April 2014
Volume 55, Issue 4
Free
Glaucoma  |   April 2014
Baseline Prognostic Factors Predict Rapid Visual Field Deterioration in Glaucoma
Author Affiliations & Notes
  • Jun Mo Lee
    Siloam Eye Hospital, Seoul, Korea
  • Joseph Caprioli
    The Jules Stein Eye Institute, David Geffen School of Medicine at University of California-Los Angeles, Los Angeles, California, United States
  • Kouros Nouri-Mahdavi
    The Jules Stein Eye Institute, David Geffen School of Medicine at University of California-Los Angeles, Los Angeles, California, United States
  • Abdelmonem A. Afifi
    Department of Biostatistics, Jonathan and Karin Fielding School of Public Health at University of California-Los Angeles, Los Angeles, California, United States
  • Esteban Morales
    The Jules Stein Eye Institute, David Geffen School of Medicine at University of California-Los Angeles, Los Angeles, California, United States
  • Meera Ramanathan
    The Jules Stein Eye Institute, David Geffen School of Medicine at University of California-Los Angeles, Los Angeles, California, United States
  • Fei Yu
    The Jules Stein Eye Institute, David Geffen School of Medicine at University of California-Los Angeles, Los Angeles, California, United States
    Department of Biostatistics, Jonathan and Karin Fielding School of Public Health at University of California-Los Angeles, Los Angeles, California, United States
  • Anne L. Coleman
    The Jules Stein Eye Institute, David Geffen School of Medicine at University of California-Los Angeles, Los Angeles, California, United States
    Department of Epidemiology, Jonathan and Karin Fielding School of Public Health at University of California-Los Angeles, Los Angeles, California, United States
  • Correspondence: Anne L. Coleman, Jules Stein Eye Institute, Glaucoma Division, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, CA 90095, USA; Coleman@jsei.ucla.edu
Investigative Ophthalmology & Visual Science April 2014, Vol.55, 2228-2236. doi:10.1167/iovs.13-12261
  • Views
  • PDF
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Jun Mo Lee, Joseph Caprioli, Kouros Nouri-Mahdavi, Abdelmonem A. Afifi, Esteban Morales, Meera Ramanathan, Fei Yu, Anne L. Coleman; Baseline Prognostic Factors Predict Rapid Visual Field Deterioration in Glaucoma. Invest. Ophthalmol. Vis. Sci. 2014;55(4):2228-2236. doi: 10.1167/iovs.13-12261.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Purpose.: To investigate baseline prognostic factors predicting rapid deterioration of the visual field in primary open-angle glaucoma patients.

Methods.: Seven hundred sixty-seven eyes of 566 glaucoma patients from the Advanced Glaucoma Intervention Study (AGIS) and the clinical database from Jules Stein Eye Institute's Glaucoma Division were included. The rates of decay at each visual field test location were calculated with pointwise exponential regression analysis (PER), and the rates were separated into faster and slower components for each series. Subjects with a faster component decay rate (≥36%/y) were defined as rapid progressors. Sex, race, age, visual acuity, intraocular pressure, mean deviation (MD), number of medications, use of diabetic or hypertension medications, and vertical cup-to-disc ratio at baseline were entered in a multivariable prognostic logistic regression model.

Results.: The average (±SD) MD was −8.02 (±6.13), and the average age was 68.64 (±11.71) years for the study group. Two hundred twenty-two eyes (28.9%) were identified as rapid progressors. The following baseline factors were predictors of faster deterioration: worse MD (P < 0.001, odds ratio [OR]: 1.11; 95% confidence interval [CI]: 1.07–1.15), larger vertical cup-to-disc ratio (P = 0.001, OR: 1.23; 95% CI: 1.09–1.39), and older age (P = 0.02, OR: 1.24; 95% CI: 1.04–1.48). After excluding the variables related to glaucoma severity at baseline (baseline MD and baseline vertical cup-to-disc ratio), the likelihood of being a rapid progressor was 54% greater in African Americans than in Caucasians (P = 0.03, OR: 1.55; 95% CI: 1.06–2.27).

Conclusions.: Patients with more severe glaucomatous damage, as measured by both visual field or optic disc cupping and older age, are at highest risk for rapid worsening of the disease, as are African Americans compared to Caucasians. More aggressive treatment of such patients should be considered to prevent visual disability.

