Nerve Fiber Atrophy

Precedes the Onset of Glaucomatous Field Loss

Clinically Detectable

Alfred Sommer, MD, MHSc; Joanne Katz, MS; Harry A. Quigley, MD; Neil R.
Alan L. Robin, MD; Ronald C. Richter, MD; Kathe A. Witt, COMT

perimetry and nerve filayer and color fundus photography

were performed annually on 1344 eyes
\s=b\ Standardized

ber

with elevated intraocular pressures. In 83

eyes, glaucomatous field defects developed that met rigid criteria on manual
kinetic and suprathreshold static perimetry. Individual nerve fiber layer photographs were read by two masked observers. The more sensitive of the two
identified nerve fiber layer defects in 88%
of readable photographs at the time field
loss first occurred; 60% (6/10) of eyes
already had nerve fiber layer defects
6 years before field loss. In contrast, the
nerve fiber layer was considered abnormal in only 11% (3/27) of normal eyes and
26% (84/327) of hypertensive eyes. The
location of nerve fiber layer and field
defects closely corresponded, but nerve
fiber layer loss was generally more widespread. Examiner experience and severity of optic nerve damage influenced results. Mild focal defects were more
readily recognized than more severe diffuse atrophy. Nerve fiber layer defects
expanded with time, often by the development and coalescence of adjacent areas of damage.

(Arch Ophthalmol. 1991;109:77-83)

"1 he definition and diagnosis of pri-

angle glaucoma

remain
Intraocular pressure
(IOP) is an important risk factor for
glaucoma but not definitive evidence of
the disease.13 At present, diagnosis
generally requires evidence of typical
glaucomatous optic nerve atrophy in
the absence of other potential causes.
Methods for reliably establishing
glaucomatous nerve atrophy are lim
ited. These include functional distur
bance, traditionally identified by typi
cal field abnormalities (eg, localized or
asymmetric depressions in retinal sen
sitivity corresponding to the arcuate
mary open

problematic.

Accepted for publication August 28,

1990.
From the Dana Center for Preventive Ophthalmology of the Wilmer Eye Institute, and the
School of Public Health, the Johns Hopkins Medical Institutions, Baltimore, Md.
Reprint requests to Wilmer 120, The Johns
Hopkins Hospital, 600 N Wolfe St, Baltimore, MD
21205 (Dr Sommer).

Miller, MD;

distribution of nerve fiber bundles),

and/or anatomical evidence suggestive
of optic nerve fiber loss (eg, large or
vertically elongated cups, notching of
the disc rim, or asymmetric cupping
between the two eyes). These ap
proaches have their shortcomings, not
the least of which is a delay until
sufficient damage for definitive diagno
sis has occurred. For example, field
defects can be preceded by significant
loss of retinal ganglion cells.4 Most
anatomical measures are nonspecific,
overlapping to a considerable degree
with findings in the normal population.
In 1972, Hoyt and coworkers"" sug
gested that slitlike defects in the ap
pearance of the nerve fiber layer
(NFL) may represent early, clinically
detectable manifestations of glaucoma
tous damage. In 1977, we confirmed, in
masked readings, that NFL defects
are more frequent among patients with
established field loss.7 Subsequent re
ports confirmed the correlation be
tween NFL abnormalities and psychophysical disturbances, including field

loss.8"10

Importantly, in a small number of

subjects photographed (suboptimally)

for different purposes, we identified

NFL defects years before the develop
ment of reproducible field abnormali
ties on routine kinetic and suprathreshold static perimetry.7 This
suggested that NFL assessment might
identify patients with progressive
glaucomatous neuropathy before the
loss of fibers required for the develop
ment of reproducible field defects.
In 1981, we initiated a prospective,
longitudinal study of normal individ
uals, subjects with ocular hypertension
(OH), and patients with glaucoma to
define the role of NFL assessment in
the diagnosis of glaucoma. An earlier
report from this study11 established
that NFL assessment identifies eyes
with established glaucomatous field
loss with relatively high sensitivity
and specificity. With further recruit
ment and follow-up, we now report the
temporal relationship between the de
velopment of visual field defects and
the prior appearance of the NFL.
These data confirm the value of NFL
assessment as an early indicator of

glaucomatous optic neuropathy.

Downloaded From: http://archopht.jamanetwork.com/ by a UQ Library User on 10/24/2014

SUBJECTS AND METHODS

After a comprehensive baseline evalua
tion, subjects were reexamined annually

following a strict protocol.11

Recruitment

Ocular hypertension was defined as IOP

above 21 mm Hg on at least two visits
before treatment and normal visual field
test results as described below. These sub
jects were recruited from the Wilmer Insti
tute, the Johns Hopkins Medical Institu
tions, Baltimore, Md; the practices of
collaborating ophthalmologists; among sub
jects originally enrolled in the Collaborative
Glaucoma Study12; and among patients iden
tified in the course of the Baltimore Eye
Survey.1'1 Ophthalmologists responsible for
their routine care remained free to treat
each patient as they thought was most
appropriate. Normal controls were recruit
ed from the same sources, from patients,
visitors, and staff of The Johns Hopkins
Hospital, and from church groups and se
nior citizen's associations. All controls were
free of a personal or family history of
glaucoma, elevated IOP, or the use of IOPlowering medication and they all had an
IOP below 22 mm Hg and entirely normal
visual fields on at least two visits.
At baseline, all subjects provided in
formed consent and a detailed history and
were examined by one of the several inves
tigators with subspecialty training in glau
coma. Ocular status was confirmed from the
baseline examination and from records pro
vided by past and current ophthalmologists
and copies of all available visual fields.
Visual Field Tests

