Objective The purposes of this feasibility study were to assess: (1) the potential utility of early brain MRI in asphyxiated newborns treated with hypothermia; (2) whether early MRI predicts later brain injury observed in these newborns after hypothermia has been completed; and (3) whether early MRI indicators of brain injury in these newborns represent reversible changes.
Patients and methods All consecutive asphyxiated term newborns meeting the criteria for therapeutic hypothermia were enrolled prospectively. Each newborn underwent one or two early MRI scans while receiving hypothermia, on day of life (DOL) 1 and DOL 2–3 and also one or two late MRI scans on DOL 8–13 and at 1 month of age.
Results 37 MRI scans were obtained in 12 asphyxiated neonates treated with induced hypothermia. Four newborns developed MRI evidence of brain injury, already visible on early MRI scans. The remaining eight newborns did not develop significant MRI evidence of brain injury on any of the MRI scans. In addition, two patients displayed unexpected findings on early MRIs, leading to early termination of hypothermia treatment.
Conclusions MRI scans obtained on DOL 2–3 during hypothermia seem to predict later brain injuries in asphyxiated newborns. Brain injuries identified during this early time appear to represent irreversible changes. Early MRI scans might also be useful to demonstrate unexpected findings not related to hypoxic–ischaemic encephalopathy, which could potentially be exacerbated by induced hypothermia. Additional studies with larger numbers of patients will be useful to confirm these results.
Statistics from Altmetric.com
Induced hypothermia has a strong safety and efficacy record for the treatment of hypoxic–ischaemic encephalopathy (HIE), suggesting decreased death and disability at 12–18 months.1,–,8 While this therapy is increasingly accepted, many questions remain unanswered,9,–,11 including issues pertaining to MRI. One such question is when the optimal timing of brain imaging is to most accurately define the degree of brain injury sustained and predict the neurological prognosis.9,–,11 The optimal earliest timing of imaging has been extensively studied in asphyxiated newborns before the cooling era,12,–,14 but it is currently not clear if expected neurological outcome extrapolated from early MRI in the precooling era can accurately be applied to the newborn treated with induced hypothermia.10 15,–,17 The minimal literature regarding brain imaging in newborns treated with induced hypothermia15,–,17 does not address this issue, because brain MRIs have usually been performed after completion of induced hypothermia.15,–,17 There are also no published studies that address the evolution of brain MRI findings within the first month of life in these newborns.
What is already known on this topic
▶ Induced hypothermia has a strong safety and efficacy record for the treatment of hypoxic–ischaemic encephalopathy.
▶ In newborns treated with induced hypothermia, the optimal earliest time of brain imaging to most accurately define the degree of brain injury is not known.
▶ MRIs have usually been performed after completion of induced hypothermia.
What this study adds
▶ This study demonstrates that MRI scans obtained on days of life 2–3 during hypothermia may predict later brain injuries in asphyxiated newborns.
▶ Brain injuries identified during this early time appear to be irreversible.
This feasibility study was designed to assess: (1) the potential utility of early brain imaging to define as early as possible the degree of brain injury sustained in asphyxiated newborns treated with induced hypothermia; (2) whether early MRI findings in term asphyxiated newborns being treated with induced hypothermia predict later brain injury observed on MRI in these patients after completion of induced hypothermia; and (3) whether or not early MRI indicators of brain injury seen in the first days of life in these newborns represent reversible changes.