Introduction
A better understanding of clinical prognostic factors for worsening of glaucoma may help us develop new strategies to improve glaucoma treatments and their outcomes. Measurement of rates of change in glaucoma can identify those patients whose condition is deteriorating rapidly and can help clinicians to better use available health care resources. 1 We recently described a method based on pointwise exponential regression (PER) analysis of threshold sensitivities in a group of patients with visual field loss 2 and demonstrated that visual field regions with faster visual field decay rates can be distinguished from the remaining test locations that have a slower rate of deterioration. This slower component frequently includes influences related to aging and media opacity, and the faster component appears to be more closely aligned with true glaucomatous progression. 3 We defined patients as rapid progressors based on the rate of the faster component of visual field decay. 2,4  
The current study used the above approach to identify a rapidly deteriorating group of patients and examined baseline prognostic factors that may forecast worsening of glaucomatous visual field damage. We confirmed the role of the prognostic factors with a widely accepted cross-validation technique. We propose that such an approach be used to identify patients at highest risk of visual loss to better allocate and monitor treatment of glaucoma. 
Methods
Patient Data, Visual Field Data, and Statistical Methods
Patient data collected during the conduct of the Advanced Glaucoma Intervention Study (AGIS) and data from the clinical database at Jules Stein Eye Institute's Glaucoma Division were combined and used as our study sample. The AGIS design and methods are described in detail elsewhere. 5,6 From the AGIS group, data from a total of 389 eyes of 309 open-angle glaucoma patients who had 6 or more years of follow-up and who underwent 12 or more visual field examinations were included. 7 Visual field tests were performed with a Humphrey Field Analyzer (HFA I; Carl Zeiss Meditec, Inc., Dublin, CA, USA) with the 24-2 test pattern, size III white stimulus, and full-threshold strategy. The inclusion criteria for the University of California-Los Angeles (UCLA) group have also been described in detail previously. 4 The latter retrospective study group consisted of 378 eyes of 257 open-angle glaucoma patients with 6 or more years of follow-up and 8 or more visual field tests. Perimetry was performed with the Humphrey Field Analyzer (HFA I or II; Carl Zeiss Meditec, Inc.) and the 24-2 test pattern, size III white stimulus, and with full-threshold, Swedish Interactive Threshold Algorithm (SITA) Standard or SITA Fast strategies. Visual field reliability criteria were defined as <30% fixation losses, <30% false positive responses, and <30% false negative responses. This study, approved by the Institutional Review Board of the University of California-Los Angeles, was performed in accordance with the tenets set forth in the Declaration of Helsinki and complied with Health Insurance Portability and Accountability Act regulations. 
A total of 10 potential candidate baseline variables for predicting visual field deterioration were explored as follows: sex, race, age, visual acuity, intraocular pressure (IOP), mean deviation (MD), number of medications, use of diabetic medications, use of hypertension medications, and vertical cup-to-disc ratio. For AGIS patients, all examiners were required to estimate the cup-to-disc ratios with stereoscopic slit-lamp examination. For UCLA patients, estimates of cup-to-disc ratios were obtained by one examiner (JC) viewing stereoscopic photographs of the optic discs. Variables associated with visual field deterioration in univariate analyses (χ 2 test, unpaired t-test, or Wilcoxon rank sum test, depending on the type of data) at a P value of <0.2 were included in the final multivariable logistic regression model. Race was entered as a binary value (African American versus Other). We examined cup-to-disc ratios both as a binary and as a continuous variable. Associations between visual field deterioration in rapid progressors and potential baseline prognostic factors were evaluated using a multivariable stepwise forward logistic regression model in the entire cohort, and the model fit was assessed by Akaike information criterion (AIC). 8,9 The associations were expressed as odds ratios (ORs). The odds of visual field deterioration are commonly used to quantify the risk of visual field deterioration, and are defined as the ratio of the probability of having visual field deterioration to the probability of not having visual field deterioration. For example, if the probability of having visual field deterioration is one-third, the odds of visual field deterioration is one-half ([1/3]/[2/3]), meaning that no visual field deterioration is two times as likely as visual field deterioration. The OR is used to compare the risk between two groups and is calculated as the relative change of odds between two groups. 10  
In order to cross-validate our prognostic factor model, we first randomly selected a training subsample consisting of two-thirds of the main study group. In this subsample, we fitted a multivariable logistic regression model with the original 10 variables. We then predicted the probability of being a rapid progressor for every eye in the remaining one-third of the main study group (test subsample) and calculated the area under the curve (AUC) for the receiver operating characteristic (ROC) curve. This procedure was repeated 1000 times, resulting in 1000 values for AUC. Each time, a different equation was derived with 10 variables using the two-thirds sample, and the AUC was estimated from the remaining one-third sample. This produced an unbiased estimate of the probability of correct classification as estimated by the AUC. The average of the 1000 values was the final cross-validated outcome for our prognostic factor model. All statistical analyses were carried out in the open programming language R. 11  
Rates of Visual Field Decay, Faster and Slower Rate Components, Rapid and Slow Progressors
Our methodology is described in detail elsewhere. 2 We performed regression analysis of the threshold sensitivity (in decibels) against time for each visual field test location with the PER model. The relationship between the response variable (threshold sensitivity) and the explanatory variable (duration of follow-up) was characterized by the equation y = ea+bx . The rate of change is represented by the coefficient b, which is the average annual rate of change in loge y. The quantity eb is interpreted as the annual percentage of sensitivity reduction. The rate of decay is defined as (1 − eb ), which is the annual percentage of remaining sensitivity. To facilitate the clinical understanding of the magnitude of severity of the rates, the b values of the exponential regressions were converted into %/y deterioration rates, where rate of decay (%/y) = (1 − eb ) × 100. 
The rate of decay (%/y) was determined for the visual field series of each eye by taking the average of the individual decay rates for all 54 test locations analyzed. Pointwise rates of decay were measured with the PER model and were plotted as a frequency distribution (the two test locations at the physiologic blind spot were excluded from all analyses). Any visual field test location with zero dB threshold sensitivity in three consecutive tests starting from the first visit was excluded from analysis. The 52 visual field test locations were ranked according to the rate of decay, and clustered into two subgroups (faster and slower components) based on the P value for the difference in mean rates between two clusters. For each partition, we computed a t-test statistic, and the corresponding P values were adjusted for multiple testing. The Benjamini-Hochberg correction was used to find the optimal P value to maximize the difference between the faster and slower component subgroups. 12 Each subgroup consisted of at least five test locations. For each eye, average faster and slower rates of decay were calculated for the partitioned components. With use of the K means cluster method in the SPSS statistical software, 13 all eyes were classified into rapid and slow progressors based on the average rate of change in the fast component. 
Results
Patient Data
Three hundred eighty-nine eyes of 309 open-angle glaucoma patients from the AGIS database and 378 eyes of 257 patients from the UCLA database were included in this study. The mean (±SD) MD was −8.02 (±6.13) dB, and the mean age was 68.64 (±11.71) years for the entire study group. The characteristics and demographic data are presented for AGIS and UCLA separately in Table 1
Table 1
 
