Neurofilament light chain and glial fibrillary acidic protein levels in metachromatic leukodystrophy.

Metachromatic leukodystrophy is a lethal metabolic leukodystrophy, with emerging treatments for early disease stages. Biomarkers to measure disease activity are required for clinical assessment and treatment follow-up. This retrospective study compared neurofilament light chain and glial fibrillary acidic protein (GFAP) levels in CSF (n = 11) and blood (n = 92) samples of 40 patients with metachromatic leukodystrophy (aged 0-42 years) with 38 neurologically healthy children (aged 0-17 years) and 38 healthy adults (aged 18-45 years), and analysed the associations between these levels with clinical phenotype and disease evolution in untreated and transplanted patients. Metachromatic leukodystrophy subtype was determined based on the (expected) age of symptom onset. Disease activity was assessed by measuring gross motor function deterioration and brain MRI. Longitudinal analyses with measurements up to 23 years after diagnosis were performed using linear mixed models. CSF and blood neurofilament light chain and GFAP levels in paediatric controls were negatively associated with age (all P < 0.001). Blood neurofilament light chain level at diagnosis (median, interquartile range; picograms per millilitre) was significantly increased in both presymptomatic (14.7, 10.6-56.7) and symptomatic patients (136, 40.8-445) compared to controls (5.6, 4.5-7.1), and highest among patients with late-infantile (456, 201-854) or early-juvenile metachromatic leukodystrophy (291.0, 104-445) and those ineligible for treatment based on best practice (291, 57.4-472). GFAP level (median, interquartile range; picogram per millilitre) was only increased in symptomatic patients (591, 224-1150) compared to controls (119, 78.2-338) and not significantly associated with treatment eligibility (P = 0.093). Higher blood neurofilament light chain and GFAP levels at diagnosis were associated with rapid disease progression in late-infantile (P = 0.006 and P = 0.051, respectively) and early-juvenile patients (P = 0.048 and P = 0.039, respectively). Finally, blood neurofilament light chain and GFAP levels decreased during follow-up in untreated and transplanted patients but remained elevated compared with controls. Only neurofilament light chain levels were associated with MRI deterioration (P < 0.001). This study indicates that both proteins may be considered as non-invasive biomarkers for clinical phenotype and disease stage at clinical assessment, and that neurofilament light chain might enable neurologists to make better informed treatment decisions. In addition, neurofilament light chain holds promise assessing treatment response. Importantly, both biomarkers require paediatric reference values, given that their levels first decrease before increasing with advancing age.

Introduction

Metachromatic leukodystrophy (MLD) (OMIM #250100) is an autosomal recessive lysosomal storage disease caused by deficient activity of arylsulfatase A. Accumulation of sulfatides, the macromolecular substrate of arylsulfatase A, in glia cells and neurons results in progressive central and peripheral demyelination. Consequently, patients show severe neurological deterioration characterized by loss of all motor and communication skills and eventually premature death.

Based on the age of symptom onset, a late-infantile, early-juvenile, late-juvenile and adult clinical phenotype can be distinguished (onset at <2.5 years, 2.5–6 years, 6–16 years and >6 years, respectively). The natural disease course can be roughly divided into four clinical stages. The first is a presymptomatic stage with normal motor and cognitive development, albeit evidence for demyelination may already be present both at brain imaging and nerve conduction studies. The second is an early plateau stage, characterized by developmental slowing and stagnation, followed by the onset of first symptoms. The third is the stage of rapid disease progression when accelerated central demyelination results in rapid loss of gross motor function. The final stage is marked by very slow deterioration or stabilization at a low functional level prior to death. The duration of these stages can be shorter (late-infantile and early-juvenile MLD) or longer and more variable (late-juvenile and adult MLD).,, Importantly, patients with late-juvenile and adult onset may present with mainly behavioural and cognitive symptoms, without the rapid motor deterioration seen in the patients with younger onset. Prediction of disease progression for these patients is particularly challenging.

During the past three decades, allogeneic haematopoietic stem cell transplantation (HSCT) has proved to be a valuable therapeutic approach for pre- or early-symptomatic patients with a juvenile or adult MLD phenotype. It aims to correct arylsulfatase A deficiency in the brain, halt sulfatide accumulation and enable remyelination. In addition, haematopoietic stem cell gene therapy (HSC-GT) has recently been approved as a treatment for presymptomatic patients with late-infantile MLD and pre- or early-symptomatic patients with early-juvenile MLD, while HSC-GT for later-onset MLD forms and intrathecal enzyme replacement therapy are still being evaluated in clinical trials (NCT04283227 and NCT01887938, respectively). Evaluation of treatment eligibility and monitoring of therapeutic effects in both clinical and research setting require reliable biomarkers (with age-matched reference data) that are ideally cheap, reproducible and observer independent. In addition, such biomarkers in MLD would preferably be able to reflect different disease stages and mark rapid disease progression during which treatment with HSCT and HSC-GT is ineffective or even detrimental., In this regard, neurofilament light chain (NfL) and glial fibrillary acidic protein (GFAP) are considered potential candidates.,

NfL and GFAP are cytoskeletal proteins that are released into CSF and eventually blood on neuroaxonal and astroglial injury, respectively. Both biomarkers can be easily and reliably quantified in CSF, plasma and serum by a Single Molecule Array, and have been proven to serve as biomarkers of disease activity in several central and, in case of NfL, peripheral demyelinating diseases. In addition, NfL and GFAP levels in CSF correlate strongly with levels in blood in most neurodegenerative diseases, highlighting the potential for NfL and GFAP in blood as minimally invasive measures for MLD disease activity.,,,, Nevertheless, data on NfL and GFAP levels in paired CSF and blood samples of paediatric patients and controls are currently lacking.

The main objective of this study was to determine the potential of NfL and GFAP levels in blood (bNfL and bGFAP) for MLD disease activity. We hypothesized that bNfL and bGFAP levels would be increased in presymptomatic and symptomatic patients with MLD compared to similar aged controls, and that higher levels would be associated with a more severe disease phenotype and clinical indicators of disease progression, reflecting higher disease activity. In addition, we presumed that bNfL and bGFAP levels would decline over time in transplanted patients, and that bNfL and bGFAP levels would strongly correlate with NfL and GFAP levels in CSF (cNfL and cGFAP). Confirmation of these hypotheses would establish bNfL and bGFAP levels as relatively non-invasive biomarkers to measure disease activity for clinical assessment and treatment follow-up for MLD.

