Association of Cognitive Functioning with Retinal Nerve Fiber Layer Thickness

Abstract

purpose. The brain areas that are responsible for cognitive functioning have the same embryonic origin as the retina. The association between cognitive functioning and retinal nerve fiber layer (RNFL) thickness was assessed in a large, population-based sample.

methods. Neuropsychological and ophthalmic examinations were performed in 1485 healthy individuals (mean age, 46 years; range, 18–85) from the Erasmus Rucphen Family (ERF) study, a study in a genetic isolate from the Netherlands. Different domains of cognitive functioning were assessed with the Dutch Adult Reading Test, the Rey Auditory Verbal Memory Test, semantic fluency, the Trail-Making Test, the Stroop Color-Word Test, and Block Design. RNFL thickness was measured with scanning laser polarimetry. The association between cognitive test scores and peripapillary RNFL thickness was studied with linear regression analyses, adjusting for age, sex, level of inbreeding, and refractive error.

results. After adjustment for confounders, a better cognitive performance was significantly associated with a thicker RNFL in all tests (P < 0.03) except for the Stroop Color-Word Test (P = 0.15). RNFL thickness explained up to 2.8% (R ² = 0.028) of the total variance in cognitive test scores. The association diminished in age groups beyond 40 years.

conclusions. The present study shows that cognitive functioning is associated with RNFL thickness in healthy young individuals. The lack of association in older individuals suggests that loss of neurons in the cerebrum and retina is not concomitant and may have different origins.

Cognitive functioning refers to the ability to attend to complex external or internal stimuli, to identify the relevant features of these stimuli, and to make appropriate responses (including storing aspects of this information). ¹ In accordance with their complexity and importance in daily life, cognitive functions take up the major part of the central nervous system (CNS). 

The eye is the only part of the body that provides a direct view of the CNS. The retina can be examined by means of direct or indirect ophthalmoscopy. Moreover, the thickness of the retinal nerve fiber layer (RNFL), which contains the axons of the retinal ganglion cells, can be objectively measured with imaging techniques such as scanning laser polarimetry (SLP). To what extent these easily quantifiable properties of the retina could provide insight into the concealed parts of the CNS, is largely unknown. 

Both the retina and the brain areas that are responsible for cognitive functioning originate from the embryonic prosencephalon. The premise of retinal involvement in cognitive functioning has been supported by studies describing an increased prevalence of glaucoma in patients with Alzheimer’s Disease (AD). ^{2 3} Other supportive evidence comes from histopathologic postmortem studies demonstrating retinal ganglion cell loss in patients with AD, ^{4 5} and from in vivo studies reporting a reduced RNFL thickness in patients with AD. ^{6 7} However, these few studies were limited to a small number of selected patients and focused on the extreme disease ends of cognitive and retinal functioning. Whether associations exist in their physiological spectrum is currently unknown. 

We had the opportunity to study the association between cognitive functioning and RNFL thickness in a large, population-based sample of healthy subjects. We assessed a broad range of cognitive functions by means of an extensive neuropsychological examination and we measured RNFL thickness with SLP. We also assessed whether cognitive functioning was associated with other ophthalmic and nonophthalmic factors. 

Methods

Study Population

Subjects were recruited as part of the Erasmus Rucphen Family (ERF) study, a family-based cohort study in a genetically isolated population in the Netherlands. This population was founded in the middle of the 18th century by fewer than 400 individuals. Eligibility for participation in the study was determined by genealogical background, not by any phenotypes of interest. Twenty-two families were selected who had at least six children baptized in the community church between 1880 and 1900. All living descendants of these families aged 18 years and older, as well as their spouses, were invited to attend a series of clinical examinations. Genetic characterization of the study population has been presented elsewhere. ^{8 9 10}  

The research adhered to the tenets of the Declaration of Helsinki and was approved by the Medical Ethics Committee of the Erasmus Medical Center in Rotterdam. Informed consent was obtained after explanation of the nature and possible consequences of the study. 

A total of 1641 subjects underwent neuropsychological and RNFL thickness assessments. Participants with visual acuity of less than 20/40 in the best eye (n = 3), poor-quality RFNL thickness measurements in both eyes (n = 21), or missing genealogical information (n = 131) or refractive error data (n = 1) were excluded from the dataset, leaving 1485 subjects for the analysis. For the Trail-Making Test, the Stroop Color-Word Test and the Dutch Adult Reading Test, 3 subjects were excluded because of illiteracy, and for the Stroop test, a further 20 subjects were excluded because of color blindness. 