The central 30 of the visual field was

tested with best distance correction and the
appropriate plus sphere for age. From 1981
through 1986, the central and peripheral
visual field was evaluated by detailed perimetric techniques following a strict protocol
employing detailed threshold-related, suprathreshold kinetic, and static stimuli on a
Goldmann perimeter examination.11 The av
erage examination required 20 to 30 min
utes per eye and included at least four
isopters encircling 360 of the field, two
additional isopters limited to the nasal pe
riphery, and literally hundreds of static

presentations.
Beginning in 1986, routine follow-up ex
aminations were performed on the Hum
phrey Field Analyzer (Allergan-Humphrey
Instruments, San Leandro, Calif) using the
C-30-2 program. All data were transferred
a central file and analyzed by our cross-

to

meridional, mirror-image technique using

specially created software.14 All subjects

Fig 1.Results in a 48-year-old white woman with normal visual field examination by Goldmann perimetry and Humphrey
C-30-2 in both eyes until 1988, when a small paracentral scotoma connecting to the blind spot developed in the right eye.
Top left, Nerve fiber layer (NFL) photograph from 1984 of the right eye. The appearance is entirely normal, with bright
striations, which cross the first- and second-order vessels, blurring the wall edges. The maculopapular bundle is dark
relative to the arcuate zones above and below. Bottom left, Corresponding Goldmann field from the same visit. Top center,
Photograph of the same eye from 1985 reveals a subtly dark, wedge-shaped area within the inferior arcuate zone, with
brighter NFL striations on either side (arrows). Bottom center, Corresponding Humphrey threshold field from the same visit.
Top right, Same eye photographed in 1987. The inferior defect has become darker and wider. The maculopapular bundle is
now brighter and more obviously striated in comparison. Vessels within the area of NFL atrophy stand out in sharp relief
(arrows). A swath of the superior arcuate zone has also become atrophie. Bottom right, Corresponding Humphrey threshold
field from the same visit.

meeting any of our criteria for field loss10 or

with suspicious clusters of depressions iden
tified by STATPAC (Allergan-Humphrey
Instruments, San Leandro)16,1'

were

re

called and reexamined on the Goldmann

perimeter, which remained the final basis
for interpretation of field status. Subjects
unreliable to threshold testing18 were rou
tinely examined on the Goldmann
instrument.
A definite, typical glaucomatous field ab
normality included one or more of the fol
lowing defects on the Goldmann perimeter,
confirmed on at least two occasions (when
the Goldmann defect was congruent with a
defect first detected by threshold peri
metry, a second, confirming Goldmann field

not required): (1) a paracentral

or arcuate scotoma (including arcuate elon
gation but not generalized enlargement of
the blind spot) at least 0.4 log units in
depth; (2) nasal step of at least 10 in width
present to at least two isopters; and central
and/or temporal islands. As this report
deals only with subjects until such time as

study

was

they first developed a field defect

annual examination, all the defects

at their
met the

mildest of our criteria.

Photographs of the posterior pole were
taken through a dilated pupil with a Zeiss
fundus camera. Stereoscopic color photo
graphs of the disc and peripapillary nerve
fiber layer (focused at the level of the
superficial vessels) were recorded on Kodachrome 25 film. A series of five monoscopic,
red-free black and white photographs were
taken of each eye: one centered on the disc
and two of each arcuate zone. These were
initially recorded on Kodak Plus X film
using a narrow band-pass filter.18 In 1983,
this approach was replaced by our improved

high-resolution technique employing a

short-pass filter and Kodak 2415 Technical
Pan film."0 Red-free negatives were printed
on positive transparency Dupont COS-7
film and the optic nerve head blocked out
with opaque metallic ink. Both red-free and
color transparencies were coded and inde
pendently evaluated by the same two ob
servers as in our previous report.11 Red-

