The Advanced Glaucoma Intervention
Study (AGIS)
13. Comparison of Treatment Outcomes within Race:
10-Year Results
The AGIS Investigators*
Objective: To present for black and white patients with medically uncontrolled glaucoma 10-year results of
treatment with 1 of 2 randomly assigned surgical intervention sequences.
Design: Randomized clinical trial.
Participants: Three hundred thirty-two black patients (451 eyes) and 249 white patients (325 eyes). Eyes had
glaucoma that could not be controlled with medications alone.
Methods: Eyes were randomly assigned to 1 of 2 sequences: argon laser trabeculoplasty (ALT)–trabeculectomy–trabeculectomy (ATT) or trabeculectomy–ALT–trabeculectomy (TAT). Second and third interventions
were offered after failure of the preceding intervention. Minimum required intraocular pressure (IOP) for intervention failure ranged upward from 18 mmHg, the value depending on whether recent optic disc or visual field
(VF) deterioration occurred, and on the magnitude of the field defect. Patients were observed every 6 months,
with total potential follow-up ranging from 8 years, 4 months to 13 years.
Main Outcome Measures: The averages over follow-up of (1) the percentage of eyes having moderate loss
of VF and (2) the percentage of eyes having moderate loss of visual acuity (VA).
Results: Race–treatment interactions in VF and VA loss are significant for the 2 main outcome measures;
therefore, results of treatment sequence differences are presented by race. In black patients the average percent
of eyes with VF loss was less in the ATT sequence than in the TAT sequence, a difference that is not statistically
significant at any visit. In white patients, conversely, after 18 months the average percent of eyes with VF loss was
less in the TAT sequence, a difference that increases and is statistically significant in years 8 to 10. In both black
and white patients, the average percent of eyes with VA loss was less in the ATT sequence; this difference is
statistically significant throughout 10 follow-up years in black patients and is statistically significant only for the
first year in white patients. In both black and white patients, average IOP reductions were greater in the TAT
sequence, though the TAT–ATT difference was substantially greater in white patients. In both black and white
patients, first-intervention failure rates were substantially lower for trabeculectomy than for trabeculoplasty.
Ten-year cumulative incidence of unilateral VF impairment comparable to legal blindness was modest in eyes of
black (ATT 11.9%, TAT 18.5%) and white (ATT 9.9%, TAT 7.3%) patients.
Conclusions: Although IOP was lowered in both sequences in black and white patients with medically
uncontrolled glaucoma, long-term visual function outcomes were better for the ATT sequence in black patients
and better for the TAT sequence in white patients. Ophthalmology 2004;111:651– 664 © 2004 by the American
Academy of Ophthalmology.
In 1998, after 7 years of follow-up, we presented the major
findings in the Advanced Glaucoma Intervention Study
(AGIS) Report 4.1 Patient follow-up has since terminated,
and we now report the complete follow-up results, classified
by race, for the 451 eyes of 332 black patients and the 325
eyes of 249 white patients. We have excluded 13 eyes of 10
Originally received: September 10, 2002.
Accepted: August 1, 2003.
Manuscript no. 220716.
Supported by grants from the National Eye Institute and the Office of
Research on Minority Health (grant nos.: U10 EY06824 through U10
EY06827, U10 EY06830 through U10 EY06835, U10 EY07057, and U10
EY09640). The 2 agencies are part of the National Institutes of Health,
U.S. Department of Health and Human Services.
Correspondence and reprint requests to Fred Ederer, MA, FACE, The
Advanced Glaucoma Intervention Study, AGIS Coordinating Center, 401
North Washington Street, Suite 700, Rockville, MD 20850. E-mail:
federer@emmes.com.
*The writing team for this report is Fred Ederer, MA, FACE, Douglas E.
Gaasterland, MD, Leonard G. Dally, MSc, Jonghyeon Kim, PhD, Paul C.
VanVeldhuisen, MS, Beth Blackwell, ScD, Bruce Prum, MD, George Shafranov,
MD, Robert C. Allen, MD, and Allen Beck, MD. A complete list of staff
in the Advanced Glaucoma Intervention Study can be found at: http://
spitfire.emmes.com/study/agi/centers.htm.
© 2004 by the American Academy of Ophthalmology
Published by Elsevier Inc.
ISSN 0161-6420/04/$–see front matter
doi:10.1016/j.ophtha.2003.09.025
651
Ophthalmology Volume 111, Number 4, April 2004
patients of other races. The visual function outcomes in
AGIS Report 41 were limited to moderate vision losses. We
have added here outcomes on vision losses akin to legal
blindness.
Patients and Methods
The AGIS design and methods, detailed elsewhere,2– 4 are briefly
summarized here.
Patients and Eyes
To be eligible for AGIS, patients had to be 35 to 80 years old, and
eyes had to be phakic and have either primary open-angle glaucoma without previous glaucoma surgery or residual open-angle
glaucoma at least 4 weeks after laser iridotomy. At least some
glaucomatous visual field (VF) loss needed to be present, and the
disease could not be adequately controlled by medications alone.
The eyes had to be on maximum medical therapy that was tolerated, effective, and acceptable to the patient, and had to meet at
least 1 of 9 specified combinations of criteria for consistently
elevated intraocular pressure (IOP), VF defect due to glaucoma,
and optic disc rim deterioration. For eyes with documented recent
optic disc or VF deterioration, the minimum required IOP for
eligibility was 18 mmHg, and for eyes without such evidence the
minimum ranged from 21 to ⬎30 mmHg, the specific minimum
depending on the magnitude of the accompanying glaucomatous
VF defect; with a minimal field defect, the IOP had to be at least
30 mmHg. Additionally, the eye had to have an AGIS VF defect
score of at least 1 and not more than 16 on the AGIS VF defect
scale, which ranges from 0 (no defect) to 20 (end stage). The
minimum required best-corrected visual acuity (BCVA) score was
56 letters on the back-lit Early Treatment Diabetic Retinopathy
Study chart5,6 (Snellen equivalent approximately 20/80).
The patients were enrolled from April 1988 through November
1992 by investigators at 11 AGIS clinical centers. Informed consent was obtained from all enrolled patients, and appropriate
institutional review boards approved the AGIS protocol. Eyes were
randomly assigned to one of two 3-step intervention sequences,
designated argon laser trabeculoplasty (ALT)–trabeculectomy–trabeculectomy (ATT) and trabeculectomy–ALT–trabeculectomy
(TAT). The methods of the surgical procedure, which followed
contemporaneous community standards, have been described in
earlier reports.1,3 The second and third treatments of a sequence
were offered only after failure (see below) of the preceding intervention.
The AGIS follow-up visits were scheduled 3 and 6 months after
enrollment, and every 6 months thereafter.
Failure of Surgical Treatment to Control
Glaucoma
Two kinds of failure of surgical glaucoma treatment were defined:
early and late. Early failure of trabeculectomy occurred within 6
weeks after the operation, and early failure of ALT within 6 weeks
after either of the 2 ALT sessions. Early failure was determined
according to study guidelines.2 Late treatment failure, occurring 6
weeks or more after the most recent glaucoma operation, was
defined by the eye again meeting the glaucoma criteria for study
eligibility.3 Late failures accounted for 95% (187/197) of failures
of black patients, and for 99% (125/126) of failures of white
patients.
