Age-Related Losses of Retinal Ganglion Cells and Axons

Ronald S. Harwerth, Joe L. Wheat, and Nalini V. Rangaswamy

PURPOSE. Age-related losses in retinal nerve fiber layer (RNFL)
thickness have been assumed to be the result of an age-dependent reduction of retinal ganglion cells (RGCs), but the published rates differ: age-related losses of RGCs of approximately
0.6%/year compared to 0.2%/year for thinning of the RNFL. An
analysis of normative data for standard automated perimetry
(SAP) sensitivities and optical coherence tomography (OCT)
measures of RNFL thickness showed that the apparent disagreement in age-dependent losses of RGCs and axons in the
RNFL can be reconciled by an age-dependent decrease in the
proportion of the RNFL thickness that is composed of axons.
The purpose of the present study was to determine whether
the mechanisms of age-related losses that were suggested by
the normative data can be confirmed with data from healthy,
normal eyes.
METHODS. Data were obtained from visual fields (normal results
in a Glaucoma Hemifield Test [GHT] on standard automated
perimetry [SAP] 24-2 fields) and RNFL thickness measurements
(standard OCT scan) of 55 patients (age range, 18 80 years;
mean, 44.5 17.3). The SAP measures of visual sensitivity and
OCT measures of RNFL thickness for one eye of each patient
were used to estimate neuron counts by each procedure.
RESULTS. The age-related thinning of RNFL was 0.27%/year
when a constant axon density was used to derive axon counts
from RNFL thickness, compared with 0.50%/year for the agerelated loss of RGCs from SAP. In agreement with the model
developed with normative clinical data, concordance between
losses of axons and soma was achieved by an age-dependent
reduction of 0.46%/year in the density of axons in the RNFL.
CONCLUSIONS. The results suggest that the proportion of RNFL
that is composed of RGC axons is not constant with age; rather,
the proportion of the total thickness from non-neuronal tissue
increases with age. If a similar compensation occurs in the
RNFL thickness with axon loss from glaucoma, then a stagedependent correction to translate OCT measurements to neuronal components is needed, in addition to the age-dependent
correction. (Invest Ophthalmol Vis Sci. 2008;49:4437 4443)
DOI:10.1167/iovs.08-1753

any visual abilities decline with age as a result of the

normal, nonpathologic losses of neurons in the peripheral and central visual pathways.1,2 A methodical loss of retinal
ganglion cells (RGCs) is part of the normal aging process,
which causes reduced visual sensitivity across the visual field in

From the College of Optometry, University of Houston, Houston,

Texas.
Supported in part by Grants R01 EY01139, P30 EY07551, T32
EY07024, and K23 EY01829 from the National Eye Institute and a John
and Rebecca Moores Professorship from the University of Houston.
Submitted for publication January 18, 2008; revised March 5 and
April 9, 2008; accepted August 11, 2008.
Disclosure: R.S. Harwerth, None; J.L. Wheat, None; N.V. Rangaswamy, None
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.
Corresponding author: Ronald S. Harwerth, 505 J. Davis Armistead
Building, University of Houston, Houston, TX 77204-2020;
rharwerth@uh.edu.
Investigative Ophthalmology & Visual Science, October 2008, Vol. 49, No. 10
Copyright Association for Research in Vision and Ophthalmology

standard automated perimetry (SAP)39 and a thinning of the

retinal nerve fiber layer (RNFL) in optical coherence tomography (OCT).10 17 In clinical practice, it is important to differentiate between the functional and structural changes that
arise from normal nonpathologic losses versus pathologic
losses, and clinical instruments for SAP4,8 and RNFL16,18 measurements therefore incorporate age-dependent normative
data for the statistical evaluation of a patients data. However,
the comparison to normative data does not allow direct interinstrument comparisons of measurements, nor does it provide
information about the amount of a patients neuronal loss from
either aging or glaucoma.
The general phenomenon of age-related losses of neurons in
the inner retina is well established. The effects of aging have
been investigated by both histologic analysis of the RGC somas19 21 or their axons in the optic nerve2125 and by optical
measurements of the RNFL.10 17 The investigations of the
number of axons in the optic nerve have shown systematic
age-related losses at rates that range from 0.3% to 0.6%/year,
whereas the age-related thinning of RNFL occurs at somewhat
lower rates of 0.2%/year.16 The aging effects found by histology have been supported by statistical analysis of the slopes
of the function for neuronal loss versus age,19 25 but the rather
low rates of RNFL thinning are problematic for the statistical
analysis of an age-dependent effect, because shallow slopes are
hard to define with confidence. For example, one study was
unable to define a statistically significant effect of age on any of
the OCT parameters for RNFL thickness,14 whereas each of the
others have demonstrated significant declines in the mean
thickness,10 13,15,16 but not always in thinning, within a quadrant, especially the inferior and nasal quadrants.17 The source
of the apparent discrepancies in age-dependent effects across
experimental and measurement methods or within a restricted
range of RNFL measurement is a likely consequence of the
optical measurement of the total RNFL thickness. Total RNFL
thickness does not separate the proportion of thickness that is
composed of RGC axons from non-neuronal support tissue26 28 and consequently the amount of thinning that was
caused by the age-related loss of RGCs is not straightforward.29,30
Potentially, neuronal losses in the retina may be quantifiable
from clinical measurements via an application of methods that
were developed for experimental glaucoma in macaque monkeys.29,31,32 If these methods are also appropriate for patients,
then the evaluation of clinical data can be achieved by translation of SAP and OCT measurements into a common parameter related to RGCs. An initial study to determine whether the
procedures could be extended to clinical patients was based
on normative age-dependent data for SAP and OCT.30 The
principal result of the analysis was that, although the general
methods seem appropriate, the quantification of neuronal populations from OCT measurements required an additional variable to reflect an age-dependent decrease in the RGC axon
density in the RNFL. The additional factor was needed to
account for the discrepancy in the rates of age-related thinning
of the RNFL and the age-related losses of RGCs. To reconcile
the effects of normal aging, opposing age-dependent alterations in the composition of the RNFL were incorporated into
the model of normal aging that is illustrated in Figure 1. By this
model, the normal age-related thinning of the RNFL of approx4437