Characteristics of the Study Groups
Table 1
 
Characteristics of the Study Groups
AGIS UCLA
Eyes, n 389 378
Patients, n 309 257
Sex, n (%)
 Male 185 (47.6%) 170 (45.0%)
 Female 204 (52.4%) 208 (55.0%)
Race, n (%)
 Caucasian 174 (44.7%) 289 (76.4%)
 African American 211 (54.3%) 24 (6.4%)
 Hispanic 4 (1.0%) 9 (2.4%)
 Other 0 (0.0%) 56 (14.8%)
Age, y 64.7 ± 9.6 65.0 ± 11.7
Baseline visual acuity 0.8 ± 0.3 0.8 ± 0.2
Baseline IOP, mm Hg 15.3 ± 5.0 15.4 ± 5.2
Baseline MD,* dB −10.8 ± 5.5 −5.3 ± 5.4
Baseline number of medications, n 2.8 ± 0.9 1.6 ± 1.2
Diabetic medication, n (%) 70 (18.0%) 39 (10.3%)
Hypertension medication, n (%) 197 (50.6%) 99 (26.2%)
Baseline vertical cup-to-disc ratio 0.8 ± 0.2 0.7 ± 0.2
Follow-up years, y 8.1 ± 1.1 9.2 ± 2.8
Number of visual fields 15.7 ± 3.0 13.7 ± 5.8
Rates of Visual Field Decay, Faster and Slower Components, Rapid and Slow Progressors
In the AGIS group, 21,006 test locations for 389 eyes were analyzed. Of the 21,006 locations, 2012 test locations (9.6%) were excluded because of consecutive zero threshold values at baseline. In the UCLA group, 20,412 visual field test locations for 378 eyes were analyzed, and 1618 test locations (7.9%) were excluded because of initial consecutive zero threshold values. 
Figure 1 demonstrates the frequency distribution for the faster component of rates of decay. The faster component shows a bimodal distribution. When the K means cluster method was applied, the faster component rates could be segregated into two separate clusters, with a cutoff point of 36%/y for separation of the two subgroups. The rate of decay (n = 222, mean [±SD]) for the rapid progressor group was 65.0%/y (±14.2%/y) and for the slow progressor group (n = 545) was 6.0 ± 7.8%/y. The rapid progressor group consisted of 29.0% of all eyes. 
Figure 1
 
Frequency distribution curve for the faster component of visual field decay for the main study group shows a bimodal distribution. If a cutoff value of 36%/y is taken to be the separation point of the two subdivisions, the rate of decay (n = 222, mean [±SD]) for the rapid progressor group was 65.0%/y (±14.2%/y) and for the slow progressor group (n = 545) was 6.0 ± 7.8%/y. The rapid progressor group consisted of 29.0% of all eyes.
Figure 1
 