Materials and methods

Subjects

In this retrospective study, we included 27 subjects from the MLD cohort (total n = 90) of the Amsterdam Leukodystrophy Center, a Dutch nationwide expertise centre, with a confirmed diagnosis of MLD; available data on the age of symptom onset and loss of gross motor function over time and at least one available CSF, plasma or serum sample. In addition, we included 13 patients from the same cohort with an already performed NfL measurement during clinical assessment in our centre, resulting in a total inclusion of 40 study participants.

Two anonymous reference cohorts were included. The first cohort consisted of children with suspected neurological disorders for whom both results of additional investigations, including brain imaging and follow-up did not confirm neurological disease. Thirty-eight of these ‘neurologically healthy’ children between the age of 0.3 and 17.5 years with available (paired) blood and CSF samples were included. The second cohort yields from a previously reported in-house reference cohort, consisting of 38 healthy volunteers between the age of 18 and 45 years with available NfL and GFAP measurements in serum. Paediatric controls older than 16 years were included in the adult group for the statistical comparison between patients and controls.

Standard protocol approvals and patient consents

The local Institutional Review Board approved the study, and appropriate written consent was obtained according to the Declaration of Helsinki.

Clinical assessment

Patient characteristics and follow-up data were collected from patient records. According to the age of symptom onset, patients were grouped into a ‘late-infantile’ (<2.5 years), ‘early-juvenile’ (2.5–6 years), ‘late-juvenile’ (6–16 years) and ‘adult’ (>16 years) clinical phenotype. Presymptomatic patients diagnosed by family screening were grouped into the same clinical phenotype as their MLD affected sibling. Presence and type of first symptoms were assessed at diagnosis and categorized as ‘presymptomatic’ (no symptoms), ‘motor phenotype’ (only motor symptoms), ‘cognitive phenotype’ (only cognitive symptoms) or ‘mixed phenotype’ (both motor and cognitive symptoms). Cognitive performance was evaluated by full-scale intelligence quotient (FSIQ) at standardized neuropsychological testing at diagnosis. Peripheral neuropathy at diagnosis was dichotomized into ‘no to mild peripheral neuropathy’ and ‘moderate to severe peripheral neuropathy’ based on neurophysiologist’s conclusions in nerve conduction study reports, since examinations were performed in multiple hospitals using different protocols and reference values. One late-juvenile patient was assigned ‘no to mild peripheral neuropathy’ based on the absence of clinical signs of peripheral neuropathy only, and one early-juvenile patient was assigned ‘moderate to severe peripheral neuropathy’ based on clinical signs and nerve pathology. Treatment characteristics included whether or not the patient proceeded to treatment with HSCT or HSC-GT based on clinical eligibility and, in the case of treatment with HSCT, the type of HSCT: ‘bone marrow transplantation’ or ‘umbilical cord blood transplantation’.

Disease activity

Disease activity was determined in light of two distinct purposes: (i) to examine whether bNfL and bGFAP levels could be used to predict the stage of rapid disease progression at diagnosis; and (ii) to examine whether higher bNfL and bGFAP levels were associated with any disease progression over time. For the first purpose, the validated Gross Motor Function Classification in MLD (GMFC‐MLD) was used to distinguish between patients with ‘rapid disease progression’ and ‘slow disease progression’ at the moment of diagnosis based on the time interval between symptom onset or, in the case of presymptomatic patients, diagnosis and entering GMFC‐MLD level 2 (loss of independent walking). Rapid disease progression and slow disease progression at diagnosis were defined as a time interval ≤27 months and >27 months, respectively, in line with previous publications., For the second purpose, a validated measure of brain demyelination, the MRI severity score, was used to quantify disease progression at any available time point during follow-up. We analysed the correlation with the crude MRI severity score at blood sampling and with ‘MRI deterioration’. For the latter, we calculated the difference in MRI severity score compared to the closest previous MRI and divided this by the time interval in-between to account for variability in follow-up duration.

Sample collection and processing

Samples of paediatric controls and patients with MLD consisted of leftover samples from venous and lumbar puncture performed in the context of routine clinical care. In addition, NfL measurements were already performed in seven CSF samples and 15 blood samples of patients (n = 13) during previous clinical assessments. When paired, blood samples were collected directly before or after lumbar puncture. CSF was collected in polypropylene collection tubes and stored at −80°C or −30°C until analyses according to previously published consensus guidelines. Blood was collected in EDTA, heparin or serum separator tubes and centrifuged for 10 min at room temperature at 1800g. Leftover plasma and serum were aliquoted in 0.5-ml volumes and stored at −80°C or −30°C until analyses according to previously published consensus guidelines. Long-term storage at −30°C is not thought to have affected NfL and GFAP concentrations., Collection and processing methods of adult control samples have been described previously. Regarding adult controls, only serum measurements were available.

Measurement of NfL and GFAP

All measurements were performed in duplicate using Single Molecule Array technology (Quanterix) with the commercial NF‐Light advantage Kit and GFAP Discovery Kit on a HD‐X platform according to the manufacturer’s instructions (www.quanterix.com/products-technology/assays). The inter-assay variation for NfL was 8%, based on three levels of internal quality control pools, measured in 45 runs. The inter-assay variation for GFAP was 15%, based on three levels of internal quality control pools, measured in 40 runs. The mean intra‐assay coefficients of variation (duplicate measurements) were <10% for both NfL and GFAP. Therefore, samples with too low volume for a duplicate measurement (n = 4) were also included in the analysis. Based on an in-house quality study, NfL values in heparin plasma, GFAP values in serum and GFAP values in heparin plasma were adjusted by −29, −13 and −18%, respectively, to allow comparison between measurements in heparin plasma, serum and EDTA plasma., Finally, the GFAP values of the adult reference cohort were multiplied by a factor of 1.3, based on internal quality control values, to correct for inter-batch variability. All measurements were performed by certified technicians at the Neurochemistry laboratory of the Amsterdam UMC location VUmc blinded to clinical information.

Statistical analyses

Categorical parameters were described by counts and percentages, and continuous and ordinal parameters by median and interquartile range (IQR). Univariable analyses were performed using a chi-squared test, Fisher's exact test, Mann–Whitney–Wilcoxon Test or Kruskal–Wallis multiple comparison with Benjamini–Hochberg stepwise adjustment as appropriate. Crude correlations and differences between NfL and GFAP in paired CSF and blood samples were assessed by either the Spearman's rank correlation method, repeated measures correlation or Wilcoxon signed-rank test.

