Neurofilament light chain as a prognostic marker in cerebral adrenoleukodystrophy

Abstract

Cerebral adrenoleukodystrophy (C-ALD) is a rapidly progressing inflammatory neurodegenerative disease with unpredictable onset in males with ALD. It must be treated at an early stage of demyelination, determined by MRI, to preserve neurocognitive function. Validation of biochemical markers that can aid in the prediction of functional outcomes is needed for optimizing disease management and therapeutic development. Our objective was to determine whether plasma neurofilament light chain (NfL), a biomarker of neuroaxonal injury, corresponds to functional outcomes after standard of care treatment with haematopoietic stem cell transplantation (HSCT).

This retrospective observational cohort study of 27 patients with C-ALD treated with HSCT at a mean age of 8.0 years (standard deviation = 2.6) examined pre-treatment biomarker levels and neurocognitive outcomes 1 year after treatment. Plasma biospecimens were collected at a median of 10 days prior to HSCT, and NfL concentrations were measured with Single-Molecule Array (SiMoA) assay. White matter lesions were characterized by Loes MRI severity scores. Following treatment, neurocognitive outcomes across six domains were measured at a median of 1 year post-HSCT using the Wechsler intelligence quotient (IQ) scales, the Beery-Buktenica Test of Visual-Motor Integration and the Purdue Pegboard test.

Pre-HSCT NfL values in C-ALD patients ranged from 5.0 to 911.0 pg/ml, with a median of 35.8 pg/ml. Higher NfL levels were associated with lower post-HSCT scores across all neurocognitive domains (P < 0.05). Large effect sizes were seen for visual reasoning, processing speed, fine motor dexterity and visual-motor integration (Pearson correlations: r = −0.74 to −0.84). Partial correlations showed a moderate association between NfL and neurocognitive outcomes, even when adjusting for MRI severity scores.

NfL has a strong correlation with neurocognitive status following treatment for C-ALD. These encouraging findings suggest an opportunity to improve our capacity to forecast outcomes and therefore improve counselling for families considering cellular therapies.

Pierpont et al. examined plasma neurofilament light chain (NfL), a marker of neuroaxonal injury, in boys with cerebral adrenoleukodystrophy. Higher NfL levels prior to haematopoietic stem cell transplantation were strongly associated with poorer cognitive outcomes one year later.

Introduction

The integration of well-validated biochemical markers into clinical practice holds tremendous promise to enhance diagnosis, management and therapeutic development for a range of progressive neurological diseases.^1–3 In paediatric neurodegenerative diseases requiring life-saving therapies, such biomarkers could also be of considerable value for prognostic forecasting: what will a child’s abilities, disabilities and quality of life look like after treatment? Adrenoleukodystrophy (ALD) is an X-linked inherited disorder that exemplifies the potential for biomarkers to assist in answering these prognostic questions.⁴ About 30%–40% of males with ALD develop the severe childhood cerebral form (C-ALD), characterized by brain lesions involving inflammation, myelin destruction and neuroaxonal injury.⁵ Once neurodegeneration begins, it can progress rapidly, with neurocognitive decline and death within a few years if untreated.^6,7 Treatment with allogeneic haematopoietic stem cell transplantation (HSCT) or gene therapy can be effective in halting advancement of brain lesions.^8,9 However, neurocognitive impairments that develop prior to treatment tend to be irreversible, and even when lesion growth is arrested, losses of function can continue in more advanced cases.^10,11 The timing of treatment relative to disease onset is therefore vital to its success.

The introduction of newborn screening (NBS) for ALD in most of the USA and several other countries has constituted a significant advancement in clinical management of this disease by enabling earlier detection and expeditious treatment of ALD-associated complications (i.e. adrenal insufficiency, C-ALD) among those identified at birth.¹² Whether or when C-ALD will emerge remains largely unpredictable, although recent reports suggest that levels of C26:0-lysophosphatidylcholine identified through NBS may assist with stratifying those at highest risk of developing early complications.¹³ Due to this variable course, clinical surveillance guidelines recommend serial anatomical brain MRIs every 6 months during the peak period of risk of lesion development (i.e. 3–12 years of age).¹⁴ However, the combination of NBS and MRI surveillance is not a complete solution. At the current time NBS is not universally available, especially outside of the USA, and follow-up programmes impose a considerable logistical and psychological burden that can complicate adherence.^15,16 Furthermore, early brain lesions can be subtle and challenging to identify on anatomical MRI, particularly outside of centres with extensive experience in ALD.¹⁷ As a result, patients with C-ALD continue to present for treatment at various stages of disease progression. Currently, the most widely used metric to determine prognosis and eligibility for therapy is the Loes MRI severity score, a scoring system that requires specialized training of reviewers and is subject to inter-reviewer variability.^18,19 Identification of additional reliable, objective metrics that can assist with predicting functional outcomes after treatment therefore offers potential to contribute to optimizing disease management.

Neurofilament proteins, particularly neurofilament light chain (NfL), have emerged as promising molecular biomarkers for neurological disease due to their established sensitivity to neuroaxonal injury. Neurofilament proteins are structural components of neurons that are released into the cerebrospinal fluid and blood when damage occurs to axons (Fig. 1). Accurate measurement of neurofilaments in the plasma across a range of concentrations has been supported by the development of Single-Molecule Array (SiMoA) technology.²⁰ Elevated plasma NfL has been linked to disease severity and progression in a spectrum of neurological conditions, including traumatic brain injury, stroke and Alzheimer’s disease.²¹ Expanding on earlier studies of adult or geriatric cohorts, recent work has demonstrated the utility of NfL for assessing disease progression in children with neurodegenerative conditions, including paediatric multiple sclerosis and leukodystrophies.^22,23 Studies of ALD have reported that NfL levels are markedly higher in C-ALD patients relative to healthy children, and are reliably associated with disease-specific MRI scores of lesion severity.^24-26 Furthermore, it has been established that elevated NfL levels decrease significantly following successful HSCT.²⁴ This growing evidence highlights the potential clinical utility of NfL in monitoring disease status and therapeutic response, including functional outcomes such as neurocognitive skills. The extent to which NfL has value for prediction of clinical outcomes after treatment for C-ALD has yet to be determined. This study investigated associations between pre-transplant NfL and neurocognitive test scores 1 year following HSCT. We hypothesized that lower NfL concentrations at the time of treatment would be associated with better neurocognitive outcomes.

Figure 1.

Neurofilament light chain analysis. Onset of cerebral adrenoleukodystrophy most often originates in the splenium of the corpus callosum. As axons are damaged, neurofilament light chain (NfL) is released into the cerebrospinal fluid and plasma. Single-Molecule Array (SiMoA) technology can reliably detect small amounts of NfL in plasma. Figure created in BioRender. Pierpont, R. (2025) https://BioRender.com/jrtqk9v

Materials and methods

Study population

We retrospectively analysed brain MRIs, plasma biospecimens and neurocognitive testing of boys (<18 years) with C-ALD who were treated with allogeneic HSCT at the University of Minnesota from March 2013 to July 2023, reflecting a period during which a consistent busulfan- and fludarabine-based HSCT conditioning protocol was used. Patients who died within the first year post-HSCT, those who underwent multiple HSCTs or those who lacked available pre-HSCT plasma samples were excluded from analysis. Characteristics of both included and excluded patients are provided in Supplementary Table 1. Kruskal–Wallis rank sum testing indicated there were no differences between the included and excluded patients for age at HSCT, nor any of the six baseline neurocognitive measures (P > 0.1 for all). The excluded patients had greater lesion severity on MRI at baseline (P = 0.034).

Standard protocol approvals, registrations and patient consents

All patients were consented on research protocols approved by the University of Minnesota’s Institutional Review Board for biospecimen collection and analysis of HSCT outcomes, including MRI studies and neurocognitive testing.

MRI severity

Disease severity in pre-HSCT MRIs was quantified with the Loes score, a 34-point scale that quantifies ALD brain lesions based on the number and extent of brain structures involved.¹⁸ MRIs were independently scored by two board-certified neuroradiologists. When scores differed across the two raters, a consensus score was determined by obtaining a score from a third neuroradiologist and utilizing the median score. Demyelination patterns (i.e. the regional distribution of disease on MRI) were also identified.²⁷

NfL assay

Plasma samples were obtained at a median of 10 days (interquartile range, IQR: 8–18 days) prior to HSCT. NfL protein levels were measured by SiMoA immunosorbent assay in batched format (Quanterix).

Haematopoietic stem cell transplantation

All boys received myeloablative conditioning prior to HSCT. Graft source was umbilical cord blood for 15 patients (56%) and bone marrow for 12 (44%). Six patients (22%) had a matched sibling donor graft, whereas 21 patients (78%) received grafts from matched unrelated donors. Established institutional protocols were followed for supportive care, including prophylaxis for graft-versus-host disease, antimicrobial therapy and blood product transfusion support. No patient had an ongoing inflammatory neurological complication at 1-year post-HSCT.

Neurocognitive outcomes

Neurocognitive testing was completed during clinical neuropsychological evaluations at 1-year post-HSCT. For two cases in which an evaluation was not completed at the 1-year time point, data from a 2-year post-HSCT evaluation were used. Age-based standard scores [mean = 100; standard deviation (SD) = 15] quantified performance across six domains; previously described procedures were used to harmonize across different versions of the Wechlser IQ scales.²⁸ Four domains were obtained from the Wechsler scales, including verbal comprehension, visual reasoning, working memory and processing speed. The Beery-Buktenica Test of Visual-Motor Integration measured visual-motor ability, and the Purdue Pegboard measured fine motor dexterity. Test scores less than −2SD below the population mean (i.e. <70) were classified as indicating clinical impairment in a particular domain; scores below −1SD (i.e. 70–84) were classified as below average. If clinical documentation indicated that a patient was unable to attempt or complete a test due to severe visual, auditory or speech impairment related to C-ALD progression, the lowest possible standard score for that test was assigned. For descriptive purposes, outcomes for each patient were categorized across three levels based on the overall impact across the set of neurocognitive measures: (i) no neurocognitive impairment: no domains with clinical impairment, and ≤1 domain with a below average score; (ii) mild to moderate neurocognitive impairment: 1–2 domains with clinical impairment, and/or 2–3 domains with below average scores; and (iii) severe neurocognitive impairment: >2 domains with clinical impairment, and/or >3 domains with below average scores. In cases where a domain score was missing (n = 4), scores in that domain from the nearest evaluations were reviewed to determine the classification.

Statistical analysis

Descriptive statistics were obtained to understand relationships between NfL, Loes scores, demyelination patterns and neurocognitive outcomes classification. Pearson correlation coefficients were computed to evaluate associations between pre-HSCT NfL levels and each of the post-HSCT neurocognitive outcomes. As a comparison, associations between pre-HSCT Loes scores and the neurocognitive outcomes were also obtained. Finally, to examine the association between NfL and neurocognitive outcomes while adjusting for the influence of the Loes score, partial correlation coefficients were calculated using analysis of covariance. The 95% confidence intervals were calculated using the percentiles of the correlation coefficient distributions across 1000 bootstrap re-samples of the actual study sample. NfL was log-transformed for all analyses to provide better model fit due to its skewed distribution. Analysis was conducted using R software v.4.4.1 (R Core Team, Vienna, Austria).

Results

Data were analysed from 27 boys with C-ALD who underwent HSCT at a mean age of 8.0 years (SD = 2.6; range, 4.4–13.3 years). Table 1 displays pre-HSCT Loes and NfL scores relative to C-ALD demyelination patterns. Across the full cohort, NfL values ranged from 5.0 to 911 pg/ml (median = 35.8 pg/ml). Most patients with parietal-occipital, frontal or mixed demyelination patterns involving the corpus callosum (n = 23) had NfL values >25 pg/ml, except for three patients with very low MRI severity (Loes score 1–1.5) who all had pre-HSCT NfL levels <10 pg/ml. Among the four patients with demyelination patterns not involving the corpus callosum (atypical patterns involving the projection fibres or temporal lobes/auditory pathways), NfL values were relatively low (<10 pg/ml) regardless of their Loes score.

Table 1.

Pre-transplant MRI severity scores and neurofilament light chain values according to demyelination pattern in 27 boys with C-ALD

Demyelination pattern	N	Loes score, median (range)	NfL, pg/ml, median (range)
Pattern 1: Parietal-occipital white matter	21	2 (1–13)	37.4 (5.9–911.0)
Pattern 2: Frontal white matter	1	4	35.9
Pattern 3: Projection fibres	3	3 (1–4)	9.2 (7.4–9.2)
Pattern 5: Mixed parietal-occipital and frontal white matter	1	17	192.0
Other: Mixed temporal white matter and optic pathway	1	4	5.0

Traditionally, five main demyelination patterns are described in C-ALD.²⁷ No patients in this cohort presented with Pattern 4 (primary cerebellar involvement). Additionally, with early and more complete imaging, additional rare patterns have surfaced, including an unusual temporal pattern observed in this study.

At 1-year post-HSCT, 12 patients in this cohort had favourable outcomes with no neurocognitive impairment; these patients had pre-HSCT NfL concentrations ranging from 5.9 to 37.4 pg/ml, with a mean of 20.0 (SD = 12.8). Patients classified as having mild to moderate neurocognitive impairment (n = 8) had pre-HSCT NfL levels ranging from 5.0 to 263 pg/ml, with a mean of 99.8 (SD = 92.6) . Those with severe neurocognitive impairment (n = 7) had pre-HSCT NfL levels ranging from 282 to 911 pg/ml, with a mean of 464.2 (SD = 227.6). There was no overlap in pre-HSCT NfL levels in the severe neurocognitive impairment group with the other two groups (Fig. 2).

Figure 2.

Relationship of baseline plasma neurofilament light chain (NfL) levels in boys with cerebral adrenoleukodystrophy to their neurocognitive status at 1 year following haematopoietic stem cell transplantation (HSCT). The boxes delineate the interquartile range for each group with a line at the median and whiskers indicating the range. To enable better visualization of individual data points, NfL values were plotted along a log-transformed scale.

Pearson correlation coefficients were calculated to assess continuous relationships between pre-HSCT variables and 1-year post-HSCT neurocognitive outcomes (Fig. 3). Higher pre-HSCT Loes scores and NfL concentrations were associated with lower post-HSCT scores across all neurocognitive domains (P < 0.05 for all). Large effect sizes were seen for the relationship between pre-HSCT NfL and visual reasoning, processing speed, fine motor dexterity and visual-motor integration (Pearson correlations: R = −0.74 to −0.84; Table 2). Both Loes scores and NfL values demonstrated the strongest associations with the processing speed measure. To assess the association between NfL and neurocognitive outcomes while accounting for the portion of variance explained by the Loes score, partial correlations were examined. Results indicated that NfL had a moderate association with most of the outcomes even after adjusting for the Loes score (Table 2).

Figure 3.

Scatter plots display associations between plasma neurofilament light chain (NfL) concentrations prior to haematopoietic stem cell transplantation (HSCT) and scores across six domains of neurocognitive function at 1-year post-HSCT.

Table 2.

Pearson correlation coefficients (R) between pre-treatment variables and neurocognitive outcomes 1 year after haematopoietic stem cell transplantation (HSCT)

	Neurocognitive outcomes (1-year post-HSCT)
Predictor(s)	Verbal reasoning	Visual reasoning	Working memory	Processing speed	Fine motor dexterity	Visual-motor integration
R Loes score (95% CI)	−0.55 (−0.77, −0.26)	−0.63 (−0.86, −0.33)	−0.34 (−0.73, 0.18)	−0.72 (−0.87, −0.49)	−0.68 (−0.95, −0.29)	−0.68 (−0.93, −0.35)
R NfL (95% CI)	−0.47 (−0.77, 0.02)	−0.74 (−0.88, −0.46)	−0.55 (−0.79, −0.17)	−0.84 (−0.94, −0.67)	−0.75 (−0.87, −0.57)	−0.75 (−0.85, −0.55)
Partial R NfL (95% CI)	−0.16 (−0.58, 0.36)	−0.57 (−0.80, −0.20)	−0.47 (−0.71, −0.09)	−0.68 (−0.87, −0.37)	−0.56 (−0.74, −0.39)	−0.53 (−0.70, −0.25)

R Loes score indicates the Pearson correlation between pre-HSCT Loes score and each neurocognitive outcome. R NfL indicates the correlation between pre-HSCT neurofilament light chain and each neurocognitive outcome. Partial R NfL indicates the correlation between pre-HSCT neurofilament light chain and each neurocognitive outcome variable, adjusted for the correlation between Loes score and each outcome variable. 95% confidence intervals are in parentheses, which were calculated using the percentiles of the distribution for R across 1000 bootstrap re-samples of the actual study sample.

Discussion

Accumulating evidence indicates that blood NfL levels reliably detect neuroaxonal injury across a range of neurologic conditions, sparking growing interest in incorporating these measurements into clinical practice.²⁹ This study presents evidence for the potential utility of NfL in predicting post-treatment functional outcomes in a challenging and unpredictable paediatric neurodegenerative disease. Specifically, we found that pre-treatment plasma NfL concentrations in C-ALD patients were robustly associated with neurocognitive outcomes 1 year following HSCT. Furthermore, computing partial correlations indicated that for most outcomes, pre-HSCT NfL explained variability in neurocognitive scores even when adjusting for pre-HSCT lesion severity as measured by Loes scores. Whereas Loes scores provide a metric to characterize the extent of brain lesions observable on structural MRI, NfL reflects degeneration at the level of neuronal microstructure. Together, they offered a more comprehensive prognostic assessment of C-ALD disease status than either predictor alone.

Since the 1990s, assessment of MRI severity using the Loes score has been a primary prognostic indicator used to inform treatment decision-making in C-ALD.^30,31 Loes scores at the time of treatment are associated with a range of clinical outcome metrics after HSCT, including survival, risk of neurological deficits, neurocognitive outcomes and independence in activities of daily living.^8,11,32 Because survival without major functional disabilities is considerably more likely for patients with Loes ≤9 than those with more extensive disease, a score cut-off of 9 has been used to determine eligibility for treatment with elivaldogene autotemcel gene therapy in published clinical trials and for approved indications.^11,33,34 However, use of cut-off scores does not reflect the reality that considerable variation in rate of progression and in clinical outcomes can be observed among patients with similar Loes scores.^35,36 Other brain MRI markers (e.g. regional cerebral blood volumes, diffusion tensor imaging metrics, intensity scores for active neuroinflammation and demyelination patterns) and metrics such as pre-treatment functional assessments (e.g. neurologic function scores, neurocognitive test scores) have also been associated with clinical outcomes and long-term functioning of patients after treatment.^28,36-40 The current study highlights the potential for a blood biomarker to contribute to prognostication when considering cellular therapies for C-ALD patients. With additional systematic validation using well-specified methodologies, NfL assays could be used as a tool to improve risk stratification and long-term supportive planning, especially for patients whose disease has progressed to more advanced stages. For example, clinicians and families may feel more informed when proceeding with treatment if NfL values are known, allowing for discussion about estimated probability of preserving neurocognitive function versus needing specialized planning.

These results not only support previous findings that NfL is linked to the size and extent of lesions in C-ALD patients^24–26 but also provide evidence that seems to suggest both NfL levels and clinical outcomes could vary depending on the location of the disease. Our group has previously reported observation of differential clinical outcomes when comparing patients with lesions in frontal versus parietal-occipital white matter, with more marked psychiatric disturbance seen in patients with frontal disease.³⁶ The current analysis included only one patient with frontal C-ALD, limiting conclusions for this cerebral pattern. Although cautious interpretation is needed due to the small sample size, results showed that the plasma NfL concentrations were relatively low (<10 pg/ml) in patients who developed lesions only within the corticospinal projection fibres (n = 3) despite having active inflammatory lesions. Interestingly, these three patients all experienced excellent stabilization of neurocognitive function with HSCT; none had clinical impairment in any of the six domains at follow-up, and only one patient had a single below average score (i.e. fine motor functioning). We speculate that lower overall volume of the damage in the projection fibres relative to large white matter bundles like the corpus callosum could have resulted in these lower NfL concentrations. If replicated with further patients, these results add important clinical insight that preservation of neurocognitive function after HSCT may be particularly likely in patients with atypical disease patterns who receive timely treatment. Future research could explore these issues further by including measures of lesion volume and/or metrics from diffusion tensor imaging.

There are several essential points for consideration regarding clinical implementation. First, this initial single-centre study identified associations between NfL and clinical outcomes at the group level; to advance the use of NfL levels to clinical decision-making for individual patients, it will be important to analyse data from other C-ALD treatment centres to clarify the ranges of NfL levels associated with neurocognitive outcomes. In the current study, although NfL levels distinguished the group with severe neurocognitive impairment from the other groups, the ranges showed overlap between the group without neurocognitive impairment and the mild to moderate impairment group. While no biomarker would be expected to sort functional outcomes completely, there remain opportunities to improve the understanding of the relationship of NfL levels to these outcomes. For example, acquisition of data from a healthy reference group and determination of age-dependent cut-offs for pathological interval changes could clarify the extent to which NfL levels have deviated from normal. Such reference data is especially essential for paediatric neurodegenerative diseases, given that very rapid brain development may be occurring in tandem with a neurodegenerative process. In addition, evaluation of plasma NfL levels in asymptomatic family members with the same ABCD1 variant would offer an important opportunity for comparison. These exercises would also enhance the opportunity to utilize NfL in serial monitoring to augment MRI acquisition to aid in early detection of disease. Given the current paradigm of screening MRI every 6 months, NfL assays between MRIs could help detect initiation of cerebral disease more quickly. Another avenue of potential study is whether increasing NfL levels will correlate with a more difficult treatment course, and whether this trajectory improves precision in predicting the ultimate functional outcomes as opposed to using only baseline values. Regardless, development of reliable NfL sample acquisition and interinstitutional methods will be needed to minimize variations in calibration, technician handling or environmental factors that could introduce errors.

A strength of the current analysis was the use of robust, domain-specific measures of neurocognitive functioning to calculate associations, demonstrating progress over most earlier studies examining the relationship between NfL and cognition.⁴¹ Aligned with previous work in C-ALD, the strongest associations of both MRI severity and NfL with neurocognitive outcomes were found in domains measuring visual reasoning, processing speed, visual-motor function and fine motor dexterity.³² Regarding limitations, the current study was unable to explore whether serial changes in NfL either prior to or following HSCT were predictive of dynamic changes in cognition (e.g. speed or extent of clinical progression). Across neurological disorders, examination of these dynamic changes could assist with the design of future clinical trials by setting benchmarks for treatment response at various time points after therapeutic intervention.⁴² An additional limitation of the current study was that we were unable to compare NfL levels based on genotype and lesion location. The infrequency of common ABCD1 pathogenic variants across patients and the scarcity of certain cerebral demyelination patterns (e.g. frontal disease) precluded meaningful statistical comparisons. Finally, given the retrospective nature of the study, most patients in this cohort were diagnosed with ALD due to family history or development of symptoms rather than by NBS. Future prospective studies following patients identified by NBS would enable examination of associations between bloodspot C26:0-lysophosphatidylcholine levels and plasma NfL concentrations or HSCT outcomes.

In conclusion, this study provides strong evidence that a recognized biomarker of neuroaxonal injury has an application in C-ALD and encourages future investigation to determine whether plasma NfL levels can predict clinical outcomes in other peroxisomal leukodystrophies (e.g. Zellweger spectrum disorder, ACOX1 deficiency). NfL testing could be a promising element to add to the decision tree to inform the complex decision-making process for treatment of these devastating conditions. Nevertheless, it will be critical that a timely multicentre investigation establish age-based standards for healthy children and those with carefully delineated ALD phenotypes to facilitate its incorporation into the decision-making paradigm.

Data availability

The data that support the findings of this study are available to qualified investigators for non-commercial academic purposes by contacting the corresponding author. The data are not publicly available to protect the privacy of research participants.

Funding

This research was supported by the National Institutes of Health’s National Center for Advancing Translational Sciences (UM1TR004405) and the National Institute for Neurological Disorders and Stroke (K23NS123258). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Funding was also provided by the University of Minnesota Foundation.

Competing interests

The authors report no competing interests.

Supplementary material

Supplementary material is available at Brain online.