652
Events and Outcome Variables
Visual function is the most important AGIS outcome. Visual field
defect scores range from 0 (no defect) to 20 (end stage).4,7 Visual
acuity (VA) scores (from the Early Treatment Diabetic Retinopathy Study scale5) range from 0 (Snellen equivalent 20/1000 or
worse) to 100 (about 20/10); the score is a count of the number of
correctly read letters on the chart.
Visual function outcome variables are defined in terms of the
following events occurring in eyes at 6-month follow-up visits.
Decrease of Visual Field (DVF). The occurrence in an eye of
(1) an increase in VF defect score of 4 or more from the preintervention baseline reference value or (2) a score of 19 or 20. (An
increase of 4 or more has a 5% chance of occurring due to
long-term fluctuation.4)
Visual Field Impairment (VFI). The occurrence in an eye of
a VF defect score of 18, 19, or 20 (an end-stage VF defect.3,4)
Decrease of Visual Acuity (DVA). The occurrence in an eye
of a decrease (worsening) in VA score of 15 or more (3 or more
lines) from the preintervention baseline reference value. (This loss
amounts to a doubling of the visual angle.)
Visual Acuity Impairment (VAI). The occurrence in an eye
of a BCVA score of less than 50 (Snellen equivalent less than
20/100).
Based on the foregoing events, outcome variables are defined
as follows.
Sustained Decrease of Visual Field (SDVF). The first occurrence in an eye of a DVF at 3 consecutive 6-month follow-up
visits. Similarly, SDVA, SVFI, and SVAI are the first occurrences
at 3 consecutive 6-month follow-up visits of DVA, VFI, and VAI,
respectively. These sustained events occur at their consummation
(i.e., at the third consecutive 6-month follow-up visit at which the
occurrence was observed). With follow-up beginning on the date
of enrollment, the earliest that these events can occur is at the
18-month visit.
Percent of Eyes with Decrease of Visual Field (PDVF). The
percent of eyes at a specified 6-month follow-up visit for which a
DVF was observed. Similarly, PDVA is the percent of eyes at a
specified 6-month follow-up visit for which DVA was observed.
Average Percent of Eyes with Decrease of Visual Field (APDVF). The unweighted average of the percent of eyes with DVF
at each 6-month follow-up visit up to and including a specified
6-month follow-up visit. The corresponding average statistic of
percent of eyes with DVA is APDVA. Whereas the sustained event
variables SDVA, SDVF, and SVAI are each the first occurrences
in an eye of that event (i.e., each event is counted as occurring at
most once in an eye during follow-up), each average percent
variable is determined at each 6-month visit, and represents the
burden of decreased vision averaged over the entire interval from
the beginning of follow-up through the specified visit. Average
percentage of eyes with DVF and APDVA are the principal AGIS
outcome variables.1
Intervention Failure. Time to failure (early or late) is measured from the date of enrollment for the first intervention, and
from the date of surgery for the second and third interventions.
Cataract Surgery. Time to cataract surgery is measured from
date of enrollment.
Statistical Methods
The interaction between race and assigned treatment group was
tested for 3 censored and 2 average percent outcome variables
(Table 1). Because some patients had both eyes enrolled, the
analyses have been adjusted for correlation between outcomes
from paired eyes. For the censored outcomes (SDVF, SDVA,
failure of initial intervention), Cox regression models were used
The AGIS Investigators 䡠 AGIS 10-Year Outcomes within Race
Table 1. P Values from Race (Black–White)–Treatment Interaction Tests for Average Percent of Eyes with Decrease of Visual Field
(APDVF) and Average Percent of Eyes with Decrease of Visual Acuity (APDVA), by Follow-up Time, and for Sustained Decrease
of Visual Field (SDVF), Sustained Decrease of Visual Acuity (SDVA), and Failure of Initial Intervention
APDVF
APDVA
Follow-up
1996*
2001†
1996*
2001†
Visit
(yrs)
Unadjusted Unadjusted Adjusted‡ Unadjusted Unadjusted Adjusted‡ Adjustment
1
2
3
4
5
6
7
8
9
10
0.4516
0.0252
0.0104
0.0040
0.0012
0.0000
0.0008
0.4558
0.0298
0.0188
0.0174
0.0192
0.0086
0.0178
0.0348
0.0216
0.0322
0.4648
0.0126
0.0102
0.0086
0.0064
0.0036
0.0084
0.0140
0.0068
0.0092
0.2832
0.0068
0.0196
0.0336
0.0128
0.0048
0.0268
0.1894
0.0064
0.0354
0.0730
0.0294
0.0208
0.0282
0.1202
0.0562
0.1032
0.2128
0.0060
0.0230
0.0340
0.0102
0.0062
0.0084
0.0424
0.0212
0.0282
Unadjusted
Adjusted‡
SDVF
1996
2001
SDVA
1996
2001
Failure of
Initial
Intervention
1996
2001
0.0003 0.0252 0.0562 0.5771 0.0016 0.0022
0.0004 0.0245 0.0111 0.3008 0.0012 0.0011
Data are P values.
*Values from AGIS Report 4 (The AGIS Investigators. The Advanced Glaucoma Intervention Study (AGIS): 4. Comparison of treatment outcomes
within race. Seven-year results. Ophthalmology 1998;105:1146 – 64), were obtained by a permutation test after 7 yrs of follow-up. Results for APDVF and
APDVA are presented unadjusted only, as in Report 4.
†
Values were obtained by a nonparametric bootstrap test after 10 yrs of follow-up.
‡
Adjusted for 6 baseline variables: age, gender, presence of hypertension, presence of diabetes, visual field defect score, and visual acuity score.
and the method of Lee, Wei, and Amato8 implemented to adjust
for paired eyes (in AGIS Report 41 we used the Wei–Lin–Weissfeld adjustment). For the average percent variables, a nonparametric bootstrap procedure9 was used to test race–treatment interactions at each annual visit, and to test treatment effects within race.
This method generates a large number of bootstrap samples by
randomly sampling with replacement and assigning the residuals
to the fitted values from a regression model that accounts for
correlation between paired eyes and for the effects of treatment,
race, and other covariates, but not for the race–treatment interaction (a nonparametric permutation test used in AGIS Report 41
does not allow adjustment for covariates). This resampling procedure determines the significance level of the interaction test by
comparing standardized test statistics obtained from bootstrap
samples with the same standardized test statistic based on the
original sample. Details of this bootstrap procedure can be found
on the AGIS Coordinating Center’s website.10
To pursue potential explanations for the significant race–treatment interactions, we included in all analyses of interaction all
variables for which blacks and whites differed at baseline: age,
gender, hypertension, diabetes, VF defect score, and VA score.
The statistical adjustments for these terms, as well as for the
interaction between each term and the treatment sequence, had
little effect on the significance of the race–treatment interactions.
We also used the aforementioned statistical methods to test for
group differences between sequences for the average percent variables (APDVF and APDVA) and for the censored variables
(SDVF, SDVA, SVFI, SVAI, intervention failure, cataract surgery). For point estimates of censored outcomes we used the
product-limit method when the date of the event was known
(cataract surgery, intervention failure); otherwise, we used the
life-table method with 6-month risk periods corresponding to the
patient follow-up visit schedule. To test mean change from baseline IOP at each visit, we used the generalized estimating equations
method of Liang and Zeger.11 To test the percentage of eyes
prescribed no medications, we used the method of Hardy for 2⫻2
tables.12 Both Liang–Zeger and Hardy take into account the
paired-eye correlation.
