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.