Choroidal evaluation of FTLD-Tau and biomarker-determined Alzheimer’s disease

Abstract

Frontotemporal lobar degeneration with tauopathy (FTLD-Tau) can present clinically similar to Alzheimer’s disease but lacks a biomarker. Alzheimer’s disease has been associated with choroidal thinning compared to controls. We compared the choroid of 25 probable FTLD-Tau (pFTLD-Tau) patients (42 eyes), 26 biomarker-determined probable Alzheimer’s disease neuropathologic change (pADNC) patients (49 eyes), and 53 normal controls (80 eyes). Cerebrospinal fluid biomarkers determined presence of ADNC. All pFTLD-Tau patients had a syndrome highly associated with FTLD-Tau. Optical coherence tomography was performed with masked manual choroidal thickness (CT) measurements. With Image J, binarized images determined the choroidal vascularity index (CVI). Linear regression with generalized estimating equations to account for inter-eye correlation was performed. For pFTLD-Tau, pADNC, and controls, the subfoveal CT was 308.9, 286.0, and 301.5 μm, and CVI was 0.72, 0.72, and 0.73, respectively (all p > 0.05 for each group comparison). Adjusting for demographics, the CT and CVI were not significantly different between groups, including 13 CT measurement locations (all p > 0.05). Among pADNC patients, an exploratory analysis found a correlation between CVI and disease duration (Pearson r = 0.32, p = 0.04). We found no significant difference of CT or CVI between pFTLD-Tau, pADNC, and controls. Additional studies are warranted to evaluate how CVI relates to ADNC.

Introduction

The neuropathology of frontotemporal dementia (FTD) is referred to as FTLD (frontotemporal lobar degeneration), and there remains an urgent need for a biomarker for the neuropathologic diagnosis of frontotemporal lobar degeneration from a tauopathy (FTLD-Tau). FTLD-Tau patients can be difficult to clinically distinguish from patients that have Alzheimer’s disease neuropathologic change (ADNC), especially in the young-onset age range where non-amnestic syndromes are more common^1–5. While advancements have been made for validated biomarkers for ADNC^6–11, cerebrospinal fluid requires an invasive lumbar puncture and the cost of positron emission tomography (PET) limits widespread use. Plasma biomarkers for ADNC hold promise but currently are not able to specifically identify FTLD-Tau^6,7. Furthermore, the clinical diagnosis of FTLD-Tau is complicated by the difficulty in clinically distinguishing tauopathies from FTD patients with neuropathology from TAR DNA-binding Protein of 43 kDa (FTLD-TDP), as about half of FTLD neuropathology cases are FTLD-Tau and half are FTLD-TDP^1,2. Thus, the absence of biomarkers specific for FTLD-Tau presents challenges for the clinical management of these patients. There is no approved drug therapy for patients with FTLD-Tau, and a biomarker could facilitate clinical trial patient categorization as mechanism-based therapies are emerging¹².

As an extension of the central nervous system that can be imaged at high-resolution, the retina is an attractive candidate biomarker tissue for dementia. Inner retina thinning has been the predominant retinal finding for clinical Alzheimer’s disease (AD) patients compared to normal controls^13,14. Interestingly, our group and others have found thinning of the photoreceptor outer nuclear layer (ONL), an outer retina layer, in patients with FTLD-Tau^15–20. Given the ability to readily image the posterior segment of the eye, studies have also evaluated the choroid, the vascular layer between the sclera and neurosensory retina that provides nourishment for the outer third of the retina. Several studies have found thinning of the choroid among clinical AD patients compared to normal controls; this has been hypothesized to be related to an amyloid-beta 42 (Aβ42) vasculopathy^21–25. On the other hand, studies comparing patients with ADNC determined by biomarkers (cerebrospinal fluid or amyloid beta PET) to normal controls have not found differences in choroidal thickness²⁶.

FTLD-Tau patients are not characterized by amyloid beta abnormalities, and there is an absence of studies evaluating the choroid of FTLD-Tau patients. We sought to evaluate the choroid of probable FTLD-Tau patients compared to biomarker-determined ADNC patients and normal controls.

Methods

Subjects

Patients were enrolled consecutively by the Penn Frontotemporal Degeneration Center (FTDC) of the University of Pennsylvania from 2014 to 2021. Each patient was discussed at a FTDC consensus conference to arrive at their diagnosis using current clinical criteria as previously described^1,15,27,28. Patients enrolled in this study were co-enrolled into an ongoing FTDC observational study (AG066597) that includes deep phenotyping with cerebrospinal fluid (CSF) biomarkers, genetic analysis, and brain donation when possible. All clinical evaluations were performed masked to ophthalmic imaging data, and clinical data was obtained from the Penn Integrated Neurodegenerative Disease Database (INDD). Disease duration was defined by the number of years from the first reported symptom to the time of the Spectral-domain optical coherence tomography (OCT) imaging.

All probable FTLD-Tau (pFTLD-Tau) patients were negative for a CSF biomarker for ADNC (see below) and also met one of the following criteria: a) diagnosis highly associated with FTLD-Tau including progressive supranuclear palsy, nonfluent primary progressive aphasia (PPA), corticobasal syndrome (with negative CSF for ADNC and no GRN pathogenic variant), b) presence of a pathogenic MAPT (OMIM 157,140) variant for FTLD-Tau, c) neuropathologic data confirming FTLD-Tau^3,28–32. The pFTLD-Tau patients had no neuropathologic data indicating significant FTLD-TDP or ADNC (i.e. intermediate or high level) co-pathology³³.

