Point/Counterpoint: Neurofilament Light

The Need for a Biomarker

In the early and late stages of multiple sclerosis (MS), inflammation-associated neuronal damage, with resultant neurodegeneration, leads to eventually permanent cognitive and physical disability. An overabundance of evidence shows the number of relapses and MRI lesions—surrogate markers of inflammatory disease activity—predict long-term disability and brain volume loss. An ardent effort to monitor and mitigate inflammatory activity to prevent neurodegeneration is needed and underway.

Clinical examination and MRI are currently used to monitor for MS disease activity. These measures, however, lack sensitivity for subclinical disease activity. Although MRI is indispensible for the diagnosis of MS, it lacks sensitivity and precision due to technique variability, difficult quantification, and deficiency in capturing a more dynamic MS pathology.

Biomarkers that are more sensitive in capturing tissue damage, are cost effective, have low variability, and could be easily used in clinical practice are highly desirable. Ideally, a biomarker meeting these criteria would assess disease activity and therapeutic response as well as having a prognostic value.

Pro: The Promise of Neurofilament as a Biomarker

By Adil Javed, MD

Neurofilaments (Nfs) are a promising new biomarker for assessing axonal damage in MS from inflammatory disease activity. The Nfs are structural components of neurons and axons, consisting of 3 main isotypes: neurofilament light 68 KDa (NfL), intermediate 150 KDa (NfM), and heavy 190-210 KDa (NfH) chain.¹ Most studies of Nfs in MS and other diseases have focused on NfL and NfH. The levels of Nfs do not correlate with one another because differences in protein stability and sensitivity of assays.² Normal levels of NfLs in the serum, plasma, or cerebrospinal fluid (CSF) are thought to reflect ongoing pruning and reparative processes in the axonal/dendritic compartment of the central nervous system (CNS). Levels of Nf in the CSF and blood increase precipitously during injury. A rise in Nf levels is thought to reflect the degree of neural damage at any given point in time. Numerous studies have shown that NfL levels increase in the CSF during MS relapses and new MRI lesion development.^3-6 Because relapses and MRI lesions both correlate with long-term disability in MS, Nf levels may also be predictive of long-term disability.^7-10

Serum NfLs correlate with CSF Nf levels, and despite the much lower concentration in serum can be reliably measured by the ultrasensitive method of single-molecular array (Simoa) technology. Using this assay, serum and CSF NfLs have been shown to be highly correlated in a large cohort of MS patients, demonstrating that serum measurements can substitute CSF sampling.¹¹ These correlations remain high during disease activity, stability, and therapeutic response.¹¹ In people with MS without disease activity, serum NfL levels are normal, but increase with the number of contrast-enhancing lesions and relapses. This relationship holds up in all forms of MS, including clinically isolated syndrome (CIS), relapsing MS (RMS), and progressive MS (PMS).

A number of clinical trial designs have incorporated NfLs as an outcome measure. In post hoc analyses, NfL levels have decreased during the therapeutic response to several disease-modifying therapies (DMTs) such as natalizumab, rituximab, fingolimod, and mitoxantrone.^12-16 A subset of participants in a phase 3 trial of fingolimod showed reduced CSF NfL levels 12 months after treatment.¹⁶ Taken together, these studies underscore the importance of NfLs as a useful biomarker for assessing axonal damage in MS and the value of serum NfL levels for assessing therapeutic response to DMTs.

In summary, there is a strong evidence supporting the use of serum NfLs as an important biomarker in MS.

1. It is an independent biomarker for assessing inflamma-tory disease activity and axonal damage in all forms of MS.

2. Measurements are fairly precise through utilization of the Simoa assay.

3. Setup of the Simoa assay is fairly simple and analysis cost is low, making it readily usable in academic and private practices.

4. Sample acquisition and storage are simple.

5. NfLs levels have prognostic value and are useful in assessing therapeutic responses to DMTs, and their longitudinal measurement could be individualized in the right context. Measurements can be done as little as every 3-6 months.

6. Most importantly, a rise in NfL levels during monitoring could alert a provider to subclinical disease activity and lead to timely intervention, perhaps preventing future disability.

