Association of Plasma Neurofilament Light
With Neurodegeneration in Patients With Alzheimer Disease
Niklas Mattsson, MD, PhD; Ulf Andreasson, PhD; Henrik Zetterberg, MD, PhD; Kaj Blennow, MD, PhD; for the
Alzheimer’s Disease Neuroimaging Initiative
IMPORTANCE
Existing cerebrospinal fluid (CSF) or imaging (tau positron emission
tomography) biomarkers for Alzheimer disease (AD) are invasive or expensive. Biomarkers
based on standard blood test results would be useful in research, drug development, and
clinical practice. Plasma neurofilament light (NFL) has recently been proposed as a
blood-based biomarker for neurodegeneration in dementias.
OBJECTIVE To test whether plasma NFL concentrations are increased in AD and associated
with cognitive decline, other AD biomarkers, and imaging evidence of neurodegeneration.
DESIGN, SETTING, AND PARTICIPANTS In this prospective case-control study, an ultrasensitive
assay was used to measure plasma NFL concentration in 193 cognitively healthy controls, 197
patients with mild cognitive impairment (MCI), and 180 patients with AD dementia from the
Alzheimer’s Disease Neuroimaging Initiative. The study dates were September 7, 2005, to
February 13, 2012. The plasma NFL analysis was performed in September 2016.
MAIN OUTCOMES AND MEASURES Associations were tested between plasma NFL and
diagnosis, Aβ pathologic features, CSF biomarkers of neuronal injury, cognition, brain
structure, and metabolism.
RESULTS Among 193 cognitively healthy controls, 197 patients with mild cognitive
impairment, and 180 patients with AD with dementia, plasma NFL correlated with CSF NFL
(Spearman ρ = 0.59, P < .001). Plasma NFL was increased in patients with MCI (mean, 42.8
ng/L) and patients with AD dementia (mean, 51.0 ng/L) compared with controls (mean, 34.7
ng/L) (P < .001) and had high diagnostic accuracy for patients with AD with dementia vs
controls (area under the receiver operating characteristic curve, 0.87, which is comparable to
established CSF biomarkers). Plasma NFL was particularly high in patients with MCI and
patients with AD dementia with Aβ pathologic features. High plasma NFL correlated with
poor cognition and AD-related atrophy (at baseline and longitudinally) and with brain
hypometabolism (longitudinally).
CONCLUSIONS AND RELEVANCE Plasma NFL is associated with AD diagnosis and with
cognitive, biochemical, and imaging hallmarks of the disease. This finding implies a potential
usefulness for plasma NFL as a noninvasive biomarker in AD.
JAMA Neurol. 2017;74(5):557-566. doi:10.1001/jamaneurol.2016.6117
Published online March 27, 2017.
Editorial page 510
Supplemental content
Author Affiliations: Author
affiliations are listed at the end of this
article.
Group Information: A list of the
Alzheimer’s Disease Neuroimaging
Initiative investigators is given at the
end of this article.
Corresponding Author: Niklas
Mattsson, MD, PhD, Clinical Memory
Research Unit, Department of Clinical
Sciences, Faculty of Medicine, Lund
University, Simrisbanvägen 14,
SE-21224 Malmö, Sweden (niklas
.mattsson@med.lu.se).
Research
JAMA Neurology | Original Investigation
(Reprinted) 557
Downloaded From: by a University College London User on 12/12/2018
A
lzheimer disease (AD) is a neurodegenerative disease
that is characterized by brain accumulation of β-amy-
loid (Aβ) and tau, progressive atrophy, and cognitive de-
cline. Biomarkers that capture biological processes in AD are
increasingly used to support the diagnosis of AD in research,
drug development, and clinical practice.
1
The most well-
established AD biomarkers include structural magnetic reso-
nance imaging (MRI); cerebrospinal fluid (CSF) biomarkers of
Aβ, tau, and neuronal injury; and positron emission tomo-
graphic imaging of Aβ, tau, and brain metabolism.
2-4
Use of these
biomarkers is hampered by a high degree of invasiveness, high
costs, or limited availability.
5
Blood-based biomarkers for AD
may allow for efficient monitoring of disease processes in AD
and could be used as a screening tool in primary care. One po-
tential blood-based biomarker for AD is the neuronal injury
marker neurofilament light (NFL)
6
because patients with AD
have increased CSF concentrations of NFL.
