Utility of neuroradiographic imaging in predicting outcomes after neonatal extracorporeal membrane oxygenation

Transcription

1 Journal of Pediatric Surgery (2012) 47, Utility of neuroradiographic imaging in predicting outcomes after neonatal extracorporeal membrane oxygenation Michael D. Rollins a,, Bradley A. Yoder b, Kevin R. Moore c, Douglas C. Barnhart a, Chris Jones d, Donald M. Null b, Robert J. DiGeronimo b a Division of Pediatric Surgery, University of Utah, Primary Children's Medical Center, Salt Lake City, UT 84113, USA b Division of Neonatology, University of Utah, Primary Children's Medical Center, Salt Lake City, UT 84113, USA c Department of Pediatric Radiology, Primary Children's Medical Center, Salt Lake City, UT 84113, USA d Primary Children's Medical Center, Salt Lake City, UT 84113, USA Received 26 September 2011; accepted 6 October 2011 Key words: Extracorporeal membrane oxygenation; Ultrasound; Magnetic resonance imaging; Neurodevelopmental outcome; Neonates Abstract Background: The need for routine neuroimaging after extracorporeal membrane oxygenation (ECMO) and the optimal radiographic study remains unclear. We sought to evaluate the correlation between findings on head ultrasound (HUS) and magnetic resonance imaging (MRI) and determine the association of these findings to neurodevelopmental outcome. Methods: A retrospective review was performed ( ) to identify neonates who had a MRI after ECMO. Each MRI was reviewed by a single pediatric neuroradiologist. Neurodevelopmental data was collected from the high-risk neonatal follow-up clinic. Results: Fifty neonates had a MRI (venoarterial 37, venovenous 13) after ECMO. HUS was abnormal in 24%, whereas MRI was abnormal in 62%. All infants with an abnormal HUS had an abnormal MRI, but an additional 50% of patients with a normal HUS had an abnormal MRI. Venoarterial ECMO was significantly associated with an abnormal MRI. Follow-up data was available for 26 neonates. The only predictor of abnormal neurodevelopment was the need for supplemental tube feeds at discharge. Conclusions: MRI identified significantly more abnormalities compared to routine HUS after neonatal ECMO. However, neither MRI nor HUS findings correlated with early neurodevelopmental outcome. Feeding ability at discharge was the overall best predictor of neurologic impairment in survivors Elsevier Inc. All rights reserved. Extracorporeal membrane oxygenation (ECMO) is a lifesaving therapy for many neonates with refractory respiratory Corresponding author. Primary Children's Medical Center, Division of Pediatric Surgery, USA. Tel.: ; fax: addresses: michael.rollins@imail.org, michael.rollins@hsc.utah.edu (M.D. Rollins). failure. Although cerebral hemorrhage and infarct are among the most feared complications while a patient is on ECMO, permanent central nervous system injury resulting in neurodevelopment impairment (NDI) in survivors is a more significant outcome measure. The incidence of abnormal intracranial neuroimaging after ECMO ranges from 0% to 52% [1-4]. ECMO survivors who suffer long /$ see front matter 2012 Elsevier Inc. All rights reserved. doi: /j.jpedsurg

2 Utility of neuroradiographic imaging in predicting outcomes term NDI ranges from 15% to 50% with the highest incidence reported in congenital diaphragmatic hernia () patients [5-7]. Head ultrasound (HUS) is routinely performed prior to initiating ECMO and daily while on ECMO. While it is useful to detect significant intracranial bleeding that would affect therapy, it is inferior to computed tomography (CT) and magnetic resonance imaging (MRI) in identifying nonhemorrhagic and small hemorrhagic lesions [4,8,9]. Bulas et al [1] found that routine neuroimaging with CT contributed additional information in 73% of patients after ECMO. The significance of additional neuroradiographic findings discovered on CT and MRI in regards to neurodevelopment remains unclear, raising the question of the utility of obtaining these additional studies routinely. We sought to evaluate the correlation between findings on HUS and MRI and determine the association of these findings with neurodevelopmental outcome. 