Visual field screening and the
development of a new screening
program
David B. Henson, Ph.D.
Abstract: The diagnosis of
chronic open-angle glaucoma
requires diligent and careful
screening. Intraocular pressure
and disc evaluation, when used
in isolation, produce an
unacceptably high number of
false positives and false negatives,
while full visual field analysis, as
conducted with kinetic or full
threshold static perimetry, takes
too long for routine use.
Screening of the visual field
may offer a viable I
alternative. The effect of :
changing certain parameters of m
visual field screening test are
-evaluated. These include the
(location of the stimuli, the
number of stimuli and their
suprathreshold increment. The
results of an optimized visual
field screening test are then
evaluated with respect to the
theoretical predictions. The
results indicate that high levels
of sensitivity and specificity
can be achieved with visual
field screening tests which are
enough to be included as of the
routine optometric examination.
words: visual fields,
^Screening, glaucoma
The increasing demands that have
been placed upon the optometric
profession to detect and diagnose
disease have encouraged practitioners
to perform more and more diagnostic procedures as part of their
routine eye examination. One such
procedure is a visual field examina-
tion. While a very good case can be
made for a full field examination to
be performed on every patient, the
time taken to perform this type of
test is so great that most practitioners
view this option as unrealistic.
One alternative to performing
a full visual field examination is to
screen part of the visual field with a
test that only takes a few minutes to
conduct, and then to perform a full
examination on those patients who
fail the screening test. A second alternative is to apply preconditions
such as raised intraocular pressure
(IOP), abnormal disc appearance or a
family history of glaucoma before
ordering a full field examination.
At present it is impossible to
say which is the best of these two
approaches to adopt, as we lack the
precise data bases necessary to calculate the relative merits of different
screening programs vs. different sets
of preconditions.
Within the literature, however,
there are several reports discussing
the parameters, such as number and
position of stimuli, which affect the
sensitivity and specificity of visual
field screening programs. This paper
reviews the data from a number of
these studies and then briefly describes a new piece of visual field
equipment with a screening program based on the results of these
research programs.
Nearly all the presented data
relates to chronic open-angle glaucoma. The reasons for this are twofold. First, glaucoma exists in a
fairly large percentage of patients
(approximately 0.5% of the population), and it is therefore feasible to
build up sufficiently large data bases
to enable certain hypotheses to be
tested. To build up data bases for
other, more rare conditions, would
require enormous research pro-
grams that, at present, have not
been conducted. The second reason
is that chronic open-angle glaucoma
is often asymptomatic. Its detection
requires diligent and careful screening. Many of the other conditions
which give rise to visual field loss
also give rise to symptoms which, in
themselves, would indicate the need
for special diagnostic procedures
and further evaluation.
Sensitivity and specificity of
IOP and optic disc evaluation
Optic disc evaluation, IOP measurement, and visual field assessment
still currently form the basis of glaucoma diagnosis. Before describing
how sensitive and specific different
visual field screening strategies are,
this review will briefly discuss IOP
and optic disc evaluation. With this
information at hand, the clinician
will be able to make a personal judgment as to the relative merits of the
three types of test.
There exists a considerable
amount of data on the sensitivity
and specificity of IOP measurement. Most of the epidemiological
studies reveal1'4 that approximately
50% of the patients with glaucoma
(usually defined as a repeatable visual
field defect allied with a cupped
disc) will, at the time of initial examination, have an IOP of less than
21 mmHg. Figure 1 gives the results
from four of these studies. It is important to emphasize that the results
given in Figure 1 are from unselected population groups. Studies
that look at selected populations,
such as those attending a glaucoma
clinic, are often biased in favor of
IOP measurements since many
practitioners use an elevated IOP as a
criterion for referral.
Volume 60, Number 12,12/89
893
Bengtsson, 1981
12 c a s e s
Cockburn, 1985
27 cas es
Framingham, 1980
71 cases
Armaly, 1980
98 cases
Figure 1: Results from four epidemiologies! studies giving the number of glaucoma cases in each study and the
proportion of glaucoma cases which, at the time of examination, had low IOP.1"343
The specificity of IOP measurements is also very low. The Framingham Eye Study3 found that 7.6%
of the population had pressures >22
mmHg, and that the vast majority
of these patients were nonglaucomatous. Put another way, if a practitioner relied entirely upon IOP
measurement as a criterion for referral, and referred all patients with
an IOP >22 mmHg, he or she would
refer 14 normals for every patient
with glaucoma and would miss 50%
of the glaucoma cases.
