Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Feb 1.
Published in final edited form as: Neurobiol Aging. 2021 Nov 1;110:27–36. doi: 10.1016/j.neurobiolaging.2021.10.017

Cerebrospinal Fluid Dynamics and Discordant Amyloid Biomarkers

Jonathan Graff-Radford a, David T Jones a, Heather J Wiste b, Petrice M Cogswell c, Stephen D Weigand b, Val Lowe c, Benjamin D Elder d,e, Prashanthi Vemuri c, Argonde Van Harten a,1, Michelle M Mielke b, David S Knopman a, Neill R Graff-Radford f, Ronald C Petersen a, Clifford R Jack Jr c, Jeffrey L Gunter g
PMCID: PMC8758540  NIHMSID: NIHMS1752904  PMID: 34844077

Abstract

Objective:

Do MRI-based metrics of a CSF-dynamics disorder, disproportionately enlarged subarachnoid-space hydrocephalus (DESH), correlate with discordant amyloid biomarkers (low CSF β-amyloid 1–42, normal Aβ-PET scan)?

Methods:

Individuals ≥50 years from the Mayo Clinic Study of Aging, with MRI, 11C-Pittsburgh compound B (Aβ) PET scans, and CSF phosphorylated-tau protein and Aβ42, were categorized into four groups: normal/ abnormal by CSF β-amyloid 1–42 and Aβ amyloid PET. Within groups, we noted MRI patterns of CSF-dynamics disorders; categorization into low-CSF β-amyloid 1–42/normal Aβ-PET scan versus normal-CSF and PET groups; and Aβ-PET accumulation-change rate.

Results:

100 participants (21%) in the abnormal-CSF/normal-PET group had highest DESH-pattern scores and lowest CSF-phosphorylated-tau levels. Among normal amyloid-PET individuals, a one-unit DESH-pattern score increase correlated with 30%-greater odds of abnormal amyloid CSF after age and sex adjustment. Mean rate over time of amyloid-PET accumulation in abnormal-CSF/normal-PET individuals approximated individuals with normal amyloid values. Adjusting for phosphorylated-tau, abnormal CSF-amyloid/normal amyloid-PET individuals had higher mean amyloid-PET accumulation rates than normal individuals.

Conclusions:

CSF dynamics disorders confound β-amyloid and phosphorylated-tau CSF-biomarker interpretation.

Keywords: Alzheimer’s disease, biomarkers, normal-pressure hydrocephalus, Mayo Clinic Study of Aging (MCSA), amyloid PET accumulation, disproportionately enlarged subarachnoid-space hydrocephalus (DESH)

1.0. INTRODUCTION

Low CSF β-amyloid 1–42 (Aβ4211) indicates β-amyloid accumulation, and increased retention of amyloid-tracer deposition on PET directly measures cortical β-amyloid load in Alzheimer’s disease (AD).(Jack et. al., 2018) Discordance of CSF Aβ42 and amyloid PET biomarkers occurs in 17% to 21% of cognitively unimpaired individuals, (Mattsson et. al., 2015, Vos et. al., 2016) which challenges clinical-trial recruitment focused on early interventions. Possibly, CSF and PET discordance occurs because low CSF Aβ42 and normal amyloid PET represent early AD below the detection level of PET (Mattsson et al., 2015, J. Reimand et. al., 2019). Concordance improves in individuals with symptomatic AD or by using the Aβ42 to β-amyloid 1–40 (Aβ40) ratio, but a minority of results remain discordant (Leuzy et. al., 2016).

As in AD, individuals with normal-pressure hydrocephalus (NPH) have low CSF Aβ42 levels. In contrast to AD, NPH patients have decreased p-tau and total tau protein (t-tau) levels (Agren-Wilsson et. al., 2007, Jeppsson et. al., 2013). However a limited number of studies show Aβ42 levels normalize 6 months after shunt surgery, suggesting that low Aβ42 is related to the CSF dynamics disorder (CDD) rather than cortical amyloidosis (N. R. Graff-Radford, 2014).

Disproportionately enlarged subarachnoid space hydrocephalus (DESH) is the combination of high-convexity tight sulci, defined as sulci compressed at the vertex, enlarged CSF spaces in the Sylvian fissure, and ventriculomegaly. DESH is an established neuroimaging phenotype detected on structural MRI of a CDD and described in NPH (Hashimoto et. al., 2010). Although DESH was defined in NPH, we recently demonstrated that this pattern occurred in 6.6% of the general population, using an automated MRI algorithm to detect DESH (J. Graff-Radford et. al., 2019, Gunter et. al., 2019).

Other population-based studies have also shown that DESH features are more common with advancing age and occur more frequently than previously believed among persons without obvious symptoms (Hiraoka et. al., 2008, Jaraj et. al., 2014). Because abnormal CSF dynamics can result in low Aβ42 levels and CDD prevalence may be more common than previously realized, we hypothesized that a subset of discrepant CSF Aβ42 and amyloid PET cases (CSF+, PET−) are attributable to altered CSF dynamics. Because p-tau increases monotonically with age and increasing AD pathology, those with low Aβ42 and normal amyloid PET scans should have normal or slightly elevated p-tau if they represent early AD pathophysiology.