7
Results from some
studies
8,9
suggest that patients with AD have increased plasma
NFL concentrations. However, those studies were performed
using standard immunoassay techniques with suboptimal
analytical sensitivity to accurately quantify low abundant
brain-specific proteins in blood samples.
10
For this reason,
our group has recently transferred the CSF NFL assay to an
ultrasensitive single-molecule array (Simoa; Quanterix Cor-
poration) platform, which provides an analytical sensitivity
of 0.6 pg/mL compared with 78.0 pg/mL for the correspond-
ing enzyme-linked immunosorbent assay (ELISA).
10
Plasma
NFL concentrations can be measured in all samples using
the ultrasensitive single-molecule array and correlate closely
with the corresponding CSF concentrations.
11,12
Herein, we
test this novel plasma NFL assay in patients with AD for
the first time, to our knowledge. We studied cognitively
healthy control individuals, patients with mild cognitive
impairment (MCI) (MCI group), and patients with AD demen-
tia (AD group) in a large prospective study. We tested the
hypotheses that the plasma NFL concentration is increased
in AD and that it correlates with impaired cognition, neuro-
imaging abnormalities, and CSF biomarkers of AD pathologic
features.
Methods
ADNI Study Design
Data were obtained from the Alzheimer’s Disease Neuroim-
aging Initiative (ADNI) database (http://adni.loni.usc.edu). The
ADNI was launched in 2003 as a public-private partnership,
led by principal investigator Michael W. Weiner, MD (the most
recent information on the ADNI is available at http://www
.adni-info.org). The ADNI participants have been recruited from
more than 50 sites across the United States and Canada. For
the present study, we used data accessed from the Laboratory
of Neuro Imaging (University of Southern California) ADNI
database on October 6, 2016. The study data and samples were
collected from September 7, 2005, to February 13, 2012.
Regional ethical committees of all participating institutions
approved the ADNI. All study participants provided written
informed consent.
ADNI Participants
Our ADNI cohort consisted of all cognitively healthy controls,
patients with MCI, and patients with AD dementia with avail-
able baseline plasma NFL samples from the ADNI-1. Inclu-
sion and exclusion criteria were described in detail previously.
13
Briefly, all ADNI-1 participants were aged 55 to 90 years, had
completed at least 6 years of education, were fluent in Span-
ish or English, and had no substantial neurological disease
other than AD. Controls had Mini-Mental State Examination
(MMSE) scores of 24 or higher, where lower scores indicate
more impairment and higher scores less impairment (range,
0-30), and a Clinical Dementia Rating (CDR) score of 0, where
lower scores indicate less impairment and higher scores
more impairment (range, 0-3). Patients with MCI had MMSE
scores of 24 or higher, objective memory loss tested by
delayed recall of the Wechsler Memory Scale (WMS) logical
memory II (>1 SD below the normal mean), a CDR score of
0.5, preserved activities of daily living, and absence of
dementia. Patients with AD dementia fulfilled the National
Institute of Neurological Communicative Disorders and
Stroke–Alzheimer Disease and Related Disorders Association
criteria for probable AD,
14
had MMSE scores of 20 to 26, and
had CDR scores of 0.5 to 1.0.
Plasma NFL
Plasma NFL concentrations were measured using an NFL kit
(NF-light; UmanDiagnostics), transferred onto the ultrasensi-
tive single-molecule array platform using a home brew kit
(Simoa Homebrew Assay Development Kit; Quanterix Corpo-
ration), as previously described.
15
In the 14 analytical runs
needed to complete the study, the relative error of the back-
calculated concentrations was below 20% for all calibrators,
run in triplicate, resulting in lower limits of quantifications of
2.2 ng/L and upper limits of quantification of 1620 ng/L. All
samples measured within the range spanned by the limits of
quantifications, and for the low-concentration quality con-
trol sample (14 ng/L), the intra-assay coefficient of variation
was 11.0% and the interassay coefficient of variation was 11.1%.
For the high-concentration quality control sample (137 ng/L),
the corresponding coefficients of variation were 8.8% and
9.6%, respectively. The measurements were performed in Sep-
tember 2016 by a board-certified laboratory technician using
a single batch of reagents.
