Introduction
Cognitive impairment is one of the most disabling symptoms of multiple sclerosis (MS), significantly hampering day-to-day functioning [
1]. In an effort to monitor cognitive functioning in MS, understanding its underlying neurobiological correlates is of utmost importance. To date, most studies investigated magnetic resonance imaging (MRI) characteristics in relation to cognitive performance, which has taught us that cognitive impairment is associated with neurodegeneration such as cortical and deep grey matter atrophy [
2,
3], as well as with functional impairment of neuronal networks [
4]. However, the clinical implementation of these prognostic biomarkers is limited, as these markers cannot fully account for the large heterogeneity found between people with MS (PwMS) [
5]. The complex pathology of MS, including inflammation, demyelination and neurodegeneration warrants a multimodal biomarker linking both molecular and imaging biomarkers [
6].
Recent studies focused on the combination of neurofilament light chain levels in serum (sNfL) and conventional imaging markers (e.g., lesion load and grey matter volume) for predicting cognitive functioning in PwMS [
7,
8]. NfL reflects the major intermediate cytoskeletal protein of axons and is considered to be a marker for neuro-axonal damage [
9]. Indeed, increased levels of NfL in the serum and cerebrospinal fluid (CSF) of PwMS have been related to cognitive impairment [
10] and decreased performance on multiple cognitive domains [
7,
11], in various disease stages [
8,
12], showing promising predictive value over time [
10]. Another potentially interesting biomarker for the assessment of neurodegeneration in MS is glial fibrillary acidic protein (GFAP), the intermediate cytoskeletal protein of astrocytes [
13,
14]. Both serum and CSF levels of GFAP have been shown to relate to disease type (i.e., increased levels of GFAP in progressive PwMS [
15]) and disease severity (i.e., increased GFAP levels were associated with higher physical disability and longer disease durations [
13,
16]). However, the association between GFAP and cognitive functioning in MS has yet to be established.
The aim of current study was to compare, confirm and combine (bio)markers of neurodegeneration (i.e., serum and CSF levels of both NfL and GFAP and conventional imaging markers) for its role in cognition in a clinical sample of PwMS that visited our outpatient clinic because of perceived cognitive complaints.
Discussion
This study investigated the relation of NfL and GFAP measured in serum and CSF and cognitive performance in PwMS presenting with cognitive complaints, and their added predictive value compared to conventional imaging markers. Based on the levels of sNfL and sGFAP we were able to distinguish cognitively preserved from cognitively impaired PwMS, albeit with limited diagnostic accuracy. Increased levels of both serum NfL and GFAP were observed in cognitively impaired PwMS compared to cognitively preserved PwMS. NfL levels (in serum and CSF) were inversely associated with processing speed, indicating that decreased processing speed was associated with increased levels of sNfL and cNfL. No correlations could be detected between GFAP (measured in either serum or CSF) and cognitive functioning in PwMS. Finally, sNfL added unique variance in the prediction of cognitive status on top of NGMV. A composite score of both measures (a multimodal marker) resulted in a fair classification of cognitive status, stressing the need for a multimodal approach when predicting cognitive functioning.
Consistent with previous literature, increased levels of sNfL were found for cognitively impaired PwMS [
7,
10,
11]. Furthermore, increased levels of sNfL and cNfL were associated with reduced processing speed. Slowed processing speed, has been hypothesized to be the major driver of cognitive impairment in MS [
1], thereby possibly explaining why correlations with this specific domain are more prevalent in studies investigating NfL and cognitive functioning in MS [
10,
12]. Yet, mixed results have been reported for increased levels of NfL and the performance in other cognitive domains [
11]. Differences in sample size, administered neuropsychological tests, study population (i.e., a focus on newly diagnosed PwMS [
12] or SPMS [
8], a combination of MS types, or PwMS with mild cognitive impairment [
39]) and specific focus on treatment are most likely explaining these differences [
40]. In our sample, the distribution of PwMS on DMT at the time of the visit (but also the distribution of PwMS on first-line DMT vs. second-line DMT) was similar between cognitive groups, thereby reducing the likelihood of impacting our findings. Nonetheless, it could have played a role on an individual level as has been shown before [
41]. Although it was beyond the scope of current research, the impact of DMTs on cognitive functioning in MS warrants further investigation [
1]. Finally, although levels of sGFAP were increased in cognitively impaired PwMS, no correlations between GFAP and the cognitive domains survived correction for multiple comparisons, hereby limiting its potential as a clinical biomarker for cognitive functioning in MS.
Jakimovski et al. demonstrated in two previous studies a relatively weaker correlation between sNfL and cognition [
10], compared to correlations between sNfL and MRI outcomes [
42]. As potential explanation they put forward the role of adaptive processes to significantly influence the relationships between the released NfL and cognitive test results. Subsequently, PwMS who demonstrate preserved functional connectivity, despite ongoing structural pathology, can maintain high levels of cognitive performance [
43]. In the current study, associations between sNfL and imaging markers were absent. However, associations of cNfL and sNfL between both processing speed and between cNfL and imaging markers were comparable in effect size with previous studies, thereby confirming aforementioned difference [
10,
42]. Interestingly, in our study, imaging markers displayed a higher number of associations with multiple cognitive domains (not only processing speed, but also verbal and visuospatial memory) compared to fluid NfL and GFAP levels, highlighting that structural pathology was present and related to several cognitive test scores. As fluid biomarkers provide a real-time evaluation of the amount of pathology compared to the less dynamic imaging markers [
9], it can be hypothesized that cognitive changes are not resulting from acute disturbances but rather from a more global effect over time on the brain in certain areas.