The unit of analysis is the eye, and in light of the multiplicity
of tests performed, we used P⬍0.01 rather than P⬍0.05 as the
criterion of statistical significance.
Results
Tables and figures additional to those in this report can be found in
a supplement to this report on the website of The EMMES Corporation, location of the AGIS Coordinating Center.10
Table 2 shows that within each race category the random
assignments to ATT and TAT created groups of patients and eyes
that were similar with respect to various personal, medical, and
ocular baseline characteristics.
Potential follow-up, the interval from enrollment to database
closure, ranges from 8 years and 4 months to 13 years. At least 10
years of potential follow-up are available on 68% (227/332) of
black patients and on 80% (200/249) of white patients. During
follow-up, 99 black patients (30%) and 68 white patients (27%)
died; of the 233 black and 181 white patients alive at the closing
date, 51 (22%) and 38 (21%), respectively, became inactive—that
is, had missed 2 or more 6-month visits after their last completed
visit (see supplement to this report10). From follow-up analyses of
VF and VA outcome variables, we excluded 15 eyes (8 ATT, 7
TAT) of black patients and 1 eye (ATT) of white patients that
either did not receive any AGIS intervention or had less than 6
months of follow-up information available; this left for analysis
436 eyes of black patients (232 ATT, 204 TAT) and 324 eyes of
white patients (157 ATT, 167 TAT).
Nearly all eyes in each of the 2 intervention sequences received
the first of the 3 AGIS interventions (black: ATT 98%, TAT 97%;
white: ATT 100%, TAT 100%), but 40% or fewer received the
second intervention (black: ATT 32%, TAT 19%; white: ATT
40%, TAT 13%), and 10% or fewer the third intervention (black:
ATT 4%, TAT 10%; white: ATT 6%, TAT 4%).10
Table 1 presents P values from black–white race–treatment
interaction tests for the most important study outcomes after potential follow-up ranging from 8 years, 4 months to 13 years
653
Ophthalmology Volume 111, Number 4, April 2004
Table 2. Baseline Characteristics of Black and White
Advanced Glaucoma Intervention Study Patients and Their
Eyes, by Treatment Assignment (Argon Laser Trabeculoplasty
[ALT]–Trabeculectomy–Trabeculectomy [ATT] or
Trabeculectomy–ALT–Trabeculectomy [TAT])
Characteristics
Patients
No. enrolled
Women (%)
Age ⱖ65 (%)
Married (%)
Completed high school (%)
With systemic hypertension (%)
With diabetes (%)
Eyes
No. enrolled
Eyes of women (%)
Mean intraocular pressure (mmHg)
Mean visual field defect score
Mean deviation (dB)
Disc hemorrhage on rim tissue (%)
Mean visual acuity score
Visual acuity scoreⱖ70 (Snellen
equivalent 20/40) (%)
Mean no. of prescribed glaucoma
medications
Black
Patients
White
Patients
332
59
58
44
51
60
27
249
48
65
66
78
39
12
ATT
TAT
ATT
TAT
240
61.3
23.7
9.2
11.1
0.8
79.7
89.2
211
60.2
23.7
9.2
11.5
1.0
78.6
82.5
158
51.9
24.1
7.8
9.7
4.0
79.3
85.4
167
49.1
24.8
7.3
9.1
3.0
80.2
88.0
2.8
2.8
2.5
2.5
(closing date: June 30, 2001); also shown are comparable results
from AGIS Report 41 based on data with potential follow-up
ranging from 4 years, 1 month to 8 years, 9 months (closing date:
December 31, 1996). The presence of a statistically significant
race–treatment interaction for a specified outcome means that the
effect of treatment sequence for that variable is different for blacks
and whites.
In Table 1, the adjusted P values for the 2001 data indicate a
significant race–treatment interaction for APDVF for years 4 to 7
and 9 to 10, and a weaker interaction for APDVA that is significant
for years 2, 6, and 7. Figure 1 clearly illustrates the interaction for
APDVF: the VF loss is less in the ATT sequence for eyes of black
patients and, after 18 months, less in the TAT sequence for eyes of
white patients. Figure 2 shows that VA loss, as represented by
APDVA, is less for the ATT sequence in both blacks and whites,
and that the magnitude of the treatment effect differs by race. In
Table 1, the interaction effect for failure of the first intervention is
statistically significant, that for SDVF is of borderline significance
(0.01ⱕP⬍0.05), and that for SDVA is not significant. The existence of significant race–treatment interactions, particularly for the
principal outcomes, justifies—indeed, mandates—separate analyses for blacks and whites.
Intervention Failure
In black and white patients, failure rates were significantly greater
for ALT than for trabeculectomy (Fig 3), whether the treatment
comparisons were made for first (P ⫽ 0.0013 for black patients,
P⬍0.0001 for white patients) or second interventions (P ⫽ 0.0063
for black patients, P ⫽ 0.0253 for white patients). There were no
significant differences between first- and second-intervention ALT
failure rates or between first- and second-intervention trabeculectomy failure rates in either black or white patients. In an analysis
that appeared in AGIS Report 11,13 after adjustment for differences in age, gender, diabetes, and hypertension, the failure rates
of third interventions (all of which were trabeculectomies) did not
significantly differ from the failure rates of trabeculectomies that
were first or second interventions. In Figure 3, the cumulative
failure probabilities for second interventions from the fourth to the
10th follow-up years are based on small numbers of failures and,
therefore, lack statistical precision. The estimated 10-year cumulative incidences of first-intervention failure were, for blacks,
Figure 1. Average percent of eyes with decrease of visual field (APDVF) in eyes of black and white patients. ATT ⫽ argon laser trabeculoplasty
(ALT)–trabeculectomy–trabeculectomy; TAT ⫽ trabeculectomy–ALT–trabeculectomy.
654
The AGIS Investigators 䡠 AGIS 10-Year Outcomes within Race
Figure 2. Average percent of eyes with decrease of visual acuity (APDVA) in eyes of black and white patients. ATT ⫽ argon laser trabeculoplasty
(ALT)–trabeculectomy–trabeculectomy; TAT ⫽ trabeculectomy–ALT–trabeculectomy.
49.2% for ALT and 31.9% for trabeculectomy and, for whites,
52.5% for ALT and 17.7% for trabeculectomy (Table 3).
Intraocular Pressure
The race–treatment interaction in mean IOP change, indicating a
greater TAT–ATT difference for whites than for blacks, is statis-
tically significant (P⬍0.01) only at 6 months, and of borderline
significance (0.01ⱕP⬍0.05) only at 1 year. In both black and
white patients, and in both treatment sequences, mean IOP decreased sharply at 6 months after the initial intervention. The mean
decrease amounted to 7 mmHg in ATT eyes and 9 mmHg in TAT
eyes in black patients, and 7 mmHg in ATT eyes and 12 mmHg in
TAT eyes in white patients; in each case the greater effect in the
Figure 3. Cumulative probability of failure of first and second interventions in eyes of black and white patients. ALT ⫽ argon laser trabeculoplasty.