Neuropsychological Assessment

All neuropsychological tests were performed during a 50-minute examination in a standardized environment. The Dutch Adult Reading Test (a validated Dutch version of the National Adult Reading Test ¹¹ ), consisting of a series of words with irregular pronunciation, was used as a measure of general cognitive ability. ¹²  

Memory function was assessed by means of the Dutch version of Rey’s Auditory Verbal Learning Test. ^{13 14} Short-term memory was defined as the number of correctly recalled words in the first trial, learning as the sum of correctly recalled words in trials 2 to 5, and delayed recall as the number of correctly recalled words after 20 minutes. Executive functioning was established with three tests: semantic fluency (animals and professions), ¹⁵ part B of the Trail Making Test as a measure of concept shifting, ¹⁶ and card III of the Stroop Color-Word Test as a measure of susceptibility to interference. ¹⁷ Visuospatial and visuoconstructive abilities were assessed with the subtest Block Design of the Wechsler Adult Intelligence Scale and were scored by the number of correctly replicated geometric designs adjusted for time of completion. ¹⁸  

RNFL Thickness Measurements

RNFL thickness was evaluated with scanning laser polarimetry (SLP). This technique, featured in the GDx VCC (Carl Zeiss Meditec, Inc., Dublin, CA), measures the retardation of a polarized scanning laser beam induced by the birefringence of retinal ganglion cell axons. It subsequently converts this retardation into RNFL thickness. 

Two GDx VCC scans were performed per eye. The first scan assessed anterior segment birefringence with the method described by Zhou and Weinreb. ¹⁹ The GDx VCC software then automatically adjusted the anterior segment compensator to each individual eye. Subsequently, the second scan was performed to measure RNFL thickness as described by Reus and Lemij. ²⁰ The cutoff for image quality was a GDx VCC scan quality score of 8 or above. Images with lower scores were excluded. 

After the boundaries of the optic disc were manually marked, the software positioned two circles centered on the disc: the first had a diameter of ∼2.5 mm (54 pixels), the second a diameter of ∼3.3 mm (70 pixels). The average RNFL thickness between these circles was used in our analyses. 

Statistical Analysis

If high-quality GDx VCC images could be obtained in both eyes, one eye was chosen randomly for inclusion in the analysis. If a high-quality image could be obtained in only one eye of a subject, this eye, rather than its fellow, was considered for analysis. The inbreeding coefficient, which represents the degree of consanguinity between a subject’s parents, was calculated by means of PEDIG software. ²¹ Statistical analyses were performed with commercial software (SPSS ver. 11.0 for Windows; SPSS, Chicago, IL). 

Characteristics of the study population were evaluated for men and women separately, and differences were tested with an independent-samples t-test (normally distributed continuous variables), a Mann-Whitney U test (not normally distributed continuous variables), or a Pearson χ² test (dichotomous variables). Multiple regression analyses were performed with cognitive test scores as outcome variables and RNFL thickness, age, sex, inbreeding coefficient, and spherical equivalent of the refractive error as predictor variables. The distribution of the multiple regression residuals was tested for normality with the nonparametric, one-sample Kolmogorov-Smirnov test. Outcome variables that were skewed were transformed by means of natural logarithm (Trail-Making Test part B, Stroop Color-Word Test card III) or square root (Rey Auditory Verbal Learning Test trial 1, Block Design). Subsequently, systolic and diastolic blood pressure, blood cholesterol level, and fasting blood glucose level were added as predictor variables. The analyses were also stratified by age tertiles (SPSS ver. 11.0 for Windows; SPSS, Chicago, IL). 

Results

Demographic and clinical characteristics of the study population are presented in Table 1 . Ages ranged from 18 to 85 years, with a mean of 45.9 years. The majority of subjects (78%) had consanguineous parents, and the median inbreeding coefficient indicated that these parents shared a common ancestor four to five generations back. Men and women differed significantly in educational level and in cognitive performance on some of the tests. Women performed significantly better than men on the Rey Auditory Verbal Learning Test (P < 0.001). Men performed significantly better on Block Design (P < 0.001). 

The results of the multiple linear regression analyses are provided in Table 2 . Better cognitive performance was significantly associated with a higher RNFL thickness in all tests (P < 0.03) except for the Stroop Color-Word Test card III (P = 0.15). These results were independent of age, sex, level of inbreeding, refractive error and also of additional possible confounders including systolic and diastolic blood pressure, serum cholesterol, and fasting glucose levels. Older age was significantly associated with a lower cognitive performance on all tests (P < 0.001). Inbreeding and spherical equivalent of any refractive error were inversely related to all cognitive tests, except for the Stroop Color-Word Test card III, although the association between inbreeding and the first trial of the 15-word test did not attain statistical significance (β = −0.016; P = 0.079). 