In all regression analyses, NfL and GFAP levels and age were log-transformed to meet the assumptions of normality. However, for clarity, regression coefficients were back-transformed to the original scales reflecting multiplicative effects (βmult) for associations with non-transformed predictive parameters and describing percentages (βperc) for associations with log-transformed predictive parameters.

Comparisons of NfL and GFAP levels between controls and patients were statistically adjusted for age at sampling and sex based on previous publications. Because of the low number of patients, only parameters that were significantly associated in univariable analyses were included in the multivariable analyses among patients at diagnosis. As FSIQ at diagnosis included many missing values (n = 10, 25%), was measured with 11 different tests, and was often not fully reliable due to the discrepancy between verbal and performance intelligence quotients in individual patients, we decided to exclude this parameter from multivariable analyses beforehand. One missing value in MRI severity score was imputed by predictive mean matching.

To analyse NfL and GFAP levels over time in patients linear mixed models were fitted including predictive patient, disease and transplantation parameters. Samples (n = 2) of two patients treated with HSC-GT were excluded from the analyses. Detailed model characteristics can be found in the Supplementary material.

All statistical tests were two-tailed and P‐values <0.05 were considered statistically significant. The R project (RStudio: Integrated Development for R. RStudio, Inc., Boston, MA) for statistical computing v.4.0.3 with the packages ‘lme4’, ‘rmcorr’ and ‘ggplot2’ was used for all analyses and for the creation of the figures.

Data availability

Unpublished anonymized data within this article are available on reasonable request from a qualified investigator.

Results

Demographics of patients and controls

The patient dataset consisted of 40 patients with MLD between the age of 0 and 42 years, including four sibling pairs. Twenty patients (50%) were male. Seven patients (17.5%) had a late-infantile clinical phenotype, seven patients (17.5%) had an early-juvenile clinical phenotype, 18 patients (45%) had a late-juvenile clinical phenotype and eight patients (20%) had an adult clinical phenotype. Eleven patients were presymptomatic at diagnosis (28%) of whom 10 proceeded to treatment. In total, 19 patients (48%) proceeded to treatment with HSCT (n = 17) or HSC-GT (n = 2). Moderate to severe peripheral neuropathy at diagnosis was significantly more often observed in symptomatic patients (69%) than in presymptomatic patients (36%, P < 0.001), and in patients with a motor phenotype (100%) than in those with a mixed or cognitive phenotype (50 and 25%, respectively, P < 0.001). By contrast, median MRI severity score at diagnosis was significantly lower in patients with a motor phenotype [11 (IQR 5–18)] than in patients with a mixed [19 (IQR 17–21), P = 0.020] or cognitive phenotype [19 (IQR 18–20), P = 0.020]. In addition, patients with a motor phenotype were on average younger [median age 5.1 (IQR 2.3–6.7) years] than patients with a mixed [median age 14.8 (IQR 11.3–19.6) years, P = 0.006] or cognitive phenotype [median age 21.4 (IQR 13.8–26.7) years, P = 0.002]. Detailed patient demographics and disease characteristics, including differences between untreated and treated patients, are presented in Table 1 (irrespective of clinical phenotype) and Supplementary Table 1 (per clinical phenotype).

Table 1

Patient demographics and disease characteristics

Variablea	All (n = 40)	Treated (n = 19)	Untreated (n = 21)	P treated versus untreatedb
Age at diagnosis, y	11.1 (4.6–19.2)	13.8 (3.2–17.4)	9.0 (5.6–20.2)	<0.001
Male	20 (50.0)	8 (42.1)	12 (57.1)	0.527
Clinical phenotype, n (L-I/E-J/L-J/A)	7/7/18/8	3/3/8/5	4/4/10/3	0.860
Symptomatic at diagnosis	29 (72.5)	9 (47.4)	20 (95.2)	0.001
Age of onset, y	7.5 (3–14)	12 (3–12)	7 (4–14)	<0.001
Type of first symptoms, n (mp/cp/mxp/p-s)	14/7/8/11	5/2/2/10	9/5/6/1	<0.001
FSIQ at diagnosisc	88 (70–99)	93 (88–105)	65 (57–77)	<0.001
GMFC‐MLD score at diagnosis	1 (0–2)	0 (0–1)	1 (0–5)	0.008
MRI severity score at diagnosis	12 (2–19)	9 (3–12)	18 (17–20)	<0.001
Moderate to severe peripheral neuropathy	24 (60.0)	10 (52.6)	14 (66.7)	0.561
Treatment, n (no, HSCT, HSC-GT)	21/17/2	0/17/2	21/0/0	—
Age at treatment, y	14.3 (4.7–17.8)	14.3 (4.7–17.8)	—	—
Rapid disease progression	15 (37.5)	4 (21.1)	11 (52.4)	0.055

A = adult; cp = cognitive phenotype; E-J = early-juvenile; GMFC‐MLD = Gross Motor Function Classification in metachromatic leukodystrophy; L-J = late-juvenile; L-I = late-infantile; MLD = metachromatic leukodystrophy; mp = motor phenotype; mxp = mixed phenotype; p-s = presymptomatic; y = years. aValues indicate median (IQR) or n (%) unless otherwise stated.

P-values were obtained with a chi-squared test, Fisher's exact test, or Mann–Whitney–Wilcoxon Test as appropriate. Significant P-values are indicated in bold.

Only 30 patients with available FSIQ measurements were included.

The control dataset included 38 unrelated neurologically healthy children (aged 0–17 years) and 38 unrelated adult healthy volunteers (aged 18–45 years). Twenty-four (63%) children were male [median age 2.8 (IQR 2.1–4.1) years] and 14 (37%) children were female [median age 3.1 (IQR 0.9–5.3) years]. Sixteen (42%) adults were male [median age 39.0 (IQR 28.5–43.3) years] and 22 (58%) adults were female [median age 37.0 (IQR 29.3–41.8) years]. One 17-year-old female paediatric control was included in the adult group (age category: >16 years) for statistical comparison of bNfL and bGFAP levels between patients and controls (Supplementary Table 1).

NfL and GFAP levels in paediatric controls

We present the results of all paediatric controls (n = 38). NfL and GFAP levels of adult controls were previously described.

The paediatric control samples consisted of 33 paired (CSF and blood) and five unpaired CSF (n = 4) and blood (n = 1) samples. The blood samples consisted of heparin plasma (n = 32) and EDTA plasma (n = 2). Baseline cNfL and bNfL levels did not differ between males and females after correcting for age (P = 0.240 and P = 0.144, respectively), nor did cGFAP and bGFAP levels (P = 0.768 and P = 0.087, respectively). NfL and GFAP levels in CSF and blood were negatively associated with age (all P < 0.001). These associations were not linear but showed a steeper decrease in NfL and GFAP levels during the first years of life. Thus, in healthy individuals NfL and GFAP levels first decrease in CSF and blood before increasing again with advancing age (Fig. 1A and B).

Figure 1

NfL and GFAP levels in controls expressed on a log (A) NfL and (B) GFAP levels in blood are visualized over age with estimated LOESS regression curves (locally weighted scatter-plot smoother) and their 95% CIs (shadows). NfL and GFAP levels in blood show a non-linear decrease during childhood with a steeper slope in the first years of life, before increasing with advancing age in adulthood. NfL and GFAP levels in CSF show similar trends (data not shown because individual adult CSF control values were unavailable). (C) Median NfL level was significantly higher in CSF (100 pg/ml) compared to blood (5.8 pg/ml) and (D) median GFAP level was significantly higher in CSF (5564 pg/ml) compared to blood (370 pg/ml) in paired samples from paediatric controls (aged 0–17 years). Boxes depict median and IQR, grey lines indicate paired measurements, and upper/lower whiskers extend from the hinge towards the largest/smallest values but no farther than 1.5 times IQR from the hinge. The P-values were obtained with the non-parametric Wilcoxon signed-rank test.

NfL levels in blood at metachromatic leukodystrophy diagnosis

NfL was quantified in 11 CSF and 92 blood samples (heparin plasma n = 1, EDTA plasma n = 76, serum n = 15), of which six and 27 were obtained at diagnosis, respectively. Median bNfL levels in patients at diagnosis were 136 (IQR 40.8–445) pg/ml for symptomatic patients and 14.7 (IQR 10.6–56.7) pg/ml for presymptomatic patients (Table 2). These levels were significantly increased in both patient groups compared to controls (βmult = 22.45, P < 0.001 and βmult = 3.23, P < 0.001, respectively), with a larger absolute and relative increase in patients with an earlier-onset clinical phenotype (Fig. 2A). Median bNfL levels were 456 (IQR 201–854) pg/ml, 291 (IQR 104–445) pg/ml, 38.6 (IQR 23.0–89.2) pg/ml and 33.8 (IQR 32.4–50.6) pg/ml for patients with late-infantile, early-juvenile, late-juvenile and adult MLD, respectively (Table 2).

Table 2

Univariable and multivariable associations between NfL levels in blood at diagnosis and clinical parameters in patients with MLD

Variable (n patients = 27)		Univariable model			Multivariable model
	bNfL, pg/mla	βb	95% CI	P c	βb	95% CI	P c
Age at sampling, y	—	−0.368	−0.845–0.110	0.126	NA	NA	NA
Sex
Female (ref)	29.0 (18.7–116)	—	—	—	—	—	—
Male	104 (50.6–291)	2.447	0.766–7.819	0.125	NA	NA	NA
Clinical phenotype
Adult (ref)	33.8 (32.4–50.6)	—	—	—	—	—	—
Late-juvenile	38.6 (23.0–89.2)	1.306	0.338–5.052	0.687	1.140	0.403–3.231	0.795
Early-juvenile	291 (104–445)	6.480	1.359–30.890	0.021	3.471	1.046–11.519	0.043
Late-infantile	456 (201–854)	8.030	1.891–34.093	0.007	3.779	1.153–12.386	0.030
Symptomatic at diagnosis
No (ref)	14.7 (10.6–56.7)	—	—	—	—	—	—
Yes	136 (40.8–445)	6.558	2.005–21.448	0.003	1.881	0.593–5.61	0.266
FSIQ at diagnosisd	—	1.000	0.970–1.032	0.981	NA	NA	NA
MRI severity score at diagnosis	—	1.021	0.936–1.114	0.625	NA	NA	NA
Peripheral neuropathy
No to mild (ref)	35.1 (16.0–87.7)	—	—	—	—	—	—
Moderate to severe	266 (60.2–464)	4.710	1.736–12.773	0.004	1.370	0.429–4.380	0.577
Rapid disease progression
No (ref)	35.1 (18.7–80.2)	—	—	—	—	—	—
Yes	291 (136–472)	6.814	2.845–16.319	<0.001	1.506	0.453–5.002	0.484
Proceeding to treatment
No (ref)	291 (57.4–472)	—	—	—	—	—	—
Yes	36.6 (18.7–95.8)	0.177	0.069–0.452	0.023	0.353	0.142–0.879	0.027

NA = not assessed; ref = reference group; y = years. aValues indicate median (IQR).

The β estimates represent multiplicative effects of parameter, holding everything else constant. However, since age at sampling was also log-transformed, this β estimate indicates the average percentage of decrease in NfL level for every percent increase in age, holding everything else constant.

P-values were obtained with univariable and multivariable linear regression models. Because of the low number of patients, only parameters that were significantly associated with bNfL level in univariable analysis were included in the multivariable analysis. These were ‘clinical phenotype’, ‘symptomatic at diagnosis’, ‘peripheral neuropathy’, ‘rapid disease progression’ and ‘proceeding to treatment’. Significant P-values are indicated in bold.

Only 22 patients with available FSIQ measurements at diagnosis were included.

Figure 2

NfL and GFAP levels in blood in patients at MLD diagnosis. (A) Levels of bNfL and (B) bGFAP expressed on a log10 scale in patients with MLD compared to similar aged controls grouped by clinical phenotype and corresponding age. Boxes depict median and IQR within a group, and dots mark individual measurements. The P-values were obtained with a linear regression model adjusted for age at sampling and sex. (C) NfL levels and (D) GFAP levels in blood expressed on a log10 scale in patients with slow versus rapid disease progression, and separately for the four clinical phenotypes (E and F). Boxes depict median and IQR within a group, and dots mark individual measurements for patients. Magenta-contoured dots indicate presymptomatic patients. The P-values were obtained with a linear regression model adjusted for clinical phenotype and presence of symptoms. *P < 0.05, **P < 0.01 and ***P < 0.001. y = years.

Univariable analyses including all patients showed that bNfL levels at diagnosis were positively associated with a late-infantile or early-juvenile clinical phenotype (reference: adult phenotype, P = 0.007 and P = 0.021, respectively), presence of symptoms (P = 0.003), severe to moderate peripheral neuropathy at diagnosis (P = 0.004) and rapid disease progression (P < 0.001), while patients who were selected for treatment had lower bNfL levels (P < 0.001). A late-infantile or early-juvenile clinical phenotype and proceeding to treatment were confirmed to be independently associated with bNfL levels in a multivariable model (Table 2). Analyses including only symptomatic patients showed comparable results (Supplementary Table 2). Notably, MRI severity score at diagnosis was not associated with bNfL level in both analyses (Supplementary Fig. 1).

GFAP levels in blood at metachromatic leukodystrophy diagnosis

GFAP was quantified in four CSF and 88 blood samples (heparin plasma n = 1, EDTA plasma n = 81, serum n = 6) of which only blood samples (n = 27) were obtained at diagnosis. Median bGFAP levels in patients at diagnosis were 591 (IQR 224–1150) pg/ml for symptomatic patients and 403 (280–440) pg/ml for presymptomatic patients (Table 3). Levels of bGFAP were significantly increased in symptomatic patients compared to controls (βmult = 2.48, P < 0.001), but not in presymptomatic patients (βmult = 1.44, P = 0.103). The absolute GFAP increase was most pronounced in patients with an earlier-onset clinical phenotype, although its relative increase was comparable among the clinical phenotypes (Fig. 2B). Median bGFAP levels were 1069 (IQR 781–1414) pg/ml, 1150 (IQR 580–1348) pg/ml, 338 (IQR 227–458) pg/ml and 192 (IQR 179–221) pg/ml for patients with late-infantile, early-juvenile, late-juvenile and adult MLD, respectively (Table 3).

Table 3

Univariable and multivariable associations between GFAP levels in blood at diagnosis and clinical parameters in patients with MLD

Variable (n patients = 27)			Univariable model			Multivariable model
		bGFAP, pg/mla	βb	95% CI	P c	βb	95% CI	P c
Age at sampling, y		—	−0.433	−0.659–0.207	<0.001	NA	NA	NA
Sex
Female (ref)		405 (190–680)	—	—	—	—	—	—
Male		580 (240–1150)	1.342	0.674–2.674	0.388	NA	NA	NA
Clinical phenotype
Adult (ref)		192 (179–221)	—	—	—	—	—	—
Late-juvenile		338 (227–458)	1.668	0.894–3.110	0.103	1.581	0.885–2.822	0.115
Early-juvenile		1150 (580–1348)	4.558	2.219–9.361	<0.001	3.875	1.949–7.704	<0.001
Late-infantile		1069 (781–1414)	5.120	2.630–9.968	<0.001	4.034	2.042–7.968	<0.001
Symptomatic at diagnosis
No (ref)		403 (280–440)	—	—	—	—	—	—
Yes		591 (224–1150)	1.614	0.733–3.551	0.223	NA	NA	NA
FSIQ at diagnosisd		—	1.011	0.993–1.029	0.212	NA	NA	NA
MRI severity score at diagnosis		—	0.976	0.929–1.025	0.318	NA	NA	NA
Peripheral neuropathy
No to mild (ref)		380 (210–454)	—	—	—	—	—	—
Moderate to severe		809 (238–1251)	1.945	1.044–3.625	0.037	0.850	0.452–1.598	0.596
Rapid disease progression
No (ref)		338 (222–440)	—	—	—	—	—	—
Yes		1069 (753–1348)	2.604	1.502–4.512	<0.001	1.455	0.734–2.881	0.267
Proceeding to treatment
No (ref)		1069 (358–1348)	—	—	—	—	—	—
Yes		360 (227–458)	0.489	0.266–0.898	0.023	0.678	0.428–1.074	0.093

NA = not assessed; ref = reference group; y = years. aValues indicate median (IQR).

The β estimates represent multiplicative effects of parameter, holding everything else constant. However, since age at sampling was also log-transformed, this β estimate indicates the average percentage of decrease in GFAP level for every percent increase in age, holding everything else constant.

P-values were obtained with univariable and multivariable linear regression models. Because of the low number of patients, only parameters that were significantly associated with bNfL level in univariable analysis were included in the multivariable analysis. These were ‘clinical phenotype’, ‘peripheral neuropathy’, ‘rapid disease progression’ and ‘proceeding to treatment’. ‘Age at sampling’ was removed from the model due to strong multicollinearity. Significant P-values are indicated in bold.

Only 22 patients with available FSIQ measurements at diagnosis were included.

Levels of bGFAP at diagnosis were positively associated with age at sampling (P < 0.001), a late-infantile or early-juvenile clinical phenotype (reference: adult phenotype, P < 0.001 and P < 0.001, respectively), severe to moderate peripheral neuropathy at diagnosis (P = 0.004) and rapid disease progression (P < 0.001), and negatively associated with proceeding to treatment (P = 0.023) in univariable analyses. Only a late-infantile or early-juvenile clinical phenotype was confirmed to be independently associated with bGFAP levels in a multivariable model (Table 3). Analyses including only symptomatic patients showed similar results and an additional positive association between GFAP level and motor phenotype compared to cognitive phenotype (P = 0.049, Supplementary Table 3). MRI severity score at diagnosis was not associated with bGFAP level (Supplementary Fig. 1).

Association between NfL and GFAP levels and rapid disease progression at diagnosis

The associations between bNfL and bGFAP levels and rapid disease progression at diagnosis were analysed within all patients (statistically adjusted for clinical phenotype and presence of symptoms), and within each clinical phenotype (statistically adjusted for the presence of symptoms). Although median bNfL level was higher in patients with rapid disease progression than in patients with slow disease progression, this difference was only statistically significant in patients with late-infantile or early-juvenile MLD (P = 0.006 and P = 0.048, respectively; Fig. 2C and D). Interestingly, median bGFAP levels were also higher in patients with late-infantile and early-juvenile MLD with rapid disease progression than in those with slow disease progression (P = 0.051 and P = 0.039, respectively), while patients with late-juvenile and adult MLD with and without rapid disease progression had similar median bGFAP levels (Fig. 2E and F).

NfL levels in blood over time

NfL levels in blood in untreated and treated patients remained elevated over time compared to age-matched controls (Fig. 3A), even 20 years after diagnosis and treatment (Fig. 3B). Median bNfL level over time was lower in treated than in untreated patients (P < 0.001). In addition, bNfL decreased with increasing age (P < 0.001) and follow-up duration (P < 0.001; Table 4). However, in treated patients an increase (up to 6.4 times higher in the first 3 months) followed by a relatively steep decrease was observed within the first year after treatment with HSCT. Individual plots indicated that the increase and decrease were larger in younger than in older treated patients (Fig. 3A). Parameters that were associated with bNfL levels in treated patients separately are shown in Supplementary Table 4. Importantly, patients who were symptomatic at the moment of treatment had higher bNfL levels over time than patients who were treated presymptomatically (motor phenotype: P = 0.019, mixed phenotype: P = 0.008 and cognitive phenotype: P = 0.015, respectively).

Figure 3

Longitudinal comparison of untreated and treated patients with MLD according to age (A and C) or since diagnosis (B and D). The left panels show the course of bNfL level (A) and bGFAP level (C) over time according to age as estimated by a linear mixed model for untreated (peach line) and treated patients (blue line). The 95% CIs are shown as shadows in a corresponding colour. Reference values are visualized in green with an estimated LOESS regression curve and 95% CI (shadow). Coloured dots (measurements) and shaded lines (course over time) reflect the individual patient data. The right panels show the same individual patient data at bNfL level (B) and bGFAP level (D), but over time since diagnosis.

Table 4

Multivariable associations between NfL levels in blood over time and clinical parameters in untreated and treated patients with MLD

Variable (n patients = 35, n measurements = 90)	Full model			Final model
	βa	95% CI	P b	βa	95% CI	P b
Follow-up since diagnosis, y	See Fig. 3A (rcs)		<0.01	See Fig. 3A (rcs)		<0.001
Age at sampling, y	−0.663	−1.519–0.194	0.127	−0.854	−1.097 to −0.612	<0.001
Sex
Female (ref)	—	—	—	—	—	—
Male	1.248	0.713–2.184	0.421	1.459	0.979–2.176	0.063
Clinical phenotype
Adult (ref)	—	—	—	—	—	—
Late-juvenile	1.003	0.366–2.753	0.995	NA	NA	NA
Early-juvenile	0.936	0.175–5.005	0.936	NA	NA	NA
Late-infantile	1.235	0.098–15.525	0.865	NA	NA	NA
Type of first symptoms
Presymptomatic (ref)	—	—	—	—	—	—
Motor phenotype	1.641	0.535–5.030	0.370	NA	NA	NA
Mixed phenotype	1.759	0.549–5.563	0.898	NA	NA	NA
Cognitive phenotype	1.062	0.409–2.753	0.898	NA	NA	NA
MRI severity score at diagnosis	0.989	0.902–1.083	0.801	NA	NA	NA
Peripheral neuropathy
No to mild (ref)	—	—	—	—	—	—
Moderate to severe	1.047	0.581–1.886	0.875	NA	NA	NA
Proceeding to treatment
No (ref)	—	—	—	—	—	—
Yes	0.409	0.179–0.933	0.035	0.388	0.256–0.588	<0.001

NA = not applicable; rcs = restricted cubic splines; ref = reference group; y = years. aThe β estimates represent multiplicative effects of parameter, holding everything else constant. However, since age at sampling was also log-transformed, this β estimate indicates the average percentage of decrease in NfL level for every percent increase in age, holding everything else constant.

P-values were obtained with a multivariable mixed-effect model including all parameters (full model) and only those selected by maximum likelihood estimation (final model). Significant P-values are indicated in bold.

GFAP levels in blood over time

Median GFAP levels in blood in untreated and treated patients remained elevated over time compared to age-matched controls, although in a few young individuals bGFAP levels decreased to normal values during the years following treatment with HSCT (Fig. 3C and D). Late-infantile, early-juvenile and late-juvenile clinical phenotypes were associated with lower bGFAP levels over time than an adult clinical phenotype (P = 0.015, P = 0.025 and P = 0.045, respectively). In addition, MRI severity score at diagnosis was positively associated with bGFAP levels over time (P < 0.001). Finally, bGFAP decreased with increasing age (P < 0.001), while follow-up duration was not a statically significant parameter (Table 5). Individual plots indicated also increase (up to 5.9 times higher in the first five months) and decrease in bGFAP level within the first year after treatment with HSCT, but this was comparable between younger and older treated patients (Fig. 3C). Parameters that were associated with bGFAP levels in treated patients separately are shown in Supplementary Table 5.

Table 5

Multivariable associations between GFAP levels in blood over time and clinical parameters in untreated and treated patients with MLD

Variable (n patients = 33, n measurements = 86)	Full model			Final model
	βa	95% CI	P b	βa	95% CI	P b
Follow-up since diagnosis, y	See Fig. 3B (rcs)		NS	See Fig. 3B (rcs)		NS
Age at sampling, y	−1.583	−2.206 to −0.956	<0.001	−1.469	−1.963 to −1.000	<0.001
Sex
Female (ref)	—	—	—	—	—	—
Male	0.854	0.604–1.207	0.354	NA	NA	NA
Clinical phenotype
Adult (ref)	—	—	—	—	—	—
Late-juvenile	0.481	0.226–1.027	0.058	0.581	0.342–0.987	0.045
Early-juvenile	0.290	0.084–1.000	0.050	0.358	0.147–0.870	0.025
Late-infantile	0.147	0.023–0.957	0.045	0.208	0.060–0.719	0.015
Type of first symptoms
Presymptomatic (ref)	—	—	—	—	—	—
Motor phenotype	1.369	0.651–2.878	0.390	NA	NA	NA
Mixed phenotype	1.162	0.562–1.429	0.673	NA	NA	NA
Cognitive phenotype	0.778	0.424–1.429	0.401	NA	NA	NA
MRI severity score at diagnosis	1.076	1.005–1.151	0.036	1.060	1.030–1.091	<0.001
Peripheral neuropathy
No to mild (ref)	—	—	—	—	—	—
Moderate to severe	0.758	0.521–1.102	0.139	NA	NA	NA
Proceeding to treatment
No (ref)	—	—	—	—	—	—
Yes	1.113	0.588–2.108	0.731	NA	NA	NA

NA = not applicable; rcs = restricted cubic splines; ref = reference group; y = years. aThe β estimates represent multiplicative effects of parameter, holding everything else constant. However, since age at sampling was also log-transformed, this β estimate indicates the average percentage of decrease in GFAP level for every percent increase in age, holding everything else constant.

P-values were obtained with a multivariable mixed-effect model including all parameters (full model) and only those selected by maximum likelihood estimation (final model). Significant P-values are indicated in bold.

Blood NfL and blood GFAP levels and disease progression on brain MRI over time

Patients with a late-infantile clinical phenotype had overall lower crude MRI severity scores during follow-up than those with an early-juvenile, late-juvenile or adult clinical phenotype (P = 0.076, P < 0.001 and P = 0.002, respectively), but showed a higher increase in MRI severity score indicating faster MRI deterioration over time (P = 0.028, P = 0.045 and P = 0.047, respectively). Higher crude MRI severity scores during follow-up were correlated with both higher bNfL and bGFAP levels (r = 0.31; P = 0.040 and r = 0.44; P = 0.003, respectively), and these associations were retained after correcting for clinical phenotype (βmult = 1.096; P < 0.001 and βmult = 1.062; P < 0.001, respectively). MRI deterioration correlated, however, only with higher bNfL levels (r = 0.45; P = 0.012), and not with bGFAP levels (r = 0.250; P = 0.192). This association retained after correcting for clinical phenotype (βmult = 1.171; P < 0.001).

Correlations between NfL and GFAP levels in CSF and blood

Analysis of paired samples of paediatric controls revealed that median NfL level in CSF [101 (IQR 77.6–135) pg/ml] was 17.3-fold higher than in blood [5.8 (IQR 4.5–9.0) pg/ml; Fig. 1C]. Median GFAP level in CSF [5564 (IQR 3401–7169) pg/ml] was 15-fold higher than in blood [370 (IQR 246–550) pg/ml, Fig. 1D]. NfL and GFAP levels in CSF showed moderate correlations with NfL and GFAP levels in blood (r = 0.52, P = 0.002 and r = 0.50, P = 0.004, respectively), but these associations were not retained after correcting for age and sex (P = 0.142 and P = 0.466, respectively).

Analysis of paired samples of patients revealed that median NfL level in CSF [7955 (IQR 2544–10 291) pg/ml] was 56-fold higher than in blood [142.0 (IQR 53.9–280) pg/ml], far beyond the multiplication factor in controls. In addition, cNfL levels showed a strong correlation with bNfL levels (r = 0.82, P = 0.034), and this association retained after correcting for age and sex, with a 10% increase in cNfL leading to a 10.6% increase in bNfL (βperc = 1.06, P < 0.001). Only one patient had paired CSF and blood results for GFAP. The cGFAP level (44 005 pg/ml) was 26.4-fold higher than bGFAP level (1664 pg/ml) in this patient.

Discussion

There is a need for objective and easy to measure biomarkers to predict MLD disease activity in patients in research and clinical settings. To determine the value of NfL and GFAP levels in CSF and blood as biomarkers for MLD activity, we measured the levels in untreated and transplanted patients over time and compared them to two combined reference cohorts of paediatric and adult controls. We found that NfL and GFAP levels in blood were significantly increased in patients with MLD compared to controls, and that higher levels in patients at diagnosis were associated with a more severe clinical phenotype and higher disease activity, thereby predicting (imminent) clinical progression. In addition, we observed that bNfL level declined over time since diagnosis but remained elevated in all transplanted patients, while bGFAP level declined to normal level in some transplanted individuals with an early- or late-juvenile clinical phenotype. In the current study, bNfL outperformed bGFAP as a biomarker for disease activity in MLD, because only bNfL level was associated with clinical eligibility for treatment at diagnosis and active demyelination on brain MRI during follow-up. Therefore, the question rises whether bGFAP level provides value as a biomarker of MLD disease activity in addition to bNfL level. Considering the exploring goal of the current study, this question still needs to be answered. It will be interesting to see whether, e.g. in clinical trial evaluation of new treatment modalities, a pattern of putative therapeutic effects across the two individual biomarkers might emerge, underlining the benefit of measuring both. The use of both biomarkers requires age-specific, paediatric reference values, since, in contrast to findings in adult controls, NfL and GFAP levels in CSF and blood in paediatric controls were negatively associated with age, showing a steep decline during the first years of life.

Value of blood NfL as a biomarker in metachromatic leukodystrophy

The data in this study indicate that especially bNfL might be useful as an enrichment and prognostic biomarker to support treatment eligibility decisions, indicate rapid disease progression at diagnosis and to monitor disease progression during long-term follow-up. The NfL level in blood was strongly correlated with NfL level in CSF in patients and has therefore the potential of reflecting disease activity in the CNS and PNS without the need for invasive testing. Although validation of this finding is needed in another cohort of patients with MLD, our results are in accordance with previous studies on cNfL and bNfL levels in paediatric neurological diseases, including juvenile neuronal ceroid lipofuscinosis, multiple sclerosis and acute disseminated encephalomyelitis., In addition, NfL levels were associated with clinical phenotype and biomarkers for disease activity and treatment response in such diseases, e.g. disease phenotype in patients with ataxia-telangiectasia; magnetic resonance spectroscopy markers and the Unified Batten Disease Rating Scale score in juvenile neuronal ceroid lipofuscinosis; lesion load on brain MRI, disease progression and time to relapse in paediatric multiple sclerosis; improvement of motor function after treatment in spinal muscular atrophy type 1; and treatment in paediatric opsoclonus–myoclonus syndrome.,,

Another important advantage of bNfL level compared to other currently used biomarkers in MLD care and research, including MRI severity scores, cognitive functioning and the GMFC‐MLD score, is that bNfL, due to its rapid increase on neuroaxonal injury and its half‐life of only a few weeks to 2 months, reflects recent or ongoing disease activity instead of earlier accumulated disease damage. This could explain why we could not establish a close correlation between bNfL and those biomarkers at diagnosis, and that patients with lower MRI severity scores already had relatively high bNfL levels and vice versa, depending on the combination of clinical phenotype and stage of the disease. In more detail, NfL blood level was increased in presymptomatic patients compared to similar aged controls, suggesting that the disease was already causing neuroaxonal injury but not clinical symptoms yet. Preclinical neuroaxonal injury is also observed in other neurodegenerative diseases, including familial amyotrophic lateral sclerosis, familial Alzheimer’s disease and familial frontotemporal dementia. Interestingly, of the two presymptomatic patients with normal brain MRI, cognitive functioning and nerve conduction study results at moment of sampling in our cohort, one, a late-juvenile patient, had increased bNfL level at age 6 years (expected age at symptom onset: 10 years), while the other, a late-infantile patient, had still normal bNfL level at age 2 months (expected age at symptom onset: 1.5–2 years). If a certain threshold of neuroaxonal and white matter damage has to be reached before symptoms and MRI abnormalities appear, NfL might be used to monitor disease activity in these patients in case immediate HSCT or HSC-GT is undesired or not possible. More research is, however, needed to examine whether this hypothesis is true, and to establish the added value of bNfL besides the currently used biomarkers.

Potential implications of blood NfL and blood GFAP levels in metachromatic leukodystrophy pathophysiology

An important finding is that bNfL level declined over time from diagnosis in all patients, including untreated patients. In untreated patients, this decline might reflect transition from the stage of rapid disease progression to the stage of slow deterioration or clinical stabilization, parallel to neuroaxonal loss, which is substantial at the beginning, but burnt out in later stages, resulting in less residual tissue to release NfL into CSF and blood. Consequently, NfL levels should always be interpreted in light of disease stage in addition to age and clinical phenotype, and comparison between patients in different disease stages should be avoided. In addition, a decline in NfL level in treated patients should be inferred as a treatment response only when it is steeper than the decline observed in untreated patients.

Contrary to bNfL, overall bGFAP levels were not increased in presymptomatic patients, and bGFAP levels continued to increase after diagnosis in most untreated patients. In addition, higher bGFAP levels over time were associated with higher MRI severity scores and an adult clinical phenotype. The mechanisms of bGFAP elevation are not yet understood. One possible explanation is that significant astrogliosis may occur only in a later symptomatic stage, especially in patients with an early-onset clinical phenotype. Pathological or imaging evidence for differences in severity of the astrogliosis between MLD clinical phenotypes and during disease evolution has however not been reported yet. A similar relation has been observed in multiple sclerosis, where bGFAP correlated with disease severity scores particularly in progressive multiple sclerosis, reflecting increasing astrogliosis in later disease stages. Alternative explanations, including differences in the mechanisms of release, turnover and kinetics of GFAP and NfL in the CNS and blood compartment, should also be considered.

NfL and GFAP blood levels in relation to transplantation

The increase in bNfL and bGFAP levels within the first months after transplantation suggests that neuroaxonal damage and astrogliosis are more prominent after HSCT, in line with the previously described progression of white matter abnormalities and atrophy on brain MRI during this period., Individual plots indicated that the increase in bNfL and bGFAP level was larger in younger than in older patients, although the longer time intervals between HSCT and blood sampling in the older patients might (partially) have caused this difference. An increase in bNfL and bGFAP levels is likewise reported in rats and humans treated with myeloablative HSCT, suggesting that these increases are, at least to some degree, the result of (transient) neurotoxicity caused by chemotherapy., Accordingly, Thebault et al. found that a higher total busulfan dose correlated with a larger increase in bNfL and bGFAP level at 3 months post-HSCT and with a greater degree of grey and white matter volume loss in patients with multiple sclerosis. However, the average increase in bNfL and bGFAP levels in that study was much lower (bNfL: 7.0 pm/ml, representing a 32.1% increase, and bGFAP: 80.3 pm/ml, representing a 74.8% increase) than the average increase observed at 3 months post-HSCT in our study (bNfL: 182 pm/ml, representing a 609% increase, and bGFAP: 245 pm/ml, representing a 187% increase). Nevertheless, since busulfan is often used as a chemotherapeutic agent in HSCT and HSC-GT for MLD, more knowledge on the increase in bNfL and bGFAP level due to busulfan neurotoxicity is needed before these levels can serve as biomarkers to monitor short-term treatment response.,

In addition, we found evidence that neuroaxonal damage and astrogliosis continue in patients treated with HSCT, despite stabilization or even improvement of white matter abnormalities on brain MRI. This might reflect progressive peripheral neuropathy after HSCT, or, more likely, the fact that enzymatic cross-correction of neurons and neuroglia after HSCT is limited if present at all, leading to a suboptimal treatment effect.,, Nevertheless, in accordance with previous findings of better treatment outcomes,,, patients who underwent transplantation at a presymptomatic stage exhibited lower bNfL levels over time than patients at a symptomatic stage. This finding again underlines the importance of early treatment in MLD.

NfL and GFAP levels in CSF and blood require age-specific, paediatric reference values

We found that NfL and GFAP levels in both CSF and blood in paediatric controls were negatively associated with age, with a steep decline during the first years of life. This is in accordance with another large study on NfL and GFAP levels in CSF of paediatric controls, although the decline during the first years of life was not explicitly mentioned there. There is no good hypothesis yet to explain this decline. NfL and GFAP turnover in infants and young children might be slower than in older children. Myelination, mainly occurring in the first 2 years of life, might also contribute to decreasing axonal neurofilament loss with older age., Other previous studies on NfL and GFAP levels in CSF or blood of paediatric controls did not find an association with age.,,, However, these studies were conducted in a small number of children,,, did not include controls younger than 3 years,, or used patients with other neurological diseases as a control group. Importantly, these physiological changes may complicate the implementation of NfL or GFAP as biomarker in childhood leukodystrophies such as MLD and require age-specific reference values.

Strengths and limitations of the study

Among the strengths of this study are the combination of extensive cross-sectional and longitudinal phenotype data in a relatively large MLD cohort, and the inclusion of a paediatric reference cohort with paired CSF and blood samples. However, there are also some limitations to be addressed. First, our results have yet to be validated in an external cohort of patients. Second, the number of paediatric controls aged 6–16 years was relatively low. Increasing the number of control samples from this age group would be of special importance since the current data indicate that NfL and GFAP levels are lowest within this age range. In addition, the number of patients limited the number of clinical parameters that could be included in our analyses regarding bNfL and bGFAP levels at diagnosis, and our study might not have had sufficient power to detect all significant associations. The groups in the analyses regarding rapid disease progression were particularly small, and these results should therefore be interpreted with caution. Finally, only a few bNfL and bGFAP measurements of untreated patients were available of during follow-up. These patients were overall older and had a slower disease course than most untreated patients.

Conclusion

To conclude, this study indicates that both NfL and GFAP levels in blood hold promise as relatively non-invasive biomarkers for MLD disease stage in both clinical and research setting, and that especially NfL could support neurologists to make informed treatment decisions and to monitor residual disease activity during clinical follow-up. However, bNfL and bGFAP thresholds and their added value with respect to current biomarkers in MLD remain to be validated in a future prospective study. Currently, data are not sufficient to consider NfL and GFAP treatment response biomarkers. Finally, age-specific reference values in children are required.

Click here for additional data file.