Data from optic disc evaluation
trials again reveal that the sensitivity
and specificity of disc evaluation is
less than perfect.5 While studies
have shown that disc changes often
precede field defects, the types of
change that occur are usually only
detected when serial photographs,
taken over a period of several years,
are analyzed.6-7 When Lichter8 presented highly skilled practitioners
with a set of photographs and asked
894
them to differentiate glaucomatous
eyes from normal eyes, he found
that 20% of them were misclassified.
Optimized visual field
screening tests
When designing an optimized visual
field screening test there are several
things that need to be considered.
Type of visual field examination
Static, suprathreshold techniques
have repeatedly been advocated for
visual field screening.9'12 They combine
a high sensitivity with a short
examination time. Kinetic tests,
such as those conducted with the
Goldmann perimeter, while being
valuable for monitoring the extent
of a defect, have a low sensitivity to
defects.13 The quantitative threshold
tests, popularized with the Octopus
and Humphrey instruments, while
furnishing more information
Journal of the American Optometric Association
concerning the depth of any defect,
take longer to perform and do not
appear to have any greater sensitivity.9 Some of the early suprathreshold
static instruments used stimuli of
constant intensity across the whole
visual field.14 It is now realized that
a reduction in errors will occur if
the intensity increases toward the
periphery at a rate which mirrors
the retina's loss in sensitivity. In this
situation the stimuli will be at a
constant suprathreshold value at all
retinal locations.
Multiple stimulus vs. single
stimulus
In the multiple stimulus technique,
which was pioneered by Harrington
and Flocks when they developed the
Harrington-Flocks Screener," patients are presented with patterns of
two, three or four stimuli. They respond verbally, stating the number
of stimuli that they see in each pres-
entation. This technique has several
advantages over the single stimulus
technique: it speeds up the examination time, it maintains the patient's
attention somewhat better, and by
being semi-automated,16 it promotes
a dialogue between the patient and
the perimetrist which has certan
psychological advantages.17 It has
also been reported as giving a lower
threshold than that of a single
stimulus technique with a smaller
standard deviation of results.18
negatives — would be at a minimum. As can be seen from Figure
4, this would occur with a test which
had approximately 20 stimuli. Thus,
the ideal number of stimuli to
have in a screening test, given that
the clinician equally weighs false
positives and false negatives, would
be 20. Many clinicians would argue,
and rightfully so, that false
positives (a normal person failing
the test) and false negatives (a glaucomatous person passing the test)
are not equal. They would be willing
to have a higher number of false
positives, providing there were fewer
false negatives. This situation can be
accommodated, providing the
clinician specifies the relative
weight given to false positives and
25
Location of stimuli
To decide where to optimally place
the stimuli in a visual field screening
test for glaucoma, it is essential to
know where glaucoma defects normally occur and whether there is
any relationship between the size of
the defects and retinal location. This
information was provided in an earlier
paper19 which examined the results
from 104 eyes with early glaucomatous visual field loss. Some of
the results from that paper are presented in Figure 2, which gives the
frequency of defects within the central
30° of the visual field. This data can
be used to derive the relationship
between the number of stimuli in the
visual field screening test and the
sensitivity of that test. Each distribution of stimuli is optimized to
the location of glaucomatous defects.20 The results of such an evaluation are given in Figure 3, where
it can be seen that the sensitivity of
the test is logarithmically related to
the number of stimuli.
Figure 2: Frequency distribution of early glaucomatous defects in 109
eyes. The highest peak, in the superior temporal region, represents 29
eyes. Data collected with Friedmann Mark II Visual Field Analyzer.19
Number of stimuli
The data presented in Figure 3 can
be used, along with data giving the
specificity of a visual field test,21 to
derive the receiver-operator characteristics of a visual field screening
test. The receiver-operator characteristics are shown in Figure 4 for a
test in which missing one nonblind
spot stimulus constitutes a failure.