Con: Not Ready for Prime-Time Clinical Use

By James Stankiewicz MD, FAAN

As noted by Dr. Javed, a number of studies have been conducted that support the use of serum NFL as a surrogate therapeutic outcome measure in trial. People with RMS who have elevated levels of NfL are more likely to convert to secondary progressive MS (SPMS) and experience worse clinical and MRI outcomes.^10,17 When treatment naïve patients are started on therapy or shifted to drugs with higher efficacy, NfL levels decrease.^11,18

Nevertheless, the evidence discussed applies to populations of people with MS. For individuals, NfL is not currently an acceptable biomarker for clinical use because it not only lacks regulatory agency approval but also lacks the required clinical validity and utility that would be required for such approvals. Moreover, the assay quantifying this protein must be analytically valid: it must prove itself accurate, precise, and reproducible.^19,20 Significant variations in measurement within the same batches have been demonstrated, and serum assay results obtained by different analysis centers will need to be calibrated and made comparable.²¹ Levels of NfL are known to increase with age, yet we do not have established normal NfL ranges at different ages. Defining the variability contributed by other known confounders including possible seasonal variation and weight will also be important.^22,23 Once we as an MS community, are in possession of a reliable translatable assay and armed with true normative values accounting for the variables cited above, then we might then explore whether there is a certain NFL value above which we might alter our clinical approach.

But any clinician knows about the heterogeneity of decisions one must make. How would this apply to NfL levels? Would a 10 pg/ml NfL increase above normal healthy control values indicate that a person with MS should be switched from the treatment they are currently on regardless of efficacy? Would this same number indicate a treatment switch if the individual had secondary progressive disease rather than relapsing-remitting MS? Would this same number again be used to initiate a higher efficacy treatment in a treatment-naïve patient? Or instead should each individual serve as their own internal control? Would we use a different value in this case and what would it be? Where is good quality trial evidence to inform any of these choices? Applying any number in the above situations assumes, of course, that a cut point with reasonable sensitivity and specificity is achievable, something which I question given high variability of measurement of NFL in MS populations even in the purest situation—serum NFL checked in the presence of gadolinium enhancing lesions.²⁴

There is also the thorny issue of how often NfL should be checked. When an insult occurs how immediately do levels increase? What is the duration of the increase? Does the localization of the lesion influence the scale of NfL increase? What is the shortest time that levels will increase after an injury? The last question may be particularly pertinent for determining optimal monitoring frequency and has not yet been defined. This is not surprising since we don’t have a definite cut point for what to consider abnormal yet.

It is also important to remember that NfL is overall a measure of neuronal damage. As such it is agnostic to the type of disease process. It will never be a good diagnostic test for MS. It will also be necessary to inculcate clinicians who manage MS to strenuously consider other possible entities, especially head trauma, when NfL is elevated.

In MS, damage might result from inflammation but also might be from other noninflammatory neurodegenerative processes. Amyotrophic lateral sclerosis (ALS) and Creutzfeld-Jacob disease are neurodegenerative but present with higher NfL levels than those seen in people with MS.^25,26 It was recently reported that after autologous hematopoietic stem cell therapy NfL continued to be elevated in some despite an absence of MRI disease activity,²⁷ and individuals with elevated levels did worse clinically. It can be posited that the NfL levels in these individuals was not related to inflammatory disease but rather to neuronal death related to other processes (eg, mitochondrial dysfunction or glutamate toxicity). However, we do not currently have a definite treatment for noninflammatory damage. As such, it does not always follow that adjusting DMTs in those with elevated NfL will achieve a better outcome for progressive MS.

In summary, interpretation of serum NFL levels will be complicated. The above concerns are not insurmountable but clearly more work needs to be done. Any clinician inclined to bring this technology to the bedside currently would be ill-advised.

Rebuttal: Nonspecific Markers Can Be Useful

By Adil Javed, MD, PhD

The comments made my good colleague Dr. Stankiewicz are convincing but rather dismissive of the surmounting evidence demonstrating usefulness of NfL assays in MS. His comments can be grouped into issues regarding assay precision and accuracy, appropriate cut-off values, clinical validity, confounding comorbidities, and NfL kinetics.

The NfL assay is more accurate and precise than many others that have been examined in MS. The intra-assay variability (reproducibility) of NfL was examined in 3 native serum samples in 22 consecutive runs on independent days.²⁸ The mean coefficient of variation (CV) of duplicate runs for the 3 samples were 5.6% (13.3 pg/ml, sample 1), 6.9% (22.5 pg/ml, sample 2), and 5.3% (236.5 pg/ml, sample 3), average 5.9%. Hence, using very different samples with a wide range of NfL levels, the intrasample variation is rather small. Similarly, Interassay CV (repeatability) for serum were 11.3% (sample 1), 9.3% (sample 2), and 6.4% (sample 3), average 9%, which is also nominal.

