Cerebrospinal fluid amyloid-β 42/40 ratio is associated with ganglion cell-inner plexiform layer thinning in individuals with and without amyloid pathology

Abstract

Background

Retinal imaging offers a noninvasive window into neurodegenerative changes, yet its relationship with established cerebrospinal fluid (CSF) biomarkers of Alzheimer's disease (AD) remains poorly understood. This study addresses a critical gap by examining whether retinal structural findings on optical coherence tomography (OCT) correlate with CSF biomarkers of AD pathology and neurodegeneration.

Objective

To determine if CSF biomarkers correlate with retinal OCT imaging findings in individuals with AD.

Methods

In this cross-sectional study, subjects underwent lumbar puncture for CSF collection and were imaged using the Zeiss Cirrus HD-5000 with AngioPlex.

Results

Forty participants (73 eyes) were included in this study. Twenty-one had normal cognition and negative CSF AD biomarkers, 12 had normal cognition and positive CSF AD biomarkers, and seven had mild cognitive impairment. Central subfield thickness (CST), retinal nerve fiber layer (RNFL), and ganglion cell inner plexiform layer (GCIPL) thicknesses were associated with CSF amyloid-β 42/40 ratio (Aβ_42/40), phosphorylated tau at threonine 181 (pTau181)/Aβ₄₀ ratio, neurofilament light chain (NfL), and glial fibrillary acidic protein (GFAP) using multivariable generalized estimating equations. There were no statistically significant associations between CSF pTau181/Aβ₄₀ ratio, NfL, or GFAP and CST, GCIPL thickness, or RNFL thickness (p > 0.05 for all). Aβ_42/40 ratio was positively associated with GCIPL thickness (p = 0.02), but not with CST or RNFL thickness (p = 0.31 and p = 0.82, respectively).

Conclusions

Decreased CSF Aβ_42/40 ratio, a biomarker of amyloid plaque pathology, is associated with decreased GCIPL thickness. GCIPL thinning may correspond with CSF abnormalities consistent with amyloid pathology that is present even prior to cognitive decline.

Keywords

Alzheimer's disease amyloid cerebrospinal fluid optical coherence tomography

Introduction

Alzheimer's disease (AD) is defined by the accumulation of extracellular amyloid-β (Aβ) plaques and intraneuronal tangles of tau neurofilaments, which together contribute to inflammation, cellular damage, and cognitive decline.^1,2 Diagnosis of AD requires postmortem brain tissue histopathology, costly antemortem brain positron emission tomography (PET), or cerebrospinal fluid (CSF) sampling and biomarker testing.³ CSF biomarkers that have been implicated in AD pathogenesis include Aβ₄₂,⁴ total tau, tau phosphorylated at threonine 181 (pTau181),⁵ neurofilament light chain (NfL),⁶ and glial fibrillary acidic protein (GFAP).⁷ The investigation of these biomarkers has primarily focused on their concentrations in CSF to aid in the diagnosis and monitoring of disease progression along the AD continuum.

Over the past decade, there has been an increasing body of literature highlighting the relationship between retinal imaging metrics and various neurodegenerative diseases associated with cognitive impairment.^8–10 Not only is multimodal retinal imaging noninvasive, relatively inexpensive, and a promising method of in vivo disease characterization, but also such imaging of the retina allows for facile visualization of neurovascular tissue even through an undilated pupil using optical coherence tomography (OCT).^3,11–14 CSF biomarker and retinal imaging abnormalities are often detectable prior to cognitive decline in patients with AD.^2,15,16 Specifically, in patients with AD, macular volume loss and ganglion cell inner plexiform layer (GCIPL) and retinal nerve fiber layer (RNFL) thinning have been noted on OCT in multiple studies.^17–22 Additionally, GCIPL and RNFL thinning have even been found to be associated with clinical progression in patients with mild cognitive impairment (MCI) and AD.¹⁸ However, no study has specifically explored the association between retinal imaging findings on OCT and CSF AD biomarkers. Thus, we sought to determine if retinal metrics, including RNFL thickness, GCIPL thickness, and CST visualized on OCT images, correlate with AD CSF biomarkers, including Aβ_42/40 ratio, pTau181/Aβ₄₀ ratio, NfL, and GFAP, in individuals with and without positive CSF AD biomarkers.

Methods

Subjects were prospectively recruited for retinal imaging from the Duke/University of North Carolina Alzheimer's Disease Research Center (Duke/UNC ADRC; Clinicaltrials.gov NCT0670354; Duke Health IRB Pro00103958). Subjects aged less than 80 years with a clinical diagnosis of normal cognition or MCI confirmed by a consensus panel comprised of neurologists, geriatric psychiatrists, and neuropsychologists were eligible for participation. The clinical and biomarker diagnosis was confirmed by the panel after review of clinical history, neurological exam, neuropsychological testing included in the National Alzheimer's Coordinating Council Uniform Data Set 3 (UDS3),²³ MRI imaging, and CSF biomarkers, including Aβ_42/40 and pTau181.