A line drawn at 45° to the axes and
tangentially to the curve touches the
curve at the point where the number
of errors — false positives and false
Figure 3: Relationship between the sensitivity of a visual field screening
test and the number of test stimuli. The failing criteria was set at one or
more missed stimuli at a suprathreshold intensity 0.5 log units above the
patient's threshold. Data was collected with a Friedmann Mark II Visual
Field Analyzer.20
Volume 60, Number 12,12/89
895
false negatives. For example, if the
clinician weighted a false positive as
being half as important as a false
negative, then rather than draw the
line at 45° to the x axis, it should be
drawn at 30°, where the ratio of the
angle between the axes equals the
relative weight given to the different
types of error. As can be seen in
Figure 4, the ideal number of stimuli
to incorporate in the screening test
has now increased to 30.
Suprathreshold increment
The extent of the suprathreshold increment will affect the overall performance of a threshold-related, suprathreshold test, one in which an
estimate of the threshold is obtained
at the beginning of the test before
the intensity is stepped up to its
suprathreshold value. If a suprathreshold increment is chosen that
is only marginally above the patient's threshold, then there would
be a large number of false positives,
as it is highly likely that the patient
will miss, by chance, one or more of
the stimuli. If the suprathreshold increment was made very large, then
the number of false negatives would
become high as patients with relative defects may, as a result of the
high intensity stimuli, see all the
stimuli and pass the screening test.
The relationship between suprathreshold increment and the
number of errors — false positives
and false negatives — can be quantified with a mathematical technique called information theory.
The results from such an analysis
indicate that the total amount of
information in a test increases with
the suprathreshold increment up to
0.6 log units, and then decreases
beyond this level. These results indicate that suprathreshold increments of approximately 0.6 log
units would be the best to choose
for an optimized screening strategy.
Quantifying visual field
defects
An aspect of visual field testing that
Figure 4: Effect upon sensitivity and specificity of varying the number of
stimuli in a visual field test. The numbers placed along the solid curve give
the number of stimuli in a visual field test, all optimally located to detect
the largest number of defects. As the number of stimuli increases,
the sensitivity of the test increases, while the specificity, probability
of a normal patient missing one or more of the stimuli, decreases. A:
The broken line, drawn at 45° to the abscissa, touches the curve at the
point where the number of stimuli in the screening test would give a
minimum number of errors, false positives and false negatives. B:
The broken line, drawn at 30° to the abscissa, touches the curve at the
point where the number of stimuli in the screening test would give twice
as many false positives as false negatives.30
896
Journal of the American Optometric Association
has received a great deal of interest
in the past few years is that of quantifying the results of an examination. There are two basic types of
quantification: the first gives you an
indication of whether a defect exists,
while the second gives you a measure of the extent of any defect. In
visual field screening tests we are
primarily concerned with the former, i.e., whether a defect exists.
The latter form of quantification is
useful once a defect has been established to ascertain whether it has
progressed or not and whether any
treatment has been successful.
Quantification systems are, in
effect, a form of data reduction.
They take the large amount of data
that is produced in a visual field
examination and reduce it to either
one or a series of different numbers.
When a single number is produced
the result is often called the visual
field score. But regardless of how
you score the visual field — be it
simply a count of the number of
missed stimuli or the result of a
more complex calculation — you
will always encounter a certain degree of variability. There is, therefore, no error-free cut-off value of a
visual field score that can be used to
classify patients as either defective
or normal. In deciding whether a
patient does or does not have a visual
field defect most researchers have
chosen to compare the patient's
visual field score with that obtained
from a normal population. They then
provide the practitioner with an
indication of the difference between
a given patient's score and that of the
normal population.
Over the years there have been
a large number of papers describing
different scoring techniques. Some
have been developed for kinetic results such as those obtained with the
Goldmann bowl perimeter.22'31 Other
techniques have been developed to
deal with the results from static
techniques in which the threshold is
measured
at
each
stimulus
location.32"35 Both of these two groups
of quantification systems have
primarily been designed to give
a measure of the extent of a defect
for either litigation purposes or for
monitoring the progress of any field
defects. They cannot be applied to
suprathreshold techniques, which, as
pointed out earlier, are the preferred
ones for visual field screening. Crick36
and Demailly and Papos37 both
described scoring systems for the
multiple stimulus suprathreshold
technique of the Mark
I Friedmann Visual Field Analyzer,
while Batko et al38 and Henson and
Dix39 have developed systems for
the Mark II instrument. The Mark
II Friedmann Visual Field Analyzer
differs from the Mark I in that it
presents more stimuli (98 rather than
46) and has a more convenient
technique for recording the data. The
technique devised by Henson and
Dix39 for the Friedmann was later
modified and incorporated into the
Henson CFS2000/ CFA3000.40'41
A characteristic of early glaucomatous visual field loss is that the
missed stimuli in a static test tend to
be clustered together. A patient who
misses four neighboring stimuli is
more likely to have a scotoma than
one who misses four stimuli that are
unconnected
and
scattered
throughout the visual field. It is important that any quantification system which is designed to detect early
defects can differentiate between
these two situations.
