Ophthalmic conditions associated with dementia risk: the Cardiovascular Health Study

Abstract

INTRODUCTION:

Ophthalmic conditions and dementia appear to overlap and may share common pathways, but research has not differentiated dementia subtypes.

METHODS:

Diagnoses of cataracts, age-related macular degeneration (AMD), diabetic retinopathy (DR), and glaucoma were based on medical histories and ICD-9 codes for 3,375 participants from the Cardiovascular Health Study Cognition Study. Dementia, including Alzheimer’s disease (AD) and vascular dementia (VaD), was classified using standardized research criteria.

RESULTS:

Cataracts were associated with AD (HR = 1.34; 95% CI = 1.01 – 1.80) and VaD/mixed dementia (HR = 1.41; 95% CI = 1.02 – 1.95). AMD was associated with AD only (HR = 1.87; 95% CI = 1.13 – 3.09), whereas DR was associated with VaD/mixed dementia only (HR = 2.63; 95% CI = 1.10 – 6.27).

DISCUSSION:

Differential associations between specific ophthalmic conditions and dementia subtypes may elucidate pathophysiologic pathways. Lack of association between glaucoma and dementia was most surprising from these analyses.

1. Introduction

Dementia is a major cause of disability and dependency among older populations [1]. The most common causes of dementia include Alzheimer’s disease (AD) and vascular dementia (VaD) [2]. Despite extensive research, many aspects of their pathophysiology and clinical course remain unknown. Better understanding of disease mechanisms, predictors, and risk factors may lead to early detection and preventive measures.

The brain and the eye share many characteristics, including embryologic origin, precise neuronal cell layers, complex neurochemistry and neurotransmitter systems, microglia, astroglia, blood barriers, and microvasculature [3]. As an extension of the central nervous system, the eye may offer opportunities to capture critical information on the health of the brain. Animal and imaging studies suggest that cataracts, glaucoma, age-related macular degeneration (AMD), and diabetic retinopathy (DR) may be associated with dementia due to shared characteristics, such as progressive neurodegeneration, characteristic β-amyloid (Aβ) deposits, and chronic microvascular lesions [4,5]. The eye, unlike the brain, can be examined physically in vivo, and may provide a window into processes in the brain.

Findings from previous epidemiologic studies that examined associations between glaucoma, AMD, or DR and dementia have been inconsistent [6–13]. Very few studies have investigated the relative associations of all these ophthalmic conditions with risk for various types of dementia, and have not included cataracts as predictors. One recent prospective cohort study found that glaucoma, AMD, and DR were associated with the clinical development of AD, but cataracts were not associated with AD or all-cause dementia [14]. Despite the contribution of vascular disease to dementia and retinal disorders [15,16], knowledge about the relationships between these ophthalmic conditions and VaD is also limited. The Cardiovascular Health Study (CHS) allowed us to quantify associations between these specific ophthalmic conditions and dementia because it had reliable methods of ascertainment. It identified diagnoses of common ophthalmic conditions by medical records, including linkage to participants’ Medicare Part B records, which allowed capture of outpatient-based diagnoses not collected by CHS. The major dementia subtypes, AD and VaD, were adjudicated using research diagnostic criteria, and included neuroimaging. We sought to evaluate associations between cataracts, glaucoma, AMD, and DR and risk of all-cause dementia, AD, and VaD/mixed dementia.

2. Methods

CHS is a population-based prospective study of 5,888 adults 65 years and older [17]. In 1989–1990, 5,201 participants were recruited into CHS during its initial wave from Medicare eligibility lists in four US communities. In 1992–1993, 687 predominantly African-American (AA) participants were recruited. Informed consent was obtained from all the participants at entry into the study and at periodic intervals. Institutional review board approval was obtained at all sites collecting and analyzing data.

2.1. Data collection

From baseline in 1989–1990 until 1998–1999, up to 10 annual clinic visits were completed. Data collected at these examinations included demographics, anthropometric measurements, vital signs, cognitive function, information from psychosocial interviews, depression level, medical history, and physical function. Genotyping of apolipoprotein E (ApoE) was performed from a blood sample.

