The evolution of NF detection technologies has followed a trajectory of progressively improved analytical sensitivity, enabling the transition from CSF-only measurement to reliable quantification in peripheral blood [Figure 3]^[16].

Accurate interpretation of NF levels requires comprehensive consideration of multiple confounding factors, which can be categorized into physiological factors, comorbid conditions, and pre-analytical variability.

AD is a neurodegenerative disorder of the CNS marked by progressive cognitive decline and behavioral dysfunction^[48]. The neuropathological hallmarks of AD are extracellular Aβ plaques and intracellular neurofibrillary tangles composed of hyperphosphorylated tau protein^[50]. Importantly, neuropathological changes begin many years before the onset of overt clinical symptoms, and even after symptoms emerge^[51], diagnosis may still be delayed by a year or more^[52]. This temporal disconnect between pathology and clinical detection underscores a critical need for biomarkers to identify neurodegeneration in the preclinical phase.

Advances in proteomic technologies have facilitated the identification of candidate biomarkers for AD^[53]. Longitudinal cohort studies confirm that blood NfL elevation occurs far in advance of clinical symptoms in autosomal dominant AD. In the DIAN cohort, the longitudinal change rate of serum NfL (sNfL) can discriminate mutation carriers approximately 16 years before estimated symptom onset^[54]. In the Colombian PSEN1 E280A mutation kindred, plasma NfL divergence between carriers and non-carriers is detectable as early as 22 years before mild cognitive impairment onset^[55]. These findings indicate that NFs alterations reflect early neurodegenerative processes rather than late-stage consequences.

Basic research indicates that calpain, a key protease involved in NF metabolism, can trigger the disintegration of the NF network upon activation, a mechanism that may contribute to axonal degeneration in AD^[14]. Meanwhile, proline-directed kinases that catalyze tau protein hyperphosphorylation can also phosphorylate the tail domains of NfM and NfH, suggesting that in AD, tau pathology and neurofilamentopathy may be driven by common upstream kinases, forming a mutually exacerbating vicious cycle^[56].

Interpretation of NfL levels in AD must account for their lack of disease specificity^[42]. Although NfL concentrations increase with normal aging, patients with neurodegenerative diseases, including AD, exhibit higher absolute levels or faster rates of increase compared with age-matched healthy individuals^[57]. Nonetheless, the ability of NfL to distinguish AD from other dementias remains limited. This is largely because NfL is elevated across multiple neurological conditions, including multiple sclerosis and frontotemporal dementia (FTD)^[58]. Consequently, while NfL reflects neuronal injury, its diagnostic utility as a standalone marker for AD is modest.

Despite these limitations, NfL may be more valuable as a marker of disease severity and progression rather than diagnosis^[16]. Elevated NfL levels in CSF and blood correlate with poorer cognitive performance, greater neurodegeneration, and accelerated disease progression^[32]. Cross-sectional studies have demonstrated associations between higher plasma NfL levels and more severe medial temporal lobe atrophy and reduced metabolic activity in AD-related cortical regions^[7]. Longitudinal analyses further show that elevated baseline NfL predicts faster hippocampal and temporal cortical atrophy and greater declines in fluorodeoxyglucose uptake over 15-30 months^[57]. While NfL alone cannot establish an AD diagnosis, its prognostic and monitoring value is significantly enhanced when combined with established plasma biomarkers such as Aβ42, Aβ40, p-tau217, and p-tau181^[56].

Other NF subunits may also contribute complementary biochemical and diagnostic information. A multi-cohort CSF proteomic study found that CSF NfM (cNfM) levels are significantly elevated in AD patients compared with cognitively healthy controls, with a trend of increase observed in the MCI stage. As an exploratory proteomic finding, its capacity to differentiate AD from MCI and its disease specificity require further validation in larger independent cohorts^[48]. pNfH, while well-established in motor neuron disease diagnostics, may also reflect neurodegenerative processes in dementia. In a memory clinic cohort of 188 patients, CSF pNfH and NfL (cNfL) levels were significantly elevated across diagnostic groups compared with controls, and plasma pNfH was also increased^[58].

Taken together, these findings suggest that in AD, NFs function less as disease-specific diagnostic markers and more as indicators of neurodegenerative intensity and progression. Their greatest clinical value may therefore lie in monitoring disease dynamics and in complementing core AD biomarkers within multimodal diagnostic frameworks.