The value of cluster analysis has
been demonstrated by both Chauhan
and Henson21 and Henson and Dix,39
and it is expected that the next
generation of computerized visual
field equipment will incorporate this
type of analysis in its screening
programs. The number and type of
calculations necessary to perform a
cluster analysis cannot realistically
be performed without the aid of a
computer.
Clinical trials with a new visual field
instrument
Much of the data obtained from the
research reviewed in this paper has
been used to design an optimized
visual field screening program which
has been incorporated in a new
visual field instrument, the Henson
CFS2000/CFA3000.
This
instrument's screening program uses
multiple stimulus patterns in a
threshold-related
suprathreshold
strategy, which is eccentrically compensated to the normal gradient of
sensitivity across the retina. This
program is divided into three stages: a
screening stage in which just 26
stimuli are presented in locations
optimized for the detection of glaucoma; an extended stage which increases the number of stimuli to 66;
and a quantification stage which increases it further to 132 stimuli. The
CFA3000 also includes a program
that uses a full threshold technique
with allied quantification techniques.
The screening stage is designed
for routine use in patients with no
suspicion of visual field loss. This
stage normally takes less than 3
minutes to perform on both eyes.
The extended stage is designed for
use on patients with some suspicion,
such as raised IOP or a family history of glaucoma, and for those patients who have missed one or more
of the screening stimuli. The quantification stage, as the name implies,
is for use on patients in whom a
defect has been found. It is used to
map the extent of the defect and to
give a numerical score that represents the extent of visual field loss.
Clinical
trials
with
the
CFS2000/CFA3000 has shown that
the screening stage has a sensitivity
to glaucomatous visual field defects
of 90% and a specificity of 88%.4M2
The extended stage, which would be
performed on all patients who failed
the screening test, has a specificity
of 100%.
Conclusion
The diagnosis of glaucoma is currently based upon three clinical
measures: an evaluation of the optic
disc, a measurement of the IOP, and
an investigation of the visual field.
While many optometrists routinely
examine the optic disc and measure
lOPs, relatively few routinely investigate the visual field.
The data reviewed in this paper
have shown that high sensitivity and
specificity levels can be reached
when visual field screening pro
grams are optimized to the detec
tion of glaucomatous damage. Op
tometrists frequently state that rou
tine investigation of the visual field
is not time-efficient. This statement
is often based on experiences gained
with instruments and programs that
were designed to quantify loss rather
than to screen the visual field. It is
hoped that the data presented in this
paper will indicate that visual field
testing has come a long way in the
past few years and that routine vis
ual field screening is now both fea
sible and desirable in optometric
practice.
••
Submitted for publication 11/88
Accepted 9/89
Dept. of Optometry
University of Wales
PO Box 905
Cardiff CF1 3YJ
Wales
Acknowledgment
The author has a proprietary interest in the
Henson CFS2000 and Henson CFA3000 visual
field analyzers.
References
1. Bengtsson B. The prevalence of glaucoma. Br J Ophthalmol 1981; 65:46-9.
2. Armaly MF. Lessons to be learned from
the collaborative glaucoma study. Surv
Ophthalmol 1980; 25:139-44.
3. Leibowitz HM, Kreuger DE, Maunder
LR, et al. The Framingham eye study
monograph. Surv Ophthalmol 1980;
24(suppl):335-610.
4. Cockburn DM. Glaucoma enigma. Am J
Optom Physiol Opt 1985; 62:913-23.
5. Gloster J. Quantitative relationship between cupping of the optic disc and visual
field loss in chronic simple glaucoma. Br J
Ophthalmol 1978; 62:665-9.
6. Pederson JE, Anderson DR. The mode
of progressive disc cupping in ocular hypertension and glaucoma. Arch Ophthalmol 1980; 98:490-5.
7. Sommer A, Pollack I, Maumenee AE.
Optic disc parameters and the onset of
glaucomatous field loss. Arch Ophthalmol 1979; 97:1444-8.
Volume 60, Number 12,12/89
897