The iMIND (Eye Multimodal Imaging in Neurodegenerative Disease) Study research protocol was approved by the Duke Health Institutional Review Board in Durham, NC (Pro00111831). Written informed consent was obtained prior to enrollment from each participant or their legally authorized representative. The study followed the tenets of the Declaration of Helsinki. In this prospective study (Clinicaltrials.gov NCT03233646), exclusion criteria included diabetes mellitus, uncontrolled hypertension, glaucoma, age-related macular degeneration, intraocular surgery other than uncomplicated cataract extraction, vitreoretinal or optic nerve pathology that could interfere with OCT analysis or corrected visual acuity worse than 20/40 on the day of iMIND study entry.

CSF biomarker acquisition

The Duke/UNC ADRC collected CSF by lumbar puncture using an established protocol to minimize adverse events and patient discomfort.²⁴ CSF Aβ₄₂, Aβ₄₀, pTau181, NfL, and GFAP concentrations were determined by electro-chemiluminescent immunoassay using the respective Lumipulse G assay (Fujirebio).

CSF biomarkers are increasingly used to rule out AD in older adults with cognitive impairment.²⁵ The Aβ_42/40 ratio is commonly used as it has a diagnostic agreement of >90% with amyloid PET imaging data.²⁵ Adjusting Aβ₄₂ values for Aβ₄₀, which does not increase in AD, improves diagnostic validity compared to the use of Aβ₄₂ alone.^26–28 An Aβ_42/40 ratio of < 0.062 is considered indicative of amyloid pathology.²⁹ Similarly, pTau181/Aβ₄₀ ratio accounts for inter-individual differences in CSF fluid dynamics and enhances the diagnostic performance of pTau181. Therefore, this ratio was used in this analysis.³⁰ Although CSF GFAP and NfL levels are elevated in AD pathology, they are not routinely used for AD diagnosis. Thus, both were included in this study solely for exploratory analysis.^6,7,31,32

We categorized subjects into three groups based on clinical and biomarker diagnosis: 1) individuals with normal cognition and negative amyloid pathology (Aβ_42/40 ratio > 0.062); 2) individuals with normal cognition and positive amyloid pathology (Aβ_42/40 ratio < 0.062); and 3) individuals with cognitive decline (MCI) and positive amyloid pathology (Aβ_42/40 ratio < 0.062). Individuals with MCI and negative amyloid pathology were excluded from the study due to potential confounding of results with other, non-AD, neurodegenerative diseases.

OCT image acquisition

Within one calendar year of the lumbar puncture to obtain CSF, all subjects were imaged using a spectral-domain OCT machine (Cirrus HD-5000 with AngioPlex; Carl Zeiss Meditec, Dublin, CA). Two images were acquired of each eye of each patient: a 512 × 128 macular cube centered on the fovea and a 200 × 200 optic disc cube centered on the optic disc. OCT images that were of poor quality (less than 7/10 signal strength) or had low resolution, segmentation error, media opacity artifact, or motion artifact were excluded. The CST and GCIPL thickness for each image were automatically calculated from the 512 × 128 macular cube image using the OCT software (Carl Zeiss Meditec, version 10.0.0.14618), while the RNFL thickness for each image was automatically calculated from the 200 × 200 optic disc cube image using the same OCT software.

Statistical analysis

To measure the association between each of the OCT parameters, including CST, GCIPL thickness, and RNFL thickness, and CSF parameters, including Aβ_42/40 ratio, pTau181/Aβ₄₀ ratio, NfL, and GFAP, multivariable generalized estimating equations (GEE) were used to account for both eyes of each patient. p-values provided were adjusted for age, sex, and history of treated hypertension. Analysis was completed using Stata (18.0, StataCorp LLC, College Station, TX). As an exploratory study, no prospective power calculation was completed as the effect size, if any exists, is unknown. An alpha of 0.05 was chosen for statistical significance; a Bonferroni correction for multiple analyses was considered, but given that this is an exploratory study and the first study of its kind, a more stringent alpha was ultimately not implemented.³³

Results

Seventy-three eyes of 40 patients were included in the study. Thirteen (16.3%) of the 512 × 128 macular cube images and 15 (18.8%) of the 200 × 200 optic disc cube scans were excluded due to poor quality. All 40 subjects had Aβ₄₀, Aβ₄₂, pTau181, and NfL data available, while only 34 of these 40 (85%) had GFAP data available due to being enrolled prior to this assay being available. Demographic data are presented in Table 1.

Table 1.

Demographic information.