PD is a common neurodegenerative disorder that predominantly affects middle-aged and elderly individuals. Its clinical features include resting tremor, bradykinesia, muscle rigidity, and postural and gait disturbances^[59]. Atypical parkinsonian syndromes (APS) are characterized by rapid disease progression, the presence of parkinsonian features accompanied by other neurological manifestations, and a poor response to conventional anti-PD medications^[33]. There is significant overlap in the clinical presentations of PD and APS, especially in the early stages of the disease, highlighting the urgent need for objective biomarkers that can differentiate these diseases based on the extent of neurodegeneration rather than clinical manifestations alone^[16].

Numerous cross-sectional and longitudinal studies have consistently shown that CSF and blood NfL levels are significantly higher in APS patients than in PD patients, with elevated levels observed across all APS subgroups^[60,61]. Receiver operating characteristic analysis revealed that the areas under the curve (AUCs) for distinguishing PD from APS using CSF and blood NfL levels reached 0.94 and 0.87, respectively^[60]. A possible mechanistic explanation is that APS is associated with faster and more extensive neurodegeneration, with more severe involvement of large-diameter myelinated axons in the subcortical brain compared to PD, and the degeneration of myelinated axons may release more NfL than neuronal soma degeneration^[59]. Furthermore, combining clinical features, neuroimaging, and biomarkers, including NfL and other more disease-specific markers such as α-syn seeding activity, may be more effective than any single modality^[61].

Currently, research on NfL as a biomarker for neurodegeneration in PD is still ongoing^[16]. Existing evidence suggests that elevated cNfL levels in certain PD subgroups may reflect a more aggressive and extensive neurodegenerative process^[60]. In the population-based New Parkinsonism in Umeå cohort, higher cNfL levels were associated with more severe PD symptoms assessed by clinical scales (e.g., Hoehn and Yahr stage and UPDRS scores) and shorter survival^[59]. In two longitudinal biomarker cohorts for PD progression, sNfL levels were elevated in PD patients compared to healthy controls and increased over time and with age^[33]. Additionally, in a cohort including PD and Parkinson’s disease dementia, both plasma NfL and cNfL levels were significantly correlated with motor dysfunction, Mini-Mental State Examination scores, and orthostatic hypotension^[61]. The relationship between NfH levels and PD clinical progression remains unclear; a post-hoc cohort study demonstrated that elevated baseline CSF NfH levels in early PD patients predicted faster motor and cognitive decline, and the association between NfH and cognitive decline was stronger than that observed for NfL^[62].

Although NfL lacks disease specificity for PD, it represents one of the most promising blood-based biomarker candidates for the disease. Further research is needed to determine whether CSF or blood NfL can be used for longitudinal monitoring of PD progression and whether combining age, more disease-specific biomarkers, and simplified clinical subtyping scoring systems can yield clinically useful prognostic tools^[63].

DLB is one of the major causes of dementia, with an incidence second only to AD and FTD^[16]. Unlike AD, there are currently no blood biomarkers that can directly reflect the underlying pathological changes in DLB, which poses a significant diagnostic challenge, especially in the early stages of the disease, when clinical symptoms overlap with other dementia syndromes^[64].

NfL is elevated in the CSF and plasma of DLB patients; however, its levels are generally comparable to those in AD patients and lower than in FTD patients, limiting its utility as a standalone tool for differential diagnosis^[65]. NfL levels are not merely passive indicators of axonal damage but also molecular signatures resulting from the combined effects of α-syn pathology, axonal transport disorders, and genetic background^[14]. Aberrant activation of calpain serves as a critical driver of neuronal injury across multiple neurodegenerative diseases. In DLB, calpain mediates site-specific cleavage of α-syn aggregation. The resulting truncated isoforms display enhanced aggregation propensity and form the core component of Lewy bodies, accelerating protein misfolding and aggregation propagation and driving dopaminergic neuron degeneration. This pathological alteration has been validated in human brain tissue specimens^[66]. Calpain also mediates broad-spectrum axonal injury and degeneration via the degradation of axonal cytoskeletal proteins: NfL, NfM and NfH, act as calpain substrates, with NfM being the most susceptible to proteolytic cleavage and undergoing degradation as early as the initial phase of calcium overload. Studies in multiple animal models of neural injury and neurodegeneration demonstrate that calpain-mediated NF hydrolysis disrupts axonal cytoskeletal integrity, subsequently triggering axonal transport dysfunction and axonal degeneration, and represents a shared downstream pathway of neuronal damage^[67].