2.0. MATERIAL AND METHODS

2.1. Aim

We sought to determine whether CDD on MRI correlated with the presence of discrepant CSF and PET amyloid biomarkers.

2.2. Participants

The MCSA is a population-based study of cognitive decline; details of the study design have been published previously (Roberts et. al., 2008). Using an age- and sex-stratified random-sampling design, the MCSA enumerated residents of Olmsted County, Minnesota, via the Rochester Epidemiology Project health records–linkage system (St Sauver et. al., 2012). The current study included MCSA participants age ≥50 years who had undergone structural MRI, amyloid PET imaging, and CSF evaluation. The first visit with all three measures was considered the baseline visit for this study.

2.3. Cognitive Evaluation

A study coordinator and physician evaluated participants, who underwent neuropsychological examination. A consensus panel (physician, study coordinator, neuropsychologist) then classified participants as cognitively impaired (mild cognitive impairment or dementia) or unimpaired (Roberts et al., 2008). The diagnosis of cognitively impaired was established by previously published criteria (Petersen, 2004, Roberts et al., 2008).

2.4. Other Covariates

Demographic information was assessed by interview and electronic health-record review (St Sauver et al., 2012).

2.5. CSF Measurements

The full details of CSF collection and validation have been recently published (Van Harten et. al., 2020) Fasting lumbar punctures were performed in the morning in the lateral decubitus position from the L3 and L4 intravertebral space using a 20 or 22 gauge Quincke needle. CSF was collected in polypropylene tubes and centrifuged. Two cc were used to evaluate routine markers (glucose, protein, cell count). The remainder was divided into 0.5 cc aliquots and stored at −80 °C.

All samples were thawed once before analysis. CSF Aβ42, t-tau, and hyperphosphorylated tau (p-tau181) were analyzed using Elecsys (Lenexa, KS) Aβ (1–42) CSF, Elecsys total tau CSF, and Elecsys phosphorylated-tau (181P) CSF electrochemiluminescence immunoassays (Roche Diagnostics, Basel, Switzerland). Based on the study by Van Harten (Van Harten et al., 2020) we used a cut point of ≤1,026 pg/mL to define CSF Aβ42 abnormality and ≥0.023 to define abnormality for the CSF p-tau181-to −Aβ42 ratio.

2.6. Neuroimaging

2.6.1. Magnetic Resonance Imaging

Magnetization-prepared rapid gradient-echo (MPRAGE) images were acquired with consistent imaging parameters and imaging performed with 3T GE MRI scanners (GE Healthcare). Complete acquisition details are available elsewhere (Jack et. al., 2008).

MPRAGE images were analyzed with probabilistic segmentation using SPM12 with the Mayo Clinic Adult Lifespan Template (MCALT)(Schwarz et. al., 2017) population-optimized priors and settings (https://www.nitrc.org/projects/mcalt/). Deformations mapping the template grayscale image to each subject scan was done by Advanced Normalization Tools (Avants et. al., 2008) toolkit applied to two atlas parcellations: one with 122 tissue regions (MCALT_ADIR122) and another with 123 named sulcal regions (adapted from BrainVISA [http://brainvisa.info]).

2.6.2. Structural MRI Metric for Measuring CDD

We developed a previously described, fully automated imaging pipeline for detecting DESH (Gunter et al., 2019). In brief, the algorithm begins with the SPM12 CSF-segmentation output probability in each sulcal region. Manual grading for DESH previously completed by an expert clinician was the standard against which a machine-learning algorithm was trained (Gunter et al., 2019). DESH-predictive CSF regions were chosen by analyzing area under the receiver-operating characteristic curve (AUROC). A support-vector machine model was then trained on the regions selected. Using the model, the methodology classified DESH scans with AUROC >0.96 with respect to the expert evaluations (Gunter et al., 2019). This machine-learning pipeline was previously validated on a dataset separate from the training dataset (Gunter et al., 2019) with a second validation using the automated technique with visual grading of DESH by a board-certified neuroradiologist blinded to the machine-learning categorization. The machine-learning had 100% sensitivity and 91% specificity relative to the expert visual grading (J. Graff-Radford et al., 2019). The algorithm measure is referred to herein as the “DESH pattern score.” The use of a support-vector machine sets the output to return negative scores if no pattern is found. Increasingly positive scores reflect a better match. For analyses with a dichotomized pattern-matching score, an abnormal cut point for the DESH pattern score was set at >1.0 (J. Graff-Radford et al., 2019, Gunter et al., 2019).