Probable ADNC (pADNC) patients were positive for a CSF biomarker for ADNC and had a diagnosis consistent with ADNC: amnestic mild cognitive impairment (MCI), AD, or an FTD spectrum clinical diagnosis often associated with ADNC (posterior cortical atrophy, logopenic variant of PPA, corticobasal syndrome). The pADNC patients also had no genetic (i.e. MAPT, C9orf72, GRN pathogenic variant) or neuropathologic data indicating FTLD-Tau or FTLD-TDP copathology.

As previously described, CSF was obtained with standard lumbar puncture^15,19. In more recently enrolled patients (N = 28), CSF was analyzed using the second generation Fujirebio Lumipulse assay¹⁰ (Fujirebio, Malvern, PA, USA) or, for legacy data, the Luminex xMAP assay (Luminex, Austin, TX, USA) was used. To categorize patients with pADNC, established cutpoints were used (Fujirebio, Aβ₄₂: Aβ₄₀ ratio ≤ 0.072; Luminex, phosphorylated-tau: Aβ₄₂ ratio > 0.08)^34,35.

All FTD patients were tested for pathogenic variants in FTD-associated genes by massively parallel sequencing (whole exome or a custom targeted sequencing panel of > 22 neurodegenerative disease associated genes including GRN, MAPT, TBK1, and VCP) and by repeat-primed PCR for C9orf72 hexanucleotide repeat expansions as previously described^36,37. Family history was classified as sporadic or familial using a three-generation pedigree history, as previously reported³⁸.

Normal controls were recruited from patients of the Scheie Eye Institute of the University of Pennsylvania as previously described¹⁵. Normal controls had no history of diabetes, neurodegenerative disease, or ocular disease that could affect the retina or choroid.

This study was approved by the University of Pennsylvania Institutional Review Board (protocol #819894, initial approval date 4/7/2014), and it followed the tenets of the Declaration of Helsinki. All subjects (or caregivers when appropriate) provided written informed consent.

Imaging protocol and image analysis

All subjects were seen at the Scheie Eye Institute and received a full dilated eye exam (BJK or TSA) to determine if there was any eye disease that may affect the choroid measurements. As pre-specified criteria, eyes were excluded if there was evidence or history of ocular disease that would affect the choroid (e.g. diabetic retinopathy, hypertensive retinopathy, central serous chorioretinopathy), high refractive error (≥ 6.00 or ≤ − 6.00 diopter spherical equivalent, or known axial length of > 26.5 mm), significant media opacity, intraocular eye surgery within 90 days of study enrollment, or poor OCT image quality. As a pre-specified image quality metric, graders also determined the percent of the choroid-sclera interface that was well-visualized. If the mean percent of well-visualized choroid-sclera interface across the scan was less than 60%, then the eye was excluded from analysis. One patient in the pFTLD-Tau group and one patient in the pADNC group had a history of diabetes with no visible diabetic retinopathy on exam.

Subjects had Spectral-domain OCT (Heidelberg Engineering, Carlsbad, CA, USA) using enhanced depth imaging (EDI) of the choroid with high-resolution (1024 A-scans per B scan) vertical and horizontal 30 degree cross-sectional images centered on the fovea³⁹. The automatic real-time (ART) was set at 100, and we required analyzed images to have an ART of at least 50, ensuring good image quality. Two masked image graders at the Scheie Image Reading Center independently measured the choroidal thickness using the Heidelberg Eye Explorer software (version 6.12.1) and then measured the choroidal vascularity index (CVI) as described below. Choroidal thickness measurements were made manually for the subfoveal choroidal thickness as well as at 500, 1000, and 1500 μm from the fovea nasally and temporally along the horizontal axis and superiorly and inferiorly along the vertical axis (Fig. 1). The choroidal thickness was measured from the outer (basal) border of the retinal pigment epithelium/Bruch’s membrane interface to the choroidal–scleral interface. If a suprachoroidal space was visible, then the measurement included the suprachoroidal space, meaning that the reference point for the measurement was where the outer boundary of the suprachoroidal space meets the inner boundary of the sclera⁴⁰. To determine the CVI, Image J (version 1.52n, National Institutes of Health) was used to determine the choroidal vessel lumen area and the total choroidal area of the horizontal B-scans as described by Agrawal et al. and Robbins et al.^41,42. Briefly, the polygon tool was used to outline the total fovea-centered choroid area that was 1500 microns in length (Fig. 2), and this was added to the region of interest manager. The image was converted into an 8-bit image, and Niblack was used with auto local threshold. Next, the color threshold function was used to delineate the vessel lumen area, and then this area was merged with the total choroid area with the “AND” function in the region of interest manager. This allowed determination of the lumen area within the total choroid area. The CVI was defined as the vessel lumen area divided by the total choroidal area.

Fig. 1.

Representative optical coherence tomography (OCT) enhanced depth imaging scan with choroidal thickness measurements. Point measurements of the choroidal thickness at the fovea and at 500, 1000, and 1500 microns from the fovea along horizontal (A) and vertical (B) fovea-centered scans.

Fig. 2.

Representative optical coherence tomography (OCT) enhanced depth imaging scan with image binarization. (A) Fovea centered OCT scan with arrows showing luminal and stromal areas within the choroid. (B) Image binarization with segmentation of a fovea-centered choroid area that is 1500 microns wide (outlined in red). (C) Segmentation of the luminal vs. stromal areas of the binarized image⁴¹. (D) Overlay of segmented luminal and stromal areas on the OCT scan.

