DIAGNOSTIC EDITION
Documenting
Progression on
Humphrey Visual Fields
Using the Glaucoma Progression Analysis software in clinical practice.
BY PAMELA A. SAMPLE, P H D
T
he Humphrey Visual Field Analyzer (HFA; Carl
Zeiss Meditec, Inc., Dublin, CA) can be equipped
with Glaucoma Progression Analysis (GPA) software, which assists clinicians in their evaluation of
serial visual fields by identifying areas where progression
may be occurring. Currently, this software is available only
for standard (white-on-white) visual fields, programs 24-2
or 30-2. It can be used with Swedish Interactive Threshold
Algorithm (SITA)-Standard1 and SITA-Fast2 testing strategies (both from Carl Zeiss Meditec, Inc.) for baseline and
follow-up examinations. It can also use full-threshold tests
as baselines, thereby facilitating a switch from one threshold strategy (full) to another (SITA-Standard) in the same
patient. SITA-Fast has not been evaluated in a clinical trial
setting for following glaucoma, and this author does not
recommend its use for progression analysis.
This article focuses on the clinical application of the
GPA software.
THE R ATI ONALE F OR THE GPA SOF TWARE
GPA is similar to the algorithm employed in the Early
Manifest Glaucoma Trial (EMGT),3 a multicenter study that
evaluated the effect of treatment versus no treatment on
eyes with early glaucoma. GPA differs from a previous HFA
algorithm, the Glaucoma Change Probability analysis. The
former uses the pattern deviation plot values rather than
the total deviation values used by the latter. The rationale is
that change in the total deviation values over time may be
due to factors other than glaucoma such as advancing
cataract or decreased pupillary size.4 By using the pattern
deviation values, the GPA software specifically targets the
localized change associated with glaucoma. If there were a
diffuse component to the glaucomatous change, it would
not be reflected in the GPA result, but the more likely localized component due to the formation of new glaucoma-
tous defects or the expansion and deepening of existing
defects would be characterized.
HOW D OE S THE GPA SOF TWARE WORK?
The GPA software compares a patient’s baseline visual
fields (two are needed and are averaged) to each subsequent visual field in a series. Every test location in each follow-up field is evaluated relative to the baseline. Change is
flagged if it is greater (at the 95% significance level) than the
variability seen in a large group of stable glaucoma patients
who were tested over a very short period of time. In addition, the GPA accounts for the known increase in variability
found with advancing disease and increased eccentricity.
Data from the EMGT and other clinical studies,3,5,6
using an algorithm similar to GPA, have suggested that
change needs to be present in three consecutive visual
fields before progression can be confirmed. When change
is first noted on a field at a given location, the GPA software will automatically assess the next visual field (and
the next) to determine if that change is repeatable.
The EMGT also required change to be repeatable in the
same three or more points. Two or three scattered locations
with change on a single visual field are not uncommon in
stable patients. The GPA software also determines if the
same three or more points are flagged on two or three fields.
RUNNING THE GPA SOF TWARE
The GPA software automatically selects the patient’s first
two tests (either full-threshold or SITA) as the baseline, and
then it evaluates each subsequent test (SITA only) relative
to those two baselines. Although the algorithm will automatically deselect baseline tests showing too many falsepositive responses, it is important that the clinician evaluate the two tests chosen to rule out excessive losses of fixation or false negatives, inattention, fatigue, or an obvious
NOVEMBER/DECEMBER 2007 I GLAUCOMA TODAY I 33
DIAGNOSTIC EDITION
A
A
B
B
Figure 1. This patient exhibits early visual field loss at baseline.