When added to the model, sNfL improved the prediction of cognitive status compared to NGMV alone. Especially when combining biomarkers, in our case NGMV and sNfL (the “multimodal marker”) a large effect was found for processing speed, whereas medium effects were reported for verbal and visuospatial memory. Even a medium sized effect for EF-verbal fluency was found when using the multimodal marker, which was absent when investigating individual markers. The current study is one of the first studies to combine both neuroimaging and fluid biomarkers of interest to detect cognitive impairment in MS. Investigating the role of a multimodal marker for cognitive functioning in PwMS is of high importance since these different modalities might reflect different aspects of neurodegeneration, which also has been reported in Alzheimer’s disease [
44] and recently in MS as well [
7,
45]. More specifically, previous studies investigating cross-modal fluid and imaging (bio)markers indeed show an “additive” effect of sNfL compared to cortical thickness [
45] or lesion load and grey matter volume [
7] in recently diagnosed PwMS. Together with our results, the added effect of sNfL highlights the necessity of using multiple sources of information to create a diagnostic marker for something as highly complex as cognitive performance, but also how these markers of neurodegeneration cannot be used interchangeably.
Nonetheless, clinical interpretation may be optimized when the full prognostic potential of sNfL for cognitive functioning will be evaluated over time, which is an important limitation of the current cross-sectional study design. The inclusion of a control group would have further aided the disentanglement between normal and abnormal levels of fluid and imaging (bio)markers. Also, contrary to measurements in serum, both NfL and GFAP measured in CSF were unable to discriminate between cognitive status. The most plausible explanation for this lack of detecting a difference is the limited power (
N CSF = 54 versus
N = 78 for serum). Performing a lumbar puncture is rather invasive and not all PwMS wanted to partake in this procedure. Importantly, without a post-contrast sequence being available in current study protocol, it was not possible to determine whether PwMS had active lesions at the time of evaluation. As a consequence, the investigation of the effect of recent disease activity on serum and CSF levels was limited and could be considered an important avenue for future research. Finally, the inclusion of a clinical, real-life sample is one of the biggest strengths, as the PwMS are reflective of our population at the outpatient clinic with perceived cognitive complaints. At the same time, being a real-life sample is also one of the main limitations. A homogenous sample is often desirable when investigating differences between groups, although data on other types of MS than RRMS is often lacking. Furthermore, given the fact that PwMS visited the outpatient clinic because of cognitive complaints, a slight bias towards cognitive impairment may have been present. The main clinical aim of the outpatient clinic is to investigate whether these complaints (or impairments) are due to MS pathology or, for instance, psychological or social factors (known to influence cognitive performance [
46]). Results on PROMS measuring mood, anxiety, fatigue and sleep were, therefore, reported in this manuscript showing similar scores between cognitively preserved and impaired PwMS. Consequently, the impact of these factors on cognition was also considered similar.
In conclusion, we provided novel insights into the relationship between fluid biomarkers of neurodegeneration and their relation to cognitive functioning and conventional imaging measures in PwMS. The main finding of this study is the result that sNfL explains additional variance in cognitive performance on top of NGMV. A novel insight that was further explored in our study was the potential for combining two (bio)markers from a different origin when predicting cognitive status, instead of focusing on single measures of NfL or imaging outcomes. Combining multimodal biomarkers may be the way forward to enable timely identification of cognitive decline in MS.
Declarations
Conflicts of interest
The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: M.v.D. is supported by a research grant from BMS. I.M.N. is supported by the Dutch MS Research Foundation, grant nr. 15-911. M.H. is supported by the Dutch MS Research Foundation, grant nr. 16-954b. J.J.G.G. has served as a consultant for or received research support from Biogen, Celgene, Genzyme, MedDay, Merck, Novartis and Teva. B.M.J.U. reports personal fees for consultancies from Biogen Idec, Genzyme, Merck Serono, Novartis, Roche, and Teva, outside the submitted work. C.E.T. has a collaboration contract with ADx Neurosciences and Quanterix, performed contract research or received grants from AC-Immune, Axon Neurosciences, Biogen, BioOrchestra, Brainstorm Therapeutics, Celgene, EIP Pharma, Eisai, Grifols, Novo Nordisk, PeopleBio, Quanterix, Roche, Toyama, Vivoryon. She serves on editorial boards of Alzheimer Research and Therapy, and Neurology. H.E.H. serves on the editorial board of Multiple Sclerosis Journal, receives research support from the Dutch MS Research Foundation and the Dutch Research Council. She has served as a consultant for or received research support from Atara Biotherapeutics, Biogen, Novartis, Celgene/Bristol Meyers Squibb, Sanofi Genzyme, MedDay and Merck BV. B.A.d.J., E.A.W., M.K., B.M., and S.d.G.D. report no disclosures relevant to the manuscript.