Downloaded From: http://archopht.jamanetwork.com/ by a UQ Library User on 10/24/2014

free and color transparencies were

circulated (and evaluated) independent of
one another. In evaluating the photo
graphs, each observer was masked as to
patient and diagnosis, all previous and sub
sequent photographs, and the visual fields.
Nerve fiber layer assessment followed a
rigorous pattern meant to elicit all available
information rather than allow pattern rec
ognition of particularly striking abnormali
ties (such as a focal wedge defect adjacent
to normal, bright retina) to obscure more
subtle and often more severe damage of a
diffuse nature. The readers focused their
attention on both the brightness and densi
ty of striations (axonal bundles) and the
clarity with which vessels could be visual
ized. Normally, major retinal vessels lie
within the NFL. Thinning of the NFL
results in a reduction in overlying striations
and greater visibility of the vessel walls
(Fig 1). Most information was obtained
from within 2 disc diameters of the optic
nerve. Comparisons were made between
the superior and inferior hemispheres and

same) photographs available at each of the

annual visits, and patients have occasionally
missed one or more visits. Thus, the num
bers of photographs read by the two ob
servers are rarely identical, and the num
ber of subjects on whom there are data
varies from interval to interval. To maxi
mize the utility of the available data, we
employ a life-table approach: where photo
graphs are available at a particular interval
in relation to the onset of field loss, the
subject is included; where they are not, the
subject is excluded.

RESULTS

Fig 3.A 51-year-old white woman was first

seen in 1984 with elevated intraocular pres
sure in both eyes. An isolated inferior para
central scotoma and small nasal step were
already present in the right eye. Within 2
years, there was generalized threshold de
pression throughout most of the inferior hemi
sphere. This nerve fiber layer (NFL) photo
graph of the right eye in 1985 reveals an
obvious inferior wedge adjacent to far more
normal NFL near the 6 o'clock position. Less
obvious is the diffuse NFL loss over a wide
area of the superior hemisphere (arrows).
The pattern from top to bottom (temporally) is
dark-light-dark-(light again), a reversal of nor
mal. Note how circumferentially oriented ves
sels at the 7 to 8 o'clock position "disappear"
as they move from the area of atrophy into
thicker NFL, whose striations overlie, thereby
masking the vessel walls.

A total of 1344 eyes with OH under

went baseline evaluation and at least

1 year of follow-up. Because of the
serial nature of enrollment, the num
ber present at subsequent visits de

clines with the duration of follow-up

(Table 1). One thousand twelve eyes
with OH remain under active investi

gation. Only 91 eyes (6.8%) belonged

patients who refused further follow-

to

up.

Eyes With Visual Field Conversion

Of the subjects in whom field defects
developed and who are the basis of this
report, 63% were black and 69% were
60 years or older. Visual acuity
20/40 (6/12) or better in 92%.

Reader A examined red-free

2. Results in a 35-year-old white wom

with diffuse inferior nerve fiber layer (NFL)
loss in the right eye (top) involving the macu
lopapular bundle (1984). Normal left eye (top
center) is shown for comparison. Bottom cen
ter, Goldmann field of right eye from 1983 is
shown. Bottom, Humphrey threshold field
from 1987 is shown.

Fig

an

the corresponding hemispheres of the two

eyes. The temporal retina was scanned for
alterations in its usual appearance. In the
normal eye, the brightest reflexes of the
NFL come from the upper and lower disc
polar areas, where the NFL is thickest.
The temporal retina (particularly the macu
lopapular bundle) has thinner NFL and is
therefore darker. Thus, the normal tempo
ral pattern from above to below is bright
reflexes superiorly, dark toward the fovea,
and bright again inferiorly. The brightdark-bright transition is relatively smooth
in eyes without atrophy (Figs 1 and 2).21"24
Alterations in this pattern, particularly dif
fuse atrophy, were often detected as one or
both of the arcuate zones being as dark as
or darker than the adjacent maculopapular
bundle (Figs 1, 3, and 4). Slit defects, which
are commonly narrower than the diameter
of adjacent retinal venules and fail to fan
out from disc to periphery, are of no diag
nostic significance since they are frequently
encountered in normal eyes.
To maintain masking in relation to previ
ous and future photographs, the two read
ers never received more than one annual
set of red-free photographs of a subject at a
time. To maintain masking as to clinical
diagnosis as well as to provide a basis for
identifying potential "drift" in the readings
or in the photographic technique, photo
graphs of normal control eyes and of eyes
with OH in which abnormal visual fields had
not yet developed, matched for the followup visit, were evaluated at the same time.
As the study is still ongoing, the two
observers have not yet reviewed all (or the

Downloaded From: http://archopht.jamanetwork.com/ by a UQ Library User on 10/24/2014

was

photo

graphs from 213 visits. Twenty (9.4%)

of the

photographs

were

considered

interpret. Reader consid

similar proportion unreadable

too poor to

ered

(15/214 [7.0%]). The oldest red-free

photographs, taken an average of 6

years before field conversion, were
most likely to be uninterpretable (Ta
ble 2). These had been obtained before
the introduction of our high-resolution

photographic technique.2"
As in our previous study,11 uninter
pretable photographs were more likely
to have been from older

subjects (Ta

ble 3). At the time of field conversion,

none of the eyes of subjects younger
than 60 years evaluated by either read
er were considered too poor to analyze
(none of 19 eyes, reader A; none of 18
eyes, reader B); in contrast, reader A
was unable to evaluate six of 46 eyes,
and reader B, five of 46 eyes from
subjects over 60 years old.
was twice as likely as
Reader
reader A to call NFL photographs
abnormal from eyes that suffered field
loss (Table 2). Reader considered the
NFL abnormal in a majority of eyes at
every visit (except one) before the
development of a field defect. Nearly
90% of eyes had an abnormal NFL at
the time of field conversion, the rate
declining gradually (and somewhat er
ratically) as the interval to field loss
lengthened. While the positivity rate

of reader A

was lower, it too was

reasonably steady over time, with
roughly one third to one half of the
readings being abnormal.
Readings by the two observers
agreed on the NFL status more often
than not (Table 4). At conversion, they
disagreed about NFL status in 22