2.2. Assessment of cognitive function and dementia

In 1998–1999, dementia was classified in 3,602 participants as part of the CHS Cognition Study (Supplementary Figure 1) [18]. Inclusion in the CHS Cognition Study cohort required completion of cranial magnetic resonance imaging (MRI) and the 100-point modified Mini-Mental State Examination (3MSE) in 1991–1994 (Figure 1). These participants were screened using data collected at the visit closest to MRI to identify those who were suspected of having cognitive impairment and who were asked to return to the clinic for additional cognitive testing. Description of the procedures used to evaluate cognition have been published previously [19]. Efforts were made by trained clinical study staff at each cognitive assessment to evaluate vision problems [20]. This information was recorded so that the effects of poor visual acuity in the participants’ cognitive performance were taken into consideration by the adjudication committee [7].

Figure 1.

Timeline of the Cardiovascular Health Study-Cognition Study

Dementia classification was completed by consensus of a panel of neurologists and psychiatrists using data from the neuropsychiatric tests and other data, which have been described elsewhere [19]. Dementia was defined using criteria similar to the one used in the Diagnostic and Statistical Manual of Mental Disorders, 4^th edition [21]. Participants who did not meet dementia criteria, but showed cognitive deficits based on neuropsychological evaluation were classified as having mild cognitive impairment (MCI), based on previously described criteria [19]. Cranial MRIs were used for classification of dementia subtype. AD was classified using National Institute of Neurological and Communicative Disorders and Stroke – Alzheimer’s Disease and Related Disorders Association criteria [22]. VaD was classified using State of California Alzheimer’s Disease Diagnostic and Treatment Centers criteria [23]. However, most VaD cases in CHS were clinical mixed dementia of Alzheimer’s and vascular type, which would not be traditionally classified as pure VaD [19]. Dementia onset was determined by review of the longitudinal data collected during the 10 years of study follow-up and by family input using the Neuropsychiatric Inventory. If the date of onset was determined to be before entry into the CHS Cognition Study cohort, the participant was determined to have prevalent dementia at baseline.

2.3. Diagnosis of ophthalmic conditions

Specific ophthalmic conditions included cataracts, glaucoma, AMD, and DR. The presence of these conditions (yes, no) was based on International Classification of Diseases, 9^th revision (ICD-9) codes (Supplementary Table 1). We also used information from medical histories collected at the first follow-up visit in CHS (and baseline visit for the AA cohort) to determine prevalent cataracts, glaucoma, and retinal diseases within the CHS Cognition Study cohort. In CHS, diagnostic codes were based on hospital discharge and inpatient services, and therefore may represent only advanced eye disease, which would result in a biased sample. To address this limitation, we used linked outpatient and physician/carrier claims data from Medicare Part B to identify diagnoses of ophthalmic conditions not available through records in CHS. CHS participants were matched to Medicare enrollment and claims data, which were available through 2015, by their social security number, sex, and date of birth.

2.4. Statistical analysis

Of the 3,602 participants in the CHS Cognition Study cohort, 227 with prevalent dementia on MRI were excluded. We tabulated descriptive statistics for demographics and comorbidities by each ophthalmic condition individually and by those who developed any ophthalmic condition compared to none. Cox proportional hazards models were used to compute hazard ratios (HRs) and 95% confidence intervals (CIs) to estimate the risk of dementia associated with cataracts, glaucoma, AMD, and DR. Participant age was used as the time axis to control more finely for age as a potential confounder. Examination of the proportionality assumption was done using Schoenfeld residuals. Time to dementia was calculated as days from baseline to dementia onset, death, or end of the CHS Cognition Study follow-up. Each ophthalmic condition was modeled as a binary time-varying covariate, equal to 0 at times prior to and 1 after the diagnosis, and was included in separate models, as well as in a combined model to adjust for the presence of other ophthalmic conditions. Time at risk was calculated separately for each covariate value (0 or 1) assumed by the participant. At any given time, participants without any of the specific ophthalmic conditions served as the reference group. All-cause dementia, probable or possible AD (without VaD), and probable or possible VaD (with or without AD)/mixed dementia were the outcomes of interest. For dementia subtype, persons were censored at onset of VaD/mixed dementia in models evaluating AD and for AD in models evaluating VaD/mixed dementia.