655
Ophthalmology Volume 111, Number 4, April 2004
Table 3. Kaplan–Meier Estimates of 10-Year Cumulative Incidence of Failure, Sustained Visual Field and Visual Acuity Outcome
Variables, and Cataract Surgery
Black Patients
ATT
Failure of first Intervention
Failure of second Intervention
SDVF
SDVA
SVFI
SVAI
SVFI, SVAI, or both*
Cataract surgery
White Patients
TAT
ATT
TAT
10-Year
Incidence
(%)
95%
Confidence
Interval
10-Year
Incidence
(%)
95%
Confidence
Interval
10-Year
Incidence
(%)
95%
Confidence
Interval
10-Year
Incidence
(%)
95%
Confidence
Interval
49.2
41.4
37.6
35.6
11.9
10.8
17.8
43.7
41.6–56.8
21.2–61.5
29.8–45.5
27.9–43.3
6.9–17.0
5.9–15.7
11.8–23.8
36.0–51.4
31.9
65.1
34.7
43.6
18.5
15.3
23.3
58.0
24.7–39.1
34.8–95.3
27.0–42.3
35.5–51.7
11.5–25.5
9.0–21.6
16.1–30.5
49.8–66.1
52.5
26.4
39.9
26.8
9.9
11.6
16.4
60.4
43.8–61.2
9.9–43.0
31.1–48.7
19.1–34.5
4.5–15.4
5.5–17.7
9.7–23.2
51.3–69.4
17.7
71.6
29.2
35.9
7.3
12.5
14.5
62.3
11.4–24.0
30.6–100.0
21.0–37.4
27.5–44.3
2.3–12.3
6.7–18.4
8.4–20.6
53.7–70.9
*First occurrence of either an SVFI or an SVAI.
ATT ⫽ argon laser trabeculoplasty (ALT)–trabeculectomy–trabeculectomy; SDVA ⫽ sustained decrease of visual acuity; SDVF ⫽ sustained decrease of
visual field; SVAI ⫽ sustained visual acuity impairment; SVFI ⫽ sustained visual field impairment; TAT ⫽ trabeculectomy–ALT–trabeculectomy.
TAT sequence at 6 months is statistically significant (P⬍0.0001)
(Fig 4). In white patients the greater decrease in the TAT
sequence remains significant (not shown) through 6 years, and
not thereafter; in black patients the TAT–ATT difference is of
only borderline significance for the first 2 years, and not significant thereafter. At 10 years the mean decreases in both races
and both treatment sequences were still substantial: between 8
and 9 mmHg in black patients and between 9 and 10 mmHg in
white patients.
Visual Field Loss
In black patients, although not statistically significant, VF losses,
as measured by APDVF, appeared greater in the TAT sequence
than in the ATT sequence. This is shown for APDVF in Figure 1
and Table 4, and for the cumulative incidence of the events SDVF
and SVFI in Figure 5 and Table 3. Estimated 10-year cumulative
incidences of SVFI, the criterion for which may approximate that
for legal blindness based on VF, are 11.9% for the ATT sequence
and 18.5% for the TAT sequence (Table 3; P ⫽ 0.25 for SVFI
throughout follow-up).
In white patients during early follow-up, VF losses as measured by APDVF were more frequent in eyes assigned to TAT
than in those assigned to ATT. After 11⁄2 years, however, the
pattern reversed, the losses becoming ever more frequent in
eyes assigned to ATT (Fig 1), with the difference significant for
years 8 through 10 (P⬍0.01; Table 4). Although not statistically
significant, the greater worsening in the ATT sequence after 11⁄2
years is also shown by the cumulative incidences of the events
SDVF and SVFI (Fig 5, Table 3). Estimated 10-year cumulative
incidences of SVFI (akin to legal blindness based on VF) are
9.9% for the ATT sequence and 7.3% for the TAT sequence
(Table 3; P ⫽ 0.15 for SVFI throughout follow-up).
Figure 4. Mean change from baseline in intraocular pressure in eyes of black and white patents. ATT ⫽ argon laser trabeculoplasty (ALT)–
trabeculectomy–trabeculectomy; TAT ⫽ trabeculectomy–ALT–trabeculectomy.
656
The AGIS Investigators 䡠 AGIS 10-Year Outcomes within Race
Table 4. Average Percent of Eyes with Decrease of Visual Field (APDVF)
Scheduled Visit
and Assigned
Intervention
Sequence
2
4
6
8
10
yrs
yrs
yrs
yrs
yrs
Black Patients
No. of Eyes
APDVF
White Patients
Difference in APDVF
No. of Eyes
APDVF
Difference in APDVF
ATT
TAT
ATT
TAT
ATT–TAT
P
ATT
TAT
ATT
TAT
ATT–TAT
P
213
187
164
142
89
193
172
151
128
80
11.3
14.6
18.3
21.3
24.6
17.2
19.5
21.4
23.0
25.5
⫺5.9
⫺4.9
⫺3.1
⫺1.7
⫺0.9
0.0136
0.0460
0.2352
0.5160
0.7488
139
127
114
98
61
156
146
126
107
73
13.5
19.3
23.7
26.8
30.1
11.6
13.6
16.1
18.4
20.5
1.9
5.7
7.5
8.4
9.6
0.4475
0.0382
0.0110
0.0090
0.0082
ATT ⫽ argon laser trabeculoplasty (ALT)–trabeculectomy–trabeculectomy; TAT ⫽ trabeculectomy–ALT–trabeculectomy.
Visual Acuity Loss
In black patients, VA losses, as measured by APDVA, progressively increased over 10 years of follow-up in both intervention
sequences, with losses significantly greater (P⬍0.01) in the TAT
sequence than in the ATT sequence for each of the 10 years (Fig
2, Table 5). Throughout follow-up, the risk of an SDVA was
greater in the TAT sequence, though not significant (Fig 6; P ⫽
0.0127); at 10 years the cumulative incidence rates were 35.6%
and 43.6%, respectively (Table 3). Ten-year cumulative incidences
of SVAI, the criterion for which may approximate that for legal
blindness based on VA, were 10.8% in the ATT sequence and
15.3% in the TAT sequence (Table 3, Fig 6; P ⫽ 0.41).
In white patients, as in black patients, APDVA progressively
increased over 10 years of follow-up in both intervention sequences (Table 5, Fig 2). The losses are greater in the TAT
sequence, but significantly greater only the first year (P ⫽ 0.0009,
data not shown). Over the entire range of follow-up, the cumulative incidences of SDVA are similar in the ATT and TAT sequences (Fig 6; P ⫽ 0.25); at 10 years the cumulative incidence
rates were, respectively, 26.8% and 35.9% (Table 3). Ten-year
cumulative incidences of SVAI, the criterion for which may approximate that for legal blindness based on VA, are also similar in
the 2 sequences: 11.6% (ATT) and 12.5% (TAT) (Table 3, Fig 6;
P ⫽ 0.91 for SVAI throughout follow-up).
Number of Prescribed Glaucoma Medications
In black patients in each sequence, the percentage of eyes not
prescribed any glaucoma medications at baseline was 2% or less,
and the percentage prescribed 2 or more medications was 85% or
more (Fig 7). Subsequently, the percentage prescribed no medications increased in the ATT sequence to 6% and 31% at 1 and 10
Figure 5. Cumulative probability of sustained decrease of visual field (SDVF) and of sustained visual field impairment (SVFI) in eyes of black and white
patients. ATT ⫽ argon laser trabeculoplasty (ALT)–trabeculectomy–trabeculectomy; TAT ⫽ trabeculectomy–ALT–trabeculectomy.