Fig

4.A

read as
eyes, all of which reader
abnormal. Those eyes that they agreed
were abnormal had the most severe
field defects; those that they agreed
were normal had the mildest field de
fects; and those about which they dis
agreed had defects between these two
extremes (Table 5). Eyes that they
both considered normal had smaller
cups and wider rims than those consid
ered abnormal by one or both observ
ers. Neither race nor iris color, factors
commonly associated with fundus pig
mentation and therefore the potential
visibility of the NFL, consistently in
fluenced the likelihood that an eye was
considered abnormal (data not given).
Nerve fiber layer abnormalities
were seemingly more extensive than
the corresponding field loss. At con
version, visual field loss (by kinetic and
suprathreshold static perimetry) in
volved both hemifields in only 14.8%
(reader A) and 13.2% (reader B) of all
eyes in which NFL defects were re
corded. In contrast, NFL abnormali
ties were noted in both hemispheres of
81.5% (reader A) and 84.9% (reader B)
of these same eyes (Table 6). The
proportion of eyes in which NFL ab
normalities involved both hemispheres
was greater for the three visits ending
with conversion than for the four pre
ceding visits.
In eyes with apparently localized
NFL defects, there was a close corre
spondence between the location of the
visual field and the NFL abnormali
ties. At conversion, field loss and NFL
defects were both limited to one hemi
sphere in seven eyes read by reader B,
and in every instance the hemispheres
were in perfect correspondence. Over-

man was first

elevated intraocular pres
A superior nasal step first
developed in the right eye in 1985, which
became denser and larger in 1987. By 1989,
this formed a sharply demarcated arcuate
scotoma superiorly. (This case was previous
ly reported for other purposes.31) Top left,
Nerve fiber layer (NFL) photograph taken in
1983 shows a disc hemorrhage at the 8
o'clock position in an area where the NFL
otherwise appears normal. Note the preexist
ing, mild loss of NFL at the 7 o'clock position
(arrows). Center left, The corresponding
Goldmann field is relatively normal. The NFL
photographs from 1985 through 1989 were
identical. Top right, The defect in the inferior
NFL has deepened and expanded superiorly
(toward the area of the disc hemorrhage in
1983) and interiorly (toward the 6 o'clock po
sition meridian) (arrows). The maculopapular
bundle is clearly spared. Center and bottom
right, Goldmann (1985) and Humphrey
threshold (1987) fields reflect the damage.

38-year-old white

seen in 1982 with

sure in both eyes.

Table 1.

Number of Ocular

Hypertensive Eyes at Each

Follow-up Examination

Table 2.NFL Assessment in

Unreadable

No. of

No. of Years
Before Onset of
Field Defect

Follow-up
Examinations
1
2

Eyes in Which Field Defects Subsequently Developed*

No. of Eyes
_

1176*

935
699

5301

5
6

672
449

One hundred forty eyes missed their first followwere seen at least once at a later follow-up.
tRepresents temporary interruption of study from
hiatus in funding.
up but

NFL indicates

nerve

Reader B,
No. (%)

Reader

A,
(%)
3/14(21.4)
2/16 (12.5)
0/21 (0.0)
0/18 (0.0)
3/21 (14.3)
6/58(10.3)
6/65(9.2)
fiber layer. Year 0
No.

Downloaded From: http://archopht.jamanetwork.com/ by a UQ Library User on 10/24/2014

3/13(23.1)
1/21 (4.8)
1/20(5.0)
0/19(0.0)
2/21 (9.5)
3/57 (5.3)
5/63(7.9)
was

year in which

Abnormal

Reader A,
No. (%)

Reader

B,

No. (%)

6/10(60.0)
15/20(75.0)
9/19(47.3)
16/19 (84.2)
15/19(78.9)
33/54(61.1)
51/58 (87.9)
visual field defect first appeared.
5/11

(45.5)
(14.3)
7/21 (33.3)
10/18 (55.6)
8/18(44.4)
19/52 (36.5)
27/59 (45.8)
2/14

Table 4.Agreement Between Readers

Table 3.Mean Age NFL

Assessment at Time of Field Loss*

Age, y

No.