Models were adjusted for demographics (age at entry as a continuous variable, sex, race, and education) and cardiovascular and dementia risk factors (body mass index (BMI), smoking, alcohol intake, physical activity, total cholesterol, diabetes mellitus, hypertension, history of cardiovascular disease, history of cerebrovascular disease, and ApoE genotype), which were all measured at baseline, as well as cohort and clinic site. We also included self-reported income and accessibility to medical care, refractive errors (presbyopia, astigmatism, hypermetropia, and myopia), inflammatory markers (C-reactive protein level and interleukin-6 level), and treatments for cataracts, glaucoma, AMD, or DR as covariates in additional models. Diagnosis of refractive errors were based on ICD-9 codes and treatment of specific ophthalmic conditions, such as cataract surgery, were based on Current Procedural Terminology codes (Supplementary Table 1). To identify potential effect modifiers of the association between dementia and each ophthalmic condition, we stratified participants based on age, sex, and ApoE genotype. Age was dichotomized as <80 years or ≥80 years as the prevalence of these ophthalmic conditions is highest among individuals aged ≥80 years [24]. To assess for possible interaction, we also included in the models, interaction terms for each ophthalmic condition by age as a continuous variable, and both sex and ApoE genotype as binary variables. Finally, to evaluate potential bias due to: (1) competing risks, we used the subdistribution hazard function [25]; and (2) missing covariate data, we used multiple imputation using chained equations to create 20 imputed datasets [26]. All P-values reported in this study were two-sided with significance at P<0.05. Analyses were conducted using Stata version 15.1 (StataCorp, College Station, Texas).

3. Results

Of the 3,375 participants included in the analyses, 480 were classified with incident all-cause dementia during an average of 4.1 years of follow-up. Of these, 245 were determined to have AD only and 213 were determined to have VaD/mixed dementia. Average age of participants was 74.8 years, ranging from 65 through 97 years. More than half of participants were women and 87.3% were white. Approximately 83% of the sample developed at least one ophthalmic condition (cataracts, glaucoma, AMD, DR), whereas 17% did not develop any of these conditions during the study period. Most participants developed cataracts (1,113 incident; 1,503 prevalent), followed by AMD (465 incident; 203 prevalent) and glaucoma (311 incident; 290 prevalent), while the least number of participants developed DR (162 incident; 39 prevalent). Table 1 describes characteristics of the study sample by each ophthalmic condition. Comparing across participants who developed cataracts, glaucoma, AMD, DR, or none of these ophthalmic conditions, significant differences were found by age, race, alcohol intake, BMI, diabetes mellitus, hypertension, and ApoE genotype. Participants who developed at least one of these ophthalmic conditions tended to be white, more educated, and have no history of smoking and no ε4 allele as compared to participants who did not develop any of these conditions (Supplementary Table 2).

Table 1.

Demographic and baseline clinical factors by each ophthalmic condition (cataracts, AMD, DR, glaucoma) at end of follow-up in the CHS Cognition Study^*