657
Ophthalmology Volume 111, Number 4, April 2004
Table 5. Average Percent of Eyes with Decrease of Visual Acuity (APDVA)
Scheduled Visit and
Assigned
Intervention
Sequence
2
4
6
8
10
yrs
yrs
yrs
yrs
yrs
Black Patients
No. of Eyes
APDVA
White patients
Difference in APDVA
No. of Eyes
APDVA
Difference in APDVA
ATT
TAT
ATT
TAT
ATT–TAT
P
ATT
TAT
ATT
TAT
ATT–TAT
P
215
186
168
146
94
194
172
156
127
82
5.7
11.4
14.7
17.9
20.3
19.7
23.6
25.6
26.1
28.2
⫺14.0
⫺12.1
⫺10.9
⫺8.3
⫺7.9
⬍0.0001
⬍0.0001
⬍0.0001
0.0004
0.0012
138
130
119
104
65
155
147
131
110
76
7.1
10.8
14.6
16.5
17.8
12.7
16.1
17.0
18.9
19.7
⫺5.7
⫺5.3
⫺2.3
⫺2.4
⫺1.9
0.0143
0.0217
0.3407
0.3403
0.4893
ATT ⫽ argon laser trabeculoplasty (ALT)–trabeculectomy–trabeculectomy; TAT ⫽ trabeculectomy–ALT–trabeculectomy.
years, respectively, and in the TAT sequence it increased to 59%
at 1 year and was 40% at 10 years.
In white patients in each sequence, the percentage of eyes
prescribed no glaucoma medications at baseline was 4% or less,
and the percentage prescribed 2 or more medications was approximately 85%. At 1 year, the percentages prescribed no glaucoma
medications increased modestly to 14% in the ATT sequence, but
rose sharply to 67% in the TAT sequence. Subsequently, these
percentages increased in the ATT sequence and decreased in the
TAT sequence, at 10 years reaching 38% in the ATT sequence and
46% in the TAT sequence.
Cataract Surgery
Cataract surgery was permitted when a VA defect was ascribed to
cataract, the defect was adversely affecting the patient’s lifestyle,
and the VA score was less than 65 letters (20/50). Special excep-
tions to the VA criterion were granted (by an ophthalmologist
member of the Operations Committee) when the cataract was
adversely affecting the patient’s lifestyle for 7 (4%) of the 178
cataract operations of black patients and for 37 (23%) of the 163
cataract operations of white patients.
In black patients, during follow-up, 79 (34%) of the 232 eyes
assigned to ATT and 97 (48%) of the 204 eyes assigned to TAT
received cataract operations.10 The risk of cataract surgery was
significantly greater for eyes assigned to TAT than for eyes assigned to ATT (P ⫽ 0.0003; Fig 8). In black patients, estimated
10-year incidence rates of cataract surgery are 44% for eyes
assigned to ATT and 58% for eyes assigned to TAT (Table 3). In
white patients, 76 (48%) of the 157 eyes assigned to ATT and 87
(52%) of the 167 assigned to TAT had cataract surgery10; this
difference in risk is not significant (P ⫽ 0.65; Fig 8). Estimated
10-year incidence rates of cataract surgery in white patients are
60% for ATT and 62% for TAT (Table 3).
Figure 6. Cumulative probability of sustained decrease of visual acuity (SDVA) and of sustained visual acuity impairment (SVAI) in eyes of black and
white patients. ATT ⫽ argon laser trabeculoplasty (ALT)–trabeculectomy–trabeculectomy; TAT ⫽ trabeculectomy–ALT–trabeculectomy.
658
The AGIS Investigators 䡠 AGIS 10-Year Outcomes within Race
Figure 7. Percentages of eyes prescribed 0, 1, or 2 or more medications, by race, intervention sequence, and year of followup. ATT ⫽ argon laser
trabeculoplasty (ALT)–trabeculectomy–trabeculectomy; TAT ⫽ trabeculectomy–ALT–trabeculectomy.
Discussion
The AGIS investigators undertook this glaucoma study in
the mid-1980’s to test whether, as was then generally be-
lieved to be the case, trabeculoplasty before trabeculectomy
was more beneficial than immediate trabeculectomy after
failure of medical therapy. Analyses in the early AGIS
years, which combined data from all races, showed signif-
Figure 8. Cumulative probability of cataract surgery in eyes of black and white patients. ATT ⫽ argon laser trabeculoplasty (ALT)–trabeculectomy–
trabeculectomy; TAT ⫽ trabeculectomy–ALT–trabeculectomy.
659
Ophthalmology Volume 111, Number 4, April 2004
Figure 9. Average percent of eyes with decrease of visual field (APDVF) and average percent of eyes with decrease of visual acuity (APDVA) in eyes
of all AGIS patients (451 eyes of 332 black patients, 325 eyes of 249 white patients, 13 eyes of 10 patients of other races). APDVF is less for the argon
laser trabeculoplasty (ALT)–trabeculectomy–trabeculectomy (ATT) sequence during the first 5 years of follow-up, and less for the ATT sequence for the
last 5 years; none of the differences are statistically significant (P⬎0.01). APDVA is less for the ATT sequence throughout follow-up, with the ATT to
trabeculectomy–ALT–trabeculectomy (TAT) sequence differences statistically significant (P⬍0.01) for each of the 10 years.
icantly greater VA loss in the TAT sequence than in the
ATT sequence, but no significant ATT–TAT treatment difference for VF loss. Figure 9 illustrates this result for
APDVA and APDVF, the principal AGIS outcome variables,1 over 10 follow-up years. We then undertook analyses to determine whether various ocular, medical, and demographic factors were predictive of outcomes of
treatment, analyses that were specified in the AGIS Manual
of Operations,2 which was prepared during the planning of
AGIS. These analyses showed that treatment outcomes depended on race (race–treatment interactions): at a number of
follow-up visits the ATT versus TAT treatment effect differed significantly between black and white patients for VF
and VA. This important AGIS finding, first published after
7 follow-up years,1 is now found to hold after 10 years.
The race–treatment interactions make it important to
analyze the 10-year between-treatment comparisons separately for black and white patients, as we did the 7-year
comparisons.1 Because of the differences at baseline between black and white AGIS patients with respect to a
number of personal, medical, and ocular characteristics (for
a summary, see Table 2; for a detailed comparison, see
AGIS Report 314), comparisons of outcomes between the
races should not be made without adjusting for the baseline
differences. Such adjusted treatment-specific comparisons,
made with 9-year follow-up data, were presented in AGIS
Report 9.15 The results of that report support the hypothesis
that, after failure of medical therapy, an initial intervention
660
with trabeculectomy retards the progression of glaucoma
more effectively in white patients. It also provides a weak
suggestion that, after an initial intervention with ALT, the
progression of glaucoma is retarded more effectively in
black patients than in white patients.