Evidence from the Amsterdam Dementia Cohort shows that combined detection of plasma p-tau181, NfL, and GFAP can improve the differential diagnosis between AD and non-AD dementias, including FTD and DLB, with AUCs of 0.8^[64]. While NFs alone cannot resolve the diagnostic complexity of DLB, they may still play a supportive role in stratifying neurodegeneration intensity and improving differential diagnosis within emerging precision neurology frameworks.

FTD encompasses multiple clinical syndromes, including behavioral variant FTD, nonfluent variant primary progressive aphasia, and semantic variant primary progressive aphasia, each of which differs in its dominant cognitive and behavioral features^[68]. This clinical heterogeneity reflects underlying pathological and genetic diversity, with common causative mutations involving chromosome 9 open reading frame 72 (C9orf72), progranulin (PGRN), and microtubule-associated protein tau (MAPT)^[16].

This heterogeneity poses challenges for disease stratification and monitoring. As a core structural element of the neuronal axonal cytoskeleton, changes in NfL levels not only passively reflect neuronal damage but are also closely related to pathological mechanisms driven by different gene mutations^[14]. Animal models and studies of molecular mechanisms provide key evidence for this. In C9orf72-associated FTD, toxic dipeptide repeat proteins derived from hexanucleotide repeat expansions disrupt cytoskeletal homeostasis and impair axonal transport by altering microtubule-associated protein regulation and compromising NF network integrity^[13,69]. In PGRN mutations, haploinsufficiency of PGRN triggers lysosomal dysfunction and TDP-43 pathology, and TDP-43, as a binding protein for NfL mRNA, can directly affect the synthesis and homeostasis of NF proteins through its abnormal aggregation^[70]. In MAPT mutations, hyperphosphorylated tau protein detaches from microtubules, secondarily disrupting the integrity of the NFs network^[14]. These gene-specific mechanisms provide a profound explanation for the NfL kinetic characteristics observed clinically.

In one FTD-focused study, cNfL concentrations in patients were more than threefold higher than in cognitively unimpaired controls, with an increase greater than that observed in AD^[64]. This pronounced elevation may reflect the more extensive frontotemporal degeneration and subcortical involvement characteristic of FTD^[17]. A strong correlation has been observed between CSF and blood NfL levels, and similar trends have been reported in plasma, where FTD patients exhibited significantly higher plasma NfL levels than all other diagnostic groups [vs. progressive supranuclear palsy (PSP), DLB, AD, and suspected non-AD physiopathology (SNAP), and corticobasal syndrome (CBS); all P < 0.001]^[65]. Blood NfL levels in FTD patients vary depending on the underlying genetic mutations, which directly reflects the heterogeneity at the mechanistic level mentioned above^[71]. The highest levels are observed in PGRN mutation carriers, reflecting intense TDP-43-driven axonal disintegration, whereas the lowest levels are observed in MAPT mutation carriers, consistent with axonal damage secondary to Tau pathology^[72].

The magnitude of presymptomatic NfL elevation has important prognostic implications^[16]. In the GENFI and ALLFTD cohort studies, the rate of change in sNfL peaked during the prodromal phase as individuals transitioned from asymptomatic to symptomatic disease^[70,73]. Higher sNfL concentrations were consistently associated with greater disease severity and poorer performance across multiple domains, including global cognition, social behavior, language, and executive function^[71].

Nevertheless, diseases within the FTD spectrum remain highly heterogeneous, and differences between familial and sporadic FTD further complicate interpretation^[68]. Consequently, it has been proposed that NfL be used in combination with other pathophysiological biomarkers to more reliably reflect disease progression^[70]. Such multimodal approaches may improve the prediction of disease trajectories and enhance the evaluation of disease-modifying therapies in clinical trials.

Huntington’s disease (HD) is a progressive autosomal dominant inherited neurodegenerative disorder characterized by motor dysfunction, psychiatric abnormalities, and cognitive impairment. The disease is caused by a CAG trinucleotide repeat expansion within the huntingtin (HTT) gene, leading to the production of mutant huntingtin protein (mHTT)^[74]. Although various CSF biomarkers have been proposed for HD, only a few, such as mHTT, have shown clear associations with clinical phenotypes^[75].

In R6/2 mouse models expressing human mHTT, researchers observed significant increases in NfL levels in CSF and blood as neuronal damage occurred, with the magnitude of the increase directly correlating with the extent of axonal degeneration^[76]. An independent CSF validation cohort from London further demonstrated a high correlation between CSF and plasma NfL concentrations, indicating that blood measurements can serve as a reliable window into CNS damage^[77].

