Zipping the DMT pyramid: What’s your sNfL Z-score? 


Neurofilament light (NfL) levels are a promising biomarker of disease activity in MS. These brain proteins are released into the cerebrospinal fluid when neurons die and they tell us when MS-related focal inflammation is still damaging the brain or spinal cord. The biggest advantage of NfL over other techniques to measure brain volume loss such as brain MRI is that they can be measured in the peripheral blood. However, the biggest hurdle to bring NfL from “bench to bedside” has always been that neurons also die in people without MS. This makes the distinction between really normal – normal – slightly increased – really increased brain volume loss very difficult. To complicate matters even further, the distinction between normal and abnormal becomes even more shady when growing older (unfortunately in everyone > 50 years old) and when having a high body weight (dilution of NfL). In short, not even Moses could split the NfL sea into two.  


Researchers from Swiss have now addressed this issue by studying a behemoth cohort of people without apparent neurological disease (4.532 samples) and by no longer looking at absolute NfL cut-offs based on percentiles (eg. 80th percentile is the value at which 20% of the NfL values lie above that value and 80% below). Instead, they calculated how much a specific NfL value deviates in terms of distance and direction from the mean NfL value at a certain age and when having a certain BMI. This results into a so called Z-score and ranged from 0, 0.5, 1.0, 1.5 to 2.0. 

In pwMS, both NfL cut-offs and Z scores indicated a gradually increased risk for future acute (eg. relapse and new brain lesions) and chronic (disability worsening) disease activity. However, increased Z scores outperformed absolute raw sNfL cutoff values when predicting whether a relapse had taken place in the last 4 months or when predicting signs of clinical/radiological disease activity in the following year. A sNfL Z-score above 1.5 resulted in a more than threefold increased risk of future relapses or new brain lesions in all pwMS and more than two and a half fold risk increase in people with no evidence of disease activity (NEDA). 

Skipping the ‘bedside’ part, these reference values have been – thankfully – developed into an internet-based app  which is freely available, and allows you to calculate Z-scores given you are aware of your serum NfL value, age and BMI. Ideally, we would flip the DMT pyramid and reduce brain volume loss to a maximal extent in all pwMS. For everyone with doubts for various reasons to climb to the top of the pyramid, we can now say ‘Zip it!’. Examples of Zipping the pyramid in practice:

  • Newly diagnosed pwMS with a Z-score of more than 1.5 could be advised to start on a second line instead of first line MS therapy
  • PwMS treated with platform or oral MS therapies with a persistent Z-score of more than 1.5 under treatment could be advised to switch to a second line therapy
  • PwMS treated with immune reconstitution therapies (IRT) who have an increasing Z-score after completing treatment could be advised to consider retreatment
  • PwMS who experience progression of their symptoms without relapses or new brain lesions and a sNfL Z-score of more than 1.5 could be advised to start treatment.

Curious about your Z-score? Ask your consultant neurologist during the next clinic visit.

Screenshot of the sNfL Shiny App for calculation of sNfL percentile and Z-score values.

Twitter: @SmetsIde

Disclaimer: Please note that the opinions expressed here are those of dr. Ide Smets and do not necessarily reflect the position of the Barts and The London School of Medicine and Dentistry nor Barts Health NHS Trust.

Lancet Neurol 2022 Mar;21(3):246-257. doi: 10.1016/S1474-4422(22)00009-6.

Serum neurofilament light chain for individual prognostication of disease activity in people with multiple sclerosis: a retrospective modelling and validation study

Pascal Benkert, Stephanie Meier, Sabine Schaedelin, Ali Manouchehrinia, Özgür Yaldizli, Aleksandra Maceski, Johanna Oechtering, Lutz Achtnichts, David Conen, Tobias Derfuss, Patrice H Lalive, Christian Mueller, Stefanie Müller, Yvonne Naegelin, Jorge R Oksenberg, Caroline Pot, Anke Salmen, Eline Willemse, Ingrid Kockum, Kaj Blennow, Henrik Zetterberg, Claudio Gobbi, Ludwig Kappos, Heinz Wiendl, Klaus Berger, Maria Pia Sormani, Cristina Granziera, Fredrik Piehl, David Leppert, Jens Kuhle, NfL Reference Database in the Swiss Multiple Sclerosis Cohort Study Group