4438

Harwerth et al.

IOVS, October 2008, Vol. 49, No. 10

sensitivity from SAP and the thinning of the RNFL with OCT
data from normal eyes of clinical patients. Some of the results
of these studies have been presented in abstract form (Harwerth RS, et al. IOVS 2007;48:ARVO E-Abstract 492).

MATERIALS

AND

METHODS

Subjects

FIGURE 1. The model for the age-related thinning of the RNFL proposed by Harwerth and Wheat.30 The RNFL thickness at each age
(solid line), represents the sum of two components of the total thickness: the age-dependent loss of neuronal tissue (dashed line) and a
compensating increase of nonneuronal tissue (dot-dash line).

imately 0.21 m/year (solid line) represents the sum of a

neuronal component, the well-established age-related loss of
RGC neurons that could cause a decline in RNFL thickness of
approximately 0.55 m/year (dashed line), and a non-neuronal
component that increases at a rate of approximately 0.32
m/year (dot-dash line). The smaller increases in non-neuronal,
compared with the decreases in neuronal tissue, result in a net
loss of RNFL thickness and an age-dependent reduction of axon
density of 0.007 axons/m2 per year. Therefore, to achieve
concordance between the subjective and objective methods of
estimating neuronal populations, the model proposes that the
thickness of the RNFL does not represent a constant proportion of RGC axons, but rather, that there are age-dependent
decreases in the density of axons that are partially compensated for by increases in non-neuronal components.
The normative databases are linear estimates of age-related
changes for specific instrumentation and measurements, but
the analytical form of the neuronal lossage function cannot be
addressed by these techniques, because the manufacturers
methods of incorporating the data are unknown, and the accuracy of extracting the data from the instruments cannot be
verified. In addition, modeling the average values for a population does not provide information about the variability and
limits of agreement for the population. Therefore, although the
analysis of normative data presents a strong suggestion for
differences in the effects of aging on SAP and OCT measurements, the results must be confirmed by empiric data that were
analyzed by the same model that was developed from the
normative data.30 Consequently, the purpose of the present
study was to test the model of age-related losses of visual

The results from data of 55 subjects are reported. The group data and
general characteristics of the population are presented in Table 1. The
subjects were students, faculty, or staff of the college, or their relatives.
They were initially identified on the basis of their self-reported normal
vision and were subsequently enrolled in the study if their SAP results
were identified as within normal limits by the Glaucoma Hemifield Test
(GHT). The refractive errors in all the subjects were between 6 D of
hyperopia and 6 D of myopia. The subjects ranged in age from 18.0 to
80.3 years (mean age, 44.5 17.3) with at least seven subjects for each
decade from the third through the seventh. The group included an
approximately equal number of men and women (27/28). One eye of
each subject was entered into the study (25 right and 30 left eyes),
with alteration of the eye selected in successive subjects. The research
adhered to the tenets of the Declaration of Helsinki and the experimental protocol was reviewed and approved by the University of
Houstons Committee for the Protection of Human Subjects. Informed
consent was obtained from each of the subjects, and they were remunerated for their participation.

SAP and OCT Data

The data for visual fields were obtained with SAP SITA standard 24-2
measurements (Humphrey Analyzer II; model 750; Carl Zeiss Meditec,
Inc., Dublin, CA), taken twice for each eye during a single experimental session. Testing was conducted by one of the authors (JW or NR),
who used standard clinical protocols for correction of refractive errors
and with undilated pupils. Initial practice data were collected for both
of the subjects eyes, and the patient was given a short rest break.
Subsequently, the visual field measurements were repeated in reverse
order (i.e., left eye right eye), and the data from the second measurement run for the eye selected for the study were used in the investigation. The point-by-point thresholds were entered into data files and,
for the analysis of RGC densities, each threshold was decreased by 1 dB
to compensate for the difference in data collected by SITA thresholding
versus the full-threshold strategy.33,34
For the RNFL thickness measurements, the subjects pupils were
dilated, and the RNFL images were acquired (StratusOCT, running ver.
4.0.4 software; Carl Zeiss Meditec, Inc.). One of the authors (JW or NR)
obtained RNFL scans using the standard RNFL thickness protocol of
512 measures in a nominal 10.87-mm scan length (circumference) of a
circle of 3.4-mm diameter that was centered on the ONH. Only scans
that were well-centered, with a signal strength of 7 or higher were
saved for analysis. Three standard scans for each eye were subsequently exported to a computer to derive the mean thickness at each
sample (pixel) of the scan for an analysis of axon densities across the
RNFL. All the patients ametropias were 6 D, or less, and neither axial