Univariate linear regression of RNFL thickness on cognitive functioning showed R ² (coefficient of determination) values ranging from 0.012 for the Stroop Color-Word Test card III to 0.028 for the Dutch Adult Reading Test. The multiple-regression model with RNFL thickness, age, sex, level of inbreeding, and spherical equivalent of refractive error as predictor variables had R ² in the range of 0.164 (semantic fluency) to 0.365 (Block Design). 

When stratified by age tertiles (Table 3)the association between cognitive functioning and RNFL thickness remained significant for six, two, and one of the eight cognitive tests in the categories 18 to 39 years of age, 39 to 53 years of age, and 53 years of age or older, respectively. 

Discussion

The present population-based study demonstrated an association between cognitive functioning and RNFL thickness in their physiological range. This association particularly manifested in young to middle-aged adults and diminished in individuals 40 years of age or older. RNFL thickness explained only a small proportion of the variance in cognitive functioning. 

The design of our study had three limitations. First, an isolated population may differ from the general population in its genetic and environmental composition. Our results may therefore not apply to other populations. However, simulation studies have shown that common allele frequencies in the ERF population do not deviate from the general population. ⁹ Including the inbreeding coefficient as a potential confounding factor in our analyses did not change the results. We therefore believe that the association between the level of cognitive functioning and RNFL thickness may be generalized to the outbred population. Second, a cross-sectional study design does not allow for inferences on causal relationships. To address these, longitudinal studies would be needed. Third, we did not determine the reproducibility of the RNFL thickness measurements in this population. Previous studies on intersession and interoperator variability in other populations have shown that the GDx VCC provides highly reproducible measurements. ^{22 23 24 25} We have no reason to suspect a lower reproducibility in our healthy and relatively young population, in which cataract and other ocular conditions that could interfere with measurement variability are presumably less prevalent. Strengths of our study are the large sample size, the extensive assessment of different cognitive domains, the objective RNFL thickness measurements, and the adjustments for a variety of confounders. 

Previous studies have concentrated on the pathophysiological spectrum of cognitive functioning and retinal properties. In addition to an increased prevalence of glaucoma in patients with AD, ^{2 3} common pathogenetic mechanisms for these neurodegenerative diseases have been suggested. These include the role of apolipoprotein E and amyloid-β. ^{26 27 28} Histopathologic and in vivo studies have reported retinal ganglion cell loss in patients with AD. ^{4 5 6 7} A study by Iseri et al. ⁶ hinted at a possible correlation between the amount of retinal ganglion cell loss and the severity of cognitive impairment in a group of 14 patients with AD, but should be interpreted cautiously, as no adjustments for age (a potentially important confounder) had been made, and inconsistencies between the different retinal measurements were reported. 

Our finding of a reduced association between cognitive functioning and RNFL thickness at older age is inconsistent with results in previous studies. Considering the particular strong effects in the young age category, we think that the relationship between a higher level of cognitive functioning and a thicker RNFL may reflect a better development of tissues that originate from the prosencephalon as a whole. We also speculate that any damage to any of these tissues, including the RNFL, would be unlikely to run an equal course. Generalized loss might, in principle, differentially affect the various neuronal tissues of prosencephalic origin. Selective loss to the RNFL, for instance, could be caused by increased intraocular pressure (glaucoma), which would be unlikely to affect cognitive functioning. Glaucoma and other ocular diseases particularly manifest in the elderly population. One might therefore expect the relationship between RNFL thickness and cognitive functioning to weaken with age, which is what we have indeed found. 

Most previous studies into neurologic diseases have used optical coherence tomography (OCT) to measure the RNFL thickness in vivo. ^{6 7 29 30} The principle of OCT is analogous to the ultrasound B-scan, but instead of sound OCT uses interference patterns of backscattered near-infrared light. ³¹ Although this technique is fundamentally different from SLP, RNFL measurements of both techniques have been shown to correlate well and to have a comparable diagnostic accuracy for glaucoma. ^{32 33} SLP measurements are based on the birefringence of the RNFL, which is assumed to be induced by the ganglion cell axons without any contribution from the supporting astrocytes and Müller cells. ^{34 35} This technique may therefore provide a more direct measure of pathology in for example AD, in which an increased astrocyte/neuron ratio has been reported. ³⁶ However, this may not apply to our study, in which we only consider the physiological spectrum of cognitive functioning. 