Frequency distribution curve for the faster component of visual field decay for the main study group shows a bimodal distribution. If a cutoff value of 36%/y is taken to be the separation point of the two subdivisions, the rate of decay (n = 222, mean [±SD]) for the rapid progressor group was 65.0%/y (±14.2%/y) and for the slow progressor group (n = 545) was 6.0 ± 7.8%/y. The rapid progressor group consisted of 29.0% of all eyes.
Figure 2 shows the histogram for MD and visual field index (VFI) rate of change, separately by the rapid and slow progressors. There were statistically significant differences in both MD and VFI rate of change between rapid and slow progressors (P < 0.0001 for both MD and VFI). The mean (±SD) MD rate of change for the rapid progressors was −0.68 (±0.61) dB/y, and the mean (±SD) MD rate of change for the slow progressors was −0.03 (±0.01) dB/y. The mean (±SD) VFI rate of change for the rapid progressors was −2.76 (±2.52) %/y, and the mean (±SD) VFI rate of change for the slow progressors was −0.15 (±0.02) %/y (Table 2). 
Figure 2
 
The histograms for visual field MD and VFI rate of change, separately by the rapid and slow progressor groups of eyes.
Figure 2
 
The histograms for visual field MD and VFI rate of change, separately by the rapid and slow progressor groups of eyes.
Table 2
 
Summary Statistics for Rates of Change With Visual Field Mean Deviation and Visual Field Index
Table 2
 
Summary Statistics for Rates of Change With Visual Field Mean Deviation and Visual Field Index
Rapid Progressors Slow Progressors P Value
MD rate of change
 Mean −0.68 −0.03 <0.0001
 SD 0.61 0.01
 Range −3.75 to 0.43 −1.6 to 1.38
VFI rate of change
 Mean −2.76 −0.15 <0.0001
 SD 2.52 0.02
 Range −13.92 to 1.12 −4.38 to 4.59
Figure 3 shows an example for a patient from the rapid progressor group, including the initial and last visual fields, the rate of decay from the PER method, and changes in the MD and VFI. Figure 4 shows another example for a patient from the slow progressor group. 
Figure 3
 
Example of a patient from the rapid progressor group, showing the raw threshold data and the gray scale of the patient's initial and last visual fields. Below are the rate of decay data for the fast component and slow component of this patient's eye analyzed using the pointwise exponential regression method, along with an example of the change of the visual field MD and VFI data for the same time period.
Figure 3
 
Example of a patient from the rapid progressor group, showing the raw threshold data and the gray scale of the patient's initial and last visual fields. Below are the rate of decay data for the fast component and slow component of this patient's eye analyzed using the pointwise exponential regression method, along with an example of the change of the visual field MD and VFI data for the same time period.
Figure 4
 
An example of a patient from the slow progressor group, showing the raw threshold data and the gray scale of the patient's initial and last visual fields. Below are the rate of decay data for the fast component and slow component of this patient's eye analyzed with the pointwise exponential regression method, along with plots of the visual field MD and VFI data for the same time period. When the fitted regression line of MD and VFI had a positive slope (improvement in visual field), the fitted line was adjusted as a flat line using the average of the first three values.
Figure 4
 
An example of a patient from the slow progressor group, showing the raw threshold data and the gray scale of the patient's initial and last visual fields. Below are the rate of decay data for the fast component and slow component of this patient's eye analyzed with the pointwise exponential regression method, along with plots of the visual field MD and VFI data for the same time period. When the fitted regression line of MD and VFI had a positive slope (improvement in visual field), the fitted line was adjusted as a flat line using the average of the first three values.
Logistic Regression Results
The characteristics of the study sample are presented in Table 3. Two hundred twenty-two eyes (28.9%) out of 767 eyes from the main study sample were identified as rapid progressors. Table 4 shows the results of the multivariable logistic regression. Any variables with a P value < 0.2 in the univariate analysis were included in the final model. Three variables were associated with visual field deterioration in the multivariable logistic regression model: baseline MD (mean ± SD: −11.4 ± 5.50), baseline vertical cup-to-disc ratio (0.81 ± 0.13), and age at the time of the first visit (65.9 ± 9.9). The OR of having visual field deterioration was 1.11/dB of worsening MD (95% confidence interval [CI]: 1.07–1.15; P < 0.001), meaning that the odds of deterioration were 11% higher for each decibel of worsening MD. The OR was 1.23 per 0.1-unit increase in vertical cup-to-disc ratio (95% CI: 1.09–1.39; P = 0.001) and 1.24 per decade of advancing age (95% CI: 1.04–1.48; P = 0.02). Our study also examined the cup-to-disc ratio as a binary variable. Using a cup-to-disc ratio cutoff point of 0.7, the OR for deterioration was 1.97 (95% CI: 1.22–3.26, P = 0.006) for eyes with a baseline cup-to-disc ratio > 0.7 compared to those with a baseline cup-to-disc ratio ≤ 0.7. Excluding the variables related to glaucoma severity at baseline (baseline MD and baseline vertical cup-to-disc ratio) from the multivariable logistic regression model resulted in race showing an association with glaucoma deterioration. The odds of having rapid visual field worsening in African Americans were 1.55 (95% CI: 1.06–2.27, P = 0.03) times as great as in Caucasians. 
Table 3
 