61 13
64 15
73 10

32

Reader A

Normal

Abnormal
Cannot evaluate
Reader
Normal
Abnormal
Cannot evaluate
NFL indicates

nerve

65 10
62 14
76 6
fiber

NFL Status*

Reader A/Reader

No. of Years
Before Onset of
Field Defect

Mean (SD)
Assessment

on

abn/abn

Agreement

Total

abn/nl

100.0

27

6
64.7

7
51
5

77.6

49

20

60.0

layer.
*

34 examinations by reader B,
field loss and NFL defects were limit
ed to a single hemisphere; in 33 (97%)
these were in perfect correspondence.
Results for reader A, who found fewer
NFL abnormalities, were less well cor
related: overall, there was 69% corre

all, in

to

Excludes photographs examined by only one reader to date or that one or both readers considered too poor
nerve fiber layer; abn, abnormal; and nl, normal.

Table 5.Agreement Between Readers and Severity of

Optic

Nerve

Damage*

Reader A/Reader
nl/abn

abn/abn

At Field
Conversion

(n

More severe
field loss,
No. (%)

19

(73.1)

(15.4)

nl/nl

(n

(n 22)

26)

14

(63.6)

7)

(57.1)

Field defect
in more than
1 hemisphere,

Eyes Without Visual Field Conversion

The rates at which the two readers
reported NFL abnormalities for nor
mal control eyes and control eyes with
OH were similar to those at the base
line examination, published 6 years
ago.* Both rates, of both readers, were
substantially below the rates for eyes
that subsequently suffered glaucoma
tous field loss (Table 8).

172

56

interpret. NFL indicates

spondence (11/16).

Once an eye was judged to have an

abnormal nerve fiber layer it was like
ly to be judged abnormal at all subse
quent visits (Table 7).

48

68

Total examinations

No.

(%)

2(9.1)

(0.0)

Horizontal

cup/disc,
mean

2 SE

Reader A
Reader

0.62 0.12
0.60 0.08

0.63 0.06
0.67 0.06

0.40 0.18
0.37 0.23

Reader A
Reader

0.15 0.06
0.16 0.05

0.13 0.05
0.14 0.05

0.27 0.13
0.32 0.12

Narrowest
rim

width,

mean

2 SE

Severe indicates
arcuate scotoma

or

more

worse;

a single defect (isolated paracentral scotoma

abn, abnormal; and nl, normal.

than

or

isolated nasal

step),

eg, full

COMMENT

Hoyt and Newman5 suggested that

slit and wedge-shaped defects might
be the earliest clinical signs of glauco
matous optic nerve atrophy. We subse
quently demonstrated that the vast
majority of subjects with glaucoma
tous field defects already had demon
strable NFL abnormalities,711 and that
in a small sample followed up longitudi
nally, such abnormalities developed in
a majority at least 3 to 5 years before
the onset of field loss.7 Aside from our
subsequent preliminary report from
the present study, z few reliable pro
spective data have appeared except for
the small group of subjects with disc
hemorrhages described by Airaksinen
et al.26

The present study confirms our orig

inal conclusions of over a decade ago7:
NFL assessment can help identify pa
tients who have already suffered optic
nerve damage and in whom it will
progress to definitive visual field loss.
The present, prospective study em
ployed improved photographic tech
niques, defined criteria for NFL dam
age, rigid masking of observers, and

Table

6.Correspondence

Between NFL and Visual Field Abnormalities*

Hemispheres

No. of

Years
Before
Onset of
Field Defect

Hemifields

Reader

Reader A

With

Goldmann ,Defects
Sup Inf

Sup

3-6

Both

Total

10

(includes
conversion)
At conversion

Sup

Both
Total

Sup

Both
Total

6
17 (77.2%) 22
32
26
12
io

45

(83.3%)

12

Sup

(27.3%)

10(18.5%)

*ln
nerve

Total

4(14.8%)

by definition,
Sup, superior; and Inf, inferior.

all instances, visual field defects are,

fiber layer;

far more subjects, who were reexamined annually following a strictly moni
tored protocol.
Nerve fiber layer abnormalities
were infrequent among normal con
trols and somewhat more frequent

Downloaded From: http://archopht.jamanetwork.com/ by a UQ Library User on 10/24/2014

16

11

6 31 (67..
10 45

46

56
30
15

14

(23.9%)

(14.9%)

84(83.2%) 101

4 26

15

22 (81.5%) 27

Total

Both

11

54

8
Both

Inf

Inf

2-0

With NFL Abnormalities

.-

13

5 45

(84.!

30
16
7
53

(13.2%)

those at the time of field conversion. NFL indicates

among eyes with OH that had not

yet

shown definite field loss. These results

are almost identical to those reported
from the baseline assessment," con
firming our previous results and dem
onstrating little drift in either the pho-

Table 7.Consistency of Subsequent

NFL Interpretations Following an
Individual's Earliest Abnormal

Table 8.NFL Assessment in Eyes

That Have Not Shown Field Defects*
Reader A,
No. (%)

Reading*
Reader

12

82(87%)

3/27

Total

47

94

layer. Results

are

57/214(26.6) 84/327(25.7)
27/59 (45.8) 51/58 (87.9)
*
NFL indicates nerve fiber layer. Field loss is prev
alence of NFL defects at the time of field loss repeated
from Table 3 for ease of comparison. Difference in
rates for ocular hypertensive vs control eyes, < .05
(reader A), P< .001 (reader B).
Field loss

of

.001).