2.6.3. Positron Emission Tomography

Amyloid PET imaging was performed with Pittsburgh compound B (PiB). Details of PET acquisition have been published previously (Jack et. al., 2008, Jack et. al., 2017, Knopman et. al., 2012). An amyloid PET meta–region of interest was computed as the voxel-number weighted average of median-PiB uptake in the prefrontal, orbitofrontal, parietal, temporal, anterior, posterior cingulate, and precuneus regions normalized by median uptake in the cerebellar crus grey matter. Abnormal amyloid was defined as PiB standard uptake value ratio (SUVR) ≥1.48.

2.7. Statistical Analysis

Individuals were categorized into four groups based on abnormality status on CSF Aβ42 and amyloid PET using the cut points described above: normal on both measures (CSF−PET−), abnormal CSF but normal PET (CSF+PET−), normal CSF but abnormal PET (CSF−PET+), and abnormal on both (CSF+PET+).

Three different models evaluated associations between CSF Aβ42, amyloid PET, and DESH pattern score. First, to evaluate associations between DESH pattern score (outcome) and the CSF/PET amyloid groups (predictors), we used quantile regression of the median adjusting for age, sex, and total intracranial volume. Next, because the CSF+PET− group was of particular interest, among the normal amyloid PET subset of individuals, we used logistic-regression models to assess which variables had greater odds of being in the CSF+PET− versus CSF−PET− group. Finally, linear mixed-effects models with log-transformed amyloid PET SUVR as the outcome were fit among all individuals to assess different amyloid PET accumulation rates by the CSF/PET amyloid groups. Age at baseline, sex, and total intracranial volume were included as predictors with terms for association with the baseline amyloid PET measure and with the rate of change in amyloid PET (i.e., variable × time interaction). The models included a per-person intercept and slope as random effects. Because the outcome (amyloid PET) was log-transformed, coefficient estimates were interpreted as the approximate annual-percent SUVR change. All three models were fit with and without CSF p-tau as a predictor.

Analyses were performed using R language (The R Foundation) and environment for statistical computing version 3.6.2. We used quantreg package version 5.54 for the quantile-regression models, and the nlme package version 3.1–145 for the mixed models.

3.0. RESULTS

3.1. Characteristics of Participants

This study included 471 MCSA participants ≥50 with results for CSF, MRI, and amyloid PET (Table 1).

Table 1.

Demographic characteristics of MCSA participants by CSF and amyloid PET statusa

Characteristic All (N=471) CSF−PET− (n=217) CSF+PET− (n=100) CSF−PET+ (n=36) CSF+PET+ (n=118)
Age, y 72 (63 to 79) 69 (61 to 76) 65 (59 to 73) 78 (72 to 82) 77 (72 to 84)
 Min, max 50, 95 50, 94 51, 92 55, 90 58, 95
Male sex, No. (%) 265 (56) 128 (59) 54 (54) 17 (47) 66 (56)
Education, y 14 (12 to16) 14 (12 to16) 15 (13 to17) 14 (12 to17) 14 (12 to16)
Clinical diagnosis, No. (%)
 Cognitively unimpaired 406 (86) 197 (91) 93 (93) 31 (86) 85 (72)
 Mild cognitive impairment 59 (13) 19 (9) 7 (7) 4 (11) 29 (25)
 Dementia 6 (1) 1 (0) 0 (0) 1 (3) 4 (3)
ApoE ε4 carrier, No. (%) 138 (29) 42 (20) 28 (28) 6 (17) 62 (53)
Amyloid PET, SUVR 1.39 (1.33 to 1.56) 1.35 (1.30 to 1.39) 1.35 (1.30 to 1.40) 1.53 (1.50 to 1.62) 1.97 (1.70 to 2.36)
CSF Aβ42, pg/mL 1,063 (746 to 1,516) 1,520 (1226 to 1,701) 820 (688 to 919) 1,278 (1,196 to 1,455) 607 (472 to 764)
CSF p-tau181, pg/mL 18 (14 to 24) 19 (16 to 23) 13 (11 to 14) 21 (18 to 27) 23 (18 to 31)
DESH score −1.71 (2.87 to −0.57) −1.98 (−3.23 to −1.03) −1.38 (−2.24 to 0.06) −1.89 (−3.55 to −0.54) −1.61 (−2.81 to −0.53)
 Abnormal,b No.(%) 31 (7) 9 (4) 10 (10) 4 (11) 8 (7)
Follow-up PET imaging, No. (%) 331 (70) 159 (73) 72 (72) 27 (75) 73 (62)
First to last amyloid PET scan, y 4.7 (2.7 to 6.2) 4.8 (2.7 to 6.1) 4.9 (2.7 to 6.1) 3.8 (2.5 to 6.0) 4.5 (2.5 to 6.3)
 Min, max 0.9, 10.0 1.1, 10.0 1.0, 8.8 1.2, 9.0 0.9, 9.4
Open in a new tab

Abbreviations: DESH, disproportionately enlarged subarachnoid space hydrocephalus; IQR, interquartile range; max, maximum; MCSA, Mayo Clinic Study of Aging; min, minimum; SUVR, standardized uptake value ratio.

a

Data are expressed as median (IQR) unless otherwise indicated.

b

A DESH score >1 was considered abnormal.