For the choroidal measurements, the mean of the grader measurements for choroidal thickness was the final value used for analysis. However, if the measurements from two independent graders differed by > 20 percent of their mean, then the measurements were adjudicated by the Reading Center Director (ED) and this adjudicated value was used for analysis.

Statistical analysis

We pre-specified the statistical comparison of choroidal thickness and each of the choroidal vascularity index parameters (choroid luminal area, choroid area, choroidal vascularity index) for: (1) pFTLD-Tau vs. pADNC; (2) pFTLD-Tau vs. normal controls; (3) pADNC vs. normal controls. Univariable and multivariable linear regression analyses were performed without and with adjustment by age, sex, and race, which can potentially affect choroidal thickness^43,44. We also performed these comparisons by restricting to those 10 pADNC patients with a clinical diagnosis of “probable AD” as a sensitivity analyses. The presentation of these 10 patients was consistent with clinical AD as opposed to FTD. For exploratory analyses of choroid measurements and clinical data (disease duration and Aβ₄₂/ Aβ₄₀ ratio), the Pearson correlation coefficient was calculated to evaluate the correlation between choroid measurements and clinical data. Since most patients contributed two study eyes, the inter-eye correlation was accounted for by using generalized estimating equations^45,46. All the statistical analyses were performed in SAS v9.4 (SAS Institute Inc., Cary, NC, USA), and two-sided p < 0.05 (without correcting for multiple comparisons) was considered statistically significant.

The study sample size of 42 eyes from 25 patients with pFTLD-Tau and 49 eyes from 26 patients with pADNC provided 80% power to detect a mean difference of 70 microns in subfoveal choroidal thickness, assuming a standard deviation of 89 microns. A difference between these groups of about 70 microns or more can be expected based on other publications of choroidal thinning observed in clinical AD patients compared to normal controls^24,25.

Results

Subject characteristics

Thirty-three pFTLD-Tau patients, 33 pADNC patients, and 58 normal controls were enrolled. After pre-specified exclusion criteria were applied, there were 25 (42 eyes) pFTLD-Tau patients, 26 (49 eyes) pADNC patients, and 53 (80 eyes) normal controls (Fig. 3). The demographics for these subjects are shown in Table 1. The pFTLD-Tau patients were older than the pADNC and normal control groups with mean ages of 66.8, 61.9, and 58.1 years for the pFTLD-Tau, pADNC, and normal controls, respectively. While the race of the pFTLD-Tau and pADNC groups was similar, there was a significant race difference for the normal controls vs. the other groups, as thirty-eight percent of the controls were African-American. There was no significant difference in sex between the groups. The median disease duration (interquartile range [IQR]) was 3.0 (2.0–4.0) years for the pFTLD-Tau patients and 4.0 (3.0–6.0) years for the pADNC patients. The median (IQR) mini-mental state exam (MMSE) was 27.5 (24.0–29.0) for the pFTLD-Tau patients and 21.5 (16.0–25.0) for the pADNC patients. The median (IQR) clinical dementia rating for FTD (CDR-FTD) sum of boxes score was 6.5 (3.5–8.5) for the pFTLD-Tau patients and 5.0 (3.3–7.5) for the pADNC patients. While eyes with high refractive error were excluded, axial lengths were also available for 16 pFTLD-Tau eyes, 30 pADNC eyes, and 15 normal control eyes. For these eyes, there was no significant difference in the mean axial length (23.7, 24.2, and 24.3 mm for the pFTLD-Tau, pADNC, and normal control eyes, respectively, p = 0.42).

Fig. 3.

Flow Diagram of Patients and Eyes Excluded. Pre-specified exclusion criteria were followed for each of the subject groups. *Unilateral eye disease included: Glaucoma (N = 2 pADNC patient eyes; N = 2 normal control eyes), Epiretinal membrane (N = 5 normal control eyes), Age-related macular degeneration (N = 3 pFTLD-Tau patient eyes; N = 3 normal control eyes), Amblyopia (N = 3 normal control eyes), Retinal tear or hole/retinal detachment (N = 4 normal control eyes), Central retinal artery occlusion (N = 1 normal control eye), Unilateral high refractive error (N = 2 normal control eyes), Cataract (N = 1 normal control eye), Subject only wanted one study eye due to time constraints (N = 3 normal control eyes). Abbreviations: AD = Alzheimer’s disease; AMD = age-related macular degeneration; CRAO = central retinal artery occlusion; ERM = epiretinal membrane; pADNC = probable Alzheimer’s disease neuropathologic change; pFTLD-Tau = probable frontotemporal lobar degeneration tauopathy.

Table 1.