A number of studies have shown that glaucoma affects
blacks more often and more severely than whites. The
Baltimore Eye Survey found the prevalence of open-angle
glaucoma in blacks to be 4 times as high as in whites,16 and
that of blindness from the disease to be 6 times as high.17
Among blacks in the Caribbean,18,19 glaucoma prevalence
was found to be even higher than in Baltimore. In AGIS
(Table 2) and other clinical studies,20,21 baseline VF defects
were more severe in black patients. To the well-established
finding that blacks are more vulnerable than whites to
glaucoma, AGIS has added the finding that, after failure of
medical therapy, black patients respond more favorably
than white patients to initial trabeculoplasty, and less favorably than white patients to initial trabeculectomy. Because
AGIS is the only study to have reported a differential racial
response to glaucoma surgery, it is important for other
studies of glaucoma surgery to investigate possible racial
differences in outcome.
The 7-year results1 showed that in eyes of black patients
both VF and VA were better preserved in the ATT sequence
than in the TAT sequence. The additional follow-up information in the present report has not materially changed the
comparative effect of the 2 treatment sequences in black
The AGIS Investigators 䡠 AGIS 10-Year Outcomes within Race
patients. Although some outcomes for black patients
throughout the 10 years were more favorable for eyes in the
TAT sequence, these benefits are outweighed by more favorable outcomes in the ATT sequence, particularly by
more favorable visual function outcomes. The more favorable outcomes for the TAT sequence are lower risk of
failure of the first intervention (P⬍0.01), greater average
pressure lowering (P⬍0.01 at 6 months and 24 months), and
a greater proportion of eyes prescribed no medications
(P⬍0.01 at 1 year, P ⫽ 0.14 at 10 years). The more
favorable outcomes in the ATT sequence are lower APDVF
(not significant) and lower APDVA (P⬍0.01 throughout the
10 years). Additionally, the risk of cataract surgery was less
in the ATT sequence (44% ATT, 58% TAT). For black
patients with medically uncontrolled glaucoma, these results continue to support use of the ATT sequence.
The trends established for white patients in the first 7
years also continued, by and large, for the next 3 years. As
for black patients, the benefits of a specific treatment sequence for white patients are not all one-sided. In particular,
VA losses were less, mostly marginally less (for APDVA
the difference was statistically significant only for the first
year), in eyes assigned to ATT. The risk of cataract surgery
was about the same in the 2 sequences (48% ATT, 52%
TAT). For white patients, benefits are greater for TAT than
for ATT: a lower risk of failure of the first intervention
(P⬍0.01), a greater proportion of eyes prescribed no medications (P⬍0.01 at 1 year, P ⫽ 0.25 at 10 years), greater
average pressure lowering (P⬍0.01 for the first 6 years),
and, after 11⁄2 years, less VF loss (P⬍0.01 in years 8 –10).
Largely on the strength of the long-term reduced progression of VF loss, many white patients with medically uncontrolled glaucoma may prefer the TAT sequence.
Thus, we find a complex set of outcomes after surgeries
for advanced glaucoma. In discussing treatment options
with black and white patients in whom medical treatment
has failed, the available research information should be
reviewed. Patients need to know about the complications of
surgery, including the increased cataract risk after trabeculectomy, and about the greater likelihood of need for both
additional medications and additional glaucoma surgery after ALT. To white patients, the benefit of reduced likelihood
of glaucomatous damage after trabeculectomy should be
pointed out, but ALT as initial surgery should not be ruled
out, even for patients with expected longevity of 2 or more
years, because some patients may prefer to postpone the
risks of filtering surgery.
Cataract formation was undoubtedly an important cause
of VA loss in AGIS patients, not only because of their
elderly age, but also because some two thirds of study eyes
received at least 1 trabeculectomy. As shown in AGIS
Report 8,22 a first trabeculectomy increased the risk of
cataract formation, compared with no trabeculectomy, by
78%, and a second trabeculectomy, compared with one
trabeculectomy, nearly tripled the risk. Of the eyes of both
black and white patients in the present analysis, virtually all
those in the TAT sequence had an initial trabeculectomy
soon after enrollment. By contrast, less than half the eyes in
the ATT sequence received a trabeculectomy, and these
trabeculectomies were performed at a later follow-up time.
The increased cataract risk after trabeculectomy undoubtedly contributed to the greater VA loss in the TAT sequence, particularly in the early follow-up years. In both
black and white patients, after 2 years the APDVA was
greater in the TAT sequence than in the ATT sequence
(Table 5, Fig 2).
In AGIS Report 6 (closing date December 31, 1998) we
reported that, on average, cataracts were extracted at an
earlier stage of development in white patients, and that after
analytically removing the effects of cataract the race differences in cataract were shown to account for 25% of the
race–treatment interaction for the VF variable APDVF and
for 45% of the VA variable APDVA.23 In a repeat analysis
of Report 6 using the data set for the present report (closing
date June 30, 2001), we found that analytically removing
the effects of cataract reduced the APDVF interaction by
4.2% and the APDVA interaction by 31.5%. Thus, it seems
that the earlier removal of cataract in whites continues to
play a part in the race–treatment interactions, though apparently a lesser part over 10 years than over 7 years. The
removal of cataracts at a more advanced stage in blacks
suggests that the effect of VA loss from early cataract
formation on quality of life may be less for blacks than for
whites. If that is so, should it affect how blacks are surgically treated for advanced glaucoma? We think not. Even if,
in blacks, the substantially greater VA loss in the TAT
sequence has little effect on the quality of life, there is still
no reason for black patients to prefer the TAT sequence,
considering that for them the ATT sequence entails a better
VF prognosis and lower complication rates.
The association between low IOP and reduced VF progression, demonstrated in AGIS Report 7,24 has been
widely recognized for many years.25,26 Two recent randomized clinical trials, the Ocular Hypertension Treatment
Study27 and the Early Manifest Glaucoma Trial,28 have
shown that this is not merely an association, but a causeand-effect relationship. In the present study, the pressurelowering effect in black and whites is greater for TAT than
for ATT, with greater TAT–ATT differences in mean IOP
change in whites (Fig 4). The greater TAT–ATT differences
in whites, indicating a race–treatment interaction for IOP
change, are significant (P⬍0.01) at 6 months, and of borderline significance (0.01ⱕP⬍0.05) at 1 year.10 Correspondingly, the TAT–ATT contrast in VF loss as measured
by APDVF, after 11⁄2 years, is also greater in whites, a
race–treatment interaction that is statistically significant in
years 4 to 7, 9, and 10, and of borderline significance in
years 2, 3, and 8 (Table 1). Thus, the larger TAT–ATT
difference in IOP reduction in whites expresses itself in a
larger TAT–ATT contrast in reduced VF deterioration in
whites relative to blacks.
The findings of AGIS Report 724 suggest that pressures
may need to be lowered to less than 14 mmHg to minimize
damage. In AGIS, the minimum required IOP for intervention failure ranged from 18 to ⬎30 mmHg, so that the
pressure-lowering goals may not have been set low enough
to achieve optimal results. We speculate that if the AGIS
design had provided for lower intervention failure pressure
thresholds, there would have been more failures of ALT,
with more eyes progressing to trabeculectomy. Presumably
661
Ophthalmology Volume 111, Number 4, April 2004
Table 6. Number of Trabeculectomies with and without Antifibrotics, by Intervention Sequence Number and Race
Black Patients
All trabeculectomies
N
%
Antifibrotics used
None
N
%
Any§
N
%
White Patients
All
Trabeculectomies
First
Interventions
Second
Interventions*
Third
Interventions†
All
Trabeculectomies
First
Interventions
Second
Interventions*
Third
Interventions†
315‡
100.00
206‡
100.00
77
100.00
32
100.00
247
100.00
167
100.00
63
100.00
17
100.00
253
80.3
204
99.0
47
61.0
2
6.2
229
92.7
167
100.0
59
93.7
3
17.6
62
19.7
2
1.0
30
39.0
30
93.8
18
7.3
0
0.0
4
6.3
14
82.4
*Argon laser trabeculoplasty (ALT) was the first intervention when trabeculectomy was the second intervention.