nature of NFL loss in the

series, where patients had

tographic technique

the criteria by
which the two readers judged the pres
ence of NFL abnormalities.
The prevalence of defects among
eyes with OH is presumably greater
than among normal controls because
some eyes with OH have already suf
fered glaucomatous optic nerve dam
age sufficient to alter NFL appearance
but insufficient to produce field loss
meeting our rigid criteria. Indeed, this
is exactly the explanation postulated in
our baseline report," before the eyes
that were then in the group with H
(many of which already had NFL de
fects) subsequently suffered the field
loss that is the basis of the present

analysis.

or

Like other anatomical measures of

nerve status, NFL assessment
has its limitations. Decreasing visibili
ty of the NFL with age, media opaci
ties, and other factors11 can reduce the
likelihood of obtaining interpretable
photographs. However, less than 10%
of photographs in this series, regard
less of observer, were too poor to read.
The sole exception was the oldest pho
tographs, obtained with the original
technique: almost a fourth were uninterpretable. Substitution of our highresolution technique2" appears to have
substantially improved visualization of
the NFL.
Interest and experience clearly im
proved performance. Five years ago,
reader A detected NFL defects in over
80% of eyes with glaucomatous field
loss of unknown duration. This was
only slightly lower than the 90% rate
of reader B. In the present series,
reader A continued to identify NFL
defects more frequently in eyes des
tined to suffer field loss than in normal
control eyes or eyes with OH that had
not shown field loss, but at only half
the rate of reader B. No doubt part of
the difference between the original re
sults and these reflects the subtler

optic

(11.1)

eyes

NFL readings at visits following the first at which the

subject's NFL was considered abnormal. Table ex
cludes unreadable photographs. The proportion sub
sequently read as abnormal was greater than ex
pected by chance (reader A, P-C.06; reader B,
P<

(3.7)

hypertensive

16

fiber

1/27

Ocular

31(66%)
nerve

controls

Reader

Normal
Abnormal
indicates

No.

Normal

Interpretation of
Subsequent
NFL Photographs

*NFL

B,
(%)

Reader

present
not yet
to show

shown or had only just begun

field loss meeting study criteria. In the
few instances where both readers
agreed the NFL was normal at the
onset of field loss, the field defect was
less severe than when one or both
readers considered the NFL ab
normal.
Reader A utilized the original crite
ria and evaluation techniques designed
in 1981. Reader B, who utilizes NFL
assessment daily in the management of
glaucoma and was free to adapt his
technique with experience, was clearly
more sensitive to subtler change (but
also slightly less specific). At the visit
at which field loss was first document
ed, NFL abnormalities were already
apparent in almost 90% of eyes, and
NFL defects were far more likely to
involve both the superior and inferior
hemispheres than were their corre
sponding field defects (85% vs 13%).
Part of this difference may be artifact.
Our criteria for early field loss exclude
mild contraction of the field (a less
specific sign of glaucomatous neuropa
thy than asymmetric variations in light
sensitivity, manifested in nasal steps,
paracentral scotomas, and the like).
Hence, some instances of early bihemispheric but symmetric field loss may
have been overlooked. It is unlikely,
however, that this would account for
the large difference in the extent of
NFL and visual field disturbance ob
served. Indeed, cross-meridional "mir
ror image" comparisons of threshold
fields, based on similar asymmetry be
tween the superior and inferior hemi
spheres, identifies 90% of cases with
glaucomatous field loss.15,16 Further
more, NFL defects became more wide
spread as field loss was approached,
additional evidence that NFL assess
ment documented progressive axonal
loss.
When the NFL abnormality was
limited to a single hemisphere, it al
most invariably corresponded to the

Downloaded From: http://archopht.jamanetwork.com/ by a UQ Library User on 10/24/2014

Fig 5.A 72-year-old black woman had ad

vanced field loss. The nerve fiber layer is
extremely thin. At this stage we see vessels in
sharp relief, irregular granularity, and widely
spaced striations.
location of the field defect, particularly
for reader B. As expected, the preva
lence of NFL abnormalities declined as
the time to field loss increased, but the
decline was less dramatic than might
have been expected. Roughly 60% of
eyes already had NFL abnormalities
4 to 6 years before field loss. Consis
tency was also high: once reader
identified an eye as abnormal, it was
far more likely than not to be consid
ered abnormal at all subsequent visits
despite unavoidable variation in the
quality and orientation of the photo
graphs from year to year.
The evolution of NFL abnormalities
varied with their initial appearance
and with the subsequent course of the
individual eye. Most began with some
degree of diffuse thinning already in
volving both hemispheres, principally
the arcuate zones. Diffuse loss was
best appreciated by the relative pauci
ty or absence of nerve fiber striations
over second-order vessels, by reversal
and other alterations of the normally
light-dark-light pattern when scanning
the temporal peripapillary area from
top to bottom (Figs 1 to 5), and by the
clarity with which vessel walls could be
distinguished (Figs 3 to 5). Compari
sons between corresponding areas of
the two hemispheres, or of the two
eyes, often revealed subtle damage.
In some eyes, localized defects easily
recognized by the sharp transition to
adjacent, more normal retina were su
perimposed on diffuse damage. These
wedge-shaped focal defects were often
more readily apparent (but less func
tionally significant) than more exten
sive diffuse atrophy (Fig 3). Presum
ably this relates to the relative ease of
pattern recognition of sharp changes in
NFL thickness (brightness) at the bor
ders of a wedge defect.2021 In monkeys,
50% loss of local NFL thickness is
readily recognized when it is adjacent