Comparison of Baseline Characteristics of the Study Sample According to Rates of Decay (Rapid Progressors Defined as Eyes With a Rate of Decay Equal to or Exceeding 36%/y)
Table 3
 
Comparison of Baseline Characteristics of the Study Sample According to Rates of Decay (Rapid Progressors Defined as Eyes With a Rate of Decay Equal to or Exceeding 36%/y)
Rapid Progressors Slow Progressors P Value
N % N %
Total 222 28.9 545 71.1
Sex
 Male 112 50.5 243 44.6 0.1623
 Female 110 49.5 302 55.4
Race
 Caucasian 120 54.1 343 62.9 0.0022
 African American 88 39.6 147 27.0
 Hispanic 2 0.9 11 2.0
 Other 12 5.4 44 8.1
Age at first visit
 Mean 65.93 64.11 0.0253
 SD 9.86 11.02
 Range 38.38–88.69 7.60–89.16
Baseline visual acuity
 Mean 0.81 0.83 0.4608
 SD 0.23 0.27
 Range 0.1–1.33 0.05–3.5
Baseline IOP
 Mean 14.88 15.54 0.1161
 SD 5.45 4.93
 Range 1–45 2–34
Baseline MD
 Mean −11.35 −6.74 <0.001
 SD 5.50 5.85
 Range −25.36 to 0.12 −23.67 to 2.29
Baseline number of medications
 Mean 2.44 2.11 0.0005
 SD 1.18 1.24
 Range 0–4 0–5
Diabetic medication
 Yes 40 18.0 69 12.7% 0.0698
 No 182 82.0 476 87.3%
Hypertension medication
 Yes 97 43.7 199 36.5 0.0766
 No 125 56.3 346 63.5
Baseline vertical cup-to-disc ratio
 Mean 0.81 0.73 <0.001
 SD 0.13 0.18
 Range 0.2–1 0–1
Table 4
 
Results of Multivariable Logistic Regression for Prediction of Visual Field Progression Using Both the Advanced Glaucoma Intervention Study and University of California-Los Angeles Databases
Table 4
 
Results of Multivariable Logistic Regression for Prediction of Visual Field Progression Using Both the Advanced Glaucoma Intervention Study and University of California-Los Angeles Databases
P Value Odds Ratio 95% CI for Odds Ratio
Lower Upper
Sex, reference, female 0.071 1.371 0.973 1.933
Race, reference, Caucasian 0.619 1.108 0.738 1.657
Age, per decade 0.019 1.236 1.037 1.479
Visual acuity, per unit change 0.072 1.943 0.933 3.977
Baseline IOP, per 1 mm Hg 0.626 0.991 0.957 1.026
Baseline MD,* per dB <0.001 1.112 1.074 1.152
Baseline number of medications 0.985 0.998 0.856 1.165
Diabetes medication, yes 0.168 1.394 0.864 2.228
Hypertension medication, yes 0.656 0.920 0.637 1.324
Baseline vertical cup-to-disc ratio, per 0.1-unit change 0.001 1.226 1.089 1.389
Cross-Validation for the Multivariable Logistic Regression Model
Figure 5 is a plot of the ROC curve for discriminating the rapid progressors based on the multivariable logistic regression model. The sensitivities and specificities were also calculated. The sensitivity describes the probability of predicting an eye to be a rapid progressor based on the model when the eye is observed to be a rapid progressor. The specificity describes the probability of predicting an eye to be a slow progressor when that eye is observed to be a slow progressor. Based on 1000 cross-validated AUC values, the prediction score for the multivariable logistic regression model was 0.72 ± 0.03, meaning a 72% probability of correct classification as either a rapid progressor or a slow progressor using the multivariable prediction model. 
Figure 5
 
The receiver operating characteristic curve for discriminating rapid progressors from slow progressors with the multivariable logistic regression model. The predictive value of the multivariable logistic regression model after cross-validation was 0.72 (±0.035). This can be interpreted as the probability of correct classification as either a rapid progressor or a slow progressor using the multivariable logistic regression model developed in the current study.
Figure 5
 