Multiple case-control and longitudinal studies have shown that cNfL levels are elevated in HD patients and correlate with clinical severity^[78]. However, compared to cNfL and mHTT, plasma NfL shows the strongest association with clinical disease severity^[77]. Baseline plasma NfL levels have been demonstrated to predict multiple aspects of subsequent disease progression, including regional and global brain atrophy rates, particularly in disease-related brain regions^[79]. Cognitive decline, assessed by measures such as the Unified Huntington’s Disease Rating Scale and the Symbol Digit Modalities Test, is associated with higher plasma NfL levels and poorer cognitive performance, as well as smaller brain volumes^[75]. Furthermore, by analyzing the temporal changes of various CSF markers, research found that the elevation of NfL precedes that of neuroinflammatory markers; this key finding supports the pathological model of HD: primary axonal damage (reflected by NfL) caused by the direct toxicity of mHTT is the initiating factor of the disease, while subsequent glial cell activation and neuroinflammatory responses are likely secondary responses to the initial neuronal damage^[80]. Longitudinal analysis indicates that in premanifest mutation carriers, baseline plasma NfL can independently predict the risk of disease conversion within three years, as well as future cognitive decline, multi-site brain atrophy, and ventricular enlargement^[75].

In clinical research settings, employing NfL as an outcome measure of treatment efficacy may significantly improve statistical power, potentially outperforming the use of CSF mHTT alone^[77]. In future clinical trials of disease-modifying therapies, a reduction in NfL levels post-intervention would directly indicate a protective effect on neuronal axons and serve as a sensitive surrogate marker for efficacy evaluation^[78]. There is still a lack of large, well-phenotyped cohorts that include both plasma and CSF samples. Therefore, long-term longitudinal data are needed to characterize the trajectory of these biomarkers throughout the disease process and to directly compare their value in predicting disease progression.

Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease characterized by the progressive degeneration of both upper and lower motor neurons^[81]. Average survival is only two to five years following diagnosis. Currently, early diagnosis remains limited, highlighting a critical unmet need for objective biomarkers that detect disease activity earlier^[82]. In this context, NfL has emerged as a promising candidate biomarker. Elevated NfL levels have even been proposed as susceptibility biomarkers, with increased concentrations observed in asymptomatic carriers of pathogenic ALS-associated mutations potentially reflecting an increased risk of disease onset^[28].

Proteomic analysis of autopsy spinal cord tissues from 8 patients found that although NF proteins content was significantly reduced in degenerated spinal cord tissues, their levels in CSF were significantly elevated. This phenomenon suggests that the elevation of NFs in CSF may primarily originate from the release of damaged neurons^[16]. Although NFs are not disease-specific biomarkers, meta-analyses of evidence indicate that their concentrations in both CSF and blood are generally higher in patients with ALS compared to other neurodegenerative diseases^[83]. This phenomenon may be explained by the selective vulnerability and rapid degeneration of large-caliber motor neurons, resulting in a substantial release of NF proteins into CSF and peripheral blood^[15].

NFs levels provide valuable prognostic insight into disease progression. Elevated NfL and NfH concentrations correlate with a more rapid decline in ALS Functional Rating Scale-Revised scores and reduced survival^[84]. Predictive modeling analyses by Witzel’s team have demonstrated that higher NFs levels are consistent predictors of accelerated disease progression and shorter life expectancy^[15]. While the precise onset of NFs elevation remains under investigation, current evidence suggests that NfL levels increase during early symptomatic stages before plateauing as the disease advances^[28].

To summarize, NFs elevation indicates neuroaxonal injury; it lacks disease specificity; therefore, it cannot independently establish a diagnosis of ALS^[85]. Longitudinal trajectories of NFs levels remain inconsistent across studies, potentially reflecting heterogeneity in disease onset site, progression rate, or genetic background^[11]. Despite these challenges, NFs have emerged as the most robust prognostic biomarkers currently available for ALS^[86]. The antisense oligonucleotide tofersen reduces SOD1 protein levels in the CNS, slows disease progression, and significantly decreases sNfL levels in SOD1 mouse models. In phase 3 clinical trials of patients with SOD1-mutated ALS, tofersen treatment also results in robust and sustained reductions in plasma NfL, supporting its value as a pharmacodynamic biomarker for therapeutic efficacy^[82,87]. These preclinical and clinical findings collectively establish NfL not only as a sensitive biomarker of axonal damage, but also as a pharmacodynamic indicator for verifying target engagement and therapeutic efficacy in ALS drug development. Future efforts aimed at assay harmonization, reference standardization, and integration with clinical parameters will further enhance their clinical utility^[88].