†
Trabeculectomy was the third intervention in both the ALT–trabeculectomy–trabeculectomy and trabeculectomy–ALT–trabeculectomy sequences.
‡
Includes 2 eyes that were excluded from further analyses because they had less than 6 mos of follow-up.
§
5-fluorouracil, mitomycin C, or both.
this would have resulted in retarded progression of optic
nerve damage and improved VF outcomes, particularly in
the ATT sequence. Consideration should be given to studying in a randomized clinical trial the question of whether the
maintenance of low target pressures preserves optic nerve
function in glaucoma, irrespective of the method used to
attain the target.
In these mostly elderly AGIS patients with advanced
glaucoma, the estimated 10-year cumulative incidence of
unilateral legal blindness based on VF, as approximated by
SVFI, has been modest in eyes of black (11.9% ATT, 18.5%
TAT) and white (9.9% ATT, 7.3% TAT) patients (Table 3).
From data on patients with both eyes enrolled in AGIS, in
which one eye was assigned to ATT and the other to TAT,
the estimated 10-year cumulative incidence of bilateral legal
blindness based on either VF or VA (SVFI or SVAI) was
also modest: 13.0% (95% confidence interval [CI]: 3.3–
22.8) in black patients and 3.5% (95% CI: 0.0 –10.1) in
white patients.
To enable comparisons of the AGIS incidence rates of
legal blindness after trabeculectomy with those of 2 published studies,29,30 we apply to the AGIS data the event VFI
or VAI or both, the first occurrence of VFI or VAI, an event
that occurs at a single visit, rather than the event SVFI or
SVAI, which is sustained over 3 consecutive visits; we note
that the cumulative rates based on the single occurrence are
considerably larger than those based on sustained occurrence, evidence of the instability of single-visit VF measurements. The 10-year cumulative incidence of unilateral
first occurrence of VFI or VAI or both in AGIS white
patients in the TAT group was 30.4% (95% CI: 20.6%–
40.2%). Investigators from a glaucoma study in the predominantly white population of Otago Province, New Zealand
have reported a similar 10-year cumulative incidence of
unilateral legal blindness after trabeculectomy of 28% (95%
CI: 21%– 40%).29 Higher unilateral legal blindness rates
after filtering surgery were reported by an earlier glaucoma
study in the predominantly white population of Olmsted
County, Minnesota: the 10-year cumulative incidence was
46% for legal blindness from glaucoma30 and 54% (95% CI:
662
33%– 68%) for legal blindness from all causes (Johnson
DH, personal communication, 2002). The filtering operations were carried out from 1988 to 2001 in AGIS, from
1975 to 1996 in Otago Province, and from 1965 to 1994 in
Olmsted County, so temporal improvements in filtering
surgery and medical treatment may have contributed to the
lower rates in AGIS and Otago Province.
The AGIS methods of trabeculectomy and ALT have
followed the standard in the United States.31 The AGIS
postsurgical management, including topical anti-inflammation medications,32 suture lysis33 or releasable sutures,34
and adjunctive antifibrotics,35 adhered to community standards at the time of the surgical intervention, standards that
have been stable since 1993. Routine use of antifibrotics in
AGIS began in 1992, about the time that all initial trabeculectomies had been completed. As shown in Table 6, adjunctive antifibrotics were used in only 1% of trabeculectomies that were first AGIS interventions, but much more
frequently, particularly in black patients, in trabeculectomies that were second interventions (after ALT failure) and
third interventions (all of which were trabeculectomies after
failure of a previous trabeculectomy). In black patients
antifibrotics were used in 39% of second-intervention trabeculectomies and 94% of third-intervention trabeculectomies; in white patients they were used in 6% of secondintervention trabeculectomies and 82% of third-intervention
trabeculectomies. Factors that may have contributed to
greater antifibrotic use in black patients are a greater failure
rate of filtering surgery in contralateral eyes and a belief by
AGIS ophthalmologists that, without adjunctive antifibrotics, these eyes were at increased risk of failure. In the
absence of information from randomized clinical trials, one
can only speculate about the effect that more extensive use
of adjunctive antifibrotics might have had on AGIS outcomes. In black patients the deleterious effects of trabeculectomy (relative to ALT) might have been reduced, and in
white patients the beneficial effects of trabeculectomy (relative to ALT) might have been enhanced, although these
benefits would likely have come at the cost of increased
The AGIS Investigators 䡠 AGIS 10-Year Outcomes within Race
complications. Our main conclusions would probably not
have been affected.
As we noted in AGIS Reports 41 and 11,13 failure rates
in AGIS have been lower than in many other studies, at least
in part due to differences in failure criteria. In eyes of both
black and white patients, the need for a second intervention
was markedly less in the TAT sequence than in the ATT
sequence. Thus, the 10-year cumulative incidence of failure
of first-intervention ALT was approximately 50% in black
and white patients, whereas the failure incidence of firstintervention trabeculectomy was only 32% in black patients
and 18% in white patients (Table 3). Although we found no
statistically significant differences between failure rates of
first- and second-intervention trabeculectomies or of firstand second-intervention ALTs (Fig 3), the statistical power
to detect differences was limited by small numbers of failures of second interventions; moreover, the findings may be
affected by new medications that became available after
first interventions were completed in 1992. Whereas nearly
all trabeculectomies that were first interventions were performed without adjunctive antifibrotics, some trabeculectomies that were second interventions (39% [30/77] of those
in blacks and 6% [4/63] of those in whites) were performed
with adjunctive antifibrotics. Topical carbonic anhydrase
inhibitors, prostaglandin analogs, and ␣-agonists, introduced into AGIS in 1995 to 1996, could have helped to
lower failure rates of both ALT and trabeculectomy as first
and second interventions, but particularly as second interventions. Over 10 years the relative frequencies of third
interventions were low in black and white patients in either
treatment sequence, ranging from 3% to 8%.10
The Advanced Glaucoma Intervention Study is one of
several ongoing or completed large, randomized trials of
glaucoma that have enrolled substantial numbers of black
and white patients. Of the other trials, 1 was of 5-fluorouracil after trabeculectomy in patients with poor prognoses,36
and 2 were of early open-angle glaucoma, one comparing
ALT with medical therapy37 and the other comparing trabeculectomy with medical therapy.38 The investigators of
these studies have not published information on tests of
race–treatment interactions. The significant race–treatment
interactions found in AGIS indicate that race is an important
variable for study in current and future clinical, epidemiologic, and pathologic investigations of glaucoma. Future
studies should be designed with large enough enrollment to
test for race differences and race–treatment interactions.10
Such tests will be particularly important in clinical trials of
laser surgery, filtering surgery, and other intervention modalities.