Our study showed that higher scores on cognitive tests were associated with a more negative refractive error. This finding is consistent with those in previous studies, which have reported that intelligence and educational attainment, both correlates of cognitive functioning, ^{37 38 39} are significantly related to myopia. ^{40 41 42 43}  

The R ² of the multiple-regression model ranged from 0.16 for semantic fluency to 0.37 for Block Design. This indicates that the predictor variables in the model together account for 16% to 37% of the total variance in cognitive test scores. RNFL thickness alone explained up to 2.8% of the variance in cognitive test scores. Hence, despite the statistically significant slope of the regression line, RNFL thickness is by no means a precise predictor of cognitive ability. The tempting idea of making inferences about cognitive functioning or decline by easily and safely measuring RNFL thickness is therefore not realistic at the present time. Technologies for in vivo imaging of the retina have recently advanced quickly. The latest devices offer an axial imaging resolution of 2 to 3 μm, thus approaching the level of detail achieved in histopathology. ⁴⁴ It would be of interest to investigate the association between cognitive functioning and these more accurate RNFL thickness measurements and to explore the applicability of the new imaging devices to neurologic practice. 

In conclusion, our study is the first to our knowledge to show a significant association between the level of cognitive functioning and RNFL thickness in a healthy population. Although any clinical implications would currently be limited, our results may warrant further investigations into the causality and future applicability of this association 

 Supported by the Rotterdam Eye Hospital Research Foundation (Stichting Wetenschappelijk Onderzoek Het Oogziekenhuis [SWOO] Prof. Dr. H. J. Flieringa, Rotterdam), the Center for Medical Systems Biology (CMSB), Swart van Essen Foundation, Prof. Dr. Henkes Foundation, an Erasmus Medical Center fellowship grant, and The Netherlands Organization for Scientific Research (NWO).

 Submitted for publication November 19, 2008; revised March 31, 2009; accepted July 8, 2009.

 Disclosure: L.M.E. van Koolwijk, None; D.D.G. Despriet, None; C.M. Van Duijn, None; B.A. Oostra, None; J.C. van Swieten, None; I. de Koning, None; C.C.W. Klaver, None; H.G. Lemij, Carl Zeiss Meditec (F)

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Table 1.

 

Characteristics of the Study Population

	Men (n = 679)	Women (n = 806)
Age, mean ± SD (y)	46.5 ± 14.0	45.4 ± 13.8
Inbreeding > 0 (%)	78.6	77.5
Inbreeding coefficient, median (2.5th–97.5th percentiles)	0.00123 (0.00000–0.02237)	0.00062 (0.00000–0.02131)
Highest level of education*
Elementary school (%)	29.8	28.0
Junior vocational training (%)	60.6	67.5
Senior vocational or academic training (%)	9.6	4.6
Systolic blood pressure, mean ± SD (mm Hg)*	141.6 ± 16.6	134.7 ± 19.9
Diastolic blood pressure, mean ± SD (mm Hg)*	81.3 ± 9.4	78.9 ± 9.9
Blood cholesterol, mean ± SD (mM)	5.5 ± 1.1	5.6 ± 1.0
Blood glucose, mean ± SD (mM)*	4.7 ± 1.0	4.3 ± 0.8
Ophthalmic
Randomized eye, right (%)	46.1	50.9
RNFL thickness, mean ± SD (μm)	57.6 ± 6.2	57.7 ± 6.2
Spherical equivalent of refraction, mean ± SD (D)	−0.2 ± 1.9	0.0 ± 1.8
Cognitive functioning median test scores (2.5th–97.5th percentiles), †
15-word test I*	4 (1–8)	5 (2–8)
15-word test II-V*	33 (14–49)	36 (18–52)
15-word test VI*	7 (2–13)	8 (2–14)
Semantic fluency	38 (21–58)	37 (20–56)
Trail-Making test B	75 (33–208)	73 (34–230)
Stroop Color-Word Test part III*	96 (66–180)	93 (61–171)
Block Design*	29 (11–65)	24 (10–63)
Dutch Adult Reading Test	66 (20–92)	64 (20–90)

 Differences between the groups were tested for statistical significance with an independent-samples t-test for normally distributed continuous variables, with the Mann-Whitney U test for non-normally distributed continuous variables, and with the Pearson χ² test for dichotomous variables.