Baseline factors†	Cataract diagnosis (n = 2,616)	AMD diagnosis (n = 668)	DR diagnosis (n = 201)	Glaucoma diagnosis (n = 601)	No diagnosis of cataract, AMD, DR, or glaucoma (n = 571)	p-value
Age‡, mean (SD), years	74.9 (4.9)	76.1 (5.3)	74.3 (4.3)	75.2 (4.8)	74.4 (4.9)	<0.001
Sex, No. (%)						0.988
Female	1,532 (58.6)	387 (57.9)	119 (59.2)	350 (58.2)	327 (57.3)
Male	1,084 (41.4)	381 (42.1)	82 (40.8)	251 (41.8)	244 (42.7)
Race, No. (%)						<0.001
White	2,284 (87.3)	623 (93.3)	154 (76.6)	501 (83.4)	410 (71.8)
Nonwhite	332 (12.7)	45 (6.7)	47 (23.4)	100 (16.6)	161 (28.2)
Education, No. (%)						0.228
Less than high school	613 (23.5)	176 (26.4)	55 (27.5)	138 (23.1)	146 (25.6)
High school graduate	736 (28.2)	185 (27.7)	62 (31.0)	158 (26.4)	181 (31.8)
Some college	640 (24.5)	166 (24.9)	45 (22.5)	153 (25.6)	126 (22.1)
College graduate	624 (23.9)	140 (21.0)	38 (19.0)	149 (24.9)	117 (20.5)
Smoking status, No. (%)						0.413
Never	1,246 (47.7)	319 (47.8)	100 (49.8)	290 (48.3)	258 (45.2)
Former	1,094 (41.9)	273 (40.9)	81 (40.3)	250 (41.7)	232 (40.6)
Current	272 (10.4)	76 (11.4)	20 (10.0)	60 (10.0)	81 (14.2)
Alcohol intake, mean (SD), drinks/week	2.8 (13.7)	2.3 (6.0)	1.6 (4.4)	2.6 (5.7)	3.0 (7.6)	<0.001
Body mass index, mean (SD), kg/m2	26.5 (4.4)	26.1 (4.3)	27.8 (4.7)	26.2 (4.2)	26.7 (4.2)	<0.001
Physical activity, No. (%)						0.678
No exercise	163 (6.2)	55 (8.2)	12 (6.0)	42 (7.0)	32 (5.6)
Low exercise	1,176 (45.0)	301 (45.1)	94 (46.8)	261 (43.4)	254 (44.5)
Moderate exercise	954 (36.5)	236 (35.4)	74 (36.8)	229 (38.1)	226 (39.6)
High exercise	321 (12.3)	75 (11.2)	21 (10.4)	69 (11.5)	59 (10.3)
Diabetes mellitus, No. (%)						<0.001
Normal	1,928 (74.3)	498 (74.9)	65 (32.7)	438 (73.4)	401 (70.7)
Impaired fasting glucose	329 (12.7)	76 (11.4)	19 (9.6)	87 (14.6)	88 (15.5)
Diabetes mellitus	339 (13.1)	91 (13.7)	115 (57.8)	72 (12.1)	78 (13.8)
Hypertension, No. (%)						0.005
Normal	1,197 (45.8)	296 (44.4)	70 (34.8)	270 (44.9)	234 (41.0)
Borderline	354 (13.6)	95 (14.2)	23 (11.4)	78 (13.0)	101 (17.7)
Hypertensive	1,062 (40.6)	276 (41.4)	108 (53.7)	253 (42.1)	236 (41.3)
Cardiovascular disease§, No. (%)	436 (16.7)	131 (19.6)	45 (22.4)	98 (16.3)	99 (17.3)	0.123
Cerebrovascular disease¶, No. (%)	103 (3.9)	26 (3.9)	11 (5.5)	20 (3.3)	29 (5.1)	0.318
Total cholesterol#, mean (SD), mg/dL	211.6 (38.2)	210.3 (38.6)	203.0 (38.0)	210.1 (37.3)	210.4 (39.3)	0.073
ApoE genotype, No. (%)						0.031
Presence of ε4 allele	564 (23.6)	124 (20.1)	50 (27.5)	125 (22.8)	140 (24.5)
No ε4 allele	1,822 (76.4)	493 (79.9)	132 (72.5)	422 (77.2)	384 (67.2)

AMD = Age-related macular degeneration; ApoE = Apolipoprotein E; DR = Diabetic retinopathy; SD = Standard deviation; kg/m² = kilograms per meter squared; mg/dL = milligrams per deciliter.