  • PMID: 35182510
  • DOI: 10.1016/S1474-4422(22)00009-6


Background: Serum neurofilament light chain (sNfL) is a biomarker of neuronal damage that is used not only to monitor disease activity and response to drugs and to prognosticate disease course in people with multiple sclerosis on the group level. The absence of representative reference values to correct for physiological age-dependent increases in sNfL has limited the diagnostic use of this biomarker at an individual level. We aimed to assess the applicability of sNfL for identification of people at risk for future disease activity by establishing a reference database to derive reference values corrected for age and body-mass index (BMI). Furthermore, we used the reference database to test the suitability of sNfL as an endpoint for group-level comparison of effectiveness across disease-modifying therapies. Methods: For derivation of a reference database of sNfL values, a control group was created, comprising participants with no evidence of CNS disease taking part in four cohort studies in Europe and North America. We modelled the distribution of sNfL concentrations in function of physiological age-related increase and BMI-dependent modulation, to derive percentile and Z score values from this reference database, via a generalised additive model for location, scale, and shape. We tested the reference database in participants with multiple sclerosis in the Swiss Multiple Sclerosis Cohort (SMSC). We compared the association of sNfL Z scores with clinical and MRI characteristics recorded longitudinally to ascertain their respective disease prognostic capacity. We validated these findings in an independent sample of individuals with multiple sclerosis who were followed up in the Swedish Multiple Sclerosis registry. Findings: We obtained 10 133 blood samples from 5390 people (median samples per patient 1 [IQR 1-2] in the control group). In the control group, sNfL concentrations rose exponentially with age and at a steeper increased rate after approximately 50 years of age. We obtained 7769 samples from 1313 people (median samples per person 6·0 [IQR 3·0-8·0]). In people with multiple sclerosis from the SMSC, sNfL percentiles and Z scores indicated a gradually increased risk for future acute (eg, relapse and lesion formation) and chronic (disability worsening) disease activity. A sNfL Z score above 1·5 was associated with an increased risk of future clinical or MRI disease activity in all people with multiple sclerosis (odds ratio 3·15, 95% CI 2·35-4·23; p<0·0001) and in people considered stable with no evidence of disease activity (2·66, 1·08-6·55; p=0·034). Increased Z scores outperformed absolute raw sNfL cutoff values for diagnostic accuracy. At the group level, the longitudinal course of sNfL Z score values in people with multiple sclerosis from the SMSC decreased to those seen in the control group with use of monoclonal antibodies (ie, alemtuzumab, natalizumab, ocrelizumab, and rituximab) and, to a lesser extent, oral therapies (ie, dimethyl fumarate, fingolimod, siponimod, and teriflunomide). However, longitudinal sNfL Z scores remained elevated with platform compounds (interferons and glatiramer acetate; p<0·0001 for the interaction term between treatment category and treatment duration). Results were fully supported in the validation cohort (n=4341) from the Swedish Multiple Sclerosis registry. Interpretation: The use of sNfL percentiles and Z scores allows for identification of individual people with multiple sclerosis at risk for a detrimental disease course and suboptimal therapy response beyond clinical and MRI measures, specifically in people with disease activity-free status. Additionally, sNfL might be used as an endpoint for comparing effectiveness across drug classes in pragmatic trials.

About the author

Ide Smets


  • V informative IDE. Pity I am prehistoric! I assume you need to persuade the neurologist you need an LP. If you ever get to see one, to use the app.

    • No, that’s the good part. The app calculates Z-scores based on serum/blood NfL. And most MS centres nowadays have the option to send your blood externally to quantify sNfL. In the Netherlands this is reimbursed by the health care insurance. In Barts-MS I think it’s also possible. So at least in theory, the Z-scores are ready for use in practice.

      • Thanks for clarifying. I am hoping to get onto Prof G’s exercise study so would have an LP for that, but meanwhile, the app can give me what I need.

  • When do you think it will be mainstream for people with MS as part of a 3,6 ,12 monthly blood screening? Especially for newly diagnosed or on a new treatment.

    • That’s a difficult question, because quantifying sNfL and knowing your Z-score does not mean neurologists know how to act upon the result. I expect sNfL to become mainstream in blood work for newly diagnosed pwMS in the next 5 years.

  • I raised this by the consultant neurologist I ‘see’ (ie don’t see face-to-face any more) a little while ago and he said he thought it was a waste of time as whatever the results were it wouldnt change his diagnosis of inactive secondary progressive MS or his treatment options, ie none. He also said there was no facility to examine the samples in the local trust (*********) and so it would mean paying a London lab to do it – and there was no money to do that either.
    It’s really frustrating to be part of the postcode lottery which means we can’t access developing techniques like this and other therapies (eg Sativex isnt available in this Trust due to budget constraints) whilst in other parts of the country people can.

    • Interesting, and frustrating. To be truly inactive, Z-score should be low. If there’s no clinical progression that’s most likely to be the case. But for pwMS who feel they are deteriorating I do think it would be useful to make the distinction. We expect that about 10% of inflammatory disease activity cannot be detected by MRI or clinical symptoms.

      • Thank you. I really feel I need to find someone more proactive to assist me. I was diagnosed in 2012 on the basis of one MRI and lumbar puncture. Since then I have had one more MRI in 2018 which led to/confirmed the ‘clincally inactive’ diagnosis. But in the last 10 years I have deteriorated steadily from EDSS 4 to 6.5 so there is clearly something still going on. Maybe it is simply ‘preprogrammed’ nerve loss but it would ‘nice’ to at least have that confirmed or otherwise.

      • I’m on Ocrelizumab and in Australia, my neuro also refuses to prescribe sNfl as he doesn’t see it would be necessary. I’m happy to pay for the tests so I can be more sure about disease activities..

By Ide Smets



Recent Posts

Recent Comments