Demographics and clinical diagnoses for pFTLD-Tau and pADNC patients.

	pFTLD-Tau* (N = 25 subjects)	pADNC** (N = 26 subjects)	Normal Controls (N = 53 subjects)	P-Value pFTLD-Tau vs. pADNC	P-Value FTLD-Tau vs. Normal Controls	P-Value pADNC vs. Normal Controls
Age (years)				0.02	0.001	0.33
Mean (SD)	66.8 (6.4)	61.9 (7.3)	58.1(12.1)
Q1, Q3	63.0, 71.0	58.0, 66.0	53.0, 66.0
Range	54.0–79.0	49.0–76.0	26.0–82.0
Sex				0.12	0.74	0.13
F	17 (68.0%)	12 (46.2%)	34 (64.2%)
M	8 (32.0%)	14 (53.8%)	19 (35.8%)
Race				0.23	0.004	0.003
Asian	0	1 (3.8%)	1 (1.9%)
Black	1 (4%)	0	20 (37.7%)
Other	2 (8%)	0	2 (3.8%)
Unknown	3 (12%)	1 (3.8%)	0
White	19 (76%)	24 (92.3%)	30 (56.6%)
Clinical diagnosis	PSP = 14	AD = 10
	CBS = 4	CBS = 3
	bvFTD = 2***	lvPPA = 2
	naPPA = 5	bvFTD = 4
		PCA = 5
		MCI (non-amnestic) = 2

* Includes 8 patients with neuropathologic confirmation of FTLD-Tau.

** Includes 4 patients with a neuropathologic confirmation of Alzheimer’s disease.

***One bvFTD had MAPT E10 + 16 C > T and the other bvFTD patient had MAPT c.1216C > T, p.R406W.

5 pFTLD-Tau patients were unable to have CSF collection for logistical reasons: 1 had both MAPT pathogenic variant and FTLD-Tau neuropathology, 1 had FTLD-Tau neuropathology, and the other 3 patients had PSP, which is highly associated with FTLD-Tau^1,28.

AD Alzheimer’s disease, pADNC probable Alzheimer’s disease neuropathologic change, bvFTD behavioral variant of frontotemporal dementia, CBS corticobasal syndrome, pFTLD-Tau probable frontotemporal lobar degeneration tauopathy, lvPPA logopenic variant of primary progressive aphasia, MCI mild cognitive impairment, naPPA nonfluent/agrammatic primary progressive aphasia, PCA posterior cortical atrophy, PSP progressive supranuclear palsy.

The 25 pFTLD-Tau patients included 8 patients with neuropathologic confirmation of FTLD-Tau. Among the patients with autopsy data, there was one patient with a MAPT c.1216C > T, p.R406W pathogenic variant and one patient with a MAPT E10 + 16 C > T pathogenic variant. The clinical diagnoses for the pFTLD-Tau patients are shown in Table 1 and included 14 patients with a diagnosis of progressive supranuclear palsy.

The 26 pADNC patients included a range of clinical syndromes often associated with underlying ADNC (Table 1). All pADNC patients had a positive CSF biomarker for ADNC, and 4 patients had neuropathologic confirmation of ADNC.

Choroidal thickness analyses

The unadjusted mean subfoveal choroidal thickness for the pFTLD-Tau, pADNC, and normal control groups was 308.9, 286.0, and 301.5 μm, respectively (all p > 0.05 for each group comparison, Tables 2, 3, 4). After adjusting for age, sex, and race, there was still no significant difference in the subfoveal choroidal thickness between each of the groups. Similarly, the choroidal thickness at twelve other locations along fovea-centered horizontal and vertical scans also showed no significant differences between the groups (all p > 0.05, Tables 2, 3, 4). There was still no significant difference for pFTLD-Tau vs. pADNC and pFTLD-Tau vs. normal controls after limiting the pFTLD-Tau patients to only those patients with a neuropathologic confirmation of tauopathy or PSP, which is highly associated with FTLD-Tau (all p > 0.05, data not shown).

Table 2.

Choroidal evaluation of pFTLD-Tau and pADNC patients.

	Unadjusted			Adjusted*
	pFTLD-Tau (n = 25; 42 eyes) Mean(SE)	pADNC (n = 26; 49 eyes) Mean(SE)	P-Value	pFTLD-Tau (n = 25; 42 eyes) Mean(SE)	pADNC (n = 26; 49 eyes) Mean(SE)	Difference (95% CI)	P-Value
Choroidal Thickness (microns)
Subfoveal	308.9 (15.6)	286.0 (19.1)	0.35	315.7 (16.6)	279.3 (22.7)	36.4 (− 17.0, 89.7)	0.18
500 microns (Inferior)	292.6 (17.2)	286.5 (19.2)	0.81	308.2 (19.9)	286.8 (21.8)	21.5 (− 27.4, 70.3)	0.39
500 microns (Nasal)	294.0 (14.8)	277.9 (17.7)	0.49	304.3 (16.3)	275.2 (20.8)	29.1 (− 18.3, 76.4)	0.23
500 microns (Superior)	295.1 (16.3)	292.5 (19.6)	0.92	310.0 (20.3)	290.4 (22.1)	19.6 (− 30.1, 69.2)	0.44
500 microns (Temporal)	289.9 (13.2)	278.4 (17.4)	0.60	301.6 (14.1)	276.2 (19.5)	25.4 (− 21.4, 72.1)	0.29
1000 microns (Inferior)	291.8 (16.8)	284.9 (19.9)	0.79	307.5 (19.7)	282.7 (21.0)	24.8 (− 23.0, 72.6)	0.31
1000 microns (Nasal)	278.5 (13.9)	268.9 (17.3)	0.66	288.1 (15.3)	263.6 (19.4)	24.5 (− 22.1, 71.2)	0.30
1000 microns (Superior)	291.9 (14.7)	294.7 (19.0)	0.91	309.3 (18.9)	296.0 (20.9)	13.3 (− 34.8, 61.4)	0.59
1000 microns (Temporal)	291.3 (13.8)	272.1 (16.3)	0.37	295.9 (13.4)	267.6 (18.8)	28.3 (− 17.8, 74.4)	0.23
1500 microns (Inferior)	293.0 (16.8)	280.3 (20.6)	0.63	308.8 (19.1)	276.1 (20.3)	32.7 (− 18.1, 83.4)	0.21
1500 microns (Nasal)	255.0 (14.9)	248.0 (16.8)	0.76	264.1 (14.7)	244.3 (18.8)	19.8 (− 27.3, 66.9)	0.41
1500 microns (Superior)	285.6 (12.1)	293.2 (18.5)	0.73	304.4 (16.5)	298.9 (20.3)	5.5 (− 39.6, 50.6)	0.81
1500 microns (Temporal)	285.2 (14.5)	267.1 (16.3)	0.41	290.3 (12.6)	266.7 (18.1)	23.6 (− 24.2, 71.4)	0.33
Average of each of the point measurements of choroidal thickness	292.5 (14.3)	286.6 (18.7)	0.80	304.7 (17.0)	287.7 (20.3)	17.0 (− 30.8, 64.8)	0.49
Choroidal Vascularity Index Parameters
Choroid luminal area (square pixels)	337,184.8 (17,443.9)	328,446.3 (21,829.7)	0.75	348,633.4 (20,771.3)	324,246.9 (25,441.4)	24,386.5 (− 33,176.5, 81,949.5)	0.41
Choroid area (square pixels)	465,682.6 (23,574.9)	458,840.7 (30,606.4)	0.86	488,878.2 (28,229.0)	460,562.7 (33,245.2)	28,315.5 (− 50,543.6, 107,175.0)	0.48
Choroidal vascularity index	0.72 (0.01)	0.72 (0.01)	0.55	0.71 (0.01)	0.71 (0.01)	0.01 (− 0.02, 0.04)	0.53