The grayscale,numerical threshold values in decibels and the
total and pattern deviation probability plots are shown for the
two baseline SITA-Standard visual fields.Also given for these
fields are the glaucoma hemifield test (GHT),the mean deviation
(MD),and pattern standard deviation (PSD) results,along with
the percentage of fixation losses,false negatives,and false positives.The last three values indicate the two baseline fields are reliable.A key for the probability symbols is given in the lower left
corner of the figure (A).The slope of the MD for the two baseline
fields averaged along with the MD for all analyzed follow-up visits is shown,and the probability that the MD slope is significant is
given.In this case,the probability is less than 1% (B).
learning effect in patients new to visual field testing. Large
differences in the pattern and the location of field loss seen
in the two baselines may indicate that one of them is not a
suitable choice. It is possible to manually override the GPA
software in these instances and specify which two tests the
clinician would like to set for the baseline. The program will
then remember the chosen baselines and automatically
add each follow-up examination to the analysis.
There are other instances when new baseline tests
should be chosen. If progression has occurred and there
is a resultant change in therapy, the clinician should
establish a new baseline so that any additional progres34 I GLAUCOMA TODAY I NOVEMBER/DECEMBER 2007
Figure 2. This patient exhibits advanced visual field loss at
baseline. Many locations are too advanced to measure for
progression, as indicated by an X. The same information as
detailed in Figure 1 is given here (A). The slope of the MD for
the two baseline fields averaged along with the MD for all
analyzed follow-up visits is shown, and the probability that
the MD slope is significant is given. In this case, the probability is less than 1%. The printout also warns that the MD value
eventually falls out of range for follow-up (B).
sion can be found. If the patient undergoes ocular surgery or develops another ocular condition, new baseline
tests after he stabilizes should be selected for use in evaluating subsequent examinations.
It is important that technicians always enter the
patient’s name and/or identification numbers in exactly
the same way at each visit so that all of his tests will be
available for automatic selection by the GPA software.
INTERPRETING THE GPA SOF TWARE
Overview
Figures 1 and 2 show the printouts giving the baseline
results from two sample patients, one with early and one with
advanced field loss. The printouts give much of the same
information found on the single field printout and allow clinicians to assess whether the chosen baselines are appropriate.
DIAGNOSTIC EDITION
A
A
B
B
C
C
Figure 3. Three consecutive follow-up fields are provided for the
baseline case example shown in Figure 1.All of the fields show
the same single field information as given in the baseline field
example.This field shows nine locations that deteriorate from
baseline for the first time (A),one that has deteriorated for the
second time in a row (B),and one for the third time (C).Some
prior fields are not shown.This field does not yet meet the GPA
software’s criteria for progression (A).Eight points continue to
show deterioration on both fields A and B.This result meets the
criteria for “possible progression,”and this statement appears
just below the false-positive percentage for this field (B).Six
points show deterioration on all three fields (A to C).This result
meets the GPA software’s criteria for “likely progression”(C).
Figure 4. Three consecutive follow-up fields are provided for
the baseline case example shown in Figure 2. Although this
field shows numerous locations that cannot be assessed for
progression designated by an X, analysis is still possible in
the lower hemifield. Five locations deteriorate from baseline
for the first time (A). These same five points continue to show
deterioration on field B. This result meets the criteria for “possible progression” (B). Four points show deterioration on all
three fields (A to C). This result meets the GPA software’s criteria for “likely progression” (C).
The grayscale, absolute threshold values, total and pattern
deviation probability plots, and mean deviation and pattern
standard deviation global indices are provided. Losses of fixation, false negatives, and false positives are also shown so that
clinicians can assess the reliability of the two fields.
At the bottom of the printouts appears a mean deviation plot for all of the examinations in the series. This
section gives the slope associated with change in the
mean deviation. It is important to remember that this
change will include anything that affects the subject’s
visual sensitivity, including advancing cataract, and that it
may not reflect change due to glaucoma.
Figures 3 and 4 show the GPA printouts for follow-up of
the aforementioned sample patients. When change is first
noted on a follow-up field at a given location, an open triangle appears at that location on the printout. The GPA
software will automatically assess the next visual field (and
the next) to determine if that change is repeatable. If it is
present on two consecutive tests, a half-filled triangle will
appear at the location. For repeatable change on three consecutive tests, a closed triangle will appear.