3.2. Relationship between Biomarkers of Amyloid and DESH Score

We established associations between CSF Aβ42 and amyloid PET with points colored by DESH score (Fig 1). CSF and PET quadrants were defined using our previously described abnormality cut points. Of note, 100 participants (21%) had abnormal CSF but normal amyloid PET (lower left quadrant, CSF+PET−). Histograms show distribution of CSF Aβ42 and amyloid PET values in the population. In the lower left quadrant, a higher DESH score was overrepresented in the CSF+PET− group.

Figure 1. Scatter plot of CSF Aβ42 pg/mL versus amyloid PET SUVR colored by DESH score.

Figure 1.

Open in a new tab

The black line shows the estimated median CSF Aβ42 for a given amyloid PET SUVR based on quantile regression. Histograms of the CSF Aβ42 and amyloid PET distributions are shown in the margins. Cut points of 1.48 SUVR for amyloid PET and 1,026 pg/mL for CSF Aβ42 are shown with straight black lines. The number (%) of individuals in each quadrant based on these cut points is shown. DESH indicates disproportionately enlarged subarachnoid space hydrocephalus; SURV, standardized uptake value ratio.

Median DESH scores were highest among the CSF+PET− quadrant (Fig 2A). After adjusting for age, sex, and total intracranial volume using quantile regression, median DESH scores were 0.7 (95% CI, 0.4–1.1), 1.1 (95% CI, 0.6–1.6), and 0.6 (95% CI, 0.2–1.1) units higher in the CSF+PET− group than in the CSF−PET−, CSF−PET+, and CSF+PET+ groups, respectively. However, the CSF+PET− group had the lowest median CSF p-tau of any group including the normal group on both amyloid measures (CSF−PET−; Fig 2B). A 20% lower p-tau score correlated with 0.15 (95% CI, 0.07–0.23) higher median DESH score. Adjusting for p-tau, median DESH scores were 0.5 (95% CI, 0.1–0.8) and 0.5 (95% CI, 0.1–1.4) higher in the CSF+PET− group than in the CSF−PET− and CSF−PET+ groups but were similar to the CSF+PET+ group (estimate: 0.2 [95% CI, −0.3 to 0.7]).

Figure 2. Boxplots of DESH score and CSF p-tau distribution.

Figure 2.

Open in a new tab

A, The horizontal line at 1 indicates an abnormal/positive DESH score or the presence of DESH-like imaging features among the four CSF Aβ42 and amyloid PET quadrants. The number (%) of individuals with abnormal DESH scores in each quadrant is shown above the boxplots. B, The boxplot shows the CSF p-tau distribution by the CSF Aβ42 and amyloid PET quadrants.

A logistic-regression model further evaluated which variables increased inclusion odds in the amyloid CSF+PET− group compared with the CSF−PET− group (Fig 3). Holding age and sex constant, a one-unit increase in the DESH score correlated with 1.3 (95% CI, 1.2–1.5) greater odds of discordance on CSF and PET amyloid biomarkers (Fig 3A). When including p-tau in the model (Fig 3B), 20% lower p-tau score correlated with 3.2 (95% CI, 2.4–4.3) greater odds of being in the CSF+PET− group versus the CSF−PET− group, while the DESH score effect was somewhat attenuated (OR, 1.2 [95% CI, 1.0–1.4]).

Figure 3: Summary of a logistic regression model used to determine odds of inclusion in the CSF+PET− group.

Figure 3:

Open in a new tab

A, Odds ratios with 95% CI with abnormal (low) CSF Aβ42 as the outcome and age, sex, and DESH score as predictors. The model was fit among individuals with normal amyloid PET (<1.48 SUVR) and can therefore be interpreted as which variables increase the odds of membership in the CSF+PET− versus the CSF−PET− group. B, Summary of odds ratios (95% CI) from a model that includes log-transformed CSF p-tau as a predictor.

3.3. Longitudinal Amyloid Accumulation

The mean amyloid PET-accumulation rate was 1.2% (95% CI, 0.4%−1.9%) for the CSF− PET+ group and 2.1% (1.5%−2.7%) per year higher for the CSF+PET+ group than the CSF+PET− group, but the rate in the CSF+PET− group did not significantly differ from the CSF−PET− group (estimate: −0.2% [95% CI, −0.6% to 0.3%]; Fig 4A). A 20% increase in CSF p-tau was associated with a 0.1% (95% CI, 0.05%−0.3%) higher annual amyloid-accumulation rate. Adjusting for p-tau gave the CSF+PET− group a higher mean rate of amyloid PET accumulation than the CSF−PET− group (estimate: 0.5%/year [95% CI, 0% to 1.0%]). DESH score was not significantly associated with the amyloid PET-accumulation rate (Fig 4B).

Figure 4: Amyloid Accumulation.