*Adjusted by age, sex, and race.

pADNC probable Alzheimer’s disease neuropathologic change, pFTLD-Tau probable frontotemporal lobar degeneration tauopathy, SE standard error.

Table 3.

Choroidal evaluation of pFTLD-Tau and normal controls.

	Unadjusted			Adjusted*
	pFTLD-Tau (n = 25; 42 eyes) Mean(SE)	Normal Controls (n = 53; 80 eyes) Mean(SE)	P-Value	pFTLD-Tau (n = 25; 42 eyes) Mean(SE)	Normal Controls (n = 53; 80 eyes) Mean(SE)	Difference (95% CI)	P-Value
Choroidal thickness (microns)
Subfoveal	308.9 (15.6)	301.5 (12.6)	0.71	314.5 (17.1)	291.6 (11.8)	22.8 (− 16.7, 62.4)	0.26
500 microns (Inferior)	292.6 (17.2)	283.0 (11.8)	0.64	299.0 (16.7)	275.1 (11.7)	23.8 (− 14.0, 61.7)	0.22
500 microns (Nasal)	294.0 (14.8)	299.7 (13.3)	0.77	300.4 (16.2)	289.7 (12.5)	10.7 (− 28.3, 49.7)	0.59
500 microns (Superior)	295.1 (16.3)	280.8 (11.3)	0.47	299.6 (16.0)	272.5 (11.0)	27.1 (− 9.8, 64.0)	0.15
500 microns (Temporal)	289.9 (13.2)	298.2 (12.5)	0.65	301.1 (14.9)	289.0 (11.3)	12.1 (− 23.9, 48.1)	0.51
1000 microns (Inferior)	291.8 (16.8)	284.8 (12.0)	0.74	303.1 (16.9)	278.0 (11.8)	25.1 (− 13.3, 63.4)	0.20
1000 microns (Nasal)	278.5 (13.9)	290.4 (13.4)	0.54	288.0 (15.5)	281.3 (12.6)	6.7 (− 31.8, 45.2)	0.73
1000 microns (Superior)	291.9 (14.7)	289.2 (11.7)	0.89	299.3 (14.9)	281.4 (10.9)	17.9 (− 17.2, 53.0)	0.32
1000 microns (Temporal)	291.3 (13.8)	294.8 (12.3)	0.85	300.6 (15.6)	285.6 (11.0)	15.1 (− 20.6, 50.7)	0.41
1500 microns (Inferior)	293.0 (16.8)	288.9 (11.8)	0.84	304.3 (16.8)	282.4 (11.6)	22.0 (− 15.8, 59.7)	0.25
1500 microns (Nasal)	255.0 (14.9)	263.1 (13.0)	0.68	262.4 (15.7)	254.6 (12.2)	7.8 (− 31.1, 46.7)	0.69
1500 microns (Superior)	285.6 (12.1)	290.6 (11.9)	0.77	299.2 (13.2)	283.2 (10.5)	16.0 (− 15.9, 48.0)	0.33
1500 microns (Temporal)	285.2 (14.5)	283.0 (11.0)	0.91	291.0 (15.8)	274.8 (10.3)	16.2 (− 19.3, 51.7)	0.37
Average of each of the point measurements of choroidal thickness	292.5 (14.3)	293.1 (11.8)	0.97	300.8 (15.4)	284.2 (10.8)	16.6 (− 19.3, 52.5)	0.36
Choroidal vascularity index parameters
Choroid luminal Area (square pixels)	337,184.8 (17,443.9)	338,560.5 (14,247.8)	0.95	350,322.8 (19,321.7)	325,122.8 (12,489.1)	25,200.0 (− 18,841.2, 69,241.2)	0.26
Choroid area (square pixels)	465,682.6 (23,574.9)	466,322.7 (19,534.8)	0.98	486,511.7 (25,832.2)	448,035.2 (17,009.6)	38,476.4 (− 21,677.2, 98,630.1)	0.21
Choroidal vascularity index	0.72 (0.01)	0.73 (0.01)	0.86	0.72 (0.01)	0.73 (0.01)	− 0.01 (− 0.03, 0.02)	0.66

*Adjusted by age, sex, and race.

pFTLD-Tau probable frontotemporal lobar degeneration tauopathy, SE standard error.