TABLE 1. Characteristics of the Subject Population

Age (Decades)

Age SD
(y)

Sex
(M/F)

Eye
(OD/OS)

RNFL SD
(dB)

MD SD
(dB)

PSD SD
(dB)

3
3
4
5
6
7
7
Mean SD Total

2
13
10
10
7
9
4
55

18.8 1.6
26.4 3.2
33.3 2.6
43.9 2.9
56.5 2.8
65.4 2.9
76.5 4.2
44.5 17.3

2/0
6/7
6/4
4/6
3/4
5/4
1/3
27/28

1/1
7/6
6/4
4/6
4/3
4/5
1/3
25/30

94.2 13.3
104.4 7.6
100.3 9.4
98.6 8.8
94.1 5.2
89.5 7.5
82.6 16.1
97.2 10.2

0.58 1.4
0.01 0.6
0.17 0.9
0.10 0.7
0.00 0.6
0.21 1.2
0.28 0.6
0.02 0.8

1.47 0.4
1.30 0.2
1.31 0.2
1.40 0.2
1.51 0.3
1.45 0.3
1.60 0.4
1.39 0.3

Age-Related Losses of RGCs and Axons

IOVS, October 2008, Vol. 49, No. 10

4439

FIGURE 2. Age-dependent functions

for RNFL thickness and total number
of RGCs and axons. Solid line: the
results of a linear regression analysis
of the data; inset: parameters. (A)
OCT measurements of the age-related thinning of the RNFL in 61 normal subjects. (B) The total number of
axons in the RNFL derived from the
OCT measurements using a constant
axon density of 1.23 axons/m2 for
each subject, to translate the area of
the RNFL to the number of axons.
(C) The number of RGCs derived
from SAP measurements of visual
sensitivity according to published
methods.31 (D) The total number of
axons in the RNFL derived from the
OCT measurements by using an axon
density that varied with age (density 0.007[age] 1.4) to translate
the area of the RNFL to the number
of axons.
length nor refractive error was considered in the analysis of the RNFL
thickness data.

Translation of SAP and OCT Data to the Number

of Neurons
The methods for quantification of the RGC populations underlying
measurements of visual field sensitivities and RNFL thicknesses have
been described in detail and will be presented only in brief.31,32 The
relationships between the number of ganglion cells and the perimetry
measurement results, which were initially derived from data obtained
by behavioral perimetry and retinal histology, are linear functions
when both variables are expressed in logarithmic (decibel) units.
However, the parameters of the linear functions vary with eccentricity
and, thus, the RGC density, in decibel units, for a given location in the
visual field was a product of the perimetric sensitivity and retinal
eccentricity. The retinal eccentricities for the perimetry data were
modified for application to data from human subjects because of the
longer axial length of the human eye. The nodal point-to-retina distance in the human eye is approximately 16.7 mm, compared with
12.5 mm in a monkeys eye, and therefore, the retinal eccentricities (in
arc degrees) for point-wise estimations of ganglion cell densities from
SAP measurements were increased by the ratio of these lengths. The
number of RGCs in linear units at a given test field location was
determined by multiplying the antilog of the cell density by the retinal
area that corresponds to the 6 6-arc deg area of the visual field (i.e.,
an area set by the sampling density of SAP).
The number of RGC axons in a section of the RNFL were estimated
from the area defined by the RNFL thickness profile and the density of
axons in the RNFL. The first factor, RNFL area, was determined by
multiplying the OCT scan height at each pixel by a pixel length of 21.2
m. The pixel areas were summed across a specific scan distance to
obtain the RNFL area of a given region or over the entire RNFL scan
around the optic nerve head (ONH) to obtain the total number of
axons. The second value, the axon density, which translates an RNFL
area to the number of axons, was either a fixed value of 1.23 axons/
m2 or an age-dependent variable, as is illustrated by the model in
Figure 1. The variable axon density in the present study was taken from
the model of normative data, with a reduction of axon density at a rate
of 0.007 axons/m2 per year of age, starting from an initial value of 1.4
axons/m2.
The topographic mapping of the visual field onto the ONH was
accomplished by an empiric strategy of dividing the RNFL scan and

visual field into 10 corresponding sectors.29,32 The RNFL data were

divided into equal sectors of 51 pixels of the OCT scan and the SAP
visual field locations, with the RGC axons entering each sector, were
assigned to that sector. The number of test field locations assigned to
a given sector varied between one SAP location near fixation corresponding to the temporal portion of the RNFL scan to 13 SAP locations
for the arcuate locations of the visual field, with axons entering the
superior and inferior poles of the ONH.