	All subjects(n = 73 eyes of 40 patients)^a
Age (years)
Mean (SD)	63.7 (10.5)
Range	(37.1–79.5)
Sex	14/26
Male/ Female	14/26
Hypertension	9/31
Y/N	9/31
Clinical Diagnosis (n)
Cognitively Normal	33
Mild Cognitive Impairment	7
Amyloid Status (n)
Cognitively Normal, Amyloid Negative	21
Cognitively Normal, Amyloid Positive	12
Cognitively Abnormal, Amyloid Positive	7
NfL (pg/mL)
Mean (SD)	447.3 (634.6)
Range	(19–2184)
GFAP (pg/mL)
Mean (SD)	9310 (27116)
Range	(347–155656)
Aβ₄₂ (pg/mL)
Mean (SD)	617.9 (237.5)
Range	(216–1171)
pTau181 (pg/mL)
Mean (SD)	50.53 (31.32)
Range	(17.4–147)
pTau181/Aβ₄₀ Ratio
Mean (SD)	0.00524 (0.00312)
Range	(.00257−.0156)
Aβ_42/40 Ratio
Mean (SD)	0.0646 (0.0218)
Range	(0.0283–0.1028)
CST (μm)^b
n	67
Mean (SD)	265.0 (23.2)
Range	(218–322)
GCIPL (μm)^b
n	67
Mean (SD)	79.9 (4.3)
Range	(71–90)
RNFL (μm)^b
n	65
Mean (SD)	91.0 (7.7)
Range	(73–109)

Because of image quality or other exclusion criteria, not all images of all eyes were included.

CST, GCIPL, RNFL summary statistics provided by individual eye.

SD: standard deviation; NfL: neurofilament light chain; GFAP: glial fibrillary acidic protein; Aβ₄₂: amyloid-β 42; pTau181: tau phosphorylated at threonine 181; Aβ₄₀: amyloid-β 40; CST: central subfield thickness; GCIPL: ganglion cell inner plexiform layer; RNFL: retinal nerve fiber layer

Multivariate GEE regression results between OCT parameters and CSF biomarkers are provided in Table 2. Because of the significant variance in CSF NfL and GFAP values, analysis was completed using a logarithmic transformation of these variables. After controlling for age, sex, and diagnosis of treated hypertension, the logarithmic transformation of CSF NfL was not significantly associated with GCIPL thickness (regression coefficient of 0.11; 95% CI of −7.98–1.02; p = 0.81), CST (regression coefficient of 0.58; 95% CI of −3.63–4.80; p = 0.79), or RNFL (regression coefficient of −0.084; 95% CI of −1.40–1.56; p = 0.91). After controlling for age, sex, and diagnosis of treated hypertension, the logarithmic transformation of CSF GFAP was not significantly associated with GCIPL thickness (regression coefficient of −0.29; 95% CI of −1.21–0.64; p = 0.54), CST (regression coefficient of −1.03; 95% CI of −5.38–3.33; p = 0.64), or RNFL thickness (regression coefficient of −0.44; 95% CI of −1.02–1.91; p = 0.55).

Table 2.

Regression coefficients from multivariate generalized estimating equationst regression model.

Regression coefficients from multivariate generalized estimating equations regression model
Dependent variable	Explanatory Variable	Estimate^a	95% CI^a	p ^a
NfL^b	CST	0.58	(−3.63–4.80)	0.79
NfL^b	GCIPL	0.11	(−7.98–1.02)	0.81
NfL^b	RNFL	0.084	(−1.40–1.56)	0.91
GFAP^b	CST	−1.03	(−5.38–3.33)	0.64
GFAP^b	GCIPL	−0.29	(−1.21–0.64)	0.54
GFAP^b	RNFL	0.44	(−1.02 –1.91)	0.55
pTau181/Aβ₄₀ Ratio	CST	−500.3	(−2459–1458)	0.62
pTau181/Aβ₄₀ Ratio	GCIPL	−277.9	(−685.0–129.2)	0.18
pTau181/Aβ₄₀ Ratio	RNFL	437.9	(−203.8–1079.7)	0.18
Aβ_42/40 Ratio	CST	−162.9	(−475.0 –149.3)	0.31
Aβ_42/40 Ratio	GCIPL	76.1	(13.8–138.3)	0.02
Aβ_42/40 Ratio	RNFL	−12.3	(−118.3–93.7)	0.82

Multivariate regression estimates, confidence intervals, and p-values provided are controlling for age, sex, and history of hypertension using a generalized estimating equations regression model.

NfL and GFAP associations are provided for the logarithmic transformation of the respective variables.