^*Participants could develop more than one ophthalmic condition, so the column totals do not represent mutually exclusive groups;

^†Information was collected around the time of the cranial MRI between 1992 and 1994, and was missing for some variables as follows: Education: n = 5 (0.15%), Smoking status: n = 4 (0.12%), Alcohol intake: n = 13 (0.38%), Body mass index: n = 9 (0.27%), Physical activity: n = 2 (0.06%), Diabetes mellitus: n = 25 (0.74%), Hypertension: n = 3 (0.09%), ApoE genotype: n = 286 (8.5%);

^‡Age = Age at MRI;

^§Cardiovascular disease defined as prevalent angina, congestive heart failure, coronary heart disease, claudication, or myocardial infarction at baseline;

^¶Cerebrovascular disease defined as prevalent stroke or transient ischemic attack at baseline;

^#To convert total cholesterol from milligrams per deciliter to millimoles per liter, multiply by 0.0259.

Baseline diagnoses of cataracts, AMD, and DR, but not glaucoma, were significantly associated with risk of all-cause dementia based on crude Kaplan-Meier estimates (Figure 2). In adjusted models for demographics and cardiovascular and dementia risk factors, as well as cohort and clinic site (Table 2), risk of all-cause dementia was associated with cataracts (HR: 1.38; 95% CI: 1.10 – 1.72; p = 0.005) and AMD (HR: 1.57; 95% CI: 1.07 – 2.30; p = 0.021) compared to those without any ophthalmic condition. Also, cataracts and AMD were associated with a 34% (HR: 1.34; 95% CI: 1.01 – 1.80; p = 0.039) and 87% (HR: 1.87; 95% CI: 1.13 – 3.09; p = 0.014) increased risk of AD, respectively. DR and glaucoma were not significantly associated with risk of all-cause dementia or AD. DR was associated with risk for VaD/mixed dementia (HR: 2.63; 95% CI: 1.10 – 6.27; p = 0.029). Cataracts were also associated with increased risk of VaD/mixed dementia (HR: 1.41; 95% CI: 1.02 – 1.95; p = 0.037). AMD and glaucoma were not significantly associated with risk of VaD/mixed dementia. These results did not change appreciably when excluding participants with MCI (Table 3). When including all four ophthalmic conditions in a combined model to evaluate the association of each condition with dementia, independent of other conditions, the magnitude of the associations was greater compared to results from models examining each ophthalmic condition individually (Supplementary Table 3).

Figure 2.

Survival curves for cataracts, AMD, DR, and glaucoma from the primary (individual) models for all-cause dementia

Table 2.

Risk of all-cause dementia, Alzheimer’s disease, and vascular/mixed dementia in the individual (each ophthalmic condition modeled separately) models, with reference group as those without cataracts, AMD, DR, and glaucoma

Outcome and Specific Ophthalmic Condition	Model 1*		Model 2†
All-cause dementia (n = 480)	Hazard Ratio (95% CI)	p-value	Hazard Ratio (95% CI)	p-value
Cataracts (n = 2,616)	1.26 (1.03 – 1.54)	0.026	1.38 (1.10 – 1.72)	0.005
AMD (n = 668)	1.49 (1.03 – 2.14)	0.033	1.57 (1.07 – 2.30)	0.021
DR (n = 201)	1.42 (0.67 – 3.00)	0.362	1.51 (0.70 – 3.28)	0.295
Glaucoma (n = 601)	1.18 (0.76 – 1.86)	0.459	1.00 (0.58 – 1.70)	0.991
Alzheimer’s disease‡ (n = 245)	Hazard Ratio (95% CI)	p-value	Hazard Ratio (95% CI)	p-value
Cataracts (n = 2,616)	1.25 (0.94 – 1.67)	0.119	1.34 (1.01 – 1.80)	0.039
AMD (n = 668)	1.83 (1.14 – 2.96)	0.013	1.87 (1.13 – 3.09)	0.014
DR (n = 201)	0.39 (0.05 – 2.79)	0.348	0.46 (0.06 – 3.33)	0.439
Glaucoma (n = 601)	0.93 (0.46 – 1.88)	0.836	0.69 (0.28 – 1.68)	0.410
Vascular/mixed dementia§ (n = 213)	Hazard Ratio (95% CI)	p-value	Hazard Ratio (95% CI)	p-value
Cataracts (n = 2,616)	1.30 (0.96 – 1.76)	0.084	1.41 (1.02 – 1.95)	0.037
AMD (n = 668)	1.08 (0.59 – 2.01)	0.795	1.25 (0.67 – 2.33)	0.474
DR (n = 201)	2.96 (1.30 – 6.73)	0.010	2.63 (1.10 – 6.27)	0.029
Glaucoma (n = 601)	1.35 (0.71 – 2.56)	0.355	1.12 (0.52 – 2.40)	0.767

