An MRI Study of Neurological Injury Before and After Congenital Heart Surgery
Background Neuorological deficits are observed in patients with congenital heart disease (CHD) before and after neonatal surgery, the etiology being multifactorial. To understand the impact of preoperative events and to characterize the evaluation of neurological injury, we performed serial magnetic resonance imaging (MRI) studies of the brain in a cohort of neonates undergoing open-heart surgery.
Methods and Results Twenty-four term neonates with CHD were studied prospectively with brain MRI: before surgery, within 2 weeks of surgery, and several months after surgery. Preoperative MRI examinations showed periventricular leukomalacia (PVL) in 4 patients (16%) and infarct in 2 subjects (8%). MR spectroscopy was performed in 19 subjects preoperatively and revealed elevated brain lactate in 53%. An early postoperative MRI (n=21) identified new PVL in 48%, new infarct in 19%, and new parenchymal hemorrhage in 33%. New lesions or worsening of preoperative lesions occurred in 67% of subjects. No patient- or procedure-related factors for the development of early postoperative lesions were identified. A late postoperative MRI (n=17) demonstrated resolution of early lesions in 8 and mild cerebral atrophy in 2.
Conclusions Mild ischemic lesions, primarily in the form of PVL, occur in a number of neonates with CHD before surgery and >50% of patients postoperatively. Resolution of these lesions is common 4 to 6 months after surgery. Longer-term follow-up is needed to determine the significance of perioperative ischemic lesions on functional outcome.
There has been longstanding concern about the potential for cerebral injury after infant heart surgery.1,2 Postoperative evaluations of cognitive function have consistently demonstrated neurological deficits.1–3 The etiology of neurocognitive impairment appears to be multifactorial, and a better understanding is needed. A variety of techniques have been used to gain insight into early neurological injury, including electroencephalography,4 brain isoenzymes,5 and measurements of cerebral metabolism.6 Imaging studies of the brain can provide valuable information regarding neurological insults. To date, most studies examining neonates with congenital heart disease (CHD) have used cranial ultrasound as opposed to magnetic resonance imaging (MRI).7–9 However, advances in MRI technology, such as MR spectroscopy (MRS), can provide significant additional information.9 Therefore, in the present study, we sought to determine the pattern and time course of neurological injury after open-heart surgery by performing serial MRI studies in neonates undergoing surgery with cardiopulmonary bypass (CPB).
A prospective observational study was carried out at the Children’s Hospital of Philadelphia with the approval of the Institutional Review Board. Full-term neonates (gestation>36 weeks, postnatal age<1 month) with CHD admitted between September 1, 2000, and January 31, 2001, for whom cardiac surgery with CPB had been planned were eligible for enrollment. Exclusion criteria were (1) named genetic anomaly associated with developmental delay, (2) birth asphyxia (5-minute Apgar score<5), (3) preoperative cardiac arrest, (4) postoperative mechanical circulatory support, or (5) reoperation requiring CPB. Of the 36 eligible patients who presented during the study period, 24 (66%) were enrolled in the study.
Serial MRI scans of the brain were performed at the following time periods: first scan (preoperative), day of surgery; second scan (early postoperative), 5 to 12 days after surgery; third scan (late postoperative), 3 to 6 months of age.
The third scan was obtain before additional cardiac surgery, such as a bidirectional Glenn procedure, or it was scheduled electively. The scans were performed on a Siemens Magnetom 1.5-T scanner. The MRI sequences were (1) sagittal T1-weighted spin-echo, (2) axial and coronal T2-weighted fast-spin-echo, (3) axial T2-weighted gradient-echo images, and (4) axial diffusion-weighted images. The follow-up studies consisted of the same sequences. In addition, preoperative MRS with long echo time (135 ms) of the basal ganglia and frontal and parietal white matter was performed on 19 study subjects. All MRI scans were reviewed by a single neuroradiologist (R.A.Z.) blinded to the subjects’ clinical status. MRI scans were reviewed for various lesions, including atrophy, periventricular leukomalacia (PVL), focal tissue loss, cerebral edema, delayed myelination, parenchymal hemorrhage, intracranial hemorrhage, and infarct. All lesions were classified as mild, moderate, or severe. Lesions were diagnosed from the combination of T1-weighted and T2-weighted images as well as diffusion and susceptibility studies, which are especially sensitive to blood products, water content, and calcium.