CI: confidence interval; NfL: neurofilament light chain; GFAP: glial fibrillary acidic protein; Aβ₄₂: amyloid-β 42; pTau181: tau phosphorylated at threonine 181; Aβ₄₀: amyloid-β 40; CST: central subfield thickness; GCIPL: ganglion cell inner plexiform layer; RNFL: retinal nerve fiber layer

After controlling for age, sex, and diagnosis of treated hypertension, CSF pTau181/Aβ₄₀ ratio was not associated with CST (regression coefficient of −500.3; 95% CI of −2459–1458; p = 0.62), GCIPL thickness (regression coefficient of −277.9; 95% CI of −685.0–129.2; p = 0.18), or RNFL thickness (regression coefficient of 437.9; 95% CI of −203.8–1079.7; p = 0.18). After controlling for age, sex, and diagnosis of treated hypertension, CSF Aβ42/40 ratio was positively associated with GCIPL thickness (regression coefficient of 76; 95% CI of 13.8–138.3; p = 0.02) (Figure 1); however, there was no association between CSF Aβ_42/40 ratio and CST (regression coefficient of −162.9; 95% CI of −475.0 −149.3; p = 0.31) or RNFL thickness (regression coefficient of −12.3; 95% CI of −118.3–93.7; p = 0.82).

Figure 1.

Scatterplot of ganglion cell inner plexiform layer thickness (GCIPL) and Aβ_42/40 ratio. Association of GCIPL in μm and CSF amyloid-β 42/40 ratio (Aβ_42/40) with a linear line of best fit.

Mean GCIPL parameters for each of the three groups based on clinical and biomarker diagnosis are provided in Table 3. In subjects with MCI and Aβ_42/40 ratio consistent with AD, mean ± SD GCIPL thickness was 76.1 ± 3.6 μm. In subjects with normal cognition and Aβ_42/40 ratio consistent with AD, mean ± SD GCIPL thickness was 79.8 ± 4.5 μm. In subjects with normal cognition and Aβ_42/40 ratio not consistent with AD, mean ± SD GCIPL thickness was 81.0 ± 3.9 μm. There was a statistically significant difference between the mean GCIPL thickness among these groups using linear mixed models to account for correlations between the two eyes of the same participants (p = 0.015). For each of these groups, the association between Aβ_42/40 ratio and GCIPL was also determined. Results are provided in Table 3 and Figure 2. In subjects with MCI and Aβ_42/40 ratio consistent with AD, the regression coefficient was −157.3 (p = 0.24). In subjects with normal cognition and Aβ_42/40 ratio consistent with AD, the regression coefficient was 225.7 (p = 0.02). In subjects with normal cognition and Aβ_42/40 ratio not consistent with AD, the regression coefficient was 65.2 (p = 0.46).

Figure 2.

Scatterplot of ganglion cell inner plexiform layer thickness and Aβ_42/40 ratio by cognitive and amyloid status. Subjects were categorized into three groups based on clinical and biomarker diagnosis: individuals with normal cognition and negative amyloid pathology (Aβ_42/40 ratio > 0.062); individuals with normal cognition and positive amyloid pathology (Aβ_42/40 ratio < 0.062); and individuals with cognitive decline (MCI) and positive amyloid pathology (Aβ_42/40 ratio < 0.062).

Table 3.

Ganglion cell inner plexiform layer thickness by clinical and biomarker diagnosis.

	MCI and Amyloid Positive^a	Cognitively Normal and Amyloid Positive^a	Cognitively Normal and Amyloid Negative^a	p^b
GCIPL Thickness (μm)n	11	19	37	0.015
Mean (SD)	76.1 (3.6)	79.8 (4.5)	81.0 (3.9)
Range	71–82	72–90	75–87
Regression Coefficient on Aβ_42/40 Ratio(p-value)^c	−157.3 (0.24)	225.7 (0.02)	65.2 (0.46)

Subjects were categorized into three groups based on clinical and biomarker diagnosis: individuals with normal cognition and negative amyloid pathology (Aβ_42/40 ratio > 0.062); individuals with normal cognition and positive amyloid pathology (Aβ_42/40 ratio < 0.062); and individuals with cognitive decline (MCI) and positive amyloid pathology (Aβ_42/40 ratio < 0.062).

p-value provided using linear mixed effects model to control for correlation between eyes.

Multivariate regression estimates and p-values provided using a generalized estimating equations regression model.

MCI: mild cognitive impairment; Aβ₄₂: amyloid-β 42; GCIPL: ganglion cell inner plexiform layer.

Discussion

These data demonstrate that there are significant associations between Aβ_42/40 ratio and GCIPL when controlling for age, sex, and history of treated hypertension. While the CSF biomarker Aβ_42/40 ratio was associated with GCIPL thickness, no significant association was observed between it and RNFL thickness or CST. These data suggest that CSF biomarkers of amyloid pathology may correspond with changes in GCIPL thickness.