Figure 4:

Open in a new tab

A, Estimated mean difference (95% CI) in the rate of amyloid PET accumulation from a linear mixed effects model with log-transformed amyloid PET as the outcome and age, sex, CSF/PET quadrant, and DESH score as predictors. Rates of accumulation are summarized as approximate annual percent change in amyloid PET. B, Summary of a model that includes log-transformed CSF p-tau as a predictor. The CSF/PET quadrant data are summarized as the contrast of the CSF+PET− group compared with the other three quadrants.

Using the CSF p-tau ratio to Aβ42 inverted the notion of abnormality along the vertical axis: abnormal CSF ratio but normal amyloid PET is, therefore, in the upper left quadrant (Fig 5). This change left only 16 participants (3%) in this group. Using this ratio, 415 (88%) were concordant. The association with DESH score appears to dissipate with use of the CSF ratio compared with CSF Aβ42 in isolation.

Figure 5: Scatter plot of CSF p-tau/Aβ42 versus amyloid PET SUVR colored by DESH score.

Figure 5:

Open in a new tab

The black line shows the estimated median CSF p-tau/Aβ42 for a given amyloid PET SUVR based on quantile regression. Histograms of the CSF p-tau/Aβ42 and amyloid PET distributions are shown in the margins. Cut points of 1.48 SUVR for amyloid PET and 0.023 for CSF p-tau/Aβ42 are shown with black lines. The number (%) of individuals in each quadrant based on these cut points is shown.

4.0. DISCUSSION

The key findings of this paper are: 1) MRI pattern of CDD correlated with discrepant biomarkers for amyloid (CSF+PET−); 2) Low CSF p-tau also discriminated the amyloid CSF+PET− group from other biomarker pairings; 3) A ratio of CSF p-tau/Aβ42 attenuated the CDD effect on discrepant biomarkers; 4) After adjusting for baseline CSF p-tau, higher amyloid-accumulation rates were seen in the CSF+PET− versus the CSF−PET− groups.

These results suggest three reasons for discrepant CSF and amyloid PET biomarkers (CSF+PET−): 1) CSF Aβ42 abnormality precedes an amyloid PET abnormality and, therefore, this group will become PET+; 2) A subset of individuals are low-protein producers (i.e., all protein levels are a scale lower); and 3) Individuals with an imaging phenotype consistent with CDD have artificially low CSF Aβ42 levels but similar progression rates to PET+ than those initially CSF−PET−.

4.1. CSF+PET−, representing early amyloid accumulators

Much prior literature emphasized that CSF+PET− represents a stage of amyloid accumulation below the threshold for amyloid PET detection. In Alzheimer’s Disease Neuroimaging Initiative, CSF+PET− individuals had greater amyloid accumulation on PET than CSF−PET− individuals (S. Palmqvist et. al., 2016, Sebastian Palmqvist et. al., 2017). Similarly, initially normal amyloid PET individuals who converted to abnormal on follow-up imaging had lower levels of CSF Aβ42 than those remaining at normal levels (Vlassenko et. al., 2016) Our data suggest that early amyloid accumulation does not completely explain the CSF+PET− discordant group.

The CSF+PET− group had lower CSF p-tau levels, suggesting that some proportion of this group are not discordant because CSF Aβ42 is a leading amyloid indicator on the AD pathway; in this case, CSF p-tau should not be lower than in the control CSF−PET− group (Fig 2B). Adjusting for p-tau level decreased the percentage of participants with discordant CSF Aβ42 and PET markers. Lower CSF p-tau and Aβ42 levels suggest other possible mechanisms: A subset of low-protein producers or CSF dynamics causing all proteins to appear low (N. R. Graff-Radford, 2014). Although early amyloid accumulation does not fully explain results, some evidence suggests that a subset of CSF+PET− are early amyloid accumulators. In the current study, after adjusting for CSF p-tau level, the smaller resulting CSF+PET− (ratio-discordant) group had marginally higher amyloid accumulation levels than the CSF–PET− group, consistent with the leading-indicator hypothesis. Discordance due to early amyloid accumulation, however, does not explain the pathophysiology in all individuals in the CSF+PET− quadrant.

Adding p-tau into the model attenuated the CDD effect and low-protein producer status. Adjusting for these factors, the residual effect was the “third-reason” group—CSF as a leading indicator of amyloid accumulation—stood out because the subset with higher baseline CSF p-tau had higher amyloid-accumulation rates (i.e., the AD pathway group). Therefore, those with discrepant amyloid CSF+PET− on the AD pathway can be identified as those with relatively higher p-tau levels despite the overall amyloid CSF+PET− quadrant having lower p-tau levels.

4.2. CSF+PET−, representing low-protein producers

Both low and high CSF Aβ42 may be attributable to total Aβ-peptide concentrations rather than an AD pathologic state (Wiltfang et. al., 2007). Therefore, using Aβ42 in isolation would result in a subset of individuals not on the AD pathway being incorrectly diagnosed as “Alzheimer’s disease pathologic change” and a subset with high Aβ42 values diagnosed as normal despite actually having “Alzheimer’s disease pathologic change.” A ratio corrects for total Aβ concentration and, therefore, provides a more accurate reflection of AD pathologic state.