In terms of appropriate cut-off values, NfL levels in healthy have been reported as ranging from 16 to 20 pg/ml.²⁹ In a longitudinal study of healthy individuals (median age 44.3), median NfL levels increased by 1.8 % over one year and 2.2% for each additional year.²⁹ Sex does not affect NfL levels. There appears to be a small contribution of body mass index (BMI), with a unit rise in BMI associated with a decrease of 0.02 pg/ml in serum NfL leves.³⁰ This is probably because of dilution and volume of distribution. Although the influence of age and BMI are significant, they contribute to less than 2.5% percent of the variability in NfL levels.

Clinical validity of NfL has been established in several diseases, including amyotrophic lateral sclerosis (ALS), Alzheimer disease, stroke, paraneoplastic disorders, and peripheral nervous system disorders.³¹ A good biomarker does not need to be disease specific. According to the widely accepted Prentice criteria for validation of surrogate markers, fundamental characteristics of a biomarker include ability to predict outcome at baseline, changes in levels over time are prognostic, and levels change in response to therapy—attributes that all fulfilled by examining NfL in MS.³²

Levels of NfL are closely linked to the appearance of new contrast-enhancing lesions, a time of intense inflammatory activity and damage. In those who develop new contrast-enhancing lesions, NfLs levels are increased for up to 3 months.³³ Variation in NfL levels within and between individuals with MS also correlate with increased risk of new MRI lesions. Within person increases of 10 pg/ml in NfL have a 1.48 odds of having a contrast enhancing lesion and 1.62 of new T2 lesion.³³ In response to treatment, NfL levels decline over time with levels becoming below cut-off at 3 to 6 months.³⁴

Taken together, a situation could be envisioned in which serum NfL levels are used as a C-reactive protein equivalent in MS, conveying an inflammatory response in the CNS, and triggering a provider to act accordingly. As such, NfL levels should be assessed within an appropriate context, being attentive to comorbidities, lifestyle changes, and age related variations. Individuals could serve as their own baseline, hence decreasing the between subject variability. The NfL assay can be easily setup, and site-specific quality control can be established aimed at reducing intra- and intersubject variability, as has been done already at numerous practices in the world. Based on site-specific receiver operating characteristics (ROC) analysis, sensitivity and specificity of the assay could be established, leading to appropriate cut-off values discriminating between disease stability, activity, or even clinical progression. In short, NfL assays are ready now for prime time, in the hands of an experienced and thoughtful clinician.

Rebuttal: The Picture Shows Too Little Specificity

By Dr. James Stankiewicz

I emphatically agree with Dr. Javed that serum NfL holds promise as an MS biomarker. There is an accumulation of correlative and trial evidence that supports the use of serum NfL for determining treatment impact. I differ with Dr. Javed, however, with respect to the current role of serum NfL for the treatment of individuals. The evidence that Dr. Javed marshals exists only on a population level. As a field, we currently lack good-quality evidence elucidating how serum NfL should be interpreted in individuals with MS. We simply cannot move from the observation that serum NFL on average rises in the presence of relapses or MRI activity to a biologically plausible understanding of how to apply this to individuals in particular situations. Dr. Javed’s arguments necessarily lack detail regarding levels and monitoring because one cannot reach definitive conclusions from the currently available information.

In his comments Dr. Javed writes, “NfLs levels have prognostic value and are useful in assessing therapeutic responses to DMTs, and their longitudinal measurement could be individualized.” I agree that when used in an MS cohort a certain cut-off might be determined above which on average we would expect patients to do worse long term. But a fundamental question of how certain we can be that a level about a cut point predicts a worse outcome for an individual in the context of highly variable measurements and nonstandardized assays? With the information we are currently armed with, would we be confident in choosing a stronger treatment on this cut point? The same question arises for assessing whether a current treatment is having an effect. Furthermore Dr. Javed writes “Measurements can be done as little as every 3 to 6 months on a routine basis” yet does not offer a reference in support of this time window. The current literature does not achieve consensus regarding the frequency with which one would need to obtain NfL to be confident one is not missing new disease based on NfL kinetics.

The old trope a picture in worth a thousand words applies when considering NfL for clinical use. Anyone examining the Figure, compiled from 2 of the seminal studies of NfL in MS^11,24 cannot avoid being impressed by the degree of overlap in the NfL measurements between individuals who either did or did not have gadolinium-enhancing lesions. This overlap occurs in carefully controlled studies. Given all the other caveats I have listed that will apply to real world practice what sort of grounding for treatment decisions does a thoughtful clinician currently have?