What influences if your CSF Neurofilament comes down?


Many of you will have now heard about neurofilament light chain (NfL) measurements. You may even have a record of your last reading.

A large bulk of the published literature on NfL now is on blood NfL, because it is easier to obtain (less invasive) allowing you to carry out large scale analyses. However, sampling the CSF has the added advantage that it is close to the brain and is therefore representative of what is happening there.

For a while now we have been offering the CSF NfL test clinically and found it to be a very useful in the risk-stratification newly diagnosed MS – high-activity vs. low activity and checking treatment efficacy. Others, across the UK (I’ve noted) have used it to check the risk of conversion from clinically isolated syndrome (CIS; the first clinical relapse) to relapsing-remitting MS (RRMS) and to check whether an individual with progressive disease still have active MS in them (aka smouldering MS). All of these uses have validity.

But, the work needs to be pushed further. The questions we have been grappling with in this field are as follows:

  1. Whether baseline patient characteristics (age, gender, type of MS, disease duration etc.) influenced CSF NfL. In blood NfL levels their is an association with age and BMI (thought to be related to larger circulating blood volume in this case).
  2. Whether reductions in CSF NfL (really if you’re on an anti-inflammatory therapy it should drop your NfL) were associated with the above mentioned patient characteristics.
  3. In treatments that lead to a drop in lymphocyte count (i.e. immunosuppressive therapies) does the depletion influence CSF NfL reduction.

To study this we had to look at those individuals that had a repeat CSF sampling, of which we located 61 (moderate size for this type of study); it is in this group we studied No. 1. From this 40 individuals were untreated to start with but then went on to treatment afterwards. It is in this latter group where we looked at the treatment effect of DMTs on CSF NfL (No. 2). In 37 individuals who either received alemtuzumab or sub-cutaneous cladribine (off label use) we checked whether lymphocyte depletion impacted on the subsequent CSF NfL measurement (No. 3).

Do baseline characteristics influence CSF NfL?

There was a trend for an association with sex (male vs female), but no association was found with age. We believe it’s because it’s the MS that determines what your CSF NfL is and therefore is the strongest determinant of what value you have at the end of the day. In healthy individuals without neurological disorders, however, you do see an increase in CSF NfL with ageing especially beyond the age of 60 (this is from the turnover of neurones that you see as you get older).

Do baseline characteristics influence what treatment response you see in your CSF NfL with treatment?

Sadly, it seems if you have relapsing-remitting MS vs progressive MS (see Figure below 1a), your EDSS score (Figure 1a) and your age (Figure 1a and b) influenced the degree of drop in CSF NfL that was observed after treatment.

With age for every 10 year increase on your age, the second CSF NfL measurement was 24% higher than if you were 10 years younger. For instance, in an individual with an average CSF NfL value of 702 pg/ml at the start, the predicted CSF NfL drop at second LP varied from 451 pg/mL in a 30-year-old to 228 pg/ml in a 60-year-old.

Likewise with disability level as depicted by the EDSS score for every one point increase in EDSS score the second CSF NfL measurement increased by 15% compared to someone with an EDSS score 1 point lower.

And not surprisingly, having RRMS was associated with 37% drop in CSF NfL at the follow-up LP than in Progressive MS. This maybe that there are other processes contributing to neuronal damage in progressive MS, such as neurodegeneration that are less prominent in RRMS.

Does lymphodepletion affect CSF NfL?

No, is the short answer to this. Drops in lymphocyte counts after 23 weeks following the first induction treatment course didn’t influence the ratio of CSF NfL at time point 2/time point 1. Therefore, using lymphodepletion as a measure of treatment response in so far as CSF NfL is concerned is not that useful.

Figure 1: Regression analysis and predictions based on log(CSF NfL at T2) as a function of log(CSF NfL T1) and individual patient characteristics (n = 40).
(A) A forest plot visualising point estimates and confidence intervals of the univariable regression analysis with log(CSF NfL at T2) as the outcome modelled as a function of log(CSF NfL at T1) and individual patient characteristics.
(B) Spaghetti plot visualising predictions of CSF NfL at T2 based on age and the median CSF NfL value at T1 (702 pg/mL) in the untreated/treated group for the regression model described in A. Light grey lines represent the raw CSF NfL data for people untreated at T1 and treated at T2 (n = 40)

If I am to critique our own work, I would say that more samples would be helpful. Moreover, we may be skewing the findings by those who decide to have their CSF NfL checked. But, the latter will become less of an issue as the sample size increases.


About the author

Neuro Doc Gnanapavan


    • Yes, lamotrigine has been shown to reduce neurofilament heavy chain levels. This is different to light chain levels by the way.

      • What’s the difference between heavy chains and light chains? Are these representing different types of axonal injury?

        • Light chain is more abundant than heavy chain within neurones and therefore easier to pick up as a marker of neuronal damage as per MD2 reply. Heavy chain however is more prevalent in the large myelinated axons, it’s larger mass and ergo structure determines how large a neuron is. When you have heavy chain loss in addition to light chain loss you’re potentially loosing the larger neurones. All of this work is from animal models and extrapolated to humans.

  • What does this mean?
    We found no evidence of a simvastatin treatment effect on serum neurofilaments.

    “Our results suggest that treatments capable of slowing the rate of whole brain atrophy in SPMS, such as simvastatin, may act via mechanisms largely independent of neuroaxonal injury, as quantified by Nf”

    Williams TE, Holdsworth KP, Nicholas JM, Eshaghi A, Katsanouli T, Wellington H, Heslegrave A, Zetterberg H, Frost C, Chataway J.Assessing Neurofilaments as Biomarkers of Neuroprotection in Progressive Multiple Sclerosis: From the MS-STAT Randomized Controlled Trial.Neurol Neuroimmunol Neuroinflamm. 2022 Jan 14;9(2):e1130. doi: 10.1212/NXI.0000000000001130.

    • Hope you get an answer.

      Is brain atrophy an inaccurate marker of MS progression. Looks good on MRIs but is it doing much 🤷‍♂️

    • I’d also like to point out that they looked at only the light chain and not the heavy chain, so you’re seeing a slice of the whole picture. Generally, it’s anti-inflammatory drugs that reduce NFL. For instance, even Ibudilast didn’t lower NFL. You really need to understand biomarkers before you start measuring them in clinical trials!



Recent Posts

Recent Comments