to normal

retina.23

Rare cases began with seemingly

focal loss, often adjacent to a notched
rim. With time, focal defects tended to
deepen and expand by merging with
newly thinned, adjacent areas of atro
phy (Fig 1). In a few cases, adjacent
focal atrophy was preceded or accom
panied by a disc hemorrhage (Fig 4).
Given the infrequent (annual) exami
nation, we may well have missed tran
sient disc hemorrhages in other cases.31
Occasionally, a focal wedge defect was
not accompanied by functional distur
bances or further damage, but this was
uncommon.

Diffuse

atrophy would often deepen

time, with further loss of stria
tions, darkening of the fundus, increas
ing definition and clarity of secondorder vessels and their walls, and
reversal of the normal temporal retinal
pattern. Instead of being the darkest
zone, the normally thin maculopapular
over

bundle

became as bright as, or

brighter than, the adjacent retina
(Figs 1 and 2). Ultimately, little if any

NFL persists, such that asymmetry

between superior and inferior hemi
spheres disappears. One is left with a
dark, granular-appearing fundus some
times displaying very fine, widely
spaced, faint striations (Figs 3 and 5)
and bright-walled vessels that stand
out in sharp relief, devoid of overlying
striations (Figs 1 and 5). Eventually,
all striations are lost.

The evolution of NFL loss is best

before the development of
substantial field loss. Once this occurs,
the NFL has practically disappeared,
at least to our recording techniques,
making current methods of NFL as
sessment most suitable for monitoring
early loss and progression.
A variety of image enhancement
techniques that alter the recorded im
age have been applied to improving
visibility of the NFL and its abnormal
ities. None has been particularly suc
cessful at improving recognition of de
fects and distinguishing normal from
abnormal NFL. Quantifying the actual
thickness of the NFL2' (or, as a surro
gate, the neurosensory retina in its
entirety) offers greater promise.28 In
the meanwhile, the clinician is best
advised to learn to recognize normal
NFL, first in healthy young eyes and
later in older individuals. Only then
should one attempt to recognize abnor
malities corresponding to glaucoma
tous damage. Red-free photographs
are extremely helpful when first learn
ing to evaluate the NFL and identify
abnormalities. They remain conve
nient, but less essential, once direct
visualization with a bright, direct oph
thalmoscope or slit lamp and contact
lens has been mastered.
We do not claim our patients were
entirely free of psychophysical evi
dence of optic nerve dysfunction before
meeting our criteria for visual field

appreciated

loss. No doubt other tests and criteria

would have suggested abnormalities
earlier in the course of disease, but
with less reliability, consistency, and
specificity than this study's long-stand

ing criteria.29,30 Indeed, referring phy

sicians had already elected to give a
substantial number of subjects thera
py, because of the height of the IOP,
size of the cup, presence of NFL ab

normalities, or the appearance of mi

nor but potentially valid defects in
threshold perimetry. If effective in
slowing progression of optic nerve

damage, such therapy would have de

layed the onset of field loss and pro
longed the apparent duration between

the onset of NFL defects and subse

quent field conversion.
The presence of NFL defects should
alert the physician to the likelihood of
past optic nerve damage and the need
for careful, detailed examination and
follow-up. A normal-looking disc and
visual field offer only partial reassur
ance and should be scrutinized regular
ly until one is certain that the patient
is not suffering progressive damage.
In other settings, an abnormal NFL
may be sufficient evidence for initiat

ing glaucoma therapy.

This study was supported by grants EY03605

and RR04060 from the National Institutes of
Health, Washington, DC.
We wish to thank Donna Gilbert, Rachel Scott,
Patty Lenane, and Felecia Keel, and all those who
referred patients to this study.

References
1. Sommer A. Intraocular pressure and glaucoAm J Ophthalmol. 1989;107:186-188.
2. Anderson DR. Glaucoma: the damage caused
by pressure: XLVI Edward Jackson Memorial Lecture. Am J Ophthalmol. 1989;108:485-495.
3. Van Buskirk EM. Glaucoma lexicon. Am J
ma.

Ophthalmol. 1989;108:730-731.