The GPA software then assesses the repeatability of
three or more points and gives a plain-language report of
“possible progression” if two consecutive fields show that
the same three or more points changed from baseline or
“likely progression” if three consecutive fields show
change at the same three or more points.
NOVEMBER/DECEMBER 2007 I GLAUCOMA TODAY I 35
DIAGNOSTIC EDITION
Is It Really Progression?
The GPA software is designed to assist clinicians in
looking for progression based on visual field results.
Although it goes a long way toward verifying change,
physicians should use the system’s findings in conjunction with other clinical information to determine if a
report of “likely progression” is reasonable and not due to
factors other than glaucoma. Again, it is important that
all of the fields in the series be reliable and that the
appropriate baselines be chosen.
THE FUTURE
The GPA software provides a type of events analysis of
progression. It does not directly provide a rate of change due
to glaucoma. Subtle, slow progression that finally exceeds the
probability limit for change is not easily differentiated from a
larger change reaching the same limit more quickly. An estimate of the rate of progression would be helpful in determining how aggressive treatment should be for a given individual. More sophisticated algorithms are under evaluation
in research settings. For example, one such algorithm-using
machine that is learning classifiers can assess the progression
of specific patterns of glaucomatous defects, provide a verification of change, and estimate the rate of change.7 These
algorithms are not yet available for clinical use.
At present, the GPA software represents a significant
improvement over previous HFA progression algorithms.
It provides clinicians with information they need to identify and verify change in serial visual fields that is likely
due to progressing glaucoma. ❏
Subscribe to
Glaucoma Today’s e-News
A biweekly newsletter delivered directly to your
e-mailbox contains news briefs and breaking news
releases to keep you up to date between our print
issues. Subscribing is easy and free. Simply email us
at [email protected], type “Subscribe e-News”
in the subject line, and include your name. You can
unsubscribe at any time by clicking on the
“unsubscribe” link in the e-Newsletter.
We look forward to hearing from you!
Pamela A. Sample, PhD, is Professor and Director of Clinical Vision Research at the Hamilton
Glaucoma Center, Department of Ophthalmology, University of California, San Diego. She
has received research support in the form of test
equipment from Carl Zeiss Meditec, Inc., Welch Allyn
Medical Products, and Haag-Streit AG. Dr. Sample may be
reached at (858) 534-6629; [email protected].
1. Bengtsson B, Olsson J, Heijl A, Rootzen H. A new generation of algorithms for computerized threshold perimetry, SITA. Acta Ophthalmol Scand. 1997;75:368-375.
2. Bengtsson B, Heijl A. SITA Fast, a new rapid perimetric threshold test. Description of
methods and evaluation in patients with manifest and suspect glaucoma. Acta Ophthalmol
Scand. 1998;76:431-437.
3. Heijl A, Leske MC, Bengtsson B, et al. Measuring visual field progression in the Early
Manifest Glaucoma Trial. Acta Ophthalmol Scand. 2003;81:286-293.
4. Bengtsson B, Lindgren A, Heijl A, et al. Perimetric probability maps to separate change
caused by glaucoma from that caused by cataract. Acta Ophthalmol Scand. 1997;75:184-188.
5. AGIS Investigators. The Advanced Glaucoma Intervention Study (AGIS): 7. The relationship between control of intraocular pressure and visual field deterioration. Am J Ophthalmol.
2000;130:490-491.
6. Kass MA, Heuer DK, Higginbotham EJ, et al. The Ocular Hypertension Treatment Study: a
randomized trial determines that topical ocular hypotensive medication delays or prevents the
onset of primary open angle glaucoma. Arch Ophthalmol. 2002;120:701-713.
7. Sample PA, Boden C, Zhang Z, et al. Unsupervised machine learning with independent
component analysis to identify areas of progression in glaucomatous visual fields. Invest
Ophthalmol Vis Sci. 2005;46:3684-3692.