In the current study, adjusting for p-tau attenuated DESH pattern score, which indicates that other factors, such as low protein production, could induce this biomarker pattern and supports using a ratio rather than Aβ42 value alone. However, low-protein producers cannot fully explain individuals with abnormally low p-tau levels in the CSF+ PET− quadrant. If they did, then normalization of low β-amyloid (Aβ) and p-tau would not occur after shunting.

4.3. CSF+PET−, representing altered CSF dynamics

Supporting the idea that not all individuals with abnormal CSF and normal amyloid PET scans are on the AD pathway, a recent study showed that five years post-baseline amyloid testing, 18F-flortaucipir PET uptake in the CSF+PET− group did not differ from the CSF−PET− group (Juhan Reimand et. al., 2020). The level of CSF Aβ42 in AD is low, whereas t-tau and p-tau are elevated. In a CDD like NPH, however, many protein levels, including Aβ42, t-tau, and p-tau, are low but normalize after shunt placement (N. R. Graff-Radford, 2014, Jeppsson et al., 2013, Kapaki et. al., 2007). In this study, a higher DESH score correlated with greater likelihood of CSF+PET− status. If low Aβ42 were due to a CDD resembling NPH, we would expect a subset of the CSF+PET− group to have low p-tau also. In contrast, if the CSF+PET− group were all on the AD pathway, the p-tau level should be normal or slightly increased. We saw lower p-tau levels in the CSF+PET− group, supporting CDD as a cause of discrepant amyloid biomarkers. We have previously shown that PET amyloid is unassociated with CDD (J. Graff-Radford et al., 2019); therefore, CDD is not related to PET− status.

4.5. AD biomarker interpretation in the setting of CDD

This is not the first study to observe that AD-biomarker interpretation can be confounded by CDD. Recently, individuals with CDD were reported to be more commonly characterized as having abnormal neurodegeneration on the basis of an abnormal AD-signature cortical thickness despite lower tau burden, indicating that the abnormal-signature cortical thickness could not be explained by tauopathy (J. Graff-Radford et al., 2019). They were overrepresented in the suspected non–AD pathophysiology group. Therefore, converging evidence suggests that identifying CDD is necessary for accurate biomarker interpretation in AD, both for CSF biomarkers and structural MRI. Despite lower levels of AD CSF biomarkers in CSF dynamics disorders, the CSF p-tau to Aβ42 ratio correlates with brain AD pathology and is a better indicator than one biomarker interpreted in isolation (Elobeid et. al., 2015). Additionally, reference limits adjusted for neuroimaging findings of CSF dynamics disorders may be appropriate.

Biomarkers are increasingly used to determine clinical-trial eligibility and to measure outcomes—thus, the importance of recognizing how CDD affects these biomarkers. For example, if a clinical trial aims to enroll cognitively unimpaired individuals and include those at early stages of amyloid accumulation, CSF Aβ42 in isolation may lead to erroneous enrollment of persons with normal amyloid and CDD and affect the results. Using a CSF Aβ42/Aβ40 ratio has improved concordance with amyloid PET (Niemantsverdriet et. al., 2017). Our study also supports a ratio to control for CDD because the CSF p-tau/Aβ42 ratio significantly decreased the number of discordant individuals. It remains unclear whether using a CSF Aβ42/Aβ40 or p-tau/Aβ42 ratio would better account for CSF dynamics but a recent study suggests they have similar performance in measuring amyloid pathology (Campbell et. al., 2021).

Recent population-based studies have shown CDD neuroimaging features to be more common than previously recognized (Hiraoka et al., 2008); notably, 5.4% frequency in a Swedish study (Jaraj et al., 2014). Therefore, accounting for CDD should be standard procedure.

CDD has been under-recognized as a contributor to AD biomarker abnormalities for several reasons. First, the DESH pattern may be confused with atrophy. Second, DESH is only detectable in vivo because CDD resolves upon removing the brain from the skull, so there is an absence of pathologic literature about DESH. Lastly, only recently have we recognized that the DESH pattern occurs frequently enough to impact interpretation of AD biomarkers (Jaraj et al., 2014)

The strengths of this study include the large number of population-sampled participants with AD biomarkers and our in-house automated CDD detection. While the current study was population-based, replication in a clinical trial like cohort will be an important next step. Limitations include our measurements of Aβ42 and p-tau; therefore, future studies investigating the influence of CSF dynamics on Aβ40 are needed. Future research should evaluate whether CSF dynamics influence the interpretation of plasma biomarkers of Aβ and p-tau and whether a ratio may similarly correct for the influence of the CDD.