Table 4.

Choroidal evaluation of pADNC and normal controls.

	Unadjusted			Adjusted*
	pADNC (n = 26; 49 eyes) Mean(SE)	Normal Controls (n = 53; 80 eyes) Mean(SE)	P-Value	pADNC (n = 26; 49 eyes) Mean(SE)	Normal Controls (n = 53; 80 eyes) Mean(SE)	Difference (95% CI)	P-Value
Choroidal thickness (microns)
Subfoveal	286.0 (19.1)	301.5 (12.6)	0.50	301.3 (19.1)	294.4 (11.6)	6.8 (− 37.2, 50.8)	0.76
500 microns (Inferior)	286.5 (19.2)	283.0 (11.8)	0.88	305.2 (17.6)	280.4 (10.9)	24.9 (− 14.4, 64.1)	0.21
500 microns (Nasal)	277.9 (17.7)	299.7 (13.3)	0.33	293.2 (18.5)	292.9 (12.3)	0.4 (− 42.8, 43.5)	0.99
500 microns (Superior)	292.5 (19.6)	280.8 (11.3)	0.60	312.2 (17.9)	277.0 (10.4)	35.2 (− 4.5, 74.9)	0.08
500 microns (Temporal)	278.4 (17.4)	298.2 (12.5)	0.36	295.7 (17.3)	291.4 (11.2)	4.3 (− 35.4, 44.0)	0.83
1000 microns (Inferior)	284.9 (19.9)	284.8 (12.0)	1.00	307.0 (17.7)	282.6 (10.9)	24.3 (− 16.0, 64.6)	0.24
1000 microns (Nasal)	268.9 (17.3)	290.4 (13.4)	0.33	287.1 (17.8)	284.0 (12.4)	3.1 (− 40.6, 46.8)	0.89
1000 microns (Superior)	294.7 (19.0)	289.2 (11.7)	0.81	317.7 (17.5)	285.6 (10.3)	32.1 (− 6.6, 70.9)	0.10
1000 microns (Temporal)	272.1 (16.3)	294.8 (12.3)	0.27	291.5 (16.6)	287.8 (10.9)	3.7 (− 33.0, 40.4)	0.84
1500 microns (Inferior)	280.3 (20.6)	288.9 (11.8)	0.72	299.8 (18.4)	287.7 (10.8)	12.1 (− 29.5 53.7)	0.57
1500 microns (Nasal)	248.0 (16.8)	263.1 (13.0)	0.48	266.5 (17.5)	257.5 (11.8)	9.0 (− 31.6, 49.7)	0.66
1500 microns (Superior)	293.2 (18.5)	290.6 (11.9)	0.91	321.8 (16.8)	287.3 (9.9)	34.5 (− 2.4, 71.4)	0.07
1500 microns (Temporal)	267.1 (16.3)	283.0 (11.0)	0.42	282.7 (16.5)	276.6 (10.2)	6.2 (− 29.1, 41.5)	0.73
Average of each of the point measurements of choroidal thickness	286.6 (18.7)	293.1 (11.8)	0.77	306.6 (17.6)	287.2 (10.5)	19.4 (− 20.2, 58.9)	0.34
Choroidal vascularity index parameters
Choroid luminal Area (square pixels)	328,446.3 (21,829.7)	338,560.5 (14,247.8)	0.70	349,300.9 (20,702.8)	330,866.4 (12,285.1)	18,434.4 (− 28,677.4, 65,546.2)	0.44
Choroid area (square pixels)	458,840.7 (30,606.4)	466,322.7 (19,534.8)	0.84	491,156.4 (28,160.3)	454,932.7 (16,681.9)	36,223.7 (− 28,614.5, 101,062.0)	0.27
Choroidal vascularity index	0.72 (0.01)	0.73 (0.01)	0.39	0.71 (0.01)	0.73 (0.01)	− 0.02 (− 0.04, 0.01)	0.17

*Adjusted by age, sex, and race.

pADNC probable Alzheimer’s disease neuropathologic change, SE standard error.

Choroidal vascularity index analyses

The unadjusted mean CVI for the pFTLD-Tau, pADNC, and normal control groups was 0.72, 0.72, and 0.73, respectively (all p > 0.05 for each group comparison, Tables 2, 3, 4). After adjusting for age, sex, and race, there was still no significant difference in the CVI between each of the groups (Tables 2, 3, 4).

Sensitivity analysis and exploratory correlation analyses

Given the different clinical syndromes composing the pADNC group, we also evaluated the choroid of those pADNC patients with a clinical diagnosis of “probable AD” (10 patients, 18 eyes) with a sensitivity analysis. These 10 patients had no significant differences in the choroidal thickness measurements or CVI compared to normal controls or the pFTLD-Tau patients (all p > 0.05, Supplementary Tables 1–4).