*   P < 0.05

†  The 15-word test I was scored by the number of correctly recalled words in the first trial; 15-word test II-V by the sum of correctly recalled words in the second to fifth trials; 15-word test VI by the number of correctly recalled words after 20 minutes; semantic fluency by the sum of correctly named items in two categories (animals and professions); Trail-Making Test B by the time in seconds to completion of the task; Stroop Color-Word Test part III by the time in seconds to completion of the task; Block Design by the number of correctly replicated geometric designs; and the Dutch Adult Reading Test by the number of correctly pronounced words.

Table 2.

 

Multiple Linear Regression Analyses of RNFL Thickness and Other Covariates on Cognitive Functioning

	RNFL Thickness	Age	Male	Inbreeding	Spherical Equivalent
15-word test I*	0.004 (0.002); P = 0.012	−0.012 (0.001); P < 0.001	−0.090 (0.021); P < 0.001	−0.016 (0.009); P = 0.079	−0.012 (0.006); P = 0.040
15-word test II-V	0.098 (0.031); P = 0.002	−0.312 (0.015); P < 0.001	−3.243 (0.383);P < 0.001	−0.342 (0.171); P = 0.046	−0.324 (0.110); P = 0.003
15-word test VI	0.024 (0.011); P = 0.026	−0.084 (0.005); P < 0.001	−0.909 (0.131); P < 0.001	−0.136 (0.058); P = 0.020	−0.110 (0.038); P = 0.003
Semantic fluency	0.099 (0.038); P = 0.009	−0.237 (0.018); P < 0.001	1.005 (0.467); P = 0.032	−0.890 (0.209); P < 0.001	−0.360 (0.134); P = 0.007
Trail-Making Test B* , †	−0.006 (0.002); P = 0.001	0.019 (0.001); P < 0.001	−0.005 (0.022); P = 0.829	0.031 (0.010); P = 0.001	0.017 (0.006); P = 0.007
Stroop Color-Word Test part III* , †	−0.002 (0.001); P = 0.150	0.010 (0.001); P < 0.001	0.016 (0.014); P = 0.239	0.001 (0.006); P = 0.915	0.004 (0.004); P = 0.295
Block Design*	0.015 (0.005); P = 0.001	−0.054 (0.002); P < 0.001	0.333 (0.059); P < 0.001	−0.087 (0.026); P = 0.001	−0.066 (0.017); P < 0.001
Dutch Adult Reading Test	0.305 (0.071); P < 0.001	−0.414 (0.034); P < 0.001	1.819 (0.880); P = 0.039	−2.669 (0.395); P < 0.001	−1.246 (0.251); P < 0.001

 Presented data are regression coefficients (SE) and P-values in a multiple-regression model including age, sex, inbreeding, and refractive error.

*  Transformed variables.

†  Higher test scores represent lower cognitive performance.

Table 3.

 

Multiple Linear Regression Analyses of RNFL Thickness on Cognitive Functioning Stratified by Age Tertiles

	Age 18.0–38.9 y	Age 38.9–52.5 y	Age 52.6–85.3 y
15 word test I*	0.006 (0.003); P = 0.047	0.004 (0.003); P = 0.196	0.001 (0.003); P = 0.676
15 word test II-V	0.120 (0.050); P = 0.016	0.047 (0.058); P = 0.423	0.101 (0.054); P = 0.061
15 word test VI	0.036 (0.017); P = 0.039	−0.002 (0.020); P = 0.912	0.026 (0.018); P = 0.155
Semantic fluency	0.109 (0.061); P = 0.075	0.094 (0.070); P = 0.178	0.082 (0.063); P = 0.192
Trail-Making Test B* , †	−0.005 (0.002); P = 0.031	−0.007 (0.003); P = 0.044	−0.007 (0.004); P = 0.060
Stroop Color-Word Test part III* , †	0.000 (0.002); P = 0.810	−0.002 (0.002); P = 0.184	−0.002 (0.002); P = 0.327
Block Design*	0.025 (0.009); P = 0.003	0.017 (0.009); P = 0.071	0.004 (0.007); P = 0.593
Dutch Adult Reading Test*	0.245 (0.085); P = 0.004	0.278 (0.139); P = 0.046	0.400 (0.146); P = 0.006

 Presented data are regression coefficients (SE) and P-values for RNFL thickness in a multiple-regression model, including age, sex, inbreeding, and refractive error.

*  Transformed variables.

†  Higher test scores represent lower cognitive performance.

The authors thank all participants in the ERF Study, and Hans Bij de Vaate, Patricia van Hilten, Margot Walter, Lidian van Amsterdam, Margriet van der Meer, Tiny Haest, Leon Testers, and all research assistants for help in data collection.