Key Points
Question What is the importance of plasma neurofilament light in
Alzheimer disease?
Findings In this case-control study of 193 cognitively healthy
controls, 197 patients with mild cognitive impairment, and 180
patients with Alzheimer disease dementia, plasma neurofilament
light was associated with Alzheimer disease and correlated with
future progression of cognitive decline, brain atrophy, and brain
hypometabolism.
Meaning Plasma neurofilament light may be a promising
noninvasive biomarker for Alzheimer disease.
Research Original Investigation Plasma Neurofilament Light and Neurodegeneration in Alzheimer Disease
558 JAMA Neurology May 2017 Volume 74, Number 5 (Reprinted) jamaneurology.com
Downloaded From: by a University College London User on 12/12/2018
CSF Measurements
Cerebrospinal fluid was sampled by lumbar puncture from a
subset of the participants, with CSF Aβ42, CSF total tau
(t-tau), and CSF phosphorylated tau (p-tau) measured using a
multiplex platform (xMAP; Luminex Corporation) with a kit
(INNO-BIA AlzBio3; Fujirebio Europe). Participants were clas-
sified as Aβ positive or Aβ negative using a previously estab-
lished cutoff (CSF Aβ42 < 192 ng/L).
16
We excluded 8 patients
with AD dementia who were Aβ negative and therefore likely
to be misdiagnosed. The CSF NFL concentrations were mea-
sured using a commercial ELISA (NF-light; UmanDiagnos-
tics) and have been reported previoushy.
7
In total, we in-
cluded CSF data from 112 controls, 189 patients with MCI, and
90 patients with AD dementia.
Cognition
Cognition was assessed by MMSE, Alzheimer Disease Assess-
ment Scale–cognitive subscale (ADAS-Cog 11), delayed recall
of the WMS logical memory II, Trail-Making test part B (TMT-
B), and Wechsler Adult Intelligence Scale–Revised (WAIS-R)
digit symbol substitution test. All tests were administered at
baseline and at 6, 12, 18, 24, 36, and 48 months, except for de-
layed recall of the WMS logical memory II, which was not as-
sessed at 18 months.
Neuroimaging
Structural brain images were acquired using 1.5-T MRI imaging
systems with T1-weighted MRI scans using a sagittal volumet-
ric magnetization-prepared rapid acquisition gradient echo
sequence (at baseline and at 6, 12, 18, 24, 36, and 48 months).
A software program (FreeSurfer; https://surfer.nmr.mgh
.harvard.edu/) was used for quantification of cortical thick-
ness and subcortical volumes.
17
We used volumetric data for
hippocampal volume and lateral ventricles (averaged be-
tween right and left sides). We used the mean cortical thick-
ness for a set of regions defined a priori based on work by Jack
et al
18
to represent AD cortex (including entorhinal, inferior
temporal, middle temporal, and fusiform cortex).
White matter hyperintensities (WMHs) were quantified at
baseline and at 6, 12, 18, 24, 36, and 48 months using a fully
automated protocol.
19
Positron emission tomography with 18F-
fluorodeoxyglucose image data were acquired at baseline and
at 6, 12, 18, 24, 36, and 48 months.
20
We created mean counts
of the lateral and medial frontal, anterior, and posterior cin-
gulate regions, as well as lateral parietal and lateral temporal
regions.
Statistical Analysis
We tested associations between plasma NFL and demo-
graphic factors using the Kruskal-Wallis test and Spearman rank
correlation. We tested associations between biochemical mark-
ers and between plasma NFL and diagnosis using linear re-
gression models. We calculated diagnostic accuracies using area
under the receiver operating characteristic curve (AUROC)
analysis with 10-fold cross-validated logistic regression mod-
els. We tested associations between plasma NFL concentra-
tions and longitudinal cognition, brain structure, and brain me-
tabolism using linear mixed-effects models. These models had
random intercepts and slopes for time and an unstructured co-
variance matrix for the random effects and included the in-
teraction between (continuous) time and plasma NFL as pre-
dictor. All outcome variables in linear mixed-effects models
were standardized to z scores to facilitate comparisons be-
tween modalities. Therefore, β coefficients refer to standard-
ized effects (β = 1 implies that an increase of 1 ng/L in plasma
NFL was associated with a 1-SD increase in the dependent
variable).