Anesthetic and CPB Management
Anesthetic and CPB management followed our standard institutional practice. For the preoperative MRI/MRS scans, anesthesia was induced with intravenous fentanyl (5 μg/kg) and pancuronium (0.2 mg/kg) followed by nasotracheal intubation and mechanical ventilation. After the initial MRI scan, patients were transported to the operating room for cardiac surgery. Anesthesia was maintained with intravenous fentanyl (20 μg/kg) and pancuronium (0.1 mg/kg). CPB used a membrane oxygenator, an arterial filter, and a nonpulsatile roller pump. Arterial blood-gas pH management followed the α-stat strategy. The CPB primer included fentanyl (30 μg/kg), pancuronium (0.2 mg/kg), methylprednisolone (30 mg/kg), Furosamide (1 mg/kg), and cefazolin (25 mg/kg). Whole blood was added to the primer to yield a hematocrit of 25% during CPB. All patients were CPB cooled to deep hypothermia (17 to 22°C), at which point total circulatory arrest or low-flow CPB was used. After surgical repair, CPB rewarming occurred with mannitol administration. No vasoconstrictors or vasodilators were used during CPB. All patients underwent modified ultrafiltration after separation from CPB. Continuous intravenous fentanyl was used for sedation in the early postoperative period. For the second MRI scan, patients received intravenous pentobarbital 2 to 4 mg/kg titrated to loss of consciousness with spontaneous ventilation. For the third MRI scan, patients scheduled for cardiac surgery immediately after the scan received intravenous fentanyl, propofol infusion, and pancuronium (0.2 mg/kg). All other patients received sedation similar to that for the second scan.
Data Collection and Statistical Analysis
Data recorded preoperatively included characteristics of gestation, labor and delivery, Apgar scores, and demographics. Duration of CPB and DHCA, cooling time, and hematocrit were included in intraoperative measurements.
Values are expressed as mean±SD for parametric data or median and range for nonparametric data, where appropriate. The primary outcome measure was the development of new or worsened parenchymal lesion (ie, PVL, infarction, hemorrhage) on the early postoperative scan when compared with the preoperative scan. Small choroid plexus hemorrhage and/or subdural hemorrhage was not included in the analysis. Univariate analysis was performed by Fisher exact test and logistic regression. Analysis was performed with STATA 6.0 (Texas A&M University, College Station, Tex). Significance was determined at a probability value of <0.05. All probability values are 2-sided, and confidence intervals are 95%.
The 24 subjects (16 male) in this study had various forms of complex CHD that necessitated neonatal surgical intervention. There were 13 patients with functional single ventricle, including 8 with the hypoplastic left-heart syndrome (Table 1). Prenatal diagnosis of CHD was made in 15 patients (63%). No patient had a preoperative seizure. Additional perinatal and preoperative data are shown in Table 2.
The patients underwent surgery at a median age of 4 days; range, 1 to 24 days. The median duration of CPB was 46.5 minutes; range, 37 to 185 minutes. DHCA was used in 21 patients (88%). The median duration of DHCA was 50 minutes; range, 2 to 78 minutes. The mean duration of DHCA was 44.3 minutes. All 24 of the enrolled subjects survived to hospital discharge. No patient had a seizure in the postoperative period. Three subjects (13%) did not have a second MRI scan. One subject was excluded due to the need for postoperative mechanical support, and 2 subjects withdrew. The median interval from the date of surgery to the second MRI scan was 7 days; range, 5 to 11 days. A third scan was performed in 17 subjects (2 withdrawal, 2 late deaths). The median interval from surgery to the third MRI scan was 161 days; range, 126 to 197 days. There were no complications related to sedation or anesthesia used for the MRI scans.
