiratory disorder and neurologic outcome in ECMO-treated neonates. 8• 12• 27• 56 Gringlas et al27 noted that WPPSI-R scores in ECMO survivors from our hospital were not significantly different among groups of children classified by their primary respiratory diagnoses; children too handicapped to test with the WPPSI-R were excluded from the analysis. D' Agostino et al8 noted a motor delay at age 1 year in survivors of CDH treated with neonatal ECMO therapy and reported that the

COMPL!CATJONS AND SEQlJELAE OF NEONATAL ECMO

669

size of the CDH at the time of the surgical repair predicted the severity of feeding difficulty, chronic lung disease, and failure to thrive during the first year of life. Desai et al1 2 compared the neurodevelopment of 18 children with CDH to that of 25 children with meconium aspiration syndrome (MAS), all of whom had been treated with neonatal ECMO at our hospital. The mortality rates in infants with CDH is relatively high, but the two groups of ECMOtreated survivors were comparable in gestational age, birthweight, and age at assessment. BSID testing at 1 and 2 years of age disclosed significantly lower PDI scores (P s 0.03) but not MDI scores in the CDH survivors compared with the MAS group. WPPSI-R scores at 5 to 6 years of age, however, were not significantly different in the two groups, indicating that the early delay in motor skills noted in children with CDH was not predictive of cognitive development at school age. Also, growth rates of height and weight but not of head circumference during early childhood were lower in survivors of neonatal ECMO who had been treated for CDH at our hospital than in those who had been treated for neonatal respiratory distress syndrome.5€• Finally, we have not observed severe spastic CP in CDH survivors treated with ECMO. Meconium-stained amniotic fluid is noted in 10°/c, to 15% of all term and near-term pregnancies, and about 5'1o of infants born through meconium-stained fluid are at signjficantly increased risk of developing MAS.71 Ninety-two of 209 survivors of neonatal ECMO treated at our hospital had MAS as their primary cause of respiratory failure, and eight of the 92 infants developed CP, including six with severe spastic quadriparesis and severe to profound mental retardation. WiswelF' has summarized in a recent review the chemical and laboratory evidence that meconium in amniotic fluid may result in vasoconstriction of placental and fetal blood vessels, increasing the risk of cerebral ischemia prior to birth. Severe cardiorespiratory failure from MAS after birth then may further compromise intracranial circulation, especially in those affected neonates who become hypotensive. Therefore, the pathogenesis of brain injury associated with neonatal ECMO is at least in part related to the primary cause of the neonatal cardiorespiratory failure. COGNITIVE AND AUDITORY DEFICITS AT EARLY SCHOOL AGE

Reports of short-term developmental outcome suggest that by the age of 1 to 2 years, most ECMO survivors (between 59'Yo and 75%) appear to function within the normal range in developmental and neurologic testsi. 2 · 5, 17• 19• 66; however, deficits, such as hearing impairment, language delays, difficulty in visual/spatial abilities, and attention and behavior disorders that were not necessarily observed during earlier assessments, have been reported in ECMO survivors at early school age. 18• 30• 53 • 04 Environmental and psychosocial risk factors unrelated to specific perinatal complications of ECMO also may contribute to classroom and academic difficulties in school-age survivors of neonatal ECMO, as discussed by Glass. 16 Most ECMO survivors (67%) from our hospital who were tested at early school age had normal intellectual functions, but serial assessments revealed a significant increase in the incidence of deficiencies in developmental and cognitive assessments with increasing age. 26 In addition, of the 12 children with a sensorineural hearing loss in our study population, three with unilateral deficits and one with bilateral deficits who were followed since birth were not identified until after 3 years of age. 20 The late detection of audiologic deficits in some of