All tests were 2-sided. Statistical significance was set at
P < .05. All regression analyses were corrected for age, sex,
educational level, diagnosis, and APOE ε4 genotype, as well
as intracranial volume for hippocampus and ventricles. All sta-
tistical analyses were performed using a software program (R,
version 3.2.3; The R Foundation).
Results
Table 1 lists demographics for the study population. In the
whole cohort, plasma NFL correlated with age (Spearman
ρ=0.35,P < .001) but not with sex (median, 36.2 ng/L for
men vs 37.4 ng/L for women; P = .98), educational level
(ρ = −0.03, P = .52), or APOE ε4 genotype (37.7 ng/L in carri-
ers vs 35.6 ng/L in noncarriers, P = .19). These results were
similar within diagnostic groups, except that plasma NFL
concentrations were higher in APOE ε4 carriers in the MCI
group (35.9 ng/L in carriers vs 39.3 ng/L in noncarriers,
Table 1. Demographics for the Study Population
a
Variable
Controls
(n = 193)
MCI
(n = 197)
AD Dementia
(n = 180) P Value
Age, mean (SD), y
75.9 (4.9) 74.7 (7.5) 75.3 (7.3) .58
Female, No. (%)
87 (45.1) 65 (33.0) 86 (47.8) .007
Educational level, mean (SD), y
16.0 (2.9) 15.8 (3.0) 14.7 (3.1) <.001
APOE ε4 genotype carriers, No. (%)
50 (25.9) 103 (52.3) 123 (68.3) <.001
Plasma NFL, mean (SD), ng/L
34.7 (21.4) 42.8 (29.0) 51.0 (26.9) <.001
MMSE score, mean (SD)
29.1 (1.0) 26.9 (1.8) 23.2 (2.1) <.001
CSF Aβ42, mean (SD), ng/L
207 (52) 165 (52) 134 (23) <.001
Aβ+, No./total No. (%)
41/112 (36.6) 138/189 (73.0) 90/90 (100) <.001
CSF t-tau, mean (SD), ng/L
68 (29) 102 (60) 126 (56) <.001
CSF p-tau, mean (SD), ng/L
25 (15) 36 (19) 44 (20) <.001
Abbreviations: Aβ, β-amyloid;
AD, Alzheimer disease dementia;
CSF, cerebrospinal fluid;
MCI, mild cognitive impairment;
MMSE, Mini-Mental State
Examination; NFL, neurofilament
light; p-tau, phosphorylated tau;
t-tau, total tau.
a
P values are from the Kruskal-Wallis
test or Fisher exact test. β-Amyloid
positivity was defined as CSF Aβ42
less than 192 ng/L.
Plasma Neurofilament Light and Neurodegeneration in Alzheimer Disease Original Investigation Research
jamaneurology.com (Reprinted) JAMA Neurology May 2017 Volume 74, Number 5 559
Downloaded From: by a University College London User on 12/12/2018
P = .049) and in the AD dementia group (41.6 ng/L in carri-
ers vs 51.5 ng/L in noncarriers, P = .03).
Plasma NFL and Other Biochemical Markers
in CSF and Plasma
Plasma NFL correlated with high CSF NFL (Spearman ρ = 0.59,
P < .001) (Figure 1) and with low CSF Aβ42, high CSF t-tau, high
CSF p-tau, and high plasma tau (Table 2). The strongest cor-
relations were seen with CSF NFL, which were also present in
all diagnostic groups. Plasma NFL also correlated with CSF
Aβ42 and CSF t-tau in the MCI group, as well as with plasma
tau concentrations in all diagnostic groups.
Plasma NFL in Different Diagnostic Groups
Plasma NFL concentrations were higher in the AD group
compared with controls and the MCI group, as well as in the
MCI group compared with controls (Figure 2A). Plasma NFL
differentiated between the AD dementia group and controls,
with an AUROC of 0.87 (Figure 2B). By comparison, the
AUROCs were 0.87 to 0.90 for CSF NFL, CSF Aβ42, CSF t-tau,
and CSF p-tau and 0.78 for plasma tau. These AUROCs were
corrected for age, sex, educational level, and APOE ε4 geno-
type. When only correcting for age, sex, and educational
level, the AUROCs were reduced to 0.79 for plasma NFL, 0.81
for CSF NFL, 0.85 for CSF t-tau, 0.81 for CSF p-tau, and 0.64
for plasma tau.