The majority of pADNC patients had the Fujirebio Aβ₄₂/ Aβ₄₀ ratio (Fujirebio) CSF biomarker (N = 14 pADNC patients). Among the pADNC patients, there was no correlation of choroidal thickness with the Aβ₄₂/ Aβ₄₀ ratio (all p > 0.05, Supplementary Table 5). However, the CVI showed a mild negative correlation that was borderline significant (Pearson correlation coefficient = − 0.36; p = 0.06).

The choroidal thickness measurements did not correlate with disease duration for the pFTLD-Tau (all p > 0.05, Supplementary Table 6) or pADNC patients (all p > 0.05, Table 5). The CVI also did not correlate with disease duration for the pFTLD-Tau patients (Pearson correlation coefficient = − 0.03, p = 0.81). However, the CVI had a positive correlation with disease duration among the pADNC patients (Pearson correlation coefficient = 0.32, p = 0.04, Table 5, Supplementary Fig. 1).

Table 5.

Correlation of choroidal thickness and choroidal vascularity index with disease duration among pADNC patients.

	pADNC (n = 26 patients; 49 eyes)
	N of eyes	Pearson correlation coefficient	P-Value *
Choroidal thickness
Subfoveal	46	− 0.11	0.64
500 microns (Inferior)	46	− 0.19	0.39
500 microns (Nasal)	45	− 0.15	0.53
500 microns (Superior)	46	− 0.23	0.28
500 microns (Temporal)	45	− 0.08	0.70
1000 microns (Inferior)	46	− 0.26	0.21
1000 microns (Nasal)	45	− 0.16	0.46
1000 microns (Superior)	46	− 0.20	0.34
1000 microns (Temporal)	45	− 0.09	0.66
1500 microns (Inferior)	46	− 0.22	0.27
1500 microns (Nasal)	45	− 0.16	0.46
1500 microns (Superior)	46	− 0.20	0.32
1500 microns (Temporal)	45	− 0.04	0.85
Average of each of the point measurements of choroidal thickness	49	− 0.22	0.28
Choroidal vascularity index parameters
Choroid luminal area	49	− 0.15	0.47
Choroid area	49	− 0.20	0.32
Choroidal vascularity index	49	0.32	0.04

*Generalized estimating equations account for the inter-eye correlation.

pADNC probable Alzheimer’s disease neuropathologic change.

Discussion

Biomarkers to distinguish FTLD-Tau pathology from ADNC and other age-associated neurodegenerative pathologies are a high priority for proper enrollment of FTLD-Tau patients into clinical trials leading to drug development. Since some studies have found choroidal thinning in clinical AD compared to normal controls potentially related to an amyloid beta vasculopathy^{21,22,24,25,47}, it is important to determine if the choroid could be a biomarker to distinguish FTLD-Tau from ADNC or from normal controls. Using deeply phenotyped patients including a CSF biomarker for all patients and neuropathologic diagnosis confirmation for almost one third of our FTLD-Tau patients, we were unable to find significant differences of the choroid in our comparison of pFTLD-Tau, pADNC, and normal controls. To our knowledge, this is the first study to evaluate the choroid of pFTLD-Tau patients and directly compare to pADNC patients. One study has evaluated the choroid of progressive supranuclear palsy patients compared to normal controls and also did not find a significant difference⁴⁸.

We hypothesized that a thinner choroid among pADNC patients may differentiate them from FTLD-Tau and normal controls. There are several reasons to explain why we did not find choroidal thinning in our pADNC group compared to these other groups. Our pADNC group was composed of atypical and younger patients that are more likely to be diagnostically confused with FTD, rather than more typical amnestic late-onset disease. It is possible that choroidal thinning may not be associated with ADNC at a younger age and that some subtypes of ADNC may not have abnormal choroids. Additionally, although our study was powered appropriately, ADNC may be associated with choroidal thinning with an effect size that may have been too small for detection. However, it is worth noting that there is no clearly elucidated mechanism for why human choroidal thinning might be associated with ADNC or for why Aβ₄₂ would be in the choroid. Although animal models of ADNC have shown amyloid beta within the choroid and some investigators have suggested an Aβ₄₂ vasculopathy mechanism^49,50, we point out that the presence of Aβ₄₂ has not been shown in the choroid of human tissue to our knowledge. Animal models may not always accurately represent the complex pathophysiology of ADNC. It is quite possible that ADNC does not uniformly associate with abnormal choroidal thinning. While some groups have found choroidal thinning in clinical AD, a significantly thin choroid or decreased CVI has not been a consistent OCT finding^26,51,52. Further, those finding choroidal thinning have not consistently seen a correlation with disease severity measured by psychometric scores²⁵. Regarding choroidal thickness studies of human tissue, Asanad et al. evaluated the choroid of 8 patients with ADNC and 11 normal controls. They found choroidal thinning nasal to the optic nerve, but there actually was a thicker choroid at the macula and the temporal periphery⁵³. This contrasted with an earlier report of choroidal tissue showing thinning of the choroid in 6 patients with ADNC compared to 6 normal controls⁴⁹. Lastly, it may be worth distinguishing OCT imaging studies with ADNC determined by biomarkers (CSF, amyloid beta PET, or genetic pathogenic variants) in contrast to studies of clinical AD. There are a very limited number of other choroid studies with biomarker determined ADNC groups, and these have not consistently found choroidal thinning compared to controls^23,26. Using amyloid-positive patients (positive CSF biomarker or amyloid PET indicating ADNC) compared to normal controls, Den Haan et al. did not observe choroidal thickness abnormalities among their patients with ADNC and suggested that the choroidal thinning seen in other studies without biomarker-determined ADNC may have represented dementia cases with a primary vascular etiology or cases with vascular co-pathology²⁶. Another study compared a group of MCI patients and ADNC patients with a positive amyloid PET scan to healthy controls. They found mild choroidal thinning at 5 of 14 choroid locations when comparing the MCI and ADNC patients together vs. healthy controls, but did not see significant choroidal thinning when comparing the ADNC patients vs. healthy controls²³. Further work with biomarker-determined populations and human tissue analysis is critical and could clarify these ongoing questions.