670

GRAZIANI et al

our subjects was more likely related to the difficulty in detecting high-frequency hearing loss (especially if unilateral) in infants and young children, though progressive sensorineural hearing loss in high-risk neonates has been reported by others. 29 Therefore, survivors of neonatal ECMO therapy classified as presumably normal when very young may be reclassified when older as having a SNHL or mild cognitive deficit, as the more subtle impairments become identifiable at school age. We have studied serially a population of 65 neonates who were treated with ECMO at our hospital and who were last tested at a mean age of 5.4 years; despite normal verbal or performance IQ scores in 46 children, 26 (56%) had a discrepancy of more than 10 points between verbal and performance scores on the WPPSI-R. Verbal scores more than 10 points higher than performance scores are noted in 12% to 15% of the normal population at age 5 years. 69 The verbalperformance discrepancy found in our ECMO survivors ranged from 11 to 30 points, and 22 of the 26 children (85%) had verbal scores more than 14 points higher than their performance scores. School-age children presenting with verbal intelligence scores significantly higher than their performance scores may be at increased risk for a nonverbal learning disability. 51 Because the right cerebral hemisphere subserves the mediation of visuospatial skills, the cause of the neuropsychologic abnormalities noted in some school-age survivors of neonatal ECMO may be related, in part, to the adverse effects of RCCA cannulation on right cerebral hemisphere perfusion. Taylor, 62 however, suggested that children who sustain brain injury early in life (and who were not treated with ECMO) exhibit their greatest impairments on tests of visual-motor skills, problem solving, memory, and learning. Alternatively or additionally, therefore, school-related disorders in ECMO survivors who are otherwise neurologically intact may reflect the effects of mild diffuse brain dysfunction rather than damage to the right cerebral hemisphere. In summary, although most children who received neonatal ECMO therapy are not severely handicapped, relatively mild cognitive abnormalities are now becoming identifiable in the older survivors. Studies have been limited in interpretative scope owing to the lack of an appropriate control group and the relatively small number of children tested at school age. We also have been unable to identify any specific perinatal complication that may explain the subtler deficiencies in brain function noted in a significant proportion of our ECMO survivors at early school age. Future research should include longerterm follow-up and further analysis of perinatal complications in relationship to the neuropsychologic and academic deficits in school-age survivors of neonatal ECMO treatment. SUMMARY

A total of 355 infants have been treated with ECMO at our hospital between 1985 and 1996, 271 of whom have been enrolled in an ongoing prospective study; of the 271 infants enrolled, 223 (82%) survived, and most function within the normal range of development. Nevertheless, handicapping sequelae, including spastic forms of CP, hearing loss, and cognitive deficiencies at school age, have been noted in a significant minority of ECMO-treated survivors. The need for RCCA cannulation during venoarterial ECMO may increase the risk of a cerebrovascular injury, and lateralized CBF abnormalities have been noted on CDI and pulsed Doppler ultrasound studies during and after venoarterial bypass.37· 39· 61 ; however, post-ECMO CT scans, HUS, MR images, or clinical evalua-