Plasma NFL and Aβ Pathologic Features
We compared plasma NFL between Aβ-negative controls, Aβ-
positive controls, Aβ-negative patients with MCI, Aβ-positive
patients with MCI, and (Aβ-positive) patients with AD demen-
tia (Figure 2C). The AD dementia group had higher plasma NFL
than Aβ-negative controls (mean, 48.8 vs 33.9 ng/L; P < .001),
Aβ-positive controls (mean, 30.9 ng/L; P < .001), Aβ-negative
MCI (mean, 38.1 ng/L; P < .001), and Aβ-positive MCI (mean,
44.5 ng/L; P = .05). There were no statistically significant dif-
ferences between Aβ-negativeand Aβ-positive controls and Aβ-
negative patients with MCI.
Plasma NFL and Progressive vs Stable MCI
Among the MCI group, 109 converted to AD dementia dur-
ing follow-up, and 65 remained stable after at least 2 years’
follow-up. Twenty-three patients in the MCI group did not
convert to AD dementia during follow-up but were observed
for less than 2 years and were not included in the stable
group. There was no difference in plasma NFL between
Aβ-positive patients with progressive MCI and Aβ-positive
patients with stable MCI, but both of these groups had
higher plasma NFL than Aβ-negative patients with progres-
sive MCI and Aβ-negative patients with stable MCI
(Figure 2D).
Plasma NFL and Cognition and Neuroimaging
Associations between plasma NFL and longitudinal cogni-
tive and imaging measures are shown in Figure 3 (coeffi-
cients and P values are listed in the eTable in the Supple-
ment). At baseline, high plasma NFL levels were associated
with worse MMSE, ADAS-COG 11, and TMT-B scores and with
larger ventricular volume, smaller hippocampal volume, and
thinner cortices in the AD cortex region. Over time, high
plasma NFL levels were associated with an accelerated
decline in all measures, except for WMHs. The strongest
influences were seen in MMSE (β = −0.073, P < .001 baseline
and β = −0.116, P < .001 longitudinally) and ADAS-COG11
scores (β = 0.101, P < .001 baseline and β = 0.106, P <.001
longitudinally) for the cognitive measures and in AD
Figure 1. Plasma Neurofilament Light (NFL) and Cerebrospinal Fluid
(CSF) NFL
9.5
9.0
8.5
8.0
7.5
7.0
6.5
6.0
1 65
Log CSF NFL, ng/L
Log Plasma NFL, ng/L
3 42
r
=
0.59
P
<.001
Controls
MCI
AD dementia
Fit lines are shown for individual diagnostic groups. The Spearman ρ and P
values are for Spearman rank correlation in the whole cohort. Table 2 lists
correlation data adjusted for covariates. AD indicates Alzheimer disease; MCI,
mild cognitive impairment.
Table 2. Correlations Between Plasma NFL and Other Biochemical Markers
a
Biomarker All Participants Controls MCI AD Dementia
β Coefficient P Value β Coefficient P Value β Coefficient P Value β Coefficient P Value
CSF NFL 0.480 <.001 0.371 <.001 0.615 <.001 0.242 .046
CSF Aβ42 −0.144 .01 0.002 .99 −0.243 <.001 −0.017 .87
CSF t-tau 0.125 .01 0.123 .21 0.170 .01 0.033 .76
CSF p-tau 0.105 .03 0.068 .49 0.113 .10 0.112 .31
Plasma tau 0.178 <.001 0.238 <.001 0.138 .03 0.188 .008
Abbreviations: AD, Alzheimer disease; CSF, cerebrospinal fluid; MCI, mild
cognitive impairment; NFL, neurofilament light; p-tau, phosphorylated tau;
t-tau, total tau.
a
Data are β coefficients (with P values) from linear regression models for
correlations between plasma NFL and other biomarkers (all standardized
to z scores), adjusted for age, sex, educational level, APOE ε4 genotype, and
diagnosis. Models were tested in the whole cohort and in individual diagnostic
groups.