Aβ₄₂ is the most implicated amyloid protein that develops into the characteristic amyloid plaque in the pathogenesis of AD.³⁴ On the other hand, pTau181 is a misfolded protein that accumulates in intracellular neurofibrillary tangles in AD and has also been implicated as a potential biomarker for the diagnosis of AD.³⁵ Although no test currently exists to clinically utilize NfL and GFAP as diagnostic biomarkers for AD, there is a substantial body of literature highlighting both as promising biomarkers for neurodegeneration.^6,7,36 Specifically, NfL is a scaffolding protein of the neuronal cytoskeleton that is expressed in the axons of neurons and released in an age-dependent manner; however, when neurons undergo apoptosis or cell death, it is released in higher amounts in neurodegenerative conditions like AD.^36,37 Similarly, GFAP is a protein found in astrocytes, and astrocytosis and increased GFAP expression is seen with inflammation related to amyloid pathology consistent with AD.^38,39 Aβ₄₂ and pTau181 are biomarkers specific to AD; whereas NfL and GFAP are non-specific for AD and could indicate several other neurodegenerative conditions associated with inflammation and cell death.⁴⁰

The neuronal layers of the retina, which can be examined with noninvasive methods such as OCT, have long held promise as a neurosensory tissue that could be utilized to diagnose and track neurodegenerative disease.⁴¹ A meta-analysis of 30 cross-sectional case-control studies found that there are significant OCT differences between cognitively normal individuals and patients with MCI or AD dementia.²¹ Specifically, subjects with AD had significantly thinner GCIPL, reduced macular volume, and thinner peripapillary RNFL thicknesses.²¹ There are two theories that potentially explain the OCT findings in patients with AD. One is that the thinner retinal structures are a non-specific response to neurodegeneration that is seen throughout the CNS in conditions like AD.⁴² Another explanation is that amyloid pathology simultaneously affects both cerebral and retinal structures, which manifests on OCT as thinner retinal layers. This latter explanation is supported by research that has identified amyloid plaques in the retinal tissue of patients with AD.^21,43 This study gives credibility to the latter proposed mechanism since a correlation between retinal structures and Aβ_42/40 strengthens the theory that both processes occur simultaneously, even in the absence of cognitive decline. This would provide a mechanistic explanation for why OCT could then be used as a noninvasive method of diagnosing and tracking neurodegenerative changes in patients with AD.

We found that decreased GCIPL thickness was associated with decreased Aβ_42/40 ratio. Since a lower Aβ_42/40 ratio is associated with higher CNS amyloid plaque burden,²⁸ and subjects with AD (who by definition have CNS amyloid plaques) have been found in the literature to have a thinner GCIPL layer,²¹ our data suggest that amyloid pathology, evidenced by decreased Aβ_42/40 ratio, may simultaneously be affecting retinal structures evidenced by thinner GCIPL. This is further supported by the subanalysis presented in Table 3, demonstrating GCIPL thickness is significantly lower in subjects with amyloid pathology and cognitive impairment (76.1 μm for MCI and amyloid positive; 79.8 μm for cognitively normal and amyloid positive; 81.0 μm for cognitively normal and amyloid negative). GCIPL thickness was lowest in the MCI group with minimal differences between the cognitively normal groups which may suggest that GCIPL thinning becomes more pronounced at the MCI stage rather than during the preclinical phase of disease.

Figure 2 highlights that the association demonstrated in the overall sample of decreased GCIPL thickness being associated with decreased Aβ_42/40 ratio is present and statistically significant in the same direction in cognitively normal subjects with Aβ_42/40 ratios consistent with amyloid pathology, a clinically important group with an unclear prognosis (p = 0.02). Though the mean GCIPL thickness between the cognitively normal groups was not significantly different, this association suggests that GCIPL thinning may have utility as a potential biomarker in tracking amyloid burden even in cognitively normal individuals. This association for the MCI group suggests a negative trend between GCIPL thickness and Aβ_42/40 ratio. However, there are only 11 eyes in this group, and given the relatively smaller sample size, there is a strong outlier bias. To determine the true effect, future studies should include more individuals with MCI in their cohorts.

The lack of significance seen between RNFL thickness and CST and CSF parameters could be due to the effect size being less pronounced in these retinal imaging findings and thus, may need a larger sample size to identify the effect; however, this could also suggest that GCIPL thinning is the layer most likely to reflect amyloid pathology as characterized by CSF AD biomarkers. Additionally, there were no significant associations between pTau181/Aβ₄₀, GFAP, and NfL with any OCT parameter. This is likely due to Aβ₄₂ being upstream of pTau181, GFAP, and NfL in the pathogenesis of AD, and thus these biomarkers may not yet be detectable as abnormal in the early stages of amyloid pathology, making it difficult to determine correlations with retinal imaging findings in a sample of mostly cognitively normal subjects.^2,36,38 Further studies with larger sample sizes of cognitively impaired individuals may yield more insight into how these downstream biomarkers may correlate with GCIPL and other retinal structures further along in disease progression. Additionally, future research with larger sample sizes would be powered to subsegment GCIPL and RNFL thickness into ETDRS grid quadrants to determine which, if any, segment is the primary modulator of the observed effect. From a statistical standpoint, no power calculation was completed for this modest-sized cross-sectional study. With a total sample size of 73 eyes, the confidence intervals were relatively wide, suggesting that studies with larger sample sizes are needed to confirm and further elucidate potential significant effects. Given that this was a modest-sized exploratory study, no multiple comparison correction was performed.³³