RESULTS
The mean thickness of the RNFL as a function of age is presented in Figure 2A. The data are well described by linear
regression (r2 0.3, P 0.0001), with an age-dependent loss
of RNFL thickness of 0.27%/year, which is in close agreement
with the published results of several studies13,14,16,17 and with
the function derived from the normative OCT data (Fig. 1).
Thus, the data for the normal age-dependent thinning of the
RNFL are consistent with both sets of published data, and the
normative databases and should provide an adequate set of
data for an evaluation of effectiveness of the model for estimating the number of RGC axons from RNFL thickness measures.
The results of a linear transformation of RNFL thickness to
axon numbers, using a constant value for the axon density
across all ages, are presented in Figure 2B. The axon density for
these conversions (1.23 axons/m2) was the value that provided a close agreement between RGCs and their axons for the
normative data of 25-year-old subjects. It is obvious that the
simple rescaling of the RNFL data does not change the slope of
the function and, as a result, it does not accurately correlate
with the rate of age-related reduction in the number of RGCs
that were estimated from the SAP data for these subjects (Fig.
2C), nor for the published data for age-related losses of RGC
somas or axons.19 26 In contrast, an application of the agedependent reduction of axon density in the RNFL produced an
agreement between the two estimates of age-related losses of
retinal neurons (Fig. 2D). Therefore, the concordant results of
the estimates of RGCs and their axons from the diverse measurements of visual sensitivities and RNFL thickness provides
strong support for the concept of an age-related decrease in the
proportion of the RNFL thickness that is composed of neuronal

Harwerth et al.

A. Subject: UHP-008 - age 19.9 years

Thickness (m)

Mean
32.1 2.4 dB

MD: -0.14 dB
PSD: 1.82 dB

IOVS, October 2008, Vol. 49, No. 10

B. Subject: UHP-041 - age 80.3 years

MD: 0.66 dB
PSD: 1.40 dB

Mean
28.3 2.4 dB

Thickness (m)

4440

Mean = 84.8 m

Mean = 99.6 m

Pixel number

Number of
somas or axons

Number of
somas or axons

Pixel number

Pixel number

Pixel number
SAP: 1,113,566 total RGCs
OCT: 975,712 total axons

SAP: 707,953 total RGCs

OCT: 778,425 total axons

tissue and for the model of aging (Fig. 1) that has been proposed to reconcile SAP and OCT measurements.
Examples of the concurrence between SAP and OCT measurements for individual subjects are illustrated in Figure 3.
These two subjects were in the youngest (Fig. 3A) and oldest
(Fig. 3B) decades of life of the subjects in the study. The global
indices (MD and PSD) of the SAP data of both subjects indicate
that the subjects have normal age-adjusted visual fields, but on
an absolute basis, the mean SAP sensitivity (shown at the
upper-right of the SAP data) for the younger subject is approximately 4 dB higher than for the older subject. As well, the OCT
measurements for both subjects are within their respective
age-adjusted normal ranges, but in this case, the older subjects
mean RNFL thickness is almost 15 m greater than the younger
subjects. These facial differences in retinal neurology were
eliminated when the age-dependent density of axons was incorporated into the translation of RNFL thickness to axon
number. The data in the bottom graph for each subject demonstrate that the SAP/OCT estimates of RGCs and their axons
are, in fact, very similar in relation to the total sums of RGCs
and axons. The two estimations of neuronal populations, approximately 1,000,000 for the younger subject and over
700,000 for the older subject, are within 12% of agreement for
each subject, and reflect an age-appropriate reduction in neurons for the older subject compared with the younger subject.
An assessment of the degree of intrasubject agreement between SAP and OCT estimates of RGCs and axons for all the
subjects is presented in Figure 4. The data represent the relationship between estimates of neurons, in decibel units, from
OCT and SAP measurements for each of 8 of the 10 OCT
sectors (2 nasal sectors were excluded32) and the corresponding SAP test locations (Fig. 4A) or summed across the 8 sectors
of the RNFL scan and related visual field locations (Fig. 4B). By
inspection, the data seem to represent a linear relationship that
is accurately modeled by the 1:1 function that is illustrated by
a solid line superimposed on each plot. The goodness-of-fit of
the model of a unity relationship between the estimates and
the empiric data was tested by two simple statistics (i.e., the
mean absolute deviation [MAD] and the coefficient of determination for the 1:1 function [r2]). The results of the analysis are
shown in the insets of the graphs and illustrate that the average
deviations from the 1:1 relationship are small (1 dB, or less) and

FIGURE 3. Examples of the translation of SAP and OCT measurements

to common parameters of neuron
numbers in two subjects. Subjects in
(A) the youngest decade and (B) the
oldest decade of the subjects in the
study.

the unity relationship is significant (P 0.0001) for both the

sector-by-sector analysis and for the whole field analysis. The
sum of the analyses clearly supports the concurrence of estimates of RGCs and axons by clinical measures of function and
structure when procedures for translating each measure to its
neuronal substrate have been implemented.
The analyses of subregions of the RNFL scan and SAP test
locations, presented in Figure 5, indicate that the effects of
aging are similar for the mean RNFL data (Fig. 4) and the
thicker RNFL regions at the superior and inferior ONH, but

FIGURE 4. The relationship between SAP and OCT estimates of RGCs

and axons in all the subjects. The data represent the correlation of
estimates of neurons, in decibels, from OCT and SAP measurements in
each of the 10 OCT sectors of the RNFL scan and the corresponding
SAP test locations (A) or summed across the whole RNFL scan and all
visual field locations (B). Inset: goodness-of-fit statistics, which are the
mean absolute deviation (MAD) with respect to a unity relationship
and the coefficient of determination (r2) for the 1:1 relationship.