Research Original Investigation Plasma Neurofilament Light and Neurodegeneration in Alzheimer Disease
560 JAMA Neurology May 2017 Volume 74, Number 5 (Reprinted) jamaneurology.com
Downloaded From: by a University College London User on 12/12/2018
cortex (β = −0.162, P < .001 baseline and β = −0.049,
P < .001 longitudinally) for the imaging measures (details are
provided in the eTable in the Supplement).
We also tested whether the influence of plasma NFL
differed between diagnostic groups. Statistically significant
interactions were found at baseline for the MCI group
and MMSE score (β = −0.107, P = .008), ADAS-COG 11 score
(β = 0.150, P = .003), delayed recall of the WMS logical
memory II (β = −0.167, P < .001), TMT-B score (β = 0.283,
P < .001), WAIS-R digit symbol substitution test score
(β = −0.253, P = .002), and AD cortex (β = −0.282, P < .001),
as well as for the AD group and MMSE score (β = −0.133,
P = .002), ADAS-COG 11 score (β = 0.171, P = .001), delayed
recall of the WMS logical memory II (β = −0.117, P = .01), and
AD cortex (β = −0.178, P = .04). These results indicate that
plasma NFL was more strongly correlated with the out-
comes in the MCI group and the AD dementia group than in
controls at baseline. Longitudinally, the only statistically
significant interaction was for the MCI group and MMSE
score (β = −0.107, P = .003), demonstrating that plasma NFL
Figure 2. Plasma Neurofilament Light (NFL) by Diagnosis and Aβ
6
5
4
3
2
1
Log Plasma NFL, ng/L
Plasma NFL by diagnosis
A
Controls AD Dementia
Diagnostic Group
1.0
0.8
0.6
0.4
0.2
0
True-Positive Rate
AUROC in AD dementia vs controls
B
1.0 0.8 0.6 0.4 0.2 0
False-Positive Rate
MCI
P
<.001
P
<.001
P
<.001
6
5
4
3
2
1
Log Plasma, ng/L
Plasma NFL by diagnosis and Aβ
C
Aβ– Controls Aβ– MCI Aβ+ MCI Aβ+ AD Dementia
Diagnostic Group
Aβ+ Controls
P
<.001
P
<.001
P
<.001
P
<.001
P
<.001
P
=
.05
P
<.001
CSF t-tau (AUROC, 0.90)
Plasma NFL (AUROC, 0.87)
Plasma tau (AUROC, 0.78)
CSF Aβ42 (AUROC, 0.88)
CSF p-tau (AUROC, 0.87)
CSF NFL (AUROC, 0.89)
6
5
4
3
2
1
Log Plasma NFL, ng/L
Plasma NFL by SMCI or PMCI and Aβ
D
Aβ– With SMCI Aβ+ With SMCI Aβ– With PMCIAβ– With PMCI
Diagnostic Group
P
=
.02
P
=
.008
P
=
.01
P
=
.002
A, Plasma NFL in controls, patients with mild cognitive impairment (MCI), and
patients with Alzheimer disease (AD) dementia. B, Area under the receiver
operating characteristic curve (AUROC) analyses for plasma NFL and other
biomarkers to differentiate between the AD dementia group and controls. CSF
indicates cerebrospinal fluid. C, Plasma NFL in controls, patients with MCI, and
patients with AD dementia, stratified by occurrence of Aβ positivity (CSF
Aβ42 < 192 ng/L). D, Plasma NFL in patients with stable MCI (SMCI) (no
progression to dementia during ⱖ2 years’ follow-up) and patients with
progressive MCI (PMCI) (conversion to dementia), with or without Aβ positivity.
The models were adjusted for age, sex, educational level, and APOE ε4
genotype. Age retained an independent statistically significant association with
higher plasma NFL (β = 0.025, P < .001 from the models shown in A). AB-
indicates AB-negative; AB+, AB-positive.
Plasma Neurofilament Light and Neurodegeneration in Alzheimer Disease Original Investigation Research
jamaneurology.com (Reprinted) JAMA Neurology May 2017 Volume 74, Number 5 561
Downloaded From: by a University College London User on 12/12/2018