This study is unique in that it is the first to explore the relationship between CSF biomarkers and OCT parameters in subjects with and without AD; however, there are inherent limitations. Specifically, the inclusion of only individuals with no retinal or neurologic pathology other than AD significantly decreases the generalizability of the findings, but excluding potential confounders in this initial work improves the clarity of our results by increasing our power to detect significant associations. Future studies that include subjects with diabetes will be tasked with addressing the strong association between AD and common ocular pathologies. In both diabetes and AD, reduced clearance of amyloid is attributed to impaired insulin signaling, and insulin resistance can accelerate progression to AD.⁴⁴ While diabetes can directly contribute to the thinning of inner retinal layers through capillary nonperfusion and subsequent ischemia, differentiating this effect from that of amyloid deposition will be an important consideration when developing a diagnostic tool using GCIPL and RNFL to diagnose and monitor AD. Similarly, GCIPL and RNFL thinning may be due to either glaucoma or AD, so determining AD-specific manifestations of inner retinal thinning in relation to glaucomatous changes will be an important consideration in future work and perhaps may ultimately be differentiated using machine learning algorithms.

This study can provide some understanding of population-level correlations, but it does not provide insights into how CSF biomarkers may vary with OCT parameters over time within individuals—an important goal for future work to understand progression. To strengthen the understanding of the relationship between OCT parameters and CSF biomarkers, future studies should correlate OCT and CSF biomarkers longitudinally and include individuals with more diverse ocular pathologies.

In conclusion, thinning in the GCIPL layer of the retina is associated with increasing amyloid burden, as evidenced by decreased Aβ_42/40 ratio. These findings suggest that OCT measurement of the GCIPL layer may be a useful correlate of positive amyloid status associated with AD pathology. In addition, it may be a useful tool in diagnosing and monitoring disease progression in cognitively normal patients at risk of developing AD or cognitively affected patients with AD. Larger studies with longitudinal analysis could help strengthen the understanding of this observed relationship and potentially develop a clinical tool to utilize OCT in the diagnosis and monitoring of AD pathology.

Footnotes

Acknowledgements

We would like to acknowledge the assistance of the Molecular Genomics Core at the Duke Molecular Physiology Institute, Duke University School of Medicine, for the generation of data for the manuscript.

ORCID iDs

Hemal Patel

Jamie Karl

Kim G. Johnson

Ethical considerations

The iMIND (Eye Multimodal Imaging in Neurodegenerative Disease) Study research (Clinicaltrials.gov NCT03233646) protocol was approved by the Duke Health Institutional Review Board in Durham, NC (Pro00111831).

Consent to participate

Written informed consent was obtained prior to enrollment from each participant or their legally authorized representative.

Consent for publication

Not applicable.

Author contribution(s)

Hemal Patel: Conceptualization; Data curation; Formal analysis; Funding acquisition; Investigation; Methodology; Project administration; Resources; Software; Supervision; Validation; Visualization; Writing – original draft; Writing – review & editing.

Alex Choi: Data curation; Investigation; Validation; Writing – original draft; Writing – review & editing.

Peter Weng: Conceptualization; Data curation; Funding acquisition; Methodology; Resources; Validation; Writing – review & editing.

Jamie Karl: Conceptualization; Data curation; Funding acquisition; Methodology; Resources; Validation; Writing – review & editing.

Suzanna Joseph: Conceptualization; Data curation; Funding acquisition; Methodology; Resources; Validation; Writing – review & editing.

Kim G. Johnson: Conceptualization; Data curation; Funding acquisition; Investigation; Project administration; Resources; Supervision; Writing – review & editing.

Miles Berger: Conceptualization; Data curation; Funding acquisition; Investigation; Project administration; Resources; Supervision; Writing – review & editing.

Heather E. Whitson: Conceptualization; Data curation; Funding acquisition; Investigation; Project administration; Resources; Supervision; Writing – review & editing.

Dilraj S. Grewal: Conceptualization; Data curation; Funding acquisition; Investigation; Project administration; Resources; Supervision; Writing – review & editing.

Sharon Fekrat: Conceptualization; Data curation; Funding acquisition; Investigation; Methodology; Project administration; Resources; Supervision; Validation; Writing – review & editing.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the following funding sources: Research to Prevent Blindness Unrestricted Grant (Duke University); NIH grant to support Duke/UNC ADRC (P30AG072958); National Institute on Aging.