Importantly, our study also found no significant difference between the choroid of pFTLD-Tau patients and normal controls. While other studies have evaluated the neurosensory retina in detail^{15,16,18–20}, there has been a need for investigations of the choroid of FTLD-Tau. Although some literature suggests that a tauopathy could cause a vasculopathy^54,55, phosphorylated tau has not been shown in the choroid of FTLD-Tau patients to our knowledge. Additional studies are needed to determine if FTLD-TDP has choroidal abnormalities, and comparisons of the choroid of FTLD-Tau and FTLD-TDP patients may be worthwhile. There is limited choroidal thickness data for amyotrophic lateral sclerosis (ALS), which is a TDP-43 proteinopathy related to FTLD-TDP; one study showed subfoveal choroidal thickening in ALS compared to healthy controls⁵⁶.

Interestingly, with exploratory analyses, our data show a mild positive correlation of CVI with disease duration among our pADNC patients (p = 0.04), and consistent with this, there also was a negative correlation of CVI with the CSF Abeta42/Abeta40 ratio that approached statistical significance (p = 0.06). The correlation of CVI with disease duration suggests that there was increased choroid vessel luminal area as the disease duration increased. This positive correlation was unexpected, but it is noteworthy that some other studies have suggested an amyloid beta vasculopathy may be more complex over time than simply a progressive loss of vasculature when patients are at an earlier stage of disease^52,57. Data from others have shown that earlier, subclinical stages of AD actually have increasing cerebral blood flow over time that later switches to a hypoperfused state of decreasing blood flow as the disease progresses⁵⁸. This is relevant in light of the young age of our pADNC patients. As noted above, there is limited data with tissue analysis, with one study demonstrating choroidal thickening with increased vascularity at the macula in advanced cases that had brain pathology showing frequent neuritic plaques and Braak stage V or VI⁵³. We believe our finding should be interpreted with caution as there were multiple comparisons, but along with these other studies, it suggests that longitudinal investigations are needed to help elucidate the relationship of CVI with ADNC.

Our study has both strengths and limitations. Of significant value is the deep phenotyping of patients conducted by our study, including dilated eye exams, the use of CSF biomarkers for ADNC to differentiate the pADNC and pFTLD-Tau groups, genetic analysis, and autopsy data for one third of the pFTLD-Tau group. Further, we performed rigorous analysis with masked manual grading of choroidal images by two independent, trained graders, followed by adjudication with the reading center director if there was a significant difference of measurements. We also measured the CVI, which is an automated image binarization technique that complements the manual choroidal thickness measurements and may have advantages over specific point measurements⁵². Several study limitations should be noted. First, p-values were not corrected for multiple comparisons, and the correlation of disease duration with CVI among pADNC patients was an exploratory analysis that is hypothesis generating. Additional investigations would be needed before making conclusions about CVI and disease duration of pADNC patients. Second, we did not have axial length data for all eyes. However, among eyes with axial lengths, there was no difference between groups, and eyes with high refractive error were excluded. Third, our normal controls did not have cognitive or genetic testing, and thus they were not phenotyped to the extent of our neurodegenerative cases. Although unlikely, we cannot exclude the possibility of a small number of preclinical neurodegenerative cases within our normal control group that could hinder our ability to see a difference between cases and controls. Fourth, our study had a modest sample size. However, FTLD-Tau patients are uncommon and our study was well-powered to find a choroid difference among the compared groups based on other studies^24,25. We would argue that a clinically useful choroid biomarker to distinguish FTLD-Tau from ADNC would likely be observed with this sample size.

Our study did not identify a choroidal measurement to distinguish FTLD-Tau or ADNC from each other or from normal controls. However, our data are consistent with other studies using biomarker determined ADNC groups. It may be worthwhile for future studies to compare the choroid of atypical ADNC groups with typical late-onset ADNC and to compare FTLD-Tau with FTLD-TDP. Finally, longitudinal choroidal thickness and CVI data, investigation of autopsy-defined groups, and ocular tissue analysis could improve our understanding of whether the choroid has abnormalities in these patients.

Funding

Supported by NIH (Bethesda, MD) grants including RF1AG083774, AG017586, NS053488, AG052943, 2-P30-EY01583-26, NS109260 (to D.J.I.), AG054519, 5P30AG072979, P01-AG066597 Penn Institute on Aging. Funding was also provided in the form of block grants for general research purposes to the Scheie Eye Institute by Research to Prevent Blindness (New York, NY) and the Paul and Evanina Mackall Foundation Trust (Chicago, IL).

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request. Raw data can be made available on request.

Declarations

Competing interest

The authors declare no competing interests.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-02020-z.