Age-Related Losses of RGCs and Axons

IOVS, October 2008, Vol. 49, No. 10

Number of Axons (dB)

B. Superior ONH

A. Temporal ONH

Goodness-of-fit
MAD = 0.67 dB
r2 = 0.16

Goodness-of-fit
MAD = 0.95 dB
r2 = 0.14

C. Inferior ONH

Goodness-of-fit
MAD = 0.62 dB
r2 = 0.21

Number of Ganglion Cells (dB)

FIGURE 5. The analyses of subregions of the RNFL scan and SAP test
locations. (AC) Data in three sectors of the RNFL scan. (A) The
temporal ONH, representing the papillomacular RNFL bundles; (B) the
superior ONH, representing the arcuate bundles of the superior retina;
and (C) the inferior ONH, representing the arcuate bundles from the
inferior retina. The statistics for a unity correlation between axons and
cell bodies for each sector were analyzed by the same statistics as the
composite data in Figure 4.

aging effects are less well correlated in the thinner temporal

region. The areas analyzed for regional effects were based on
the RNFL sectors developed for the correlation of OCT and
RNFL data, but were close to the standard clock-hour sectors.
Using the OCT scan circle as a reference, we analyzed three
sectors: (1) the temporal ONH data (Fig. 5A) representing the
papillomacular RNFL bundles in a 72-arc deg region of the
ONH (324 36 arc deg or 8 10 oclock), (2) the superior
ONH data (Fig. 5B) including the arcuate bundles of the superior retina over a 108-arc deg sector of the ONH (36 144 arc
deg, or 10 2 oclock) and, (3) the inferior ONH data (Fig. 5C)
representing the arcuate bundles from the inferior retina in a
108-arc deg sector of the ONH (from 216 324 arc deg, or
4 8 oclock). The correlations between axons and cell bodies for each sector were analyzed by the same statistics as the
composite data in Figure 4. For the superior and inferior
arcuate bundles, the MAD values (0.7 dB in both regions) are
similar to the data for the full visual field and, although the r2
values for the regional analyses are lower than those for the full
field, the 1:1 relationships are statistically significant (P
0.004) in both cases. In contrast, the data for the papillomacular bundles are more scattered (MAD 1.0 dB) and the unity
relationship is not significant (P 0.12). However, in all three
cases, the data clustered about the 1:1 line and the lower
correlations are probably a result of the smaller range of data to
define the relationship with confidence, and thus they were
considered to be generally supportive of the structurefunction models proposed for the study.

DISCUSSION
The present investigation was undertaken to determine how
the age-dependent decrease in visual sensitivity for SAP measurements and thickness of the RNFL by OCT measurements
are related to a persons age-related losses in RGCs. Specifically,
the present study was a test of a proposed model of RNFL

4441

thickness that incorporates a remodeling of the nerve fiber

layer by an increase in non-neuronal tissue, which appears to
be initiated by the axonal loss from normal aging.26 The model
was tested by an application of procedures to derive RGC
density from SAP sensitivity and RGC axons from RNFL thickness, by using the data from normal eyes of subjects between
approximately 18 and 80 years of age. These empiric results
provided strong support for the model developed with normative SAP and OCT data,30 and, thus, the present study is additional evidence that visual sensitivity measurements by SAP are
an accurate reflection of RGC density, without an age-dependent variable, whereas the translation of RNFL thickness to
RGC axons includes an age-dependent variable for axon density to produce an accurate estimate of the neuronal population.
Compensations for the effects of normal aging have been
incorporated in clinical measurements by SAP and OCT
through an evaluation of the significance of an individuals data
with respect to the mean and range of values for an ageequivalent normal population.4,8,18 This method of deriving
statistical probabilities is an effective diagnostic procedure, but
it does not provide information for comparison across diagnostic procedures or for an assessment of whether pathologic
alterations of the structural and functional components of
vision are in agreement. The present study suggests that the
comparison of the results of objective and subjective testing
for a given patient can be achieved by the translation of
structural and functional measurements to a common parameter that is related to the population of RGCs, which in turn
requires an age-dependent variable for RNFL thickness. The
application of these procedures produced consistent estimates
of RGC populations from SAP and OCT measurements for
normal eyes of patients across five decades of life, with equal
accuracy and precision for each decade (Fig. 4). Thus, the
normal age-dependent variation in neuronal populations sets a
baseline for the evaluation of the pathologic alterations caused
by glaucoma.
The function for normal age-related thinning of the total
RNFL thickness is shallow, with a decrease of the mean thickness of 3 m per decade (Fig. 2A) and standard deviations of
measurements for each decade of 5 to 8 m (Table 1). These
properties of the aging function are similar to published
data,10 16 which have called into question the clinical significance of an aging effect and the age-adjustment for normative data.16 However, the true neuronal aging effect is not
reflected in the OCT measurement of RNFL total thickness
because of the non-neuronal component of the measurement and, in fact, a 3-m reduction in RNFL thickness
represents a loss of approximately 60,000 RGCs (Fig. 2C). In
addition, the variability of the OCT data may not be an
indication of measurement variability, but rather may represent intersubject variability in retinal neurology. With incorporation of a correction for the non-neuronal component of
RNFL thickness, the age-related loss of neurons has a steeper
slope (Fig. 2D) and a high degree of intrasubject correspondence between SAP and OCT measurements. For example,
the data for the youngest and oldest subjects (Fig. 3) illustrate the concordance of neuronal estimates, which is confirmed by the group data (Fig. 4) showing that the average
difference between procedures is less than 1.0 dB.
The correlation between RNFL thickness and perimetric
sensitivity for normal, healthy eyes that was found in the
present study is stronger than the relationship reported recently by Hood and Kardon.35 The contrasting results are
probably a consequence of substantive methodological differences in (1) the translation of the visual sensitivity to a neural
substrate, (2) the portion of the scan used in the analysis (cf.,
Figs. 4, 5), and (3) the consideration of age as an independent