AMD = Age-related macular degeneration; DR = Diabetic retinopathy; CI = Confidence interval.

^*Adjusted for demographics (age (years), sex (female vs. male), race (White vs. non-White), education (years));

^†Adjusted for demographics (age (years), sex (female vs. male), race (White vs. non-White), education (years)), cardiovascular and dementia risk factors (body mass index (kg/m²), smoking status (never vs. former vs. current), alcohol intake (drinks/week), physical activity level (none vs. low vs. moderate vs. high), total cholesterol level (mg/dL), diabetes mellitus status (normal vs. impaired fasting glucose vs. diabetes mellitus), hypertension status (normal vs. borderline vs. hypertensive), history of cardiovascular disease (yes vs. no), history of cerebrovascular disease (yes vs. no), and apolipoprotein E ε4 allele (0 ε4 vs. ≥1 ε4 allele)), cohort (original vs. African-American), and clinic site (UC Davis vs. Johns Hopkins vs. Pittsburgh vs. Wake Forest);

^‡Alzheimer’s disease without VaD using the National Institute of Neurological and Communicative Disorders and Stroke criteria (possible or probable) [22];

^§Vascular dementia with or without AD using the Alzheimer’s Disease Diagnostic and Treatment Centers criteria (possible or probable) [23].

Table 3.

Associations between specific ophthalmic conditions and risk of dementia, excluding those with mild cognitive impairment

Outcome and Specific Ophthalmic Condition	Cognitively normal only (n = 2,798)
All-cause dementia	Hazard ratio (95% confidence interval)*	p-value
Cataracts	1.33 (1.07 – 1.66)	0.011
AMD	1.49 (1.01 – 2.18)	0.042
DR	1.62 (0.75 – 3.50)	0.223
Glaucoma	1.12 (0.65 – 1.91)	0.688
Alzheimer’s disease	Hazard ratio (95% confidence interval)*	p-value
Cataracts	1.29 (1.00 – 2.04)	0.047
AMD	1.80 (1.09 – 2.99)	0.023
DR	0.53 (0.07 – 3.86)	0.532
Glaucoma	0.79 (0.32 – 1.92)	0.600
Vascular/mixed dementia	Hazard ratio (95% confidence interval)*	p-value
Cataracts	1.39 (1.01 – 1.93)	0.044
AMD	1.22 (0.65 – 2.26)	0.538
DR	2.76 (1.16 – 6.58)	0.022
Glaucoma	1.19 (0.56 – 2.56)	0.647

AMD = Age-related macular degeneration; DR = Diabetic retinopathy.

^*Adjusted for demographics (age (years), sex (female vs. male), race (White vs. non-White), education (years)), cardiovascular and dementia risk factors (body mass index (kg/m²), smoking status (never vs. former vs. current), alcohol intake (drinks/week), physical activity level (none vs. low vs. moderate vs. high), total cholesterol level (mg/dL), diabetes mellitus status (normal vs. impaired fasting glucose vs. diabetes mellitus), hypertension status (normal vs. borderline vs. hypertensive), history of cardiovascular disease (yes vs. no), history of cerebrovascular disease (yes vs. no), and apolipoprotein E ε4 allele (0 ε4 vs. ≥1 ε4 allele)), cohort (original vs. African-American), and clinic site (UC Davis vs. Johns Hopkins vs. Pittsburgh vs. Wake Forest).