5.0. CONCLUSION

In this study, we provided evidence that individuals with discrepant amyloid CSF and PET biomarkers (CSF+PET−) have three different pathophysiologic phenotypes: 1) early accumulators of amyloid; 2) low-protein producers and 3) CDD. Using a CSF Aβ42 ratio to Aβ40 or CSF p-tau apparently controls for the latter two groups. Yet, these data add to growing literature calling for measurement of CDD for biomarker interpretation in AD, which should be a future focus of the field.

HIGHLIGHTS.

  • Discordant CSF/ PET amyloid biomarkers occur in cognitively unimpaired individuals

  • Low CSF β-amyloid 1–42 (Aβ42) may indicate a CSF dynamics disorder (CDD)

  • CSF p-tau:Aβ42 ratio controls for presence of CDD more accurately than Aβ42 alone

Acknowledgments

The authors thank Lea Dacy, Department of Neurology, for proofreading and formatting assistance.

Funding

This work was supported the National Institute on Aging (K76 AG057015, RF1 AG069052-01A1, R37 AG011378, R01 AG041851, U01 AG006786, and P50 AG016574), the National Institute of Neurological Disorders and Stroke (NS097495), the Elsie and Marvin Dekelboum Family Foundation, the Alexander Family Professor of Alzheimer’s Disease Research - Mayo Clinic, the Liston Award, the Schuler Foundation, the GHR Foundation, and the Mayo Foundation for Medical Education and Research. This study was made possible using the resources of the Rochester Epidemiology Project, which is supported by the National Institute on Aging of the National Institutes of Health under Award Number R01AG034676. The funders had no role in the preparation of this article.

Competing interests

JGR, CRJ, DSK, RCP, and PV are investigators on grants that supported this research. The other authors declare that they have no competing interests.

Footnotes

Ethics approval and consent to participate

The Mayo Clinic Institutional Review Board and the Olmsted Medical Center Institutional Review Board approved all study protocols. All participants provided written informed consent according to the Declaration of Helsinki.

Availability of data and materials

Data from the MCSA, including from this study, are available upon reasonable request.

1

Abbrevations: Aβ, β-amyloid; Aβ40, β-amyloid 1–40; Aβ42, β-amyloid 1–42; ADNI, Alzheimer’s Disease Neuroimaging Initiative; AUROC, area under the receiver-operating characteristic curve; CDD, CSF dynamics disorder; DESH, disproportionately enlarged subarachnoid space hydrocephalus; PiB, Pittsburgh compound B; MCSA, Mayo Clinic Study of Aging; MCALT, Mayo Clinic Adult Lifespan Template; MPRAGE, magnetization-prepared rapid gradient-echo; NPH, normal-pressure hydrocephalus; p-tau, phosphorylated-tau protein; SUVR, standard uptake value ratio; t-tau, total tau protein