Declaration of conflicting interests

The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Kim G. Johnson MD: Primary investigator of Eisai Inc. AHEAD clinical trial on lecanemab therapy for cognitively normal participants; the primary investigator of ALZ-NET at Duke; the primary investigator of LEXEO Therapeutics gene therapy trial; a speaker for Eisai at the 2024 Alzheimer's Association International (AAIC) annual meeting; a consultant with University of Southern California; Lilly Preclinical Diagnosis Advisory Board member. Dr. Johnson is an Editorial Board Member of this journal but was not involved in the peer-review process of this article nor had any access to the information regarding its peer-review.

Miles Berger MD PhD: Acknowledges additional funding from NIH grants R01-AG076903 (PI- Berger) and R01-AG073598 (PI- Berger).

Dilraj S. Grewal MD FASRS: Zeiss (consultant)

Sharon Fekrat MD FASRS: Optos (grant and research support)

The remaining authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability statement

The data supporting the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.

References

Scheltens

Blennow

Breteler

, et al. Alzheimer's disease. Lancet 2016; 388: 505–517.

Bloom

. Amyloid-beta and tau: the trigger and bullet in Alzheimer disease pathogenesis. JAMA Neurol 2014; 71: 505–508.

Ono

Saji

. SPECT Imaging agents for detecting cerebral beta-amyloid plaques. Int J Mol Imaging 2011; 2011: 543267.

Nakamura

Kaneko

Villemagne

, et al. High performance plasma amyloid-beta biomarkers for Alzheimer's disease. Nature 2018; 554: 249–254.

Ossenkoppele

van der Kant

Hansson

. Tau biomarkers in Alzheimer's disease: towards implementation in clinical practice and trials. Lancet Neurol 2022; 21: 726–734.

Vermunt

Otte

Verberk

IMW

, et al. Age- and disease-specific reference values for neurofilament light presented in an online interactive support interface. Ann Clin Transl Neurol 2022; 9: 1832–1837.

Kim

Shin

Chang

. GFAP As a potential biomarker for Alzheimer's disease: a systematic review and meta-analysis. Cells 2023; 12: 1309.

Yoon

Thompson

Polascik

, et al. Correlation of OCTA and volumetric MRI in mild cognitive impairment and Alzheimer's disease. Ophthalmic Surg Lasers Imaging Retina 2019; 50: 709–718.

Robbins

Grewal

Stinnett

, et al. Assessing the retinal microvasculature in individuals with early and late-onset Alzheimer's disease. Ophthalmic Surg Lasers Imaging Retina 2021; 52: 336–344.

10.

Grewal

Polascik

Hoffmeyer

, et al. Assessment of differences in retinal microvasculature using OCT angiography in Alzheimer's disease: a twin discordance report. Ophthalmic Surg Lasers Imaging Retina 2018; 49: 440–444.

11.

Hardy

Selkoe

. The amyloid hypothesis of Alzheimer's disease: progress and problems on the road to therapeutics. Science 2002; 297: 353–356.

12.

Yoon

Grewal

Thompson

, et al. Retinal microvascular and neurodegenerative changes in Alzheimer's disease and mild cognitive impairment compared with control participants. Ophthalmol Retina 2019; 3: 489–499.

13.

Hui

Zhao

, et al. Detection of retinal changes with optical coherence tomography angiography in mild cognitive impairment and Alzheimer's disease patients: a meta-analysis. PLoS One 2021; 16: e0255362.

14.

Cheung

Ong

Hilal

, et al. Retinal ganglion cell analysis using high-definition optical coherence tomography in patients with mild cognitive impairment and Alzheimer's disease. J Alzheimers Dis 2015; 45: 45–56.

15.

Palmqvist

Mattsson

Hansson

, et al. Cerebrospinal fluid analysis detects cerebral amyloid-beta accumulation earlier than positron emission tomography. Brain 2016; 139: 1226–1236.

16.

O'Bryhim

Apte

Kung

, et al. Association of preclinical Alzheimer disease with optical coherence tomographic angiography findings. JAMA Ophthalmol 2018; 136: 1242–1248.

17.

Lopez-de-Eguileta

Cervero

Ruiz de Sabando

, et al. Ganglion cell layer thinning in Alzheimer's disease. Medicina (Kaunas) 2020; 56: 553.

18.

Choi

Park

Kim

. Macular ganglion cell -inner plexiform layer thickness is associated with clinical progression in mild cognitive impairment and Alzheimer's disease. PLoS One 2016; 11: e0162202.

19.

Iseri

Altinas

Tokay

, et al. Relationship between cognitive impairment and retinal morphological and visual functional abnormalities in Alzheimer disease. J Neuroophthalmol 2006; 26: 18–24.

20.

Cunha

Proenca

Dias-Santos

, et al. OCT In Alzheimer's disease: thinning of the RNFL and superior hemiretina. Graefes Arch Clin Exp Ophthalmol 2017; 255: 1827–1835.

21.

Chan

VTT

Sun

Tang

, et al. Spectral-domain OCT measurements in Alzheimer's disease: a systematic review and meta-analysis. Ophthalmology 2019; 126: 497–510.