Age, sex, and ApoE genotype modified associations between cataracts and dementia, as well as between AMD and dementia (Supplementary Tables 4–6). Except for an interaction between cataracts and sex (p = 0.045), significant statistical interactions were lacking between age, sex, and ApoE genotype and each ophthalmic condition on risk of dementia. Additional covariates included in the primary models, such as income level and access to medical care, refractive errors, inflammatory markers, and treatment for cataracts, glaucoma, or AMD, showed more pronounced results compared to the primary analyses (Supplementary Table 7). Due to missing data, we conducted multiple imputation for the observed missing data, which showed that the results did not change substantially compared to the primary results from the complete case analysis (Supplementary Table 8). Finally, results were similar between the competing risk analyses and primary analyses (Supplementary Table 9).

4. Discussion

The differential associations we found between each ophthalmic condition and dementia have implications on potential disease pathways shared between the eye and the brain. We found that cataracts, which included different types and causes, were associated with a significantly increased risk for all-cause dementia and AD. Some studies have detected Aβ in the human lens with a high degree of accuracy, indicating that AD-associated pathologic mechanisms may be related to the development of age-related cataracts, but others have not been able to replicate the results using different methods [27,28]. Evidence from population-based studies is limited, with one cohort study finding no association between cataracts and dementia [14]. Given the discrepancies among the published results concerning Aβ signatures in the lens and paucity of epidemiologic evidence, our results add knowledge in supporting the connection between AD-related changes and the human lens. Our finding of a significant association between cataracts and risk of VaD/mixed dementia may be due to inclusion of cases with both VaD and AD, but given that the strength of association was higher for VaD/mixed dementia than AD, the possibility of a shared vascular pathology is intriguing and would benefit from replication in future studies.

Similar to results for cataracts, we found a significantly increased risk for all-cause dementia and AD associated with AMD. Aβ has been found in drusen among patients with AMD, which are extracellular deposits that are characteristic of this disease [5]. Aβ in drusen may contribute to AMD progression by mediating oxidative stress, uncontrolled inflammation and imbalanced angiogenesis [5], similar to its role in AD. AMD and AD may also share a genetic link through ApoE, in which the ε4 allele is known to increase AD risk. A previous study in CHS found that the ε2 allele was associated with 2.5 times greater odds of late age-related maculopathy, whereas the ε4 allele may confer some protection [29]. Other studies suggest that despite the biochemical and histological similarities between AMD and AD, these diseases have separate genetic risk profiles and distinct pathophysiology [30,31]. A meta-analysis found a significant association between AMD and dementia, including AD [32]. Furthermore, other epidemiologic studies suggest that the dry form of AMD is more strongly related to cognitive impairment than the wet form of AMD [33,34]. Differences in dry versus wet AMD may explain the lack of association with VaD/mixed dementia in our study as most cases of AMD are the dry type, which targets the macula, whereas wet AMD represents a smaller proportion of AMD cases and is typified by choroidal neovascularization under the retina [35].

In contrast to the findings with AMD, we observed that DR was associated with significantly increased risk for VaD/mixed dementia, but not AD, though these estimates should be interpreted cautiously as the small number of dementia cases in this group resulted in reduced precision. While diabetes mellitus (DM) has been established to increase risk of dementia, including AD and VaD [36], the association between dementia and DR is unclear. Some studies have found DR to be associated only with AD [14,37], while other studies have found associations only with VaD [38,39]. In CHS, retinopathy, which included microaneurysms and retinal hemorrhages, was found to be associated with 2.1 times greater odds of dementia among participants with hypertension [40], which share similar effects on structural cerebral microvascular alterations as with diabetes. A recent autopsy study found that DR was significantly associated with increased risk of deep cerebral microinfarcts, but not with any other neuropathology, including AD neuropathologic change [41]. Considering that some hypothesize DR to be a primary neurodegeneration disorder [42], further study is necessary to elucidate whether microvasculopathy causes neurodegeneration or neurodegeneration causes microvasculopathy, as it is unclear from our findings with VaD/mixed dementia that DR is a primary vasculopathy.