4442

Harwerth et al.

variable. Therefore, because the methods of analyzing OCT and

SAP data were substantially different, the results of the two
studies cannot be compared directly. Nevertheless, although
the methodological issues must be resolved, it seems reasonable that the function that relates OCT and SAP measurements
to RGC losses from glaucoma should also relate these clinical
measurements to RGC losses from aging.
The methods used for deriving axon counts from RNFL
thickness measures were developed to achieve correspondence between axon counts from OCT scans and the number
of RGCs estimated by SAP data29,30,32 and therefore are dependent on the veridical translation of visual sensitivity measures
to RGC densities. The methods used for these translations were
based on the general psychophysiological finding that visual
thresholds are nonlinearly correlated to their neural substrate,36 39 which is supported by the relationship between
visual sensitivity and ganglion cell density across the retinal
area tested by SAP.31 The RGC densities for retinal areas that
correspond to the central test locations are 10 dB (10 times)
greater than for the more peripheral test locations (50,000
cells/mm2 at a 4-arc deg eccentricity vs. 5000 cells/mm2 at a
21-arc deg eccentricity), whereas the visual sensitivities vary by
5 dB, or threefold (35 dB vs. 30 dB) across this region of the
visual field. More important, the psychophysiological link was
strongly supported by empiric studies of point-wise SAP measurements and corresponding histologic counts of RGC densities in monkeys with experimental glaucoma40,41 and in humans with clinical glaucoma.42 Thus, although neither
alternative models35,43 45 nor a role for RGC dysfunction44,46,47 was tested, the nonlinear structurefunction relationship for SAP seems sound. On the other hand, small deviations in the parameters for estimating axon numbers from
RNFL thickness would affect only the magnitude and not the
form of the relationships presented in the present study.
The normal aging-related losses of neurons in the inner
retina also may help to explain why age is a significant risk
factor for the development of visual defects associated with
glaucomatous neuropathy. If the occurrence of clinically significant visual field defects represent the time when neuronal
losses reach a critical number, then a visual defect will occur
when the pathologic and nonpathologic losses sum to that
number. Consequently, a young person will have a neural
reserve that is depleted with age, making the elderly patient
more susceptible to the development of visual field defects
from pathologic losses. For example, the functions presented
in Figures 2C and 2D show that, on average, the retina of a
25-year-old has twice as many RGCs as that of a 95-year-old
patient. Similarly, although regional or eccentricity variations
of age-dependent losses of RGCs have not been defined, the
large neural reserve for the central retina may partially explain
why the central defects occur at later stages of glaucoma than
do peripheral visual field defects.
It is implicit in the proposed model of age-related thinning
of the RNFL (Fig. 1) that remodeling of the nerve fiber layer, to
compensate partially for axonal loss, is a direct response to
axon loss. Further, if this form of remodeling is a general
response to axonal loss, then it is likely that the pathologic loss
of RGC axons in glaucoma would also initiate an increase in the
non-neuronal components of the RNFL. An increase in nonneuronal support tissue has been demonstrated histologically
by increased glial cell density in human donor eyes with glaucoma, compared with control eyes.27,48,49 Also, preliminary
data from patients with glaucoma, obtained by using the procedures described in this study, have suggested a proportional
decrease in axonal density in the RNFL as a function of visual
field defect depth (Wheat JL, et al. IOVS 2007;48:ARVO E-Abstract 491). It is likely, therefore, that the residual RNFL thickness for OCT measurements in eyes with advanced or end-

IOVS, October 2008, Vol. 49, No. 10

stage glaucoma is a reflection of this process, and the residual
thickness is actually a greater amount of glial tissue than would
be present in a normal healthy eye. For this reason, RNFL
thickness measurements on eyes that are blind from either
glaucomatous13,35,50 or nonglaucomatous51 neuropathy may
retain approximately 50% of the thickness of a healthy eye in
some regions around the optic nerve. However, if it is determined that remodeling in glaucoma is systematic, as with
aging, then the neuronal and non-neuronal components of total
thickness can be separated computationally.
In conclusion, the study of the normal age-related losses of
RGCs and their axons provided additional support for a model
proposing that the proportion of RNFL that is composed of
RGC axons is not constant across ages, but rather, greater
proportions of the total thickness are non-neuronal tissue in
older patients.30 The present analysis of empiric data from
normal eyes of subjects between 18 and 80 years of age
strongly supported the model of age-dependent thinning of the
RNFL that had been developed with normative SAP and OCT
data and, together, the results imply that the dynamic range for
RNFL thickness measurements decreases with age. Further,
because glaucoma can be considered as accelerated aging,52 a
similar remodeling of RNFL thickness should occur with glaucomatous neuropathy and, thus, stage-dependent corrections
are needed to translate OCT measurements to neuronal components in addition to the age-dependent corrections derived
in the present study.

