Imaging after therapeutic hypothermia: when to do the MRI (if at all)

I had thought it was well settled that imaging soon after rewarming was as good for predicting outcomes in HIE as later imaging, but it seems that the timing, and the methodology are still uncertain.

Tabacaru C, et al. Diffusion Tensor Imaging to Predict Neurodevelopmental Impairment in Infants After Hypoxic-Ischemic Injury. Am J Perinatol. 2023. I find the different sequences of MRI increasingly difficult to follow, as far as I can tell, DTI is a subtype of Diffusion Weighted Imaging that can be used to map the white matter tracts to derive a tractograph. My confusion is not helped by paragraphs such as this one from the methods:

The diffusion weighted MR images were preprocessed using software from the FMRIB Software Library (FSL) and MRtrix3. Preprocessing began with extraction of the b0 images from the diffusion weighted images using MRtrix3, followed by N4 bias correction of the b0 images. Post-bias correction, b0 images then underwent brain extraction via FSL’s BET to remove non-brain tissue, and to create a binary mask of the brain. Eddy current distortion correction and volume-to-volume motion correction was performed via FSL’s eddy current and motion correction using five iterations with the following parameters: outlier replacement, spline interpolation, least-squares reconstruction resampling, with full-width half maximums = [10,5,0,0,0] for each iteration. Post-eddy current and motion correction, FA values were estimated using FSL’s DTI FIT. Voxelwise statistical analysis of the FA data was carried out using Tract Based Spatial Statistics, part of FSL.22-23 These analyses were corrected for multiple comparisons using Threshold-Free Cluster Enhancement (FSL v1.2).

You might as well write “we did some stuff and got some pretty images of white matter”, it would be just as understandable for me! (and I think the vast majority of readers). However, one clear limitation of this method is that they were only able to get the pretty images in 20 of the 41 babies. The babies also had Prechtl standardised General Movements Assessment at 3 to 4 months of age, and 3 of those 20 were lacking the normal “fidgety movements”, all 3 of whom developed Cerebral Palsy (GMFCS 1,2 and 4) and 2/17 with such movements also had CP (GMFCS 1). There seemed to be a correlation between the infants who had absent fidgety movements and DTI images with abnormal Fractional Anisotropy in the Corpus Callosum, Posterior Limb of the Internal Capsule, and thalamic and optic radiations. The same changes were also found in the babies with CP who did have fidgety movements. In the results it is not at all clear of there was an association between DTI findings and long term outcomes, there is a brief statement that the same regions were different between those with and without CP, but no other mention of the results of the Bayleys, that all the babies had. The study might give an indication of which tracts are involved in the abnormal motor development that leads to absence of fidgety movements, and has research relevance, but certainly not any reason to add to clinical imaging.

Garvey AA, et al. Differences between early and late MRI in infants with neonatal encephalopathy following therapeutic hypothermia. Pediatr Res. 2023. This is a review from Boston of the images and outcomes of 94 babies who had both early, <7 days, and late, >7 days, MRI. The early images were obtained at a median of 4 days (IQR 4-4, or in other words more than half of them on day 4) and the late ones at a median of 16 days (IQR 12- 25 days). The second scan was obtained in order to satisfy their protocol which was to have a 2nd scan at 10 to 21 days, and it seems most were obtained from babies who were still inpatients, there were a concurrent 156 babies who were cooled and only had one scan (2 died before, 148 were discharged before, and 4 transferred).
Most of the normal scans (21/24) were normal at the later date, and most of the abnormal ones (59/70) were also abnormal later, with mostly very similar findings, although 4 were worse. Three of the normal scans were abnormal later, and 11 of the 70 abnormal scans were normal at the second scan. You can see a summary of the results in this table:

The authors of this study refer to the systematic review of Ouwehand S, et al. (Predictors of Outcomes in Hypoxic-Ischemic Encephalopathy following Hypothermia: A Meta-Analysis. Neonatology. 2020;117(4):411-27), which claimed that the early MRI was more predictive of outcomes (variable outcome depending on the study) than later images. But there are very few studies in that review that actually scanned the babies twice, most of the included studies used either early or late MRI, not both; and with the variability between cohorts, variability in outcomes analyzed, and variability in interpretation, I am not all sure that is a reliable conclusion. Standard MRI images may not show injury early, so some of that difference in predictive power may be because different sequences (diffusion weighted) will find abnormalities in early scans.

Parmentier CEJ, et al. Additional value of 3-month cranial MRI in infants with neonatal encephalopathy following perinatal asphyxia. J Pediatr. 2023:113402. Maybe we should just keep repeating the MRI? (facetious comment). This retrospective review of MRI repeated at 2 to 4 months, apparently they all had neonatal MRI abnormalities. The babies were also followed up to 18 to 24 months, with developmental testing and a neurological exam. (BSID III motor or cognitive score <85 considered abnormal). 41 of the infants were seen again at 5.5 years of age, and had further intellectual and neurological assessments. There was some evolution, as you might expect, between the neonatal and 3 months images, but as the authors themselves note, there seemed to be little additional prognostic value with the later scans. The scores derived from the 3 month MRI were not closely associated with 5.5 year outcomes, mostly not being significantly different between those with normal outcomes and those with abnormal outcomes, defined as a WPSSI IQ scores <85, and subscale scores of <85 or epilepsy, deafness, blindness, or CP.

Wu YW, et al. How well does neonatal neuroimaging correlate with neurodevelopmental outcomes in infants with hypoxic-ischemic encephalopathy? Pediatr Res. 2023. This is an analysis of data collected during the HEAL trial, of erythropoietin in addition to cooling in term HIE (compared to cooling alone). As you probably remember, the trial had a negative result, with no additional benefit of erythropoietin. As all the babies had MRI, and all had neurological and developmental follow up the trial gives a real world evaluation of the predictive capacity of MR imaging in a multicenter prospective trial. All the babies had T1 and T2 and Diffusion Weighted images, and MR spectroscopy of left thalamus. The main outcome of interest, I think, is what is referred to as “severe NDI”, which seems to have been BSID cognitive score <70 or a severe motor outcome, which is defined in a unique way, with a modified GMFCS, but using a matrix determined by the GMFCS and a diagnosis of CP, and its type. So an infant with a clinical diagnosis of spastic tetraparesis is considered a severe outcome even if the GMFCS is 1. An infant with hemi- or di-plegia is considered severe at a GMFCS of 3.

The scoring system for the MRI was one previously validated, which produced 3 grades of abnormality.

As this figure shows, there was no difference in the BSID scores between normal, mild or moderately abnormal 5 day MRI, only severely abnormal MRI was associated with much lower scores. The range of scores with normal or less severely abnormal MRI is worrying though, with some babies at each stage having very low scores, and some having normal or even high scores. Even the babies with severely abnormal MRI had a wide range, extending up to scores in the normal range. The MRS results show a similar pattern. The figure that I found most useful was the summary of the PPV, NPV, sensitivity and specificity of the various findings. Unfortunately they only produced this figure for death or “NDI”, which included BSID cognitive score <85, and mild motor dysfunction.

In this cohort who all had moderate or severe HIE, the PPV and NPV of almost all findings are poor. Clinical severity of the HIE is just as good as most findings on the MRI, apart from “Global Injury”. What I get from this is that the MRI score, and MRI severity of injury are quite poor at predicting medium term outcomes, but the pattern of injury, and specifically the presence of Global Injury (which is defined as damage affecting >75% of the brain, including Central Gray Matter, White Matter, and Cortex. Although not shown in this figure, in the text it states “with all 34 affected infants either dying or developing severe NDI”. The sensitivity for that finding is poor, as there were many infants with “NDI” who did not have global injury, and therefore the NPV is only around 50%, but the specificity and PPV are very high, and in this case, the outcomes are of major clinical importance.

I wonder about the clinical evolution in those babies; most of the babies I have seen who have the most extensive MRI injuries also have a very unfavourable clinical findings, with poor or absent feeding and seriously abnormal clinical exam.

I know I will come across as an “imaging nihilist”, but I really wonder about the added value of MRI imaging in these babies for our families. I know all the real experts will say that routine MRI, just after rewarming, with T1, T2, and Diffusion Weighted imaging is the “standard of care”.

The PPV and NPV of most findings, as shown in that last figure, are around 50%, so when you talk to parents after the MRI, all you can really say is with these findings, your baby has about a 50:50 chance of having some developmental delay, and some motor problems. Which is what you also have to say if the MRI is normal!

This fascinating article, from colleagues and friends at McGill, addresses some of those dilemmas (Cascio A, et al. Discussing brain magnetic resonance imaging results for neonates with hypoxic-ischemic encephalopathy treated with hypothermia: A challenge for clinicians and parents. eNeurologicalSci. 2022;29:100424). The study included 16 clinicians (physicians who sometimes have to discuss MRI results and NNPs) and parents of 28 infants. The article includes 4 real patient histories, with the actual outcomes of the babies. The histories and MRI findings were shown to the clinician respondents, but not the real follow up data. Clinicians were asked to predict the outcomes, how confident they were in their prediction, and how much the MRI helped them.

If you read the first vignette, then the clinicians predictions, you will see that almost all thought the child would have moderate or severe disability, and many were very or extremely confident in their predictions, although there was a wide spread in how much they thought the MRI helped them. In fact the child is doing OK, with an isolated language delay. In another vignette, the baby had a catastrophic start to life, and was extubated day 5 with ongoing severe encephalopathy and lack of a gag reflex, the MRI on day 10 showed relatively minor abnormalities, but the baby had a very poor long term outcome. Most clinicians felt the MRI didn’t help them very much (without knowing the long term result).

As for the parents responses, I would urge you to read the quotes, the only one that surprised me was one parent who found the doctors rude; knowing the gang at McGill fairly well, I can only assume that someone was having a bad day, and perhaps also that in this incredibly stressful environment, even a single misplaced word can have real impacts. The clinicians all felt they needed more training, both in how to interpret MRI scan reports and in how to communicate the results. As someone who trained before MRI was invented, (I remember the first magnetically isolated room being constructed in Edmonton) and who has gradually learned on the job without formal training, I can only echo those desires. The reports of the MRIs in the vignettes illustrate how they are somewhat opaque. I try to always discuss the MRI results alongside the neurologist, but I am much more uncertain about the predictive utility of the MRI than they tend to be.

Indeed, the results of the studies I have reviewed above confirm that we are right to be uncertain! I really question the added value of the MRI to families. The only finding which is reliable enough to make a definitive statement about prognosis is global injury, which mostly occurs in babies with severe clinical findings. All other grades, patterns, and extent of injury have very poor PPV and NPV. Before writing this I thought the NPV of a normal scan was good enough that we could be reassuring to parents, and that they could relax somewhat. But as the HEAL study shows, if you start off with moderate to severe HIE, even with a normal MRI about 20% of the babies will have moderate or severe problems with development or motor skills.

About Keith Barrington

I am a neonatologist and clinical researcher at Sainte Justine University Health Center in Montréal
This entry was posted in Neonatal Research and tagged , , , , . Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.