Finally, we did not find evidence of a significantly increased risk of dementia associated with glaucoma. Previous studies found the frequency of glaucoma to be higher in patients with AD or dementia compared to those cognitively normal individuals [43], as well as some epidemiologic studies showing that individuals with glaucoma to be at higher risk for developing dementia [6,12], but not all [8,11]. Aβ and phosphorylated tau have also been detected in the vitreous of patients with glaucoma, which may lead to axonal loss and neuronal cell death of retinal ganglion cells [5]. However, findings from other studies argue against glaucoma representing a primary neurodegeneration based on features of glaucoma that are not consistent with a neurodegenerative process. These include differences in the specificity of neuronal populations affected in glaucoma and AD, as well as the roles of protein misfolding and intraocular pressure for both conditions [44]. Our finding of a null association between glaucoma and dementia may also reflect limitations in the data sources used, such as differential recording of glaucoma diagnoses, or residual confounding from differences in access to care. Despite the characteristics that glaucoma shares with neurodegenerative processes, such as AD, several features set glaucoma apart, which should motivate further studies to clarify this complex relationship.

An implication of our results applies to the potential use of ocular biomarkers for early detection of dementia. Several biomarkers have been proposed to visualize changes in the retina. Measurement of retinal fluorescence in drusen by ultra-wide-field fluorescence imaging and fluorescence lifetime imaging ophthalmoscopy may serve as a biomarker for AD [45]. Retinal Aβ imaging in drusen by spectral imaging may also be used as an AD biomarker, while changes to the retinal vasculature can be measured by fundus photography, as well as by confocal laser scanning microscopy [46,47]. Optical coherence tomography (OCT) angiography is a recent, non-invasive technique also used to visualize the microvasculature of the retina and the choroid [47]. Retinal thickness of the nerve fiber layer, which has been associated with AD progression and degree of cognitive impairment, can be measured by OCT as well [45]. Compared to these possible retinal biomarkers, though, the development of lens biomarkers has not been as robust. A small clinical study used fluorescence imaging to detect Aβ in the lens and predicted the clinical diagnoses of patients with AD with 85% sensitivity and 95% specificity [48]. While the eye holds promise to provide unique insights and opportunities for research and clinical interventions, further studies are needed to validate these potential ocular biomarkers of dementia and develop methods to differentiate AD pathology from other common ophthalmic and systemic diseases.

The large sample and well-characterized, prospectively collected CHS data are strengths of the study, but several limitations exist. Ophthalmic diagnoses relied on ICD-9 codes and diagnostic criteria for these conditions may have changed over time, which could result in significant variation. This may, in part, explain the null association between glaucoma and dementia observed in our study. The methods used to ascertain dementia and date of onset in the CHS Cognition Study were nontraditional and may have resulted in misclassification. This type of error would have attenuated associations toward the null, though, and would not have changed the overall conclusions. Even though participants were followed up for up to nine years, given the long preclinical stage of dementia, which can last for decades before obvious symptoms appear [49], we were unable to clearly establish the temporal sequence between the onset of each ophthalmic condition and development of dementia. Additional research is needed to determine whether these predictors of dementia are also possible risk factors, and if so, to elucidate potential causal pathways, such as mediation by vision loss. Although we adjusted for income and education, as well as access to medical care as proxies for socioeconomic status and availability of medical services, unmeasured or residual confounding may still exist. Finally, findings related to VaD/mixed dementia risk may not be generalizable given the criteria used to classify VaD, as well as having all participants receive a cranial MRI in order to be included in the CHS Cognition Study [50]. Results of this study are also relevant only to those who live beyond age 65 years without dementia, and generalizations should be made only to this group.

The ability to evaluate some of the most common ophthalmic conditions among older adults in the CHS cohort provides insight into the connection between the brain and the eye. Cataracts, AMD and DR were independently associated with increased risk of dementia. These associations may reflect different mechanisms between the eye and brain that lead to dementia. As the public health burden of dementia is expected to grow over the next decades, evaluation of the eye in older adults may help identify patients at elevated risk of developing dementia.