References
1. Spear PD. Neural basis of visual deficits during aging. Vision Res.
1993;33:2589 2609.
2. Neufeld AH, Gachie EN. The inherent age-dependent loss of retinal
ganglion cells is related to the lifespan of the species. Neurobiol
Aging. 2003;25:167172.
3. Heijl A, Lindgren G, Olsson J. Perimetric threshold variability and
age. Arch Ophthalmol. 1988;106:450 452.
4. Heijl A, Lindgren G, Olsson J. A package for the statistical analysis
of visual fields. Doc Ophthalmol. 1987;49:153168.
5. Johnson CA, Adams AJ, Lewis RA. Evidence for a neural basis of
age-related visual field loss in normal observers. Invest Ophthalmol Vis Sci. 1989;30:2056 2064.
6. Lachenmayr BJ, Kojetinsky S, Vivell PM. Is there and accelerated
loss at older age for normal sensitivity in the central visual field? In:
Wall M, and Mills RP. Perimetry Update 1994/95. Amsterdam:
Kugler Publications; 1995;49 56.
7. Spry PGD, Johnson CA. Senescent changes in the normal visual
field: an age-old problem. Optom Vis Sci. 2001;78:436 441.
8. Heijl A, Patella VM. Essential Perimetry. 3rd ed. Dublin, CA: Carl
Zeiss Meditec; 2002.
9. Gardiner SK, Johnson CA, Spry PGD. Normal age-related sensitivity
loss for a variety of visual functions throughout the visual field.
Optom Vis Sci. 2006;83:438 443.
10. Alamouti B, Funk J. Retinal thickness decreases with age: an OCT
study. Br J Ophthalmol. 2003;87:899 901.
11. Kanamori A, Escano MTF, Eno A, et al. Evaluation of the effect of
aging on retinal nerve fiber thickness measured by optical coherence tomography. Ophthalmologica. 2003;217:273278.
12. Varma R, Bazzaz S, Lau M. Optical tomographymeasured retinal
nerve fiber layer thickness in normal Latinos. Invest Ophthalmol
Vis Sci. 2003;44:3369 3373.
13. Sony P, Sihota R, Tewari Hem K, et al. Quantification of the retinal
nerve fibre layer thickness in normal Indian Eyes with optical
coherence tomography. Indian J Ophthalmol. 2004;52:303309.
14. Ramakrishnan R, Mittal S, Sonal A, et al. Retinal nerve fibre layer
thickness measurements in normal Indian population by optical
coherence tomography. Indian J Ophthalmol. 2006;54:1115.
15. Hougaard JL, Ostenfeld C, Heijl A, et al. Modelling the normal
retinal nerve fiber layer thickness as measured by Stratus optical
coherence tomography. Graefes Arch Clin Exp Ophthalmol. 2006;
244:16071614.

IOVS, October 2008, Vol. 49, No. 10

16. Budenz DL, Anderson DR, Varma R, et al. Determinants of normal
retinal nerve fiber layer thickness by Stratus OCT. Ophthalmology.
2007;114:1046 1052.
17. Parikh RS, Parikh SR, Sekhar GC, et al. Normal age-related decay of
retinal nerve fiber layer thickness. Ophthalmology. 2007;114:921
926.
18. Patella VM. StratusOCT: Establishment of Normative Reference
Values for Retinal Nerve Fiber Layer Thickness Measurements.
Dublin CA: Carl Zeiss Meditec; 2005.
19. Blanks JC, Torigoe Y, Hinton DR, et al. Retinal pathology of
Alzheimers disease. I. Ganglion cell loss in foveal/parafoveal retina. Neurobiol Aging. 1996;17:377384.
20. Harman A, Abrahams B, Moore S, et al. Neuronal density in the
human retinal ganglion cell layer from 16 77 years. Anat Rec.
2000;260:124 131.
21. Kerrigan-Baumrind LA, Quigley HA, Pease ME, et al. Number of
ganglion cells in glaucoma eyes compared with threshold visual
field tests in the same persons. Invest Ophthalmol Vis Sci. 2000;
41:741748.
22. Balazsi AG, Rootman J, Drance SM, et al. The effect of age on the
nerve fiber population of the human optic nerve. Am J Ophthalmol. 1984;97:760 766.
23. Mikelberg FS, Drance SM, Schulzer M, et al. The normal human
optic nerve: axon count and axon diameter distribution. Ophthalmology. 1989;96:13251328.
24. Repka MX, Quigley HA. The effect of age on normal human optic
nerve fiber number and diameter. Ophthalmology. 1989;96:26
32.
25. Jonas JB, Schmidt AM, Muller-Bergh JA, et al. Human optic nerve
fiber count and optic disc size. Invest Ophthalmol Vis Sci. 1992;
33:20122018.
26. Ogden TE. Nerve fiber layer of the primate retina: thickness and
glial content. Vision Res. 1983;23:581587.
27. Wang L, Cioffi GA, Cull G, et al. Immunohistologic evidence for
retinal glial cell changes in human glaucoma. Invest Ophthalmol
Vis Sci. 2002;43:1088 1094.
28. Tezel G, Wax MB. Glial modulation of retinal ganglion cell death in
glaucoma. J Glaucoma. 2003;12:63 68.
29. Harwerth RS. Charles F. Prentice Award Lecture 2006: a neuron
doctrine for glaucoma. Optom Vis Sci. 2008;85(6):436 44.
30. Harwerth RS, Wheat JL. Modeling the effects of aging on retinal
ganglion cell density and nerve fiber layer thickness. Graefes Arch
Clin Exp Ophthalmol. 2008;246:305314.
31. Harwerth RS, Carter-Dawson L, Smith EL III, et al. Neural losses
correlated with visual losses in clinical perimetry. Invest Ophthalmol Vis Sci. 2004;45:31523160.
32. Harwerth RS, Vilupuru AS, Rangaswamy NV, et al. The relationship
between nerve fiber layer and perimetry measurements. Invest
Ophthalmol Vis Sci. 2007;48:763773.
33. Bengtsson B, Heijl A, Olsson J. Evaluation of a new threshold visual
field strategy, SITA, in normal observers. Swedish Interactive
Thresholding Algorithm. Acta Ophthalmol Scand. 1998;76:165
169.
34. Artes PH, Iwase A, Ohno Y, et al. Properties of perimetric threshold estimates from full threshold, SITA standard, and SITA fast
strategies. Invest Ophthalmol Vis Sci. 2002;43:2654 2659.