22.

Doustar

Torbati

Black

, et al. Optical coherence tomography in Alzheimer's disease and other neurodegenerative diseases. Front Neurol 2017; 8: 701.

23.

Besser

Kukull

Knopman

, et al. Version 3 of the National Alzheimer's Coordinating Center's uniform data set. Alzheimer Dis Assoc Disord 2018; 32: 351–358.

24.

Nobuhara

Bullock

Bunning

, et al. A protocol to reduce self-reported pain scores and adverse events following lumbar punctures in older adults. J Neurol 2020; 267: 2002–2006.

25.

Bouwman

Frisoni

Johnson

, et al. Clinical application of CSF biomarkers for Alzheimer's disease: from rationale to ratios. Alzheimers Dement (Amst) 2022; 14: e12314.

26.

Delaby

Estelles

Zhu

, et al. The Abeta1-42/Abeta1-40 ratio in CSF is more strongly associated to tau markers and clinical progression than Abeta1-42 alone. Alzheimers Res Ther 2022; 14: 20.

27.

Doecke

Perez-Grijalba

Fandos

, et al. Total Abeta(42)/Abeta(40) ratio in plasma predicts amyloid-PET status, independent of clinical AD diagnosis. Neurology 2020; 94: e1580–e1591.

28.

Hansson

Lehmann

Otto

, et al. Advantages and disadvantages of the use of the CSF amyloid beta (Abeta) 42/40 ratio in the diagnosis of Alzheimer's disease. Alzheimers Res Ther 2019; 11: 34.

29.

Campbell

Ashrafzadeh-Kian

Petersen

, et al. P-tau/Abeta42 and Abeta42/40 ratios in CSF are equally predictive of amyloid PET status. Alzheimers Dement (Amst) 2021; 13: e12190.

30.

Guo

Korman

La Joie

, et al. Normalization of CSF pTau measurement by abeta(40) improves its performance as a biomarker of Alzheimer's disease. Alzheimers Res Ther 2020; 12: 97.

31.

Fukuyama

Izumoto

Fushiki

. The cerebrospinal fluid level of glial fibrillary acidic protein is increased in cerebrospinal fluid from Alzheimer's disease patients and correlates with severity of dementia. Eur Neurol 2001; 46: 35–38.

32.

Agnello

Gambino

Del Ben

, et al. Establishing decisional cutoff values of neurofilament light chains in cerebrospinal fluid measured by fully automated chemiluminescent enzyme immunoassay. J Clin Lab Anal 2025; 39: e25152.

33.

Armstrong

. When to use the Bonferroni correction. Ophthalmic Physiol Opt 2014; 34: 502–508.

34.

Hampel

Hardy

Blennow

, et al. The amyloid-beta pathway in Alzheimer's disease. Mol Psychiatry 2021; 26: 5481–5503.

35.

Fagan

Mintun

Shah

, et al. Cerebrospinal fluid tau and ptau(181) increase with cortical amyloid deposition in cognitively normal individuals: implications for future clinical trials of Alzheimer's disease. EMBO Mol Med 2009; 1: 371–380.

36.

Gaetani

Blennow

Calabresi

, et al. Neurofilament light chain as a biomarker in neurological disorders. J Neurol Neurosurg Psychiatry 2019; 90: 870–881.

37.

Bacioglu

Maia

Preische

, et al. Neurofilament light chain in blood and CSF as marker of disease progression in mouse models and in neurodegenerative diseases. Neuron 2016; 91: 56–66.

38.

Pereira

Janelidze

Smith

, et al. Plasma GFAP is an early marker of amyloid-beta but not tau pathology in Alzheimer's disease. Brain 2021; 144: 3505–3516.

39.

Wilhelmsson

Bushong

Price

, et al. Redefining the concept of reactive astrocytes as cells that remain within their unique domains upon reaction to injury. Proc Natl Acad Sci U S A 2006; 103: 17513–17518.

40.

Thijssen

Verberk

IMW

Kindermans

, et al. Differential diagnostic performance of a panel of plasma biomarkers for different types of dementia. Alzheimers Dement (Amst) 2022; 14: e12285.

41.

London

Benhar

Schwartz

. The retina as a window to the brain-from eye research to CNS disorders. Nat Rev Neurol 2013; 9: 44–53.

42.

Erskine

Herrera

. Connecting the retina to the brain. ASN Neuro 2014; 6: 1759091414562107.

43.

Koronyo-Hamaoui

Koronyo

Ljubimov

, et al. Identification of amyloid plaques in retinas from Alzheimer's patients and noninvasive in vivo optical imaging of retinal plaques in a mouse model. Neuroimage 2011; 54: S204–S217.

44.

Nguyen

QTH

Nguyen

TKO

, et al. Type 3 diabetes and its role implications in Alzheimer's disease. Int J Mol Sci 2020; 21: 3165.