4. Quigley HA, Dunkelberger GR, Green WR.

Studies of retinal ganglion cell atrophy correlated
with automated perimetry in human eyes with
glaucoma. Am J Ophthalmol. 1989;107:453-464.
5. Hoyt WF, Newman NM. The earliest observable defect in glaucoma? Lancet. 1972;1:692-693.
6. Hoyt WF, Frisen L, Newman NM. Funduscopy of nerve fiber layer defects in glaucoma. Invest Ophthalmol Vis Sci. 1973;12:814-829.
7. Sommer A, Miller NR, Pollack I, Maumenee
AE, George T. The nerve fiber layer in the diagnosis of glaucoma. Arch Ophthalmol. 1977;95:2149\x=req-\
2156.
8. Iwata K, Naba K, Abe H. Typical slit-like
retinal nerve fiber layer defect and corresponding
scotoma. Acta Soc Ophthalmol Jpn. 1981;85:1791\x=req-\
1803.
9. Airaksinen PJ, Drance SM, Douglas GR,
Mawson DK, Nieminen H. Diffuse and localized
nerve fiber loss in glaucoma. Am J Ophthalmol.

1984;98:566-571.

10. Airaksinen PJ, Drance SM, Douglas GR,

Schulzer M, Wijsman K. Visual field and retinal
nerve fiber layer comparisons in glaucoma. Arch

Ophthalmol. 1985;103:205-207.

11. Sommer A, Quigley HA, Robin AL, Miller

NR, Katz J, Arkell S. Evaluation of nerve fiber

layer

assessment. Arch

1766-1771.

Ophthalmol. 1984;102:

12. Armaly MF, Krueger DE, Maunder L, et al.

Biostatistical analysis of the collaborative glaucoma study, I: summary report of the risk factors for
glaucomatous visual-field defects. Arch Ophthalmol. 1980;98:2163-2171.
13. Tielsch JM, Sommer A, Witt K, Katz J, Royall RM. Blindness and visual impairment in an
American urban population: the Baltimore Eye
Survey. Arch Ophthalmol. 1990;108:286-290.
14. Canner JK, Sommer A, Katz J, Enger C.
ICEPAC: a user-friendly software package for processing Humphrey Field Analyzer diskettes.
Invest Ophthalmol Vis Sci. 1988;29(suppl):356.
Abstract.
15. Sommer A, Enger C, Witt K. Screening for
glaucomatous field loss with automated threshold
perimetry. Am J Ophthalmol. 1987;103:681-684.
16. Enger CL, Sommer A. Recognizing glaucomatous field loss with the Humphrey STATPAC.
Arch Ophthalmol.
17. STATPAC User's Guide. San Leandro,
Calif: Allergan Humphrey; 1986.
18. Katz J, Sommer A. Reliability indexes of
automated perimetric tests. Arch Ophthalmol.

1987;105:1355-1357.

1988;106:1252-1254.
19. Miller NR, George TW. Monochromatic
(red-free) photography and ophthalmoscopy of the

peripapillary retinal nerve fiber layer. Invest

Ophthalmol Vis Sci. 1978;17:1121-1124.
20. Sommer A, D'Anna SA, Kues HA, George T.
High-resolution photography of the retinal nerve
fiber layer. Am J Ophthalmol. 1983;96:535-539.
21. Sommer A. Nerve fiber layer in glaucoma.

Ophthalmology. 1989;96:639-641.

22. Quigley HA, Sommer A. How to use nerve

fiber layer examination in the management of glau-

Downloaded From: http://archopht.jamanetwork.com/ by a UQ Library User on 10/24/2014

Trans Am Ophthalmol Soc. 1987;85:254-272.

Quigley HA. Quantitative studies of retinal
nerve fiber layer loss in monkey and human glaucoma. Trans Am Ophthalmol Soc. 1986;84:920-966.
24. Quigley HA, Addicks EM. Quantitative
studies of retinal nerve fiber layer defects. Arch
Ophthalmol. 1982;100:807-814.
25. Sommer A, Miller NR, Quigley H, Robin A,
coma.

23.

Katz J, Burkhard K. Assessment of the nerve fiber

as a predictor of glaucoma. Invest Ophthalmol Vis Sci. 1985;26(suppl):122. Abstract.
26. Airaksinen PJ, Mustonen E, Alanko HI. Optic disc haemorrhages precede retinal nerve fibre
layer defects in ocular hypertension. Acta Ophthalmol. 1981;51:627-641.
27. Weinreb RN, Dreher AW, Coleman A, Quigley H, Sharo B, Reiter K. Histopathologic validation of Fourier-Ellipsometry measurements of retinal nerve fiber layer thickness. Arch Ophthalmol.

layer

1990;108:557-560.
28. Caprioli J, Miller JM. Measurement of relative nerve fiber layer surface height in glaucoma.
Ophthalmology. 1989;96:633-639.
29. Enger C, Sommer A. Automated threshold
perimetry before glaucomatous field loss in eyes

with ocular hypertension. Invest Ophthalmol Vis

Sci. 1990;31(suppl):432. Abstract.
30. Katz J, Sommer A. Similarities between the
visual fields of ocular hypertensive and normal
eyes. Arch Ophthalmol. 1986;104:1648-1651.
31. Diehl DLC, Quigley HA, Miller NR, Sommer A, Burney EN. Prevalence and significance of
optic disc hemorrhage in a longitudinal study of
glaucoma. Arch Ophthalmol. 1990;108:545-550.