The receiver operating characteristic curve for discriminating rapid progressors from slow progressors with the multivariable logistic regression model. The predictive value of the multivariable logistic regression model after cross-validation was 0.72 (±0.035). This can be interpreted as the probability of correct classification as either a rapid progressor or a slow progressor using the multivariable logistic regression model developed in the current study.
Discussion
Worse baseline MD, larger cup-to-disc ratios, and older age were associated with rapid deterioration of the visual field loss over an average follow-up of 8.7 years. Compared to Caucasians, African Americans were more likely to have rapid visual field deterioration when baseline MD and baseline vertical cup-to-disc ratio were excluded. 
More advanced visual field loss or worse MD at baseline has been reported as a prognostic factor in several other cohorts of glaucoma patients. 14,15 In the Early Manifest Glaucoma Trial (EMGT), 14 a worse MD was reported to predict future glaucoma progression. In eyes with MD ≤ −0.4 dB, the hazard ratio (HR) for progression was 1.38 (95% CI: 1.00–1.91). In the Collaborative Initial Glaucoma Treatment Study (CIGTS), 15 advanced visual field loss was studied using the visual field scoring algorithm described by Lichter et al. 16 A 1-unit increase in the baseline visual field score was predictive of a 0.74-unit increase in the follow-up visual field scores. However, in the Ocular Hypertension Treatment Study (OHTS), 17 baseline MD was reported to have no association with the risk of developing primary open-angle glaucoma (POAG; P value = 0.89). The OHTS 18 and Medeiros et al. 19 also reported that for every 0.2-dB decrease in the baseline pattern standard deviation, the HR was 1.16 (95% CI: 0.95–1.41) and 1.25 (95% CI: 1.06–1.48), respectively. In our study, we observed an OR of having visual field deterioration of 1.11/dB of worsening baseline MD (95% CI: 1.07–1.15; P < 0.001), meaning that the odds of visual field deterioration were accelerated by 11% for every 1-dB worsening in baseline MD. In other words, if the odds of visual field deterioration are one-half (0.5) for a patient with baseline MD of −6 dB, the odds of visual field deterioration are 0.555 (0.5 × 1.11) for a patient with baseline MD of −7 dB. This finding differs from that of AGIS, in which a more advanced visual field score appeared to be protective. 20 Because AGIS patients had fairly advanced glaucoma, there may be a ceiling effect in that population. Ernest et al. 21 reviewed 85 articles investigating prognostic factors for visual field progression in patients with open-angle glaucoma and reported 103 different prognostic factors. Greater baseline visual field loss was identified as a prognostic factor that is probably associated with more visual field progression. 
Several studies have reported the association of a larger cup-to-disc ratio with glaucomatous progression. 14,17,22 In OHTS, 17 vertical cup-to-disc ratio was a statistically significant predictor for the development of POAG (HR: 1.32; 95% CI: 1.19–1.47). The EMGT 14 and the Collaborative Normal-Tension Glaucoma Study (CNTGS) 22 also explored the prognostic value of baseline cup-to-disc ratios (with a cutoff of 0.7) but did not report baseline cup-to-disc ratios to be a statistically significant predictive factor for glaucomatous progression. The CIGTS 15 and the Low-pressure Glaucoma Treatment Study (LGTS) 23 also found no statistically significant association between baseline cup-to-disc ratios and visual field progression. In contrast, we found that eyes with cup-to-disc ratios > 0.7 at baseline had a 97% greater rate of decay in the faster visual field component compared to those eyes with cup-to-disc ratios ≤ 0.7. We also observed that for every 0.1-unit increase in the baseline cup-to-disc ratio, the faster component rate increased by 23%. 
Individuals who were older at baseline had a 24% greater likelihood of rapid visual field progression than those who were a decade younger. Many of the other longitudinal cohort studies have found a similar effect of age on visual field progression. 14,15,17,23 In CIGTS, 15 older age was associated with an increasing frequency of substantial visual field loss (every 10-year increment in age increased the risk of visual field loss by 40%), while individuals greater than 68 years old at baseline had an increased risk of progression (HR: 1.51; 95% CI: 1.11–2.07) in EMGT 14 than individuals 68 years old or younger. Older age was an even stronger prognostic factor (HR: 1.86; 95% CI: 1.16–2.97) for subjects with lower baseline IOP in that study. 14 In LGTS, 23 the risk of progression increased by 43% for each decade older the subjects were at baseline. However, AGIS 18 and CNTGS 22 reported that older age at baseline was not associated with an increased sustained decrease of visual field (SDVF) risk. In the meta-analysis 21 on prognostic factors for visual field progression, older baseline age was the only definite prognostic factor identified. 