Age-Related Losses of RGCs and Axons

4443

35. Hood DC, Kardon RH. A framework for comparing structural and
functional measure of glaucomatous damage. Prog Retin Eye Res.
2007;26:688 710.
36. Barlow HB. Single units and sensation: a neuron doctrine for
perceptual psychology. Perception. 1972;1:371394.
37. Nachmias J. On the psychometric function for contrast detection.
Vision Res. 1981;21:215223.
38. Tolhurst DJ, Movshon JA, Dean AF. The statistical reliability of
signals in single neurons in cat and monkey visual cortex. Vision
Res. 1983;23:775788.
39. Parker AJ, Newsome WT. Sense and the single neuron: probing the
physiology of perception. Annu Rev Neurosci. 1998;21:227277.
40. Harwerth RS, Crawford MLJ, Frishman LJ, et al. Visual field defects
and neural losses from experimental glaucoma. Prog Retin Eye
Res. 2002;21:91125.
41. Harwerth RS, Carter-Dawson L, Smith EL, et al. Scaling the structure-function relationship for clinical perimetry. Acta Ophthalmol
Scand. 2005;83:448 455.
42. Harwerth RS, Quigley HA. Visual field defects and retinal ganglion
cell losses in patients with glaucoma. Arch Ophthalmol. 2006;124:
853 859.
43. Garway-Heath DF, Capriolo J, Fitzke FW, et al. Scaling the hill of
vision: the physiological relationship between light sensitivity and
ganglion cell numbers. Invest Ophthalmol Vis Sci. 2000;41:1774
1782.
44. Swanson WH, Felius J, Pan F. Perimetric defects and ganglion cell
damage: interpreting linear relations using a two-stage linear
model. Invest Ophthalmol Vis Sci. 2004;45:466 472.
45. Hood DC, Anderson SC, Wall M, et al. Structure versus function in
glaucoma: an application of a linear model. Invest Ophthalmol Vis
Sci. 2007;48:36623668.
46. Ventura LM, Sorokac N, De Los Santos R, et al. The relationship
between retinal ganglion cell function and retinal nerve fiber layer
thickness in early glaucoma. Invest Ophthalmol Vis Sci. 2006;47:
3904 3911.
47. Whitmore AV, Libby RT, John SWM. Glaucoma: thinking in new
waysa role for autonomous axonal self-destruction and other
compartmentalized processes. Prog Retin Eye Res. 2005;24:639
662.
48. Varela HJ, Hernandez MR. Astrocyte responses in human optic
nerve head with primary open-angle glaucoma. J Glaucoma. 1997;
6:303313.
49. Grieshaber MC, Orgul S, Schoetzau A, et al. Relationship between
retinal glial cell activation in glaucoma and vascular dysregulation.
J Glaucoma. 2007;16:215219.
50. Sihota R, Sony P, Gupta V, et al. Comparing glaucomatous optic
neuropathy in primary open-angle and chronic primary angle closure glaucoma by optical coherence tomography. Ophthalmic
Physiol Opt. 2005;25:408 415.
51. Chan CKM, Miller NR. Peripapillary nerve fiber layer thickness
measured by optic coherence tomography in patients with no light
perception from long-standing nonglaucomatous optic neuropathies. J Neuroophthalmol. 2007;27:176 179.
52. Tezel G, Luo C, Yang X. Accelerated aging in glaucoma: immunohistochemical assessment of advanced glycation end products in
the human retina and optic nerve head. Invest Ophthalmol Vis Sci.
2007;48:12011211.