References
1. The AGIS Investigators. The Advanced Glaucoma Intervention Study (AGIS): 4. Comparison of treatment outcomes
within race. Seven-year results. Ophthalmology 1998;105:
1146 – 64.
2. Advanced Glaucoma Intervention Study (AGIS) Manual of
Operations. Springfield, VA: National Technical Information
Service; 1993. Accession No. PB93-220192.
3. The AGIS Investigators. The Advanced Glaucoma Interven-
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
tion Study (AGIS): 1. Study design and methods and baseline
characteristics of study patients. Control Clin Trials 1994;15:
299 –325.
The Advanced Glaucoma Intervention Study Investigators.
Advanced Glaucoma Intervention Study. 2. Visual field test
scoring and reliability. Ophthalmology 1994;101:1445–55.
Ferris FL, Kassoff A, Bresnick GH, Bailey I. New visual
acuity charts for clinical research. Am J Ophthalmol 1982;94:
91– 6.
Ferris FL, Sperduto RD. Standardized illumination for visual
acuity testing in clinical research. Am J Ophthalmol 1982;94:
97– 8.
Katz J, Sommer A, Gaasterland DE, Anderson DR. Comparison of analytic algorithms for detecting glaucomatous visual
field loss. Arch Ophthalmol 1991;109:1684 –9.
Lee EW, Wei LJ, Amato DA. Cox-type regression analysis for
large numbers of small groups of correlated failure time observations. In: Klein JP, Goel PK, eds. Survival Analysis:
State of the Art. Dordrecht, Netherlands: Kluwer Academic
Publishers; 1992:237– 47.
Davison AC, Hinkley DV. Bootstrap Methods and Their Application. Cambridge, United Kingdom: Cambridge University Press; 1997.
The AGIS Investigators. The Advanced Glaucoma Intervention Study (AGIS). 13. Comparison of treatment outcomes
within race: 10-year results. Supplemental text, tables and
figures. Available at: http://spitfire.emmes.com/study/agi/
report13.htm. Accessed September 10, 2002.
Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika 1986;73:13–22.
Hardy RJ, Davis BR, Palmer EA, Tung B, Cryotherapy for
Retinopathy of Prematurity Cooperative Group. Statistical
considerations in terminating randomization in the Multicenter Trial of Cryotherapy for Retinopathy of Prematurity.
Control Clin Trials 1991;12:293–303.
The AGIS Investigators. The Advanced Glaucoma Intervention Study (AGIS): 11. Risk factors for failure of trabeculectomy and argon laser trabeculoplasty. Am J Ophthalmol 2002;
134:481–98.
The AGIS Investigators. The Advanced Glaucoma Intervention Study (AGIS): 3. Baseline characteristics of black and
white patients. Ophthalmology 1998;105:1137– 45.
The AGIS Investigators. The Advanced Glaucoma Intervention Study (AGIS): 9. Comparison of glaucoma outcomes in
black and white patients within treatment groups. Am J Ophthalmol 2001;132:311–20.
Tielsch JM, Sommer A, Katz J, et al. Racial variations in the
prevalence of open-angle glaucoma. The Baltimore Eye Survey. JAMA 1991;266:369 –74.
Sommer A, Tielsch JM, Katz J, et al. Racial differences in the
cause-specific prevalence of blindness in East Baltimore.
N Engl J Med 1991;325:1412–7.
Mason RP, Kosoko O, Wilson MR, et al. National survey of
the prevalence and risk factors of glaucoma in St. Lucia, West
Indies. Part I. Prevalence findings. Ophthalmology 1989;96:
1363– 8.
Leske MC, Connell AM, Schachat AP, Hyman L. The Barbados Eye Study. Prevalence of open angle glaucoma. Arch
Ophthalmol 1994;112:821–9.
Glaucoma Laser Trial Research Group. The Glaucoma Laser
Trial (GLT): 5. Subgroup differences at enrollment. Ophthalmic Surg 1993;24:232– 41.
Grant W, Burke JF Jr. Why do some people go blind from
glaucoma? Ophthalmology 1982;89:991– 8.
AGIS (Advanced Glaucoma Intervention Study) Investigators.
The Advanced Glaucoma Intervention Study: 8. Risk of cat-
663
Ophthalmology Volume 111, Number 4, April 2004
23.
24.
25.
26.
27.
28.
29.
30.
aract formation after trabeculectomy. Arch Ophthalmol 2001;
119:1771–9.
The AGIS Investigators. The Advanced Glaucoma Intervention Study 6. Effect of cataract on visual field and visual
acuity. Arch Ophthalmol 2000;118:1639 –52.
AGIS Investigators. The Advanced Glaucoma Intervention
Study (AGIS): 7. The relationship between control of intraocular pressure and visual field deterioration. Am J Ophthalmol
2000;130:429 – 40.
Anderson DR. Glaucoma: the damage caused by pressure.
XLVI Edward Jackson Memorial Lecture. Am J Ophthalmol
1989;108:485–95.
Sommer A. Intraocular pressure and glaucoma. Am J Ophthalmol 1989;107:186 – 8.
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;120:701–13.
Heijl A, Leske MC, Bengtsson B, et al. Reduction of intraocular pressure and glaucoma progression. Results from the
Early Manifest Glaucoma Trial. Arch Ophthalmol 2002;120:
1268 –79.
Molteno ACB, Bosma NJ, Kittelson JM. Otago glaucoma
surgery outcome study. Long-term results of trabeculectomy—1976 to 1995. Ophthalmology 1999;106:1742–50.
Parc CE, Johnson DH, Oliver JE, et al. The long-term outcome
of filtration surgery. Am J Ophthalmol 2001;132:27–35.
664
31. American Academy of Ophthalmology Preferred Practice Patterns Committee Glaucoma Panel. Preferred Practice Pattern:
Primary Open-Angle Glaucoma. San Francisco: American
Academy of Ophthalmology; 2000.
32. Skuta GL, Beeson CC, Higginbotham EJ, et al. Intraoperative
mitomycin versus postoperative 5-fluorouracil in high risk
glaucoma filtering surgery. Ophthalmology 1992;99:438 – 44.
33. Spaeth GL. Glaucoma surgery. In: Spaeth GL, ed. Ophthalmic
Surgery. Principles and Practice. Philadelphia: WB Saunders;
1982.
34. Hoskins HD Jr, Migliazzo C. Management of failing filtering blebs with the Argon laser. Ophthalmic Surg 1984;15:
731–3.
35. Nouri-Mahdavi K, Brigatti L, Weitzman M, Caprioli J. Outcomes of trabeculectomy for primary open-angle glaucoma.
Ophthalmology 1995;102:1760 –9.
36. The Fluorouracil Filtering Surgery Group. Five-year follow-up of the Fluorocil Filtering Surgery Study. Am J Ophthalmol 1996;121:349 – 66.
37. Glaucoma Laser Trial Research Group. The Glaucoma Laser
Trial (GLT) and Glaucoma Laser Trial Follow-up Study: 7.
Results. Am J Ophthalmol 1995;120:718 –31.
38. Lichter PR, Musch DC, Gillespie BW, et al, CIGTS Study
Group. Interim clinical outcomes in the Collaborative Initial
Glaucoma Treatment Study comparing initial treatment randomized to medications or surgery. Ophthalmology 2001;108:
1943–53.