COMPLICATIONS AND SEQUELAE OF NEONATAL ECMO

671

tions have not indicated selective or greater injury to the right, compared with the left, cerebral hemisphere in our survivors, nor was there a significant predilection for right, rather than left, cerebral hemispheric EEG abnormalities during or following venoarterial bypass. 4 • 20 • 22 • 23 • 57 Although we routinely repair the RCCA following venoarterial ECMO, the long-term consequences of a permanently ligated artery have not as yet been demonstrated. We have noted the ominous predictive value of two or more recordings that disclose ES and BS EEG abnormalities before or during venoarterial ECMO and found that the need for vigorous CPR before or during RCCA cannulation significantly increased the risk of these two markedly abnormal bioelectric patterns.2 2 Because 85% of infants with severe respiratory failure have moderate to marked EEG abnormalities (including 23% who have BS or ES patterns) before ECM0, 57 we believe that fetal and neonatal complications related to the occurrence and treatment of severe cardiorespiratory failure are responsible in large part for the neurologic sequelae in ECMO survivors. The risk for CP was significantly increased in survivors of neonatal venoarterial ECMO treated at our hospital who required CPR or who independently had a systolic BP below 39 mm Hg before or during ECM0. 20• 55 We also noted that the risk for hearing loss was increased significantly in surviving neonates who had a PaC02 below 14 mm Hg before ECM0. 20 The possibility that undetected confounding variables were, in part, responsible for the neurologic, audiologic, and cognitive sequelae in ECMO survivors could not be excluded entirely by our data analyses. Although the pathogenesis of severe brain damage has not been defined fully in neonates treated with ECMO, focal, multifocal, or diffuse cerebral ischemia is the most likely final common pathway; thrombosis, infarction, or hemorrhage may follow and contribute to the brain injury. The cause of isolated SNHL is unknown in most affected ECMO survivors, but in some very likely is associated with the complications and treatment of severe cardiorespiratory failure, including profound hypocarbia prior to ECMO. The results of our studies to date are consistent with the following conclusions: (1) hypotension before or during ECMO and the need for CPR before ECMO contribute to the pathogenesis of CP, probably through the mechanism of cerebral ischemia; (2) profound hypocarbia before ECMO and delayed ECMO treatment are associated with a significantly increased risk of hearing loss; (3) hypoxemia without hypotension does not result in CP; (4) the type and severity of neurologic and cognitive sequelae in ECMO survivors depends, in part, on the primary cause of the neonatal cardiorespiratory failure; (5) early neurodevelopment, except for severe deficits, may not predict school-age performance; and (6) abnormally low or borderline WPPSI-R IQ scores and academic deficiencies at early school age, without evidence of a congenital abnormality of brain or CP or SNHL, remain unexplained. The criteria for initiating ECMO in the neonate with severe cardiorespiratory failure include decreasing oxygenation despite mechanical hyperventilation with 100% oxygen. The decision to initiate neonatal ECMO often is determined in part by a high predicted mortality, usually late in the course of cardiorespiratory failure, a therapeutic approach that remains controversial. The neonatologist often is confronted with the dilemma of either initiating ECMO or continuing intensive therapies in the hope that marked deterioration with hypotension, cardiopulmonary arrest, and irreversible brain injury does not occur, or that profound hypocarbia and prolonged mechanical hyperventilation do not become necessary before cardiorespiratory failure improves. Venovenous ECMO offers significant advantages compared with venoarterial bypass, especially if initiated early in the course of neonatal respiratory failure. Additional research is neces-