When baseline optic nerve and visual field parameters were excluded, African Americans were more likely to be rapid progressors during follow-up than Caucasians were. African Americans had a 55% increased risk of visual field deterioration relative to Caucasians. This finding is similar to the result reported by CIGTS 15 that nonwhites are more likely to have visual field worsening (HR: 1.50; 95% CI: 1.08–2.07). The OHTS 17 reported that in the univariate model, African Americans had a 59% increased risk of developing POAG (HR: 1.59; 95% CI: 1.09–2.32) compared to other participants (white, Hispanic). In OHTS, African American participants had larger baseline vertical cup-to-disc ratios and thinner central corneas than other participants. The inclusion of either baseline vertical cup-to-disc ratio or central corneal thickness resulted in race becoming statistically nonsignificant in the multivariable model. Our findings are in contrast to the findings reported by the AGIS 20 investigators; however, the AGIS investigators did not exclude baseline visual field parameters from the analyses. 
As for sex and visual acuity, both were of borderline statistical significance in our study, which may be related to the fact that AGIS patients accounted for 51% of our study sample. The AGIS reported an HR of 2.23 for males (95% CI: 1.54–3.23) compared to females and an HR of 0.96 for baseline visual acuity (95% CI: 0.94–0.98), 20 while we reported that the odds of being a rapid progressor were 37% higher for males than for females (OR = 1.37; P = 0.071, 95% CI: 0.97–1.93) and 94% higher per unit change in visual acuity (OR = 1.94; P = 0.072, 95% CI: 0.93–3.98). In contrast, CNTGS 22 reported that the risk of glaucomatous progression in women was 1.85 times higher relative to that in men (P value = 0.062). The EMGT, 14 CIGTS, 15 OHTS, 17 and LGTS 23 reported that there was no association between sex and risk of visual field loss. 
Many cohort studies have found baseline IOP to be a predictive factor for glaucomatous progression. 14,17 The EMGT 14 reported higher baseline IOP (≥21 mm Hg) to be associated with glaucomatous progression (HR: 1.77; 95% CI: 1.29–2.43), while OHTS reported an approximately 10% increased risk of glaucomatous progression with each mm Hg higher baseline IOP. 17 Although Ernest et al. 21 identified higher baseline IOP as a prognostic factor that is probably associated with visual field progression, in our study baseline IOP was not a prognostic factor for glaucomatous progression (OR = 0.99; P = 0.626, 95% CI: 0.96–1.03). One reason may be that our baseline IOPs were all treated IOPs, in contrast to OHTS or EGTS, in which the baseline IOP was not a treated IOP. 
Although our study included the use of diabetic medications as a surrogate for the presence of diabetes, we did not find an association between the use of diabetic medication and rapid visual field progression in the 109 (14%) patients using diabetic medications (P = 0.17). Self-reporting of diabetes was associated with open-angle glaucoma progression in CIGTS 15 (HR: 1.59; 95% CI: 1.07–2.38) but not in CNTGS or EMGT. 14,22  
Although our study includes 296 (39%) patients who used systemic antihypertensive medications, we did not find an association between the use of antihypertensive medications and rapid visual field progression (P = 0.66). However, LGTS, 21 which included 56 (44%) hypertensive patients, reported an association between the use of systemic antihypertensive medications and glaucomatous progression (HR: 2.53; 95% CI: 1.32–4.87). 
Currently there is no gold standard for defining glaucomatous visual field deterioration. The EMGT 24 and OHTS 25 used pattern deviation glaucoma change probability maps and determined significant progression at test locations where P < 0.05. The CIGTS 15 used the visual field score, and AGIS 18,26 used the visual field defect score. The LGTS 23 defined visual field progression using a pointwise linear regression method. In the report by De Moraes et al., 27 progression is defined as occurring when the rate of sensitivity decline exceeds 1.0 dB/y on the PROGRESSOR software (Medisoft Ltd., Leeds, UK) that is based on pointwise linear regression analysis. In contrast, we used PER analyses to estimate visual field deterioration and the rate of decay. There is a difference in the mean baseline MD values between the rapid progressors and slow progressors (Table 3). This suggests that patients with more advance baseline VF damage are more likely to have rapid VF progression. 
To reduce the effect of cataracts on the visual field results, visual field progression was based on pattern deviation data in EMGT, 24 whereas in CNTGS, 22 patients with cataracts were excluded. To control for the effect of cataract progression, we classified the raw threshold values (dB) of each location into faster and slower components and included only the faster component in this study. The faster component of visual field loss is considered to be mostly unaffected by the progression of media opacity (cataracts), which appear to make up the majority of the slow component. 2,4 The faster component also seems to be independent of changes in the rate of MD in contrast to the slower component, with which MD has a strong linear relationship. 2,4 This suggests that test locations can be identified where local glaucomatous deterioration is occurring, largely independent of diffuse visual field loss. 2  
