Diagnostic and prognostic value of CSF neurofilaments in a cohort of patients with motor neuron disease: A cross‐sectional study

Abstract Motor neuron disease (MND) is a rare group of disorders characterized by degeneration of motor neurons (MNs). The most common form of MND, amyotrophic lateral sclerosis (ALS), is an incurable disease with a variable rate of progression. The search of robust biomarkers able to discriminate among different ALS forms is paramount to properly stratify patients, and to identify those who could most likely benefit from experimental therapies. Phosphorylated‐neurofilament heavy chain (p‐NfH) and neurofilament light chain (NfL) are neuron‐specific components of the cytoskeleton and may represent reliable markers of neuronal injury in neurological disorders. In this study, we described our cohort of ALS patients in order to investigate whether and how cerebrospinal fluid (CSF) p‐NfH and NfL levels may reflect progression rate, MN involvement and the extent of neurodegeneration. CSF p‐NfH and NfL were significantly increased in ALS compared with healthy and disease controls, including patients with other forms of MND, and were higher in patients with more aggressive disease course, reflecting progression rate. We also evaluated neurofilament diagnostic accuracy in our centre, identifying with high sensitivity and 100% specificity cut‐off values of 0.652 ng/mL for CSF p‐NfH (P < .0001) and of 1261 pg/mL for NfL (P < .0001) in discriminating ALS from healthy controls. CSF neurofilaments were significantly correlated with ALS progression rate. Overall, CSF neurofilaments appear to reflect the burden of neurodegeneration in MND and represent reliable diagnostic and prognostic biomarkers in ALS.

Significant clinical, genetic and pathological heterogeneity in MND has made early identification and prediction of disease course extremely complex. In this context, biomarkers are urgently needed not only for earlier diagnosis and prognosis estimation, but also for monitoring therapeutic trial efficacy.
Neurofilaments (Nfs) are neuron-specific components of the cellular cytoskeleton, with a fundamental role in the stabilization and polarization of neurons. 4 Following axonal injury, Nfs are released in the extracellular space and can be detected in the blood and cerebrospinal fluid (CSF). As Nfs levels have been shown to correlate with the severity of axonal damage, 5 they could be a reliable marker of neuronal injury in neurological disorders. Because of its proximity to the central nervous system (CNS), the CSF contains greater Nf concentrations than serum, providing the most suitable source for Nf measurement and study.
In this work, we investigated whether CSF p-NfH and NfL levels may reflect progression rate, MN involvement and burden of neurodegeneration in our cohort of ALS patients. Moreover, we evaluated the diagnostic accuracy of CSF p-NfH and NfL levels in these patients, exploring their correlation with disease course and severity. Finally, we included our previously described cohort of SMA type 3 patients 6 in order to analyse the differences in Nfs levels between these two extremes of the MND spectrum. Neurofilament levels were also assessed on CSF samples derived from patients ultimately diagnosed with functional or not neurological diseases. Moreover, patients with different neurological conditions and MND other than ALS were included. As previously reported, 6 the values of Nfs below the lower limit of quantification were approximated to half of the concentration of the lowest calibrator (50 pg/mL and 0.0625 ng/mL for NfL and p-NfH, respectively).

| Statistical analysis
Baseline characteristics were analysed through descriptive statistics.
Continuous variables were reported as mean ± standard deviation (SD), and categorical variables were represented as relative frequencies and percentages. Between-group comparisons were performed with the Mann-Whitney and Kruskal-Wallis tests. Spearman's correlation coefficient was used to assess the association between variables.
Binomial logistic regression was performed to evaluate the predictive potential of CSF Nfs. Receiver operating characteristic (ROC) curves were generated to assess the diagnostic value of p-NfH and NfL in ALS patients compared with controls. Best cut-off values were calculated with Youden's Index. Univariable binomial logistic regression was used to measure, in ALS patients, the association of fast progression compared with non-fast progression (dependent variable) with either p-NfH or NfL levels (independent variable); odds ratios (OR) and 95% confidence intervals (95% CI) were computed from regression coefficients and standard errors. Statistical analyses were performed with Prism (GraphPad) 8.3.1 Version.

| Demographic and clinical features
We enrolled 32 patients with ALS with mean age at presentation of  Table 2. Diagnoses in the DC group included central and peripheral nervous system inflammatory diseases (n = 15), peripheral neuropathies (n = 17), vascular disorders (n = 6) and neurodegenerative conditions other than MND (n = 29). Age of HC and DC was 42.2 (±20.7) and 63 (±14.9) years, respectively. Baseline features of patients with SMA type 3 were already described by Faravelli et al. 6
Higher CSF p-NfH and NfL levels were observed in ALS patients with fast progression rate compared to those with intermediate (P = .0149 for p-NfH) and slow progression rate (P < .0001 for both p-NfH and NfL) ( Figure 1C, D). Moreover, both CSF p-NfH and NfL levels were significantly lower in SMA type 3 patients compared with intermediate-progressing (3.19 ng/mL, P = .0008 and 4641.29 pg/ mL, P < .0001) and fast-progressing ALS patients (5.827 ng/mL, P < .0001 and 6369.44 pg/mL, P < .0001).

| CSF neurofilaments are reliable diagnostic biomarkers in ALS
We employed ROC analysis to assess to what extent CSF Nfs were able to discriminate between the ALS and control groups. Both CSF p-NfH and NfL displayed a high accuracy for ALS diagnosis when compared to HC (AUC 0.939, P < .0001, and AUC 0.995, P < .0001, respectively) ( Figure 3A, B), and the best cut-off values were 0.652 ng/mL for CSF p-NfH (sensitivity 87.5%, specificity 100%) and 1261 pg/mL for CSF NfL (sensitivity 93.75%, specificity 100%).
Comparing ALS with DC and SMA type 3 patients, CSF p-NfH and NfL maintained a high diagnostic accuracy (AUC 0.867, P < .0001, and AUC 0.853, P < .0001, respectively) ( Figure 3C, D). In this setting, a cut-off value of 1.069 ng/mL for p-NfH discriminated with 87.3% sensitivity and 84.4% specificity ALS from DC and SMA type 3 patients. Similarly, a cut-off value of 1454 pg/mL for NfL discriminated with 64.6% sensitivity and 90.6% specificity ALS from the other diagnoses.

| CSF neurofilaments are associated with progression rate in ALS patients
We investigated whether CSF Nfs reflected the progression rate in ALS patients; we found a significant correlation between both p-NfH (r = 0.443, P = .0125) and NfL levels (r = 0.4574, P = .0097) and progression rate in ALS cohort ( Figure 4A, B).

TA B L E 1 Demographic and clinical features of ALS patients
Moreover, we performed logistic regression analyses using fast progression and non-fast progression as a binary outcome. Both higher CSF p-NfH and NfL levels were significantly associated with fast-progressing ALS (OR 1.834, 95% CI 1.246-3.203, P = .008, and OR 1.006, 95% CI 1.002-1.015, P = .033, respectively).  Increasing evidence has suggested that p-NfH and NfL hold diagnostic and prognostic potential in MNDs, particularly in ALS. 10 Recent studies have demonstrated that CSF p-NfH and NfL levels can discriminate ALS patients from HC and mimics. [11][12][13][14][15] Specifically, CSF p-NfH appears to be highly specific for MND and might be useful complementary tools in aiding early diagnosis. In addition, both CSF p-NfH and NfL seem to correlate with the extent of clinical MN involvement [14][15][16] and their levels at baseline predict disease progression and survival, 12   Overall, this work corroborates the role of CSF Nfs as reliable and promising biomarkers in ALS, aiding clinical diagnosis and providing an accurate estimation of disease progression.