672

GRAZIANI et al

sary, however, to understand the impact of more liberal criteria than generally are used to begin neonatal ECMO therapy. Whether the neurologic, auditory, and school-related handicaps among ECMO survivors can be reduced by modifying the management of severe cardiorespiratory failure in newborn infants, including the use of experimental neuroprotective medications or early initiation of bypass therapy, requires further study.

ACKNOWLEDGMENT The authors wish to thank Gretchen T. McCawn for her editorial assistance.

References 1. Adolph V, Ekelund C, Smith C, et al: Developmental outcome of neonates treated with extracorporeal membrane oxygenation. J Pediatr Surg 25:43-46, 1990 2. Andrews AF, Nixon CA, Cilley RE, et al: One-to-three year outcome for 14 neonatal survivors of extracorporeal membrane oxygenation. Pediatrics 78:692-698, 1986 3. Bartlett RH, Roloff DW, Cornell RG, et al: Extracorporeal circulation in neonatal respiratory failure: a prospective randomized study. Pediatrics 76:479-487, 1985 4. Baumgart S, Graziani LJ, Streletz LJ, et al: Right common carotid artery reconstruction after extracorporeal membrane oxygenation: Vascular imaging, cerebral circulation, electroencephalographic, and neurodevelopmental correlates to recovery. J Pediatr 125:295-304, 1994 5. Bayley N: Bayley Scales of Infant Development, ed 2. San Antonio, Psychological Corporation, 1993 6. Campbell LR, Bunyapen C, Gangarosa ME, et al: Significance of seizures associated with extracorporeal membrane oxygenation. J Pediatr 119:789-792, 1991 7. Campbell LR, Bunyapen C, Holmes GL, et al: Right common carotid artery ligation in extracorporeal membrane oxygenation. J Pediatr 113:110-113, 1988 8. D' Agostino JA, Bernbaum JC, Gerdes M, et al: One year outcome of infants with congenital diaphragmatic hernia requiring extracorporeal membrane oxygenation (ECMO). J Pediatr Surg 30:10-15, 1995 9. DeAngelis GA, Mitchell DG, Merton DA, et al: Right common carotid artery reconstruction in neonates after extracorporeal membrane oxygenation: Color Doppler imaging. Radiology 182:521-525, 1992 10. Del Toro J, Louis PT, Goddard-Finegold J: Cerebrovascular regulation and neonatal brain injury. Pediatr Neurol 7:3-12, 1991 11. Desai S, Kollros PR, Graziani LJ, et al: Sensitivity and specificity of the neonatal brainstem auditory evoked potential (BAEP) for hearing and language deficits in extracorporeal membrane oxygenation (ECMO) survivors. J Pediatr 1997, in press 12. Desai S, Stanley C, Wiswell TE, et al: Long-term neurodevelopmental outcome following ECMO in congenital diaphragmatic hernia survivors: Improvement over time. Pediatr Res 37:253A, 1995 13. Desai SA, Stanley CW, Merton DA, et al: Four year follow-up Doppler ultrasound in children who had right common carotid artery reconstruction after neonatal extracorporeal membrane oxygenation. Pediatr Res 37:253A, 1995 14. Evans MJ, McKeever PA, Pearson GA, et al: Pathological complications of non-survivors of newborn extracorporeal membrane oxygenation. Arch Dis Child 71:88-92, 1994 15. Extracorporeal Life Support Organization: ELSO Registry Report. Ann Arbor, MI, ELSO, 1996 16. Glass P: Patient neurodevelopmental outcomes after neonatal ECMO. In Arensman RN, Cornish JD (eds): Extracorporeal Life Support. Boston, Blackwell Scientific Publications, 1993 17. Glass P, Miller MK, Short BL: Morbidity for survivors of extracorporeal membrane oxygenation: Neurodevelopmental outcome at 1 year of age. Pediatrics 83:72-78, 1989

COMPLICATIONS AND SEQUELAE OF NEONATAL ECMO

673

18. Glass P, Wagner AE, Papero PH, et al: Neurodevelopmental status at age five years of neonates treated with extracorporeal membrane oxygenation. JPediatr 127:447-57, 1995 19. Gleason CA: ECMO and the brain. In Arensman RM, Cornish JD (eds): Extracorporeal Life Support. Boston, Blackwell Scientific Publications, 1993 20. Graziani LJ, Baumgart S, Desai S, et al: Clinical antecedents of neurologic and audiologic abnormalities in survivors of neonatal ECMO. J Child Neurol 1997, in press 21. Graziani LJ, Spitzer AR, Mitchell DG, et al: Mechanical ventilation in preterm infants: Neurosonographic and developmental studies. Pediatrics 90:515-522, 1992 22. Graziani LJ, Streletz LJ, Baumgart S, et al: The predictive value of neonatal electroencephalograms before and during extracorporeal membrane oxygenation. J Pediatr 125:969-75, 1994 23. Graziani LJ, Streletz LJ, Kornhauser MS, et al: Electroencephalographic and cranial ultrasound studies in neonates treated with extracorporeal membrane oxygenation. Pediatric Rev Commun 4:71-75, 1989 24. Graziani LJ, Streletz LJ, Mitchell DC, et al: Electroencephalographic, neuroradiologic and neurodevelopmental studies in infants with subclavian steal during ECMO. Pediatr Neurol 10:97-103, 1994 25. Grigg-Damberger MD, Coker SB, Halsey CL, et al: Neonatal burst suppression: Its developmental significance. Pediatr Neurol 5:84--92, 1989 26. Gringlas MB, Stanley C, McKee L, et al: Developmental assessment of survivors of extracorporeal membrane oxygenation at early school age: A longitudinal analysis. Dev Med Child Neurol 37:21, 1995 27. Gringlas MB, Wiswell TE, Stanley C, et al: Primary diagnoses as predictors of longterm developmental outcome in school age survivors of extracorporeal life support (ECLS): A longitudinal study. Pediatr Res 39:266, 1996 28. Halm JS, Vaucher Y, Bejar R, et al: Electroencephalographic and neuroimaging findings in neonates undergoing extracorporeal membrane oxygenation. Neuropediatrics 24:1924, 1993 29. Hendricks-Munoz KD, Walton JP: Hearing loss in infants with persistent fetal circulation. Pediatrics 81:650-656, 1988 30. Hofkosh 0, Thompson AE, Nozza RJ, et al: Ten years of extracorporeal membrane oxygenation: Neurodevelopmental outcome. Pediatrics 87:549-555, 1991 31. Holmes GI, Rowe J, Hafford J: Significance of reactive burst suppression following asphyxia in full term infants. Clinical Electroencephalography 14:138-141, 1983 32. Holmes GL, Lombroso CT: Prognostic value of background patterns in the neonatal EEG. J C!in Neurophysiol 10:323-352, 1993 33. Korinthenberg R, Kachel W, Koelfen W, et al: Neurological findings in newborn infants after extracorporeal membrane oxygenation, with special reference to the EEG. Dev Med Child Neural 35:249-257, 1993 34. Kornhauser MS, Desai SA, Stanley CW, et al: Neurodevelopmental outcome among patients developing bronchopulmonary dysplasia following extracorporeal membrane oxygenation. Pediatr Res 37:262, 1995 35. Lago P, Rebsamen S, Clancy RR, et al: MRI, MRA, and neurodevelopmental outcome following neonatal ECMO. Pediatr Neural 12:294-304, 1995 36. Lott IT, McPherson D, Towne B, et al: Long-term neurophysiologic outcome after neonatal extracorporeal membrane oxygenation. J Pediatr 116:343-349, 1990 37. Mitchell DG, Merton D, Desai H, et al: Neonatal brain: Color Doppler imaging. Part II: Altered How patterns from extracorporeal membrane oxygenation. Radiology 167:307-310, 1988 38. Mitchell DC, Merton D, Needleman L, et al: Neonatal brain: Color Doppler imaging: I. Technique and vascular anatomy. Radiology 167:303-306, 1988 39. Mitchell DC, Merton DA, Graziani LJ, et al: Right carotid artery ligation in neonates: Classification of collateral flow with color Doppler imaging. Radiology 175:117-123, 1990 40. Mullen EM: Mullen Scales of Early Learning. Circle Pines, MN, American Guidance Service, 1993 41. O'Rourke PP, Crone RK, Vacanti JP, et al: Extracorporeal membrane oxygenation and