The only congenital structural abnormality of the brain that was detected on preoperative MRI studies was an open operculum, which was identified in 4 subjects (17%). Ischemic lesions were identified in 6 patients (Tables 3 and 4⇓), consisting of a small cortical watershed infarct seen as a high signal intensity on diffusion-weighted images in 1, a small infarct of the caudate in 1, and a PVL in 4. One subject had a small hemorrhage in the left temporal lobe. The other 17 MRI examinations were negative for parenchymal lesions. There were 12 children with blood along the tentorium, the falx, and in the choroid plexus related to vaginal delivery (Figure 1A). Of the 7 subjects with preoperative ischemic lesions detected by T1-weighted, T2-weighted, or diffusion imaging, 6 (86%) had elevated lactate on spectroscopy. There was a significant association between elevation of brain lactate and preoperative MRI lesions, P<0.02. Of the 19 patients who had an MRS, 9 had normal findings, 3 had elevated lactate with a diffuse distribution, 6 had lactate localized to the basal ganglia, and 1 had lactate in the peri-insular region (Figure 2).
Early Postoperative Scan
On the early postsurgical scan (n=21), 14 subjects (67%) demonstrated new or worsened focal lesions compared with preoperative MRI scans. These lesions included new PVL in 9 subjects (42%) (Table 3). PVL was detected primarily in the frontal or parietal regions (Figure 1B) and was bilateral in 7 of 9 cases (78%). In 1 subject, the early postoperative MRI demonstrated resolution of a mild PVL that had been detected preoperatively. Worsening of a PVL (from mild to moderate) was noted in 2 subjects. After surgery, the MRI remained normal in 6 patients (29%). Four patients had significant postoperative infarcts (mild in 2, moderate in 2). These infarcts occurred in the parietal lobe (n=1), frontal-parietal (n=1), and parietal-occipital (n=1) lobe and in the caudate (n=1). Diffusion-weighted imaging was positive in 2 of these 4 subjects. Diffusion-weighted images performed 5 to 11 days postoperatively were not able to detect a PVL in postsurgical follow-up. Mild subdural and/or choroid plexus hemorrhage was identified in 9 patients on the early postoperative scan.
There was no association between patient- and procedure-related factors and the development of new or worsened lesions on the early postoperative MRI scan. New or worsened lesions were detected in 7 of 12 (58%) patients who underwent the Norwood procedure and who had a second MRI scan. This was not statistically different from the remaining subjects, in whom lesions were detected in 7 of 9 (78%), P<0.34. The durations of DHCA and of CPB were not related to the development of early postoperative lesions, P<0.44 and P<0.56, respectively.
Late Postoperative Scans
The late-follow-up MRI demonstrated mild enlargement of the ventricles, indicating cerebral atrophy in 2 subjects (13%). Both of these subjects had lesions detected on the early postoperative scan. An old infarct of the left-parietal and occipital lobes was noted in 1 subject. One subject had developed a new cortical infarct. Resolution of a PVL was noted in all of the subjects with a PVL on the early postoperative scan (Figure 1C).
The present study demonstrates that parenchymal brain lesions occur in a number patients with CHD before neonatal surgical intervention and that these lesions are often associated with an elevation in brain lactate. In the early postoperative period, lesions are detected in over half of the patients, primarily in the form of a PVL, although more significant lesions, such as hemorrhagic infarcts, are also detected. The majority of lesions detected in the early postoperative period resolve within the next several months.
Neurolgoical deficits have been described in a large proportion of neonates with CHD.2,3,10,11 The incidence of brain lesions present in neonates with complex CHD who have not undergone surgical intervention is not well defined. Two retrospective studies have reported an incidence of cranial ultrasound abnormalities to be between 5% and 59%.7,8 This is the first study that has used MRI to evaluate neonates with CHD before surgery. Data from the low-birth-weight population suggests that MRI is more sensitive than cranial ultrasound for detecting injury in the periventricular white matter, which could not be identified on cranial ultrasound.12 Moreover, the patients with white-matter injury on MRI have an increased incidence of perceptual impairments, attention deficit disorder, and developmental delay at later evaluation, indicating that the lesions are clinically important. In the current series, preoperative ischemic lesions, primarily PVL, occurred in <25% of this patient population with complex heart disease. This is less than the incidence of abnormalities (59%) on cranial ultrasound in neonates with various forms of CHD before surgical intervention reported by Van Houten et al.7 In the study by Van Houten et al, however, the most common finding was linear echodensities in the basal ganglia and thalamus. We did not detect an MRI correlate to these echodensities in any study patient.
MRS is a novel technique not previously applied in the CHD population. Spectroscopy has been used to evaluate neonates after birth asphyxia, and elevations of brain lactate in the basal ganglia have been closely correlated with subsequent neurological deficits.13 In preterm neonates, low concentrations of brain lactate in the basal ganglia are thought to be normal. However, brain lactate is not usually present in term neonates. In this series, elevations in brain lactate were noted in >50% of subjects in the preoperative period, many of whom did not have accompanying ischemic lesions on T1 or T2 imaging. It is not surprising that lactate is present in the brain of many neonates with CHD. Previous investigators have shown that cerebral oxygen saturation is significantly lower than in the normal population.14 This is due to a combination of cyanosis and low cardiac output. Interestingly, the presence of lactate on spectroscopy did not correlate with metabolic acidosis measured as the lowest recorded pH, suggesting that vulnerability of the brain to ischemia may differ from that of other end-organs such as the kidneys or liver. It should be noted, however, that the elevation in brain lactate after birth asphyxia can greatly exceed that noted in the current series.15
The most common lesion detected in the postoperative period was PVL. Several previous investigations have identified PVL after CHD surgery.1,16 Glauser and colleagues16 identified PVL in 25% of subjects with hypoplastic left-heart syndrome. The pathogenesis of PVL is thought to be related to hypoxia/ischemia to immature oligodendroglia in the process of myelination, which are most vulnerable to injury.17 Collectively, these data and the current study suggest that the periventricular white matter is particularly susceptible to ischemic insult in the term infant with CHD. This may be due to hypoperfusion during the surgical procedure or in the early postoperative period. Studies in fetal and neonatal lambs have shown that during the phase of reperfusion after ischemia, a decrease in cerebral blood flow to white matter persists, despite blood flow recovery in all other brain areas.18
Diffusion-weighted MR is a technique used for relating image intensities to the relative mobility of endogenous tissue water molecules. Diffusion imaging has proven valuable in the detection of cerebral ischemia early after a cerebrovascular accident.19 In our study with diffusion imaging performed 5 to 11 days after surgery, this technique was unable to detect PVL in any patient but was able to detect infarcts in 50% of patients. This limited ability of diffusion to identify white-matter ischemia has previously been reported.19 Differences between the water content of neonatal and adult brain may explain the lower sensitivity of diffusion to detect ischemic damage to the deep white matter. The neonatal brain has a much higher water content than the adult brain, with increased water diffusion and faster changes in the apparent diffusion coefficient.20 It is also likely that the follow-up at 5 to 12 days postoperatively may be too late for the diffusion abnormalities to remain as a high signal intensity.
A variety of patient and procedure-related risk factors for neurological injury have been identified in the CHD population. Most attention has been focused on intraoperative support—particularly the use of DHCA and CPB.1,2,21 In this study, there was no relationship between duration of DHCA and/or CPB and ischemic brain lesions. The lack of a significant association may be related to a relatively small patient population. Alternatively, it is possible that other factors related to neonatal heart surgery, such as borderline postoperative cardiac output, may play a greater role. Previous studies have suggested that patients with hypoplastic left-heart syndrome or its variants are at particular risk for neurological injury.3,16 In this series, patients with hypoplastic left-heart syndrome did not demonstrate a higher incidence of postoperative lesions. Perhaps other factors, such as abnormalities of in utero cerebral blood flow or hemodynamics associated with a single-ventricle circulation, may be more important risk factors to later neurological deficits.
The late postoperative studies demonstrated resolution of PVL in all patients studied. This may be related in part to the relatively low-white matter content of the 4- to 6-month-old infant.22 Studies in preterm infants have found that PVL, with characteristic glial scarring, is best detected after 1 year of age.23 In addition, the degree of PVL was mild in most of the patients in this series. With more severe PVL, we would expect to find characteristic cystic changes, even at 3 to 6 months of age. Ventriculomegaly and cerebral atrophy have been reported in 2 previous studies of children with CHD who underwent previous open-heart surgery.7,24 In our series, 2 of the 17 patients had cerebral atrophy on the late postoperative scan. Interestingly, both of these patients with prominent ventricles had a PVL in the early postoperative period. Unfortunately, more quantitative measures of brain mass, as can be performed with three-dimensional MRI, were not undertaken.
The major limitation to this study is that longer functional assessment is not yet available. Although the literature supports a relationship between acquired brain lesions in subsequent deficits in the preterm population, such data are lacking in the CHD population. An additional limitation is that the second MRI scan was not obtained until 5 to 12 days after surgery. It is difficult to separate operative factors from postoperative factors that may have contributed to acquired brain lesions.
Mild ischemic lesions, primarily in the form of a PVL, occur in a number of neonates with CHD before surgery and >50% of patients postoperatively. Resolution of these lesions is common 4 to 6 months after surgery. Longer-term follow-up is needed to determine the significance of perioperative ischemic lesions on functional outcome.
This work was supported by a grant from Baxter Pharmaceuticals.
The data of this study have been presented at the annual meeting of the AHA 2001
Mahle WT, Clancy RR, Moss E, et al. Neurodevelopmental outcome and lifestyle assessment in school-age and adolescent children with hypoplastic left heart syndrome. Pediatrics. 2000; 105: 1085–1092.
Rappaport LA, Wypij D, Bellinger DC, et al. Relation of seizures after cardiac surgery in early infancy to neurodevelopmental outcome: Boston Circulatory Arrest Study Group. Circulation. 1998; 97: 773–779.
Kramer HH, Majewski F, Trampisch HJ, et al. Malformation patterns in children with congenital heart disease. Am J Dis Child. 1987; 141: 789–795.
Olsen P, Vainionpaa L, Paakko E, et al. Psychological findings in preterm children related to neurologic status and magnetic resonance imaging. Pediatrics. 1998; 102: 329–336.
Limperopoulos C, Majnemer A, Shevell MI, et al. Neurologic status of newborns with congenital heart defects before open heart surgery. Pediatrics. 1999; 103: 402–408.
Forbess JM, Visconti KJ, Bellinger DC. Neurodevelopmental outcomes in children after the Fontan operation. Circulation. 2001; 104 (suppl 1): I-127–I-32.
Barkovich AJ, Baranski K, Vigneron D, et al. Proton MR spectroscopy for the evaluation of brain injury in asphyxiated, term neonates. AJNR Am J Neuroradiol. 1999; 20: 1399–1405.
Glauser TA, Rorke LB, Weinberg PM, et al. Acquired neuropathologic lesions associated with the hypoplastic left heart syndrome. Pediatrics. 1990; 85: 991–1000.
Szymonowicz W, Walker AM, Yu YH, et al. Regional cerebral blood flow after hemorrhagic hypotension in the preterm, near term and term lamb. Pediatr Res. 1990; 162: 977–982.
Phillips MD, Zimmerman RA. Diffusion imaging in pediatric hypoxic ischemia injury. Neuroimaging Clin North Am. 1999; 9: 41–52.
Forbes KNP, Pipe JG, Bird R. Neonatal hypoxic-ischemic encephalopathy: detection with diffusion-weighted MR imaging. AJNR. 2000; 21: 1490–1496.
Miller G, Mamourian AC, Tesman JR, et al. Long-term MRI changes in brain after pediatric open heart surgery. J Child Neurol. 1994; 9: 390–397.