1. Methods A retrospective review approved by the institutional review board to identify neonates who had a MRI after ECMO was performed from January 2003 to September Venovenous (VV) or venoarterial (VA) ECMO support was used. One patient required conversion from VV to VA and was included in the VA group. The vessels were not reconstructed after ECMO. Each available MRI was reviewed by a single pediatric neuroradiologist (KRM). The MRI was obtained in each patient before hospital discharge. MRI findings of slow flow or occlusion of the right internal carotid artery (VA ECMO patients), a slight increase in the subarachnoid space and mild ventriculomegaly were considered expected or normal findings after ECMO. Any other lesion identified which was consistent with an ischemic or hemorrhagic event was considered abnormal. Each patient's most recent HUS was used for comparison. In 18 patients, a follow-up MRI was performed at another institution which was not available for review and the reported findings were used for analysis. Neurodevelopmental follow-up data were collected from outpatient clinic visits in our high-risk neonatal follow-up clinic. Bayley Scales of Infant Development were performed by neurodevelopmental specialists and parameters defined as normal (N85), mild delay (71-84), and moderate/severe delay (b70) as previously described [5]. Statistical analyses were performed using SPSS (version 18.0). Continuous data were analyzed by nonparametric tests and categorical data by two-tailed Fisher exact test. Univariate analysis was performed to evaluate the association of patient demographics, ECMO mode, presence of, and need for supplemental tube feeds at discharge. Variables that approached significance on univariate analysis (P =.1) were included in a logistic model to evaluate neurodevelopmental outcomes as a function of radiographic findings. 2. Results During the study period, 101 neonates were treated with ECMO with an overall survival rate of 63%. VA ECMO was used in 80 patients and VV in 21 with survival rates of 63% and 67% respectively. Of these patients, 39 required ECMO for management of respiratory failure secondary to with a 54% survival rate. Fifty neonates had a MRI after ECMO (24, 26 non- ). VA ECMO with ligation of the carotid artery was performed in 37 patients, while 13 were managed with VV ECMO (Table 1). Gestational age, birth weight, and gender were similar between and non- groups. Significantly more neonates with were managed using VA ECMO whereas VV ECMO was commonly used in non- patients. patients were on ECMO significantly longer, spent more days on the ventilator and required supplemental oxygen longer. HUS was abnormal in 24% of patients on ECMO whereas MRI after decannulation was abnormal in 62% (Table 2). All patients who had an abnormal HUS had an abnormal MRI but an additional 50% of patients with a normal HUS had an abnormal MRI. There was not an association between finding an abnormal HUS and an abnormal MRI (Pearson R = 0.440). Patients treated with VA ECMO had an almost 4- fold increase in odds of having an abnormal MRI (O.R. 3.78). The spectrum of individual lesions seen on MRI as well as their distribution based on cerebral vascular supply is reported in Table 3. Enlargement of the subarachnoid space was a common finding in our patients (62% overall), with no statistically significant difference seen between the VA group (25/37, 68%) and the VV group (6/13, 46%). In addition, we found no significant effect on NDI related to enlargement of the subarachnoid space. Table 1 Demographics (n=24) Non- (n=26) P value Gestation (weeks) 38.3 ± ± Birthweight (g) 3194 ± ± Gender (% male) 15 (63) 16 (62) 1.0 ECMO mode VA 22 (92) 15 (58).009 VV 2 (8) 11 (42) ECMO time (days) 9.8 ± ± Ventilator (days) 30.9 ± ± O2 (days) 67.5 ± ± Discharge age (days) 76.5 ± ± Survive to discharge 22 (91) 24 (92) 1.0 Data as mean ± SD or number. P b

3 78 M.D. Rollins et al. Table 2 Neuroimaging results Non- Total Abnormal head ultrasound VA ECMO 5/22 (23) 5/15 (33) 10/37 (27).48 VV ECMO 0/2 2/11 (18) 2/13 (15) Abnormal MRI VA ECMO 15/22 (68) 11/15 (73) 26/37 (70).05 VV ECMO 1/2 (50) 4/11 (36) 5/13 (38) P =.05 total abnormal MRI VA ECMO vs VV ECMO. Follow-up data was available for 26 neonates ( 6 months, n = 26; 12 months, n = 20; 18 months, n = 15; 24 months, n = 11; 30 months, n = 11). Findings on HUS and MRI, mode of ECMO, and diagnosis were not predictive of NDI (Table 4). However, neonates that required a feeding tube at discharge were more likely to have abnormal neurodevelopment compared to those who did not. Infants that required tube feeds at discharge were 12 times more likely to have moderate to severe NDI (OR 12.0, 95% CI ). Table 3 Findings, location, and vascular distribution of brain injury on MRI Abnormal brain MRI findings (no. of lesions) Abnormal flow right internal Basal ganglia carotid artery (26) encephalomalacia (1) Mild ventriculomegally (19) Periventricular white matter injury (6) Moderate to severe Periventricular white matter ventriculomegally (11) thinning (1) Mildly increased subarachnoid Cystic periventricular space (7) leukomalacia (1) Moderate to severely increased Cerebellum hemorrhage (4) subarachnoid space (24) Effaced subarachnoid space (2) Cerebellum atrophy (1) Choroid plexus hemorrhage (11) Parenchymal atrophy (3) Intraventricular hemorrhage (7) Parenchymal infarct (3) Germinal matrix hemorrhage (1) Cortical hemorrhage (3) Subdural hematoma (4) Focal white matter injury (3) Thalamic injury (4) White matter cystic encephalomalacia (1) Basal ganglia hemorrhage (3) Location of parenchymal injury Lateralization Right Left Bilateral Anterior circulation Hemisphere 7 (14%) 0 7 (14%) Basal ganglia 1 (2%) 2 (4%) 0 Posterior circulation Cerebellum 1 (2%) 0 1 (2%) Thalamus 1 (2%) 0 3 (6%) MRI images available for review right- vs. left-sided lesions in the anterior circulation distribution, nonsignificant (P =.24) P Table 4 Neurodevelopmental outcome in patients with follow up 12 months n = 20 Neurodevelopment P Normal Abnormal Abnormal head ultrasound 2/11 (18%) 4/9 (44%).34 Abnormal MRI 7/11 (64%) 7/9 (78%).64 ECMO mode Venoarterial (n = 15) 8 (53%) 7 (47%) 1.0 Venovenous (n = 5) 3 (60%) 2 (40%) YES (n = 10) 6 (60%) 4 (40%) 1.0 NO (n = 10) 5 (50%) 5 (50%) Feeding tube dependent 2/11 (18%) 7/9 (78%).022 P =.02 by Fisher exact test, feeding tube dependent at discharge. Cognitive, motor, hearing, and vision were normal in 83%, 50%, 80%, and 100%, respectively, of our patients with a normal MRI compared with 71%, 53%, 100%, and 62% of those with an abnormal MRI. No significant differences in neurodevelopment were identified between the and non- groups. As the only variable on univariate analysis which correlated with neurodevelopmental outcome was tube feed dependence this was included in a logistic regression model. Head ultrasound and MRI were also included in the logistic model as they were primary variables of interest. Neither significantly correlated with neurodevelopmental outcomes. 3. Discussion The utility of performing routine neuroimaging after ECMO decannulation in neonates remains unclear. Owing to the superior sensitivity of CT and MRI over HUS in identification of non-hemorrhagic and small hemorrhagic lesions, it has been recommended to obtain follow-up imaging with these modalities after ECMO in neonates [8]. Various reports in the literature support that a high percentage of these scans will yield additional information compared to HUS alone [1,8,9]. We found in our study that all neonates that had an abnormal HUS also had an abnormal MRI, while 50% of those with a normal HUS additionally had an abnormal MRI. Overall, the HUS was abnormal in 24% during ECMO, whereas MRI after decannulation showed significantly more abnormalities with findings in 62%. While previous studies have noted a slightly lower incidence of central nervous system (CNS) lesions after ECMO ranging from 0% to 42%, most of these reports used combined modes of CNS imaging and not just solely MRI [2,3,9,10]. More recent data in neonates imaged with MRI have shown significantly higher rates of CNS abnormalities, with an incidence as high as 80% to 90% [11,12]. Additional factors that may contribute to differences in abnormal imaging findings between studies in neonates

4 Utility of neuroradiographic imaging in predicting outcomes supported with ECMO include the definition of an abnormal MRI, study size, and differences in patient population. Questions remain regarding the long-term implications on neurodevelopment of neuroimaging findings after ECMO. Glass et al [10] reported their findings in 152 neonates treated exclusively with VA ECMO using head CT performed after decannulation. One or more major disability conditions occurred in 17% of patients at 5-year follow-up. Interestingly, 43% of children with severe and 67% with moderate brain injury on neuroimaging were later noted to have no disability at 5-year follow-up. Lazar et al [4] examined neonates who required ECMO for respiratory failure using follow up CT scan after 1 week of ventilator independence and MRI as an outpatient. They found 13.5% had delayed neurological development despite no evidence of anatomic injury on serial HUS or follow-up imaging. These reports suggest that neuroimaging results after ECMO should be interpreted with caution in regards to prediction of outcome. To date, definitive differences in neurodevelopmental outcome in neonates supported with VA versus VV ECMO have not been reported. A review of the Extracorporeal Life Support Organization (ELSO) database [13] of all neonates from 1991 to 2006 found neurologic complications including seizures and cerebral infarcts to be more common with VA vs VV ECMO (15.7% vs 11.5%). In our study, we also found a significant increase in abnormal MRI findings in all patients managed with VA as compared to VV ECMO; however, this result did not correlate with differences in neurodevelopmental outcome. Logistic regression analysis of our data failed to find a correlation between abnormal MRI findings and the majority of outcome variables, although prognostic significance of a patient's dependence on tube feeds on neurologic outcome was demonstrated. Infants who required a feeding tube at the time of discharge were 12 times more likely to have abnormal neurodevelopment compared to those who were meeting their caloric needs orally. While feeding problems are known to commonly occur in surviving neonates after ECMO, with and without [14,15], this is the first report that we know of that has correlated this finding with neurodevelopmental outcome in this population of patients. There are several limitations to this study. Our definition of a normal MRI included findings of slow flow or occlusion of the right internal carotid artery, a slight increase in the subarachnoid space and mild ventriculomegaly. It could be argued that this is a liberal definition in which case we would have had significantly more patients with imaging abnormalities. Our patient population is also limited with only 20 patients having at least 12 months follow up. It is possible that neurodevelopmental abnormalities may become more or less apparent with time. Finally, owing to the retrospective nature of this study, we were only able to evaluate the recorded data from follow up visits, which was incomplete in some of our patients. Nevertheless, our data do not support the utility of routine MRI for prognostication in infants after ECMO. 4. Conclusion Although MRI identified significantly more radiographic abnormalities compared to routine HUS after neonatal ECMO, neither MRI nor HUS findings correlated with subsequent neurodevelopmental outcome. In contrast, an inability to take adequate oral feeds was a strong predictor of poor neurodevelopment. Larger studies with rigorous longterm follow up are needed before the true value of MRI in this patient population can be evaluated. References [1] Bulas DI, Taylor GA, O'Donnell RM, et al. Intracranial abnormalities in infants treated with extracorporeal membrane oxygenation: update on sonographic and CT findings. AJNR Am J Neuroradiol 1996;17: [2] Griffin MP, Minifee PK, Landry SH, et al. Neurodevelopmental outcome in neonates after extracorporeal membrane oxygenation: cranial magnetic resonance imaging and ultrasonography correlation. J Pediatr Surg 1992;27:33-5. [3] Lago P, Rebsamen S, Clancy RR, et al. MRI, MRA, and neurodevelopmental outcome following neonatal ECMO. Pediatr Neurol 1995;12: [4] Lazar EL, Abramson SJ, Weinstein S, et al. Neuroimaging of brain injury in neonates treated with extracorporeal membrane oxygenation: lessons learned from serial examinations. J Pediatr Surg 1994;29: [5] Danzer E, Gerdes M, Bernbaum J, et al. Neurodevelopmental outcome of infants with congenital diaphragmatic hernia prospectively enrolled in an interdisciplinary follow-up program. J Pediatr Surg 2010;45: [6] Glass P, Wagner AE, Papero PH, et al. Neurodevelopmental status at age five years of neonates treated with extracorporeal membrane oxygenation. J Pediatr 1995;127: [7] Graziani LJ, Gringlas M, Baumgart S. Cerebrovascular complications and neurodevelopmental sequelae of neonatal ECMO. Clin Perinatol 1997;24: [8] Bulas D, Glass P. Neonatal ECMO: neuroimaging and neurodevelopmental outcome. Semin Perinatol 2005;29: [9] Gannon CM, Kornhauser MS, Gross GW, et al. When combined, early bedside head ultrasound and electroencephalography predict abnormal computerized tomography or magnetic resonance brain images obtained after extracorporeal membrane oxygenation treatment. J Perinatol 2001;21: [10] Glass P, Bulas DI, Wagner AE, et al. Severity of brain injury following neonatal extracorporeal membrane oxygenation and outcome at age 5 years. Dev Med Child Neurol 1997;39: [11] Hunt RW, Kean MJ, Stewart MJ, et al. Patterns of cerebral injury in a series of infants with congenital diaphragmatic hernia utilizing magnetic resonance imaging. J Pediatr Surg 2004;39:31-6. [12] Tracy S, Estroff J, Valim C, et al. Abnormal neuroimaging and neurodevelopmental findings in a cohort of antenatally diagnosed congenital diaphragmatic hernia survivors. J Pediatr Surg 2010;45: [13] Guner YS, Khemani RG, Qureshi FG, et al. Outcome analysis of neonates with congenital diaphragmatic hernia treated with venovenous vs venoarterial extracorporeal membrane oxygenation. J Pediatr Surg 2009;44: [14] Bernbaum J, Schwartz IP, Gerdes M, et al. Survivors of extracorporeal membrane oxygenation at 1 year of age: the relationship of primary diagnosis with health and neurodevelopmental sequelae. Pediatrics 1995;96:

5 80 M.D. Rollins et al. [15] D'Agostino JA, Bernbaum JC, Gerdes M, et al. Outcome for infants with congenital diaphragmatic hernia requiring extracorporeal membrane oxygenation: the first year. J Pediatr Surg 1995;30:10-5. Discussion Discussant: A nice study correlating with developmental outcomes. Based on the fact that you cannot correlate outcomes with the use of MRI and the increasing concern over cost of medical care, has that changed your follow up protocol in terms of imaging? And what is your protocol? Rollins (response): That was actually one of the reasons we chose to look at this because it is a definite change in our practice over the years or our neonatologists' practice to routinely obtain an MRI in any patient after ECMO. I think the literature that is out there as well as what we have seen in our series would question whether that is of any value. Patients have been shown to have neurologic compensation with time and it is unknown exactly what these lesions mean. Philip Frykman, MD (Los Angeles, CA): I think that is really great work. One of the questions I had is methodological and that is at what point did you decide after the patients had completed an ECMO run to perform your MRIs? The second question or a correlate to that would be how did you decide when you were going to do this? Rollins (response): MRIs were obtained prior to discharge so there was a varying amount of time that elapsed after the ECMO run but before they went home. All patients after ECMO have begun to receive MRIs. Some of the patients in the early part of the study had MRIs obtained owing to symptoms or just at the discretion of the treating neonatologist. Over the past 2 years they acquire an MRI on all patients after ECMO.