Patients and methods
We conducted a prospective cohort study of consecutive term asphyxiated newborns admitted to the neonatal intensive care unit (NICU) from June 2008 to August 2009 meeting the criteria for induced hypothermia: (1) gestational age ≥36 weeks and birth weight ≥2000 g; (2) evidence of fetal distress, for example, history of acute perinatal event, biophysical profile <6/10 within 6 h of birth, or cord pH ≤7.0; (3) evidence of neonatal distress, such as Apgar score ≤5 at 10 min or postnatal blood gas pH obtained within the first hour of life ≤7.0, or continued need for ventilation initiated at birth and continued for at least 10 min; (4) evidence of neonatal encephalopathy by physical examination; and (5) abnormal amplitude-integrated EEG (aEEG) background pattern. Newborns who met all five of the above criteria received whole-body cooling to an oesophageal temperature of 33.5°C, initiated ideally by 6 h of life, continued for 72 h (unless contraindications developed), and were then slowly rewarmed.3
Neonates were categorised according to their initial background pattern of aEEG into two categories: moderately or severely abnormal.2 18 The initial background pattern of aEEG was assessed during a recording of at least 20 min within 7 h of life.2 aEEG recording was started as soon as the patient was admitted to the NICU; to avoid delay, hypothermia treatment was usually initiated slowly at the same time. Seizures evident clinically or identified by aEEG or standard EEG were also recorded. Other variables associated with neonatal brain injury, including resuscitation score,19 encephalopathy score20 and seizure score,19 were also collected prospectively.
Sequential MRI studies were planned in order to clarify the time course of brain changes during the first month of life. Specifically, the newborns underwent one or two ‘early’ MRI scans while they were receiving hypothermia, including a first scan on day of life (DOL) 1 and a second on DOL 2–3. The timing of these MRI scans was chosen to ensure that there was no older brain injury (DOL 1) and according to the expected time of peak visibility of brain injury on diffusion-weighted imaging (DWI) and spectroscopy in the precooling era (DOL 2–3). Then they underwent one or two ‘late’ MRI scans, including a third scan on DOL 8–13 and a fourth at 1 month of age, in order to evaluate definitive T2-weighted MRI changes. Each enrolled newborn underwent all scans unless determined to be too unstable to tolerate the study safely. Patients receiving hypothermia treatment had the therapy maintained during the MRI scan without any adverse events. Any ventilation, pressor support or sedation was also maintained during the MRI scanning process; additional sedation was avoided. MRI scans were performed using a 3T Siemens Symphony (Siemens, Erlangen, Germany) scanner. Each MRI study included anatomic T1- and T2-weighted imaging, DWI and spectroscopy. Induced hypothermia was continued during the early brain MRI scans. The protocol was approved by the institutional review board and parental consent was obtained for each MRI scan. MRI images were interpreted by neuroradiologists, who were blinded to the clinical condition of the infants. Each MRI was scored using an MRI scoring system,21 consisting of a basal ganglia injury scale and a watershed pattern injury scale. This MRI assessment provided a quantitative description of the distribution and extent of injury.
A focused neurological examination assessing changes of tone and/or reflexes, power in the trunk and extremities, and presence or absence of cranial nerve involvement permitted calculation of a neuromotor outcome score22 on DOL 10 or at discharge from NICU, whichever came first. It was calculated again at each neurology follow-up visit through 1 year of age, typically at 2, 6 and 12 months of age. This score has been validated in asphyxiated newborns not treated with induced hypothermia. Although it has not been specifically revalidated in those treated with induced hypothermia, we selected it to provide a feasible, objective clinical comparison between the newborns. The scores were calculated from neurologists' evaluations by a single investigator.
Our relatively small sample size does not permit statistical testing. Therefore our results are reported descriptively.
Twelve asphyxiated term neonates receiving therapeutic hypothermia were enrolled in the study (table 1), four of whom had an initial moderately abnormal aEEG and eight who had an initial severely abnormal aEEG. Whole-body cooling was initiated at an average of 4.4 h of life (range: 2.6–6.6 h of life). All the newborns had very similar elevated encephalopathy scores. Three patients (nos. 10, 11 and 12) died (on DOL 4, 4 and 13, respectively, after 68.5, 48.6 and 72 h of hypothermia, respectively) from complications of HIE with no improvement in their severely abnormal neurological clinical examination and EEG. This represents 25% of the total patients and 38% of the patients with initial severely abnormal aEEGs.
Thirty-seven MRI scans were obtained in these 12 patients. Among the four patients with initial moderately abnormal aEEGs, none developed significant MRI evidence of brain injury by early or late brain MRI scans (table 2). Among the eight patients with initial severely abnormal aEEGs, four did not develop any clear MRI evidence of HIE by early or late brain MRI scans and four developed MRI evidence of brain injury, already visible on early MRI scans (table 3). Three of the four patients who developed brain injury displayed severe injury of the basal ganglia (nos. 9, 10 and 11). One patient (no. 12) developed more extensive injury, with lesions of the basal ganglia, white matter, cortex, pons and cerebellum (figure 1).
Due to clinical instability, not all newborn infants were able to receive the full sequence of four scans. MRIs on DOL 1 were performed in nine of 12 patients, at a mean of 12.2 h of life (range: 6.0–24.0 h) and at a mean of 8.0 h after initiation of induced hypothermia (range: 1.4–20.1 h). In the four patients developing MRI evidence of brain injury, these first MRIs were initially read as negative; subtle abnormalities were only retrospectively identified.
MRIs on DOL 2–3 were performed in all 12 patients, at a mean of 39.2 h of life (range: 26.0–54.0 h) and at a mean of 34.6 h after initiation of induced hypothermia (range: 21.6–47.4 h). Brain lesions were unequivocal in four patients (nos. 9, 10, 11 and 12), visible on all MRI modalities, especially DWI and spectroscopy. The remaining patients did not demonstrate any MRI evidence of brain injury on their MRIs obtained on DOL 2–3.
MRIs on DOL 8–13 were performed on nine of the 12 patients, and MRIs at 1 month of age were performed on seven of the 12 patients. These late brain imaging studies did not reveal any new MRI evidence of brain injury that had not been seen on early MRI scans. Late imaging confirmed the extent of brain lesions identified early in patient no. 9 (figure 2), despite having completed the full 72 h induced hypothermia. An autopsy was performed in patient no. 11, and also confirmed the extent of the brain injury identified early on DOL 2.
Two patients displayed unexpected findings not related to HIE on early MRIs, leading to early termination of induced hypothermia. On his MRI on DOL 2, patient no. 1 demonstrated a left parietal epidural haematoma, bilateral parietal and left frontal subpial haemorrhages, cerebellar haemorrhages and a large subgaleal haematoma (figure 3). On his MRI on DOL 1, patient no. 8 demonstrated a non-occlusive dural venous sinus thrombosis involving the bilateral transverse and superior sagittal sinuses (figure 4). Furthermore, both patients demonstrated abnormal haematological and coagulation studies while receiving induced hypothermia, requiring red blood cells, fresh frozen plasma and platelets transfusions. Because of the haemorrhages and thromboses seen on MRI in the setting of the clinical coagulopathy, induced hypothermia was terminated early in these two patients, after 38.2 and 24.7 h of treatment for patients nos. 1 and 8, respectively. Neither of these two patients developed significant MRI evidence of brain hypoxic–ischaemic injury, despite receiving only a partial course of induced hypothermia.
Of note, on his MRI scan on DOL 2, patient no. 1 also displayed a unilateral focus of restricted diffusion involving the left thalamus, with a concomitant normal magnetic resonance spectroscopy study in that area (figure 5). This lesion was still visible on DOL 10, but completely resolved at 1 month of age (figure 5), and was no longer visible on an MRI scan at 9 months of age. The patient was ultimately diagnosed with a venous infarct because of the unilaterality and the resolution of the lesion.
All patients without MRI evidence of HIE had a low neuromotor score on DOL 10 or at discharge from NICU. In these patients, the score remained unchanged or slightly improved (in patient no. 2) on follow-up. In contrast, patients with evidence of brain injury on MRI (patient nos. 9 and 12) had a high score on DOL 10 and the score remained elevated in patient no. 9 on follow-up. Patient no. 12 did not have any further follow-up as he died on DOL 13. Patients nos. 10 and 11 who died on DOL 4 prior to formal neuromotor scoring would likely have had elevated scores given their documented abnormal neurological examinations including severe encephalopathy.
Minimal literature exists regarding brain imaging in newborns treated with induced hypothermia. It mainly addresses the incidence of brain tissue injury following treatment with induced hypothermia and the predictive value of MRIs performed after completion of hypothermia for subsequent neurological impairment.15,–,17 Therapeutic hypothermia decreases brain tissue injury in asphyxiated newborns,15,–,17 as demonstrated by less cortical gray matter lesions,15–16 and less basal ganglia and thalamic lesions on MRI,16 especially in infants with initial moderately abnormal aEEG. The predictive value of these MRI scans for subsequent neurological impairment does not seem to be affected by therapeutic hypothermia.17
However, it is currently not known when is the optimal timing of brain imaging for term asphyxiated newborns treated with induced hypothermia to accurately define their brain injuries as early as possible and predict their neurological function.9,–,11 In this population of patients, a brain MRI has typically been performed on DOL 4–7.15,–,17 This timing usually falls in the practical window after induced hypothermia is complete, and before transfer to another care centre. However, in the era before induced hypothermia was widely offered, this day 4–7 window was not considered the ideal time to obtain brain imaging in newborns with HIE; abnormalities of DWI may be less evident than in the first few days of life and abnormalities on anatomical imaging may still be difficult to visualise, especially in mild forms of neonatal HIE.13–14
In our feasibility study, we found that asphyxiated infants treated with induced hypothermia develop some of the same sequential MRI changes described in those not treated.12,–,14 In cases of severe HIE, diffusion-weighted changes are subtle in the first 24 h of life but become more apparent and better defined on DOL 2–3. Proton MRI spectroscopy remains a sensitive technique for the identification of brain injury, showing an elevation of lactate in the injured areas in the first 24 h. We also found that all brain injuries are already visible on early MRI scans. The late brain imaging studies do not reveal any new brain injuries that were not seen on DOL 2–3, but also do not show that the brain injuries, if present, were underestimated on these early scans. Induced hypothermia does not appear to mask the appearance of brain injury lesions on MRI. Our findings suggest the value of adding early brain MRIs to hypothermia protocols, in order to optimise the prompt understanding of potential brain injury. MRI on DOL 2–3 may be effective in accurately defining brain injuries as early as possible in asphyxiated newborns treated with induced hypothermia. This may be helpful from a clinical perspective, especially in severe cases where some parents choose palliative care rather than pursuing intensive care treatment. From a research perspective, this may also in the future enable early identification of newborns who might benefit from adjunctive neuroprotective therapies.
However, the numbers of studied patients in the present feasibility study is too small to give enough power to these results or recommend routine early brain MRI in this population of newborns. Additional studies with larger numbers of patients will be useful to confirm these results. In clinical settings where only one brain imaging can be obtained, it is certainly reasonable and practical to delay the imaging to the second week after delivery, when the lesions are clearly visible on conventional imaging as described in the era before induced hypothermia.13–14
Questions remain regarding whether changes seen in the first days of life in asphyxiated infants treated with induced hypothermia are reversible changes in evolution.9,–,11 Our study does not support this idea. Severe hypoxic–ischaemic injuries did not resolve over time in our four patients with brain injury detected on early MRIs. Late MRIs in patient no. 9 confirm the pattern and extent of injury identified on early MRI scans, despite completion of a 72 h course of induced hypothermia. Autopsy was performed in patient no. 11, and also confirmed the extent of the brain injury identified early on DOL 2. Although he did not survive long enough to receive late MRIs, patient no. 12 with extensive brain injury on early MRI scans never showed any signs of improvement in his neurological clinical examination or his EEG, even after finishing 72 h of induced hypothermia. The only brain lesion that resolved in these 12 patients was the unilateral focus of restricted diffusion involving the left thalamus observed on early MRI scans in patient no. 1, ultimately diagnosed to be a venous infarct. However, again considering the small sample size, additional studies with larger numbers of patients will be needed to confirm this finding.