A limitation of our study is that we were not able to include disc size as a parameter because the AGIS investigators did not report disc size measurements and did not obtain stereoscopic disc photographs. Although our analysis does not include time-dependent factors that develop during follow-up and risk factors such as central corneal thickness because they were not available in AGIS, we do have a greater range of baseline glaucomatous visual field loss than AGIS had. 20 The initial MD for the baseline visual fields ranged from −25.4 to 2.3 dB for our study sample (−8.1 ± 6.1) vs. −23.7 to 1.0 dB for AGIS (−10.8 ± 5.5). 20 Even though we defined rapid progressors by a specific cutoff point of 36%/y, the specific cutoff point can be varied depending upon the configuration of the patient group. Since the faster component showed a bimodal distribution, we used the K means cluster method to distinguish the two clusters and to help choose our cutoff point of 36%/y. 
In conclusion, glaucoma patients with worse baseline MD, larger baseline vertical cup-to-disc ratios, and older age at baseline are more likely to have glaucomatous visual field deterioration over the next 8 to 9 years than patients who have better baseline MD, smaller vertical cup-to-disc ratios, and younger age. When baseline MD values and baseline vertical cup-to-disc ratios were excluded from the analysis, African Americans were more likely to be rapid progressors than Caucasians were. Although these prognostic factors are similar to those reported in other studies, our method of defining visual field progression emphasizes focal changes that are more often secondary to glaucomatous damage than is diffuse visual field loss. 2,4 Because we found prognostic factors similar to those found when diffuse visual field loss is the outcome, baseline MD values, baseline cup-to-disc ratios, older baseline age, and African American ethnicity appear to be robust predictive factors that clinicians caring for open-angle glaucoma patients can use to assess an individual's risk of glaucomatous visual field deterioration. Additional analyses including IOP parameters during follow-up and the effects of surgical intervention may provide additional information on the effectiveness of treatment in progressive visual field loss. 
Acknowledgments
Supported by institutional grants from the UCLA Center for Eye Epidemiology (FY). 
Disclosure: J.M. Lee, None; J. Caprioli, Allergan (C); K. Nouri-Mahdavi, Allergan (C); A.A. Afifi, None; E. Morales, None; M. Ramanathan, None; F. Yu, None; A.L. Coleman, None 
References
Caprioli J. The importance of rates in glaucoma. Am J Ophthalmol . 2008; 145: 191–192. [CrossRef] [PubMed]
Caprioli J Mock D Bitrian E A method to measure and predict rates of regional visual field decay in glaucoma. Invest Ophthalmol Vis Sci . 2011; 52: 4765–4773. [CrossRef] [PubMed]
Artes PH Chauhan BC Keltner JL Longitudinal and cross-sectional analyses of visual field progression in participants of the Ocular Hypertension Treatment Study. Arch Ophthalmol . 2010; 128: 1528–1532. [CrossRef] [PubMed]
Azarbod P Mock D Bitrian E Validation of point-wise exponential regression to measure the decay rates of glaucomatous visual fields. Invest Ophthalmol Vis Sci . 2012; 53: 5403–5409. [CrossRef] [PubMed]
Ederer F Gaasterland DE Sullivan EK; Investigators AGIS The Advanced Glaucoma Intervention Study (AGIS): 1. Study design and methods and baseline characteristics of study patients. Control Clin Trials . 1994; 15: 299–325. [CrossRef] [PubMed]
The AGIS Investigators. The Advanced Glaucoma Intervention Study (AGIS): 4. Comparison of treatment outcomes within race: seven-year results. Ophthalmology . 1998; 105: 1146–1164. [CrossRef] [PubMed]
The AGIS Investigators. Advanced Glaucoma Intervention Study. 2. Visual field test scoring and reliability. Ophthalmology . 1994; 101: 1445–1455. [CrossRef] [PubMed]
Burnham KP Anderson DR. Multimodel inference: understanding AIC and BIC in model selection. Sociol Methods Res . 2004; 33: 261–304. [CrossRef]
Hastie T Tibshirani R Friedman J. The Elements of Statistical Learning: Data Mining, Inference, and Prediction. 2nd ed. New York: Springer; 2009: 219–257.
Jewell NP. Risk comparisons. Am J Ophthalmol . 2009; 148: 484–486. [CrossRef] [PubMed]
R Development Core Team. R: A Language and Environment for Statistical Computing . Vienna, Austria: R Foundation for Statistical Computing; 2010.
Thissen D Steinberg L Kuang D. Quick and easy implementation of the Benjamini-Hochberg procedure for controlling the false positive rate in multiple comparisons. J Educ Behav Stat . 2002; 27: 77–83. [CrossRef]
SPSS Inc. PASW Statistics Base 18 . Chicago: SPSS Inc.; 2011.
Leske MC Heijl A Leske M Predictors of long-term progression in the Early Manifest Glaucoma Trial. Ophthalmology . 2007; 114: 1965–1972. [CrossRef] [PubMed]
Lichter P Musch D Gillepie B Interim clinical outcomes in the Collaborative Initial Glaucoma Treatment Study comