Childhood Cerebral Adrenoleukodystrophy: MR Perfusion Measurements and Their Use in Predicting Clinical Outcome after Hematopoietic Stem Cell Transplantation

Discussion

Because childhood cALD is typically a progressive and severe disorder in the absence of HSCT, the primary goals of this study were the following: 1) to determine whether MRP parameters differ from those in controls in various regions of the cALD-affected brain, 2) to describe a variety of DSC-MRP parameters both pre- and post-HSCT, and 3) to determine whether any baseline (pre-HSCT) MRP markers portend the clinical outcome following HSCT.^{6⇓⇓⇓–10} Regarding the former goal, both this study and that of Musolino et al¹² seem to be in accord, because this study confirms that in untreated cALD, rCBV within the “core” of nonenhancing POWM but not the LEE is significantly lower than that in controls. In our transplanted cohort, this difference in rCBV from controls persisted at 1 year following HSCT. Such findings indicate that the core of abnormal POWM (termed “Zone A” in the prior study by Musolino et al¹²) likely comprises severely injured and perhaps nonviable tissue. Accordingly, slowly progressing parieto-occipital regional atrophy typically develops in this region, which was also observed in the current study.¹² Meanwhile, the arrest of enhancement at the LEE, along with the preserved rCBV and rCBF at 1 year post-HSCT, would imply viable tissue therein. Notably, in our cohort, the POWM rCBV was, on average, about 50% of that of the NAFWM, whereas the rCBV was about 20% of that of the NAFWM in the prior study.¹² This difference in rCBV could relate to the higher mean Loes score (13.4 versus 6.4 in this study) in the prior study, perhaps reflecting more severe tissue injury overall.¹²

The mechanism of how HSCT attenuates the neuroinflammation in cALD is unknown; however, the normal rCBV within the LEE that persists after HSCT suggests a correction of the blood-brain barrier abnormality, along with preservation of regional parenchyma.^{3,5,10⇓–12} Similarly, the study by Musolino et al¹² found that rCBV was preserved in the LEE (termed “Zone B”) but was lower in adjacent, centrifugally located areas (Zone C); hence, the current study corroborates the Zone B findings of that study. Such an arrest in demyelination may have a histopathologic precedent, in which murine studies have shown that certain microglial cells are absent initially at the demyelinating edge, but there is a subsequent slow return of microglial guard cells by about 1 year post-HSCT as progression of the disease halts.^{22⇓⇓⇓–26} Thus, 1 theory, which remains speculative, is that blood-brain barrier dysfunction is actually helpful or is required for the marrow-derived precursor cells to return into the cerebrum and differentiate into microglia because a LEE is nearly always present before HSCT in those patients who have undergone a successful HSCT and ultimately have stabilized cerebral disease.^{22⇓⇓⇓–26} Thus, this theory suggests that microglia reflect an overall return of hematopoietic precursor stem cells to the cerebrum, resulting in the observed disease arrest, along with the subsequent repair.^{22⇓⇓⇓–26} Zonal measurements within and adjacent to the LEE were not the focus of the current study but could be considered in future ones.

Regarding the SCC, this aforementioned theory may, at least in part, explain the lack of a significant difference in the large majority of the various MRP measures at 30–60 days post-HSCT, even though there were significant differences at baseline and at 1 year post-HSCT (especially in rCBV and rCBF).^{22⇓⇓⇓–26} Thus, we surmise that there could be both components of irreversible injury and healing of the blood-brain barrier occurring macroscopically in the SCC, which may be an “inflection point” in the disease process. For example, according to the above-mentioned theory, HSCT can correct the underlying microglial dysfunction by way of bone marrow resident and progenitor cells entering the cerebrum through a patent and abnormal blood-brain barrier via capillary recruitment and resultant increased microperfusion; we opine that this might be represented by a transient relatively decreased rTTP and relatively increased rCBV and CBF.^{22⇓⇓⇓–26} In the current study, such transient changes (ie, increased rCBF and rCBV) were indeed noted in the SCC, where the rTTP transiently decreased at 30–60 days post-HSCT and then rose again to above baseline at 1 year post-HSCT. However, by 1 year post-HSCT, both SCC rCBF and rCBV decreased to below baseline, suggesting that such increased perfusion at 30–60 days post-HSCT was indeed only a temporary phenomenon.

Hence, the measurements at 30–60 days post-HSCT might reflect an attempt for the cerebrum to undergo healing by temporarily increasing perfusion to the SCC by a vascular response, though there is also likely a component of underlying irreversible injury, as evidenced by the lower 1-year rCBV and rCBF values. If so, then this combination of findings might explain the SCC being the 1 site where the baseline DSC-MRP measurements predict the neurocognitive outcomes (hence, the term “inflection point”) because it is a region that has both viable tissue that can potentially be saved, along with nonviable tissue. In contrast, the POWM appears to be a region where the tissue injury is largely irreversible, as evidenced by the lack of such transient findings at 30–60 days post-HSCT. Accordingly, within the POWM, rTTP remained higher and rCBV and rCBF both progressively decreased after HSCT. Thus, the POWM seems mostly unaffected by successful transplantation.

This study also found that the baseline SCC rCBV and rCBF values strongly correlated with various measures of neurologic function at 1 year after successful HSCT, perhaps due to the reasons described above. While the Loes MR imaging score is known to correlate with the gross neurologic outcome after HSCT, the findings in the current study suggest that regional DSC-MRP values could augment the Loes score in predicting outcomes better.^{7⇓⇓⇓–11} However, this study was limited, in that most boys within the cohort had generally favorable baseline Loes scores. Hence, these findings may not be applicable to those patients with more severe initial disease. Thus, the utility of MRP in predicting outcomes in patients with cALD with more severe disease needs to be further studied.

There has been an ongoing search for the best baseline (pre-HSCT) imaging marker of clinical outcome following transplantation (HSCT). Regarding such biomarkers, to our knowledge, the Loes score is the most reproducible and strongest predictor of outcome, being proved across multiple centers.^{6⇓⇓⇓–10} While the Loes score is a “continuous” marker of radiographic severity between 0.5 (minimum nonzero score) and 34 (maximum score), its predictive value mostly rests on a binary correlation with quite broad outcomes. Specifically, boys with a score of <10 generally do well after HSCT, while those with a Loes ≥10 generally do poorly after transplantation. Thus, there is a need for imaging biomarkers within both subgroups of less affected (ie, Loes of <10) and more severely affected (ie, Loes of ≥10) patients at baseline. Thus, to our knowledge, there has been no baseline imaging biomarker within the less affected subgroup shown to correlate with eventual posttransplantation neurocognition; the current data suggest that DSC-MRP could be of some utility in this subgroup. Meanwhile, within a subgroup of patients with higher Loes scores (≥10), a recent study found that the intensity of visible enhancement on postcontrast T1WI may lend predictive value to clinical outcome, and this was found to be relatively simple to measure, as well as reproducible between observers.²⁷

Limitations of this study include the relatively small size of the patient cohort with cALD. Thus, the expense of travel, care, and MR imaging for multiple trips is a known factor contributing to this limitation.¹⁹ Similarly, it is difficult to obtain a large number of age-matched controls for MRP because most pediatric indications for MR imaging do not necessitate contrast. A third potential limitation was the use of relative rather than absolute perfusion measurements; however, relative measurements of CBV or CBF are typically more reproducible than absolute ones.^15,16 Additionally, a limitation could be the lack of interobserver reliability calculations between readers regarding the 3+ ROI placements (6+ if the measurement was bilateral) and subsequent MRP values obtained. However, it was not feasible to have 2 readers, given the expertise needed in interpretation of cALD along with familiarity with the postprocessing software for coregistration and MRP calculation; such postprocessing steps including the MRP measurements and calculations were too exhaustive and time-consuming for multiple readers. Particularly, the numerous steps in this process could inherently cause variability among observers.

Finally, a limitation may be that the patient cohort of this study demonstrated relatively low pretransplantation Loes scores (eg, compared with a study by Musolino et al¹²), while there was no significant difference in scores between the excluded (n = 37) and included cohort (n = 10); this could raise the question of an underlying selection bias in our patient pool.^{6⇓⇓⇓⇓⇓–12} Thus, factors that decreased the mean Loes score in the excluded group were that 6 patients had no cerebral disease (Loes score of 0), while 4 were stable (having very low Loes scores near zero); however, this apparent lowering in mean Loes score at baseline may, to some degree, be counterbalanced because some of the excluded patients had disease too severe to be transplanted (n = 8) or they did not travel back after transplantation for each MR imaging due to severe dysfunction (n = 13) (Fig 1).

Meanwhile, among the 10 included patients, the mean Loes score was relatively low because those with very high scores are very unlikely to undergo transplantation, for example, compared with the prior study by Musolino et al,¹² which focused on patients with higher mean Loes scores. Hence, there may be an inherent selection bias in this study because those with much higher Loes scores are not likely to undergo HSCT and thus would be excluded, while those patients who did undergo HSCT but were too sick or dysfunctional to travel and return for an adequate follow-up MR imaging were also likely to be excluded. Therefore, as mentioned previously, the findings of this study regarding the clinical response to HSCT cannot be generalized to the subset of patients with cALD who present with much higher Loes MR imaging severity scores on initial evaluation for HSCT. The prediction of outcomes, based on MRI, may be better addressed by gadolinium-intensity measurements, which has been a focus of a recent study by Miller et al.²⁷