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. Agren-Wilsson A, Lekman A, Sjöberg W, et al. CSF biomarkers in the evaluation of idiopathic normal pressure hydrocephalus. Acta Neurol Scand 2007;116:333–339. [DOI] [PubMed] [Google Scholar]
  2. Avants BB, Epstein CL, Grossman M, Gee JC. Symmetric diffeomorphic image registration with cross-correlation: evaluating automated labeling of elderly and neurodegenerative brain. Med Image Anal 2008;12:26–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Campbell MR, Ashrafzadeh-Kian S, Petersen RC, et al. P-tau/Abeta42 and Abeta42/40 ratios in CSF are equally predictive of amyloid PET status. Alzheimers Dement (Amst) 2021;13:e12190. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Elobeid A, Laurell K, Cesarini KG, Alafuzoff I. Correlations between mini-mental state examination score, cerebrospinal fluid biomarkers, and pathology observed in brain biopsies of patients with normal-pressure hydrocephalus. J Neuropathol Exp Neurol 2015;74:470–479. [DOI] [PubMed] [Google Scholar]
  5. Graff-Radford J, Gunter JL, Jones DT, et al. Cerebrospinal fluid dynamics disorders: Relationship to Alzheimer biomarkers and cognition. Neurology 2019;93:e2237–e2246. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Graff-Radford NR. Alzheimer CSF biomarkers may be misleading in normal-pressure hydrocephalus. Neurology 2014;83:1573–1575. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Gunter NB, Schwarz CG, Graff-Radford J, et al. Automated detection of imaging features of disproportionately enlarged subarachnoid space hydrocephalus using machine learning methods. Neuroimage Clin 2019;21:101605. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Hashimoto M, Ishikawa M, Mori E, Kuwana N, Study of Ioni. Diagnosis of idiopathic normal pressure hydrocephalus is supported by MRI-based scheme: a prospective cohort study. Cerebrospinal Fluid Res 2010;7:18. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Hiraoka K, Meguro K, Mori E. Prevalence of idiopathic normal-pressure hydrocephalus in the elderly population of a Japanese rural community. Neurol Med Chir (Tokyo) 2008;48:197–199; discussion 199–200. [DOI] [PubMed] [Google Scholar]
  10. Jack CR Jr., Bennett DA, Blennow K, et al. NIA-AA Research Framework: Toward a biological definition of Alzheimer’s disease. Alzheimers Dement 2018;14:535–562. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Jack CR Jr., Bernstein MA, Fox NC, et al. The Alzheimer’s Disease Neuroimaging Initiative (ADNI): MRI methods. J Magn Reson Imaging 2008;27:685–691. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Jack CR Jr., Lowe VJ, Senjem ML, et al. 11C PiB and structural MRI provide complementary information in imaging of Alzheimer’s disease and amnestic mild cognitive impairment. Brain 2008;131:665–680. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Jack CR Jr., Wiste HJ, Weigand SD, et al. Defining imaging biomarker cut points for brain aging and Alzheimer’s disease. Alzheimers Dement 2017;13:205–216. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Jaraj D, Rabiei K, Marlow T, Jensen C, Skoog I, Wikkelso C. Prevalence of idiopathic normal-pressure hydrocephalus. Neurology 2014;82:1449–1454. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Jeppsson A, Zetterberg H, Blennow K, Wikkelso C. Idiopathic normal-pressure hydrocephalus: pathophysiology and diagnosis by CSF biomarkers. Neurology 2013;80:1385–1392. [DOI] [PubMed] [Google Scholar]
  16. Kapaki EN, Paraskevas GP, Tzerakis NG, et al. Cerebrospinal fluid tau, phospho-tau181 and beta-amyloid1–42 in idiopathic normal pressure hydrocephalus: a discrimination from Alzheimer’s disease. Eur J Neurol 2007;14:168–173. [DOI] [PubMed] [Google Scholar]
  17. Knopman DS, Jack CR Jr., Wiste HJ, et al. Short-term clinical outcomes for stages of NIA-AA preclinical Alzheimer disease. Neurology 2012;78:1576–1582. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Leuzy A, Chiotis K, Hasselbalch SG, et al. Pittsburgh compound B imaging and cerebrospinal fluid amyloid-β in a multicentre European memory clinic study. Brain : a journal of neurology 2016;139:2540–2553. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Mattsson N, Insel PS, Donohue M, et al. Independent information from cerebrospinal fluid amyloid-β and florbetapir imaging in Alzheimer’s disease. Brain 2015;138:772–783. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Niemantsverdriet E, Ottoy J, Somers C, et al. The Cerebrospinal Fluid Aβ1–42/Aβ1–40 Ratio Improves Concordance with Amyloid-PET for Diagnosing Alzheimer’s Disease in a Clinical Setting. J Alzheimers Dis 2017;60:561–576. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Palmqvist S, Mattsson N, Hansson O. Cerebrospinal fluid analysis detects cerebral amyloid-β accumulation earlier than positron emission tomography. Brain 2016;139:1226–1236. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Palmqvist S, Schöll M, Strandberg O, et al. Earliest accumulation of β-amyloid occurs within the default-mode network and concurrently affects brain connectivity. Nature Communications 2017;8:1214. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Petersen RC. Mild cognitive impairment as a diagnostic entity. J Intern Med 2004;256:183–194. [DOI] [PubMed] [Google Scholar]
  24. Reimand J, Collij L, Scheltens P, Bouwman F, Ossenkoppele R. Association of amyloid-β CSF/PET discordance and tau load five years later. Neurology 2020: 10.1212/WNL.0000000000010739. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Reimand J, de Wilde A, Teunissen CE, et al. PET and CSF amyloid-β status are differently predicted by patient features: information from discordant cases. Alzheimers Res Ther 2019;11:100. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Roberts RO, Geda YE, Knopman DS, et al. The Mayo Clinic Study of Aging: design and sampling, participation, baseline measures and sample characteristics. Neuroepidemiology 2008;30:58–69. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Schwarz CG, Gunter JL, Ward CP, et al. The Mayo Clinic Adult Life Span Template: Better Quantification across the Life Span. Alzheimers Dement 2017;13:P93–P94. [Google Scholar]
  28. St Sauver JL, Grossardt BR, Yawn BP, et al. Data resource profile: the Rochester Epidemiology Project (REP) medical records-linkage system. Int J Epidemiol 2012;41:1614–1624. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Van Harten AC, Wiste HJ, Weigand SD, et al. CSF biomarkers in Olmsted County: Evidence of 2 subclasses and associations with demographics. Neurology 2020;95:e256–e267. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Vlassenko AG, McCue L, Jasielec MS, et al. Imaging and cerebrospinal fluid biomarkers in early preclinical alzheimer disease. Annals of Neurology 2016;80:379–387. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Vos SJB, Gordon BA, Su Y, et al. NIA-AA staging of preclinical Alzheimer disease: discordance and concordance of CSF and imaging biomarkers. Neurobiol Aging 2016;44:1–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Wiltfang J, Esselmann H, Bibl M, et al. Amyloid beta peptide ratio 42/40 but not A beta 42 correlates with phospho-Tau in patients with low- and high-CSF A beta 40 load. J Neurochem 2007;101:1053–1059. [DOI] [PubMed] [Google Scholar]

RESOURCES