We also found early MRI to be useful in this population to demonstrate unexpected findings which could potentially be exacerbated by induced hypothermia. We discovered such findings in two patients during induced hypothermia: one (patient no. 1) with intracranial haemorrhage and the other (patient no. 8) with a non-occlusive cerebral venous sinus thrombosis. Both of these injuries most likely occurred perinatally,23,–,25 given the accompanying coagulopathies. We do not have clear evidence that these lesions were caused by the induced hypothermia.26 Yet since induced hypothermia has documented side effects including coagulopathy, the overall risk:benefit analysis argued in favour of early termination of the hypothermia.6 27–28 These patients did not demonstrate early or late MRI evidence of HIE despite their incomplete treatment.
Areas of mild T2 prolongation were noted in the white matter of some of the newborns enrolled in the study. Some authors have questioned the pathological nature of these signal abnormalities observed on T2-weighted imaging in other populations of newborns.29,–,32 In our study, when these T2 signal abnormalities were noted in the white matter, they remained present on all subsequent MRI scans (DOL 1, 2–3, 8–13 and at 1 month of age) without showing any evolution. They were never associated with abnormalities of T1-weighted imaging, DWI or spectroscopy. Thus, we do not consider them to be MRI evidence of perinatal hypoxic–ischaemic injury. It is also possible that these signal abnormalities may be signal artifact due to field inhomogeneity in term newborns scanned with the 3T MRI scanner. Further studies are needed to correlate these T2 signal abnormalities with the mild abnormalities of tone detected in a few of these patients and long-term neurodevelopmental outcomes. If they represent an antenatal brain injury, they may identify a patient population vulnerable to additional perinatal injury.
In this study, the main outcome variable used to confirm brain injury in these newborns was late MRI evidence of brain hypoxic–ischaemic injury. However, we recognise that neurodevelopmental outcome for survivors and autopsy confirmation of injury for infants who died would have been the gold standard. Autopsy results were available only in one of the three patients who died, but confirmed the early MRI results in this patient. We used the neuromotor outcome score22 to permit objective clinical comparisons of the neurological status and to correlate with the sequential MRI findings. The degree to which this score can predict future neurodevelopmental outcome needs to be validated with a larger number of asphyxiated infants treated with hypothermia.
In conclusion, despite the current practice for induced hypothermia during which brain MRIs are obtained on DOL 4–7, MRIs obtained on DOL 2–3 appear to be a powerful indicator of long-term brain injuries in asphyxiated newborns, especially in severe cases. The brain injuries identified on early MRI scans of asphyxiated newborns treated with induced hypothermia do not appear to represent reversible changes, and seem to follow the same MRI evolution pattern described in these patients before the cooling era. As information accumulates regarding the timing and interpretation of MRIs in newborns who have been treated with induced hypothermia, the degree to which MRI findings within the first month of life correlate with and are predictive of neurodevelopmental outcome will be better understood. Patients with moderately abnormal aEEG pattern may have especially subtle MRI changes that require more detailed scans.
The authors thank the families and their newborns for participating in this study. Special thanks are also due to the NICU nurses and MRI technicians who have made this study possible.
Funding Pia Wintermark receives research grant funding from the Thrasher Research Fund Early Career Award Program and the William Randolph Hearst Fund Award. The work of Simon K. Warfield is supported by NIH grants R01 RR021885, R01 GM074068, R03 EB008680 and P30 HD018655.
Competing interests None.
Ethics approval This study was conducted with the approval of the Institutional Review Board of the Children's Hospital Boston, Boston, USA.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.