674

42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64.

GRAZIANI et al

conventional medical therapy in neonates with persistent pulmonary hypertension of the newborn: A prospective randomized study. Pediatrics 84:957-963, 1989 Page J, Frisk V, Whyte H: Developmental outcome of infants treated with extracorporeal membrane oxygenation (ECMO) in the neonatal period: Is the evidence all in? Paediatric and Perinatal Epidemiology 8:123-139, 1994 Panerai RB, Kelsall AWR, Rennie JM, et al: Estimation of critical closing pressure in the cerebral circulation of newborns. Neuropediatrics 26:168-173, 1995 Park C, Spitzer AR, Desai HJ, et al: Brain SPECT in neonates following extracorporeal membrane oxygenation: Evaluation of technique and preliminary results. Journal of Nuclear Medicine 33:1943-1948, 1992 Pryds 0: Control of cerebral circulation in the high-risk neonate. Ann Neurol 30:321329, 1991 Pryds 0: Regulation of cerebral perfusion and brain damage. In Lou HC, Greisen G, Larsen JF (eds): Brain Lesions in the Newborn. Copenhagen, Munksgaard, 1994 Pryds 0, Edwards AD: Cerebral blood flow in the newborn infant. Arch Dis Child 74:63-69, 1996 Pryds 0, Greisen G, Lou H, et al: Vasoparalysis associated with brain damage in asphyxiated term infants. J Pediatr 117:119-125, 1990 Radack DM, Baumgart S, Gross GW: Subependymal (grade I) intracranial hemorrhage in neonates on extracorporeal membrane oxygenation: Frequency and patterns of evolution. Clin Pediatr 33:583-587, 1994 Robertson CMT, Finer NN, Sauve RS, et al: Neurodevelopmental outcome after neonatal extracorporeal membrane oxygenation. Can Med Assoc J 152:1981-1988, 1995 Rourke BP: Nonverbal Learning Disabilities: The Syndrome and the Model. New York, The Guilford Press, 1989 Schumacher RE, Barks JDE, Jolmston MV, et al: Right-sided brain lesions infants following extracorporeal membrane oxygenation. Pediatrics 82:155-161, 1988 Schumacher RE, Palmer TW, Roloff DW, et al: Follow-up of infants treated with extracorporeal membrane oxygenation for newborn respiratory failure. Pediatrics 87:451-457, 1991 Short BL, Walker LK, Bender KS, et al: Impairment of cerebral autoregulation during extracorporeal membrane oxygenation in newborn lambs. Pediatr Res 33:289-294, 1993 Stanley C, Baumgart S, Graziani L: Review of cerebral palsy in children surviving severe neonatal respiratory failure and treated with extracorporeal membrane oxygenation. Dev Med Child Neurol 37:22, 1995 Stanley CW, Wiswell TE, Desai SA, et al: Long-term growth rate of infants with congenital diaphragmatic hernia compared to those with meconium aspiration syndrome following extracorporeal membrane oxygenation. Pediatr Res 37:274, 1995 Streletz LJ, Bej MD, Graziani LJ, et al: Utility of serial EEGs in neonates during extracorporeal membrane oxygenation. Pediatr Neurol 8:190-196, 1992 Takeuchi T, Watanabe K: The EEG evolution and neurological prognosis of neonates with perinatal hypoxia. Brain Dev 11:115-120, 1989 Taylor GA, Fitz CR, Glass P, et al: CT of cerebrovascular injury after neonatal extracorporeal membrane oxygenation: Implications for neurodevelopmental outcome. Am J Roentgenol 153:121-126, 1989 Taylor GA, Glass P, Fitz CA, et al: Neurologic status in infants treated with ECMO: Correlation of imaging findings with developmental outcome. Radiology 165:679682, 1987 Taylor GA, Short BL, Glass P, et al: Cerebral hemodynamics in infants undergoing extracorporeal membrane oxygenation: Further observations. Radiology 168:163-167, 1988 Taylor HG: Childhood sequelae of early neurological disorders: A contemporary perspective. Developmental Neuropsychology 3:153-164, 1987 Tharp BR, Laboyrie PM: The incidence of EEG abnormalities and outcome of infants paralyzed with neuromuscular blocking agents. Crit Care Med 11:926-929, 1983 Towne BH, Lott IT, Hicks DA, et al: Long-term follow-up of infants and children treated with extracorporeal membrane oxygenation (ECMO): A preliminary report. J Pediatr Surg 20:410-414, 1985

COMPLICATIONS AND SEQUELAE OF NEONATAL ECMO

675

65. Yannucci RC: Heterogeneity of hypoxic-ischemic thresholds in experimental animals. In Lou HC, Greisen G, Larsen JF (eds): Brain Lesions in the Newborn. Copenhagen, Munksgaard, 1994 66. Vaucher YE, Dudell GG, Bejar R, et al: Predictors of early childhood outcome in candidates for extracorporeal membrane oxygenation. J Pediatr 128:109-117, 1996 67. 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 68. Walsh-Sukys MC, Bauer RE, Cornell DJ, et al: Severe respiratory failure in neonates: Mortality and morbidity rates and neurodevelopmental outcomes. J Pediatr 125:104110, 1994 69. Wechsler D: The Wechsler Preschool and Primary Scale of Intelligence, Revised. New York, Psychological Corporation/Harcourt Brace Jovanovich, 1989 70. Wildin SR, Landry SH, Zwischenberger JB: Prospective, controlled study of developmental outcome in survivors of extracorporeal membrane oxygenation: The first 24 months. Pediatrics 93:404--408, 1994 71. Wiswell TE: Meconium staining and the meconium aspiration syndrome. In Stevenson DK, Sunshine P (eds): Fetal and Neonatal Brain Injury. London, Oxford University Press, 1997 72. Wiswell TE, Gannon CM, Desai S, et al: Outcomes in a large population of infants following venovenous extracorporeal life support (ECLS). Pediatr Res 39:284, 1996

Address reprint requests to Leonard J. Graziani, MD Division of Child Neurology and Development Jefferson Medical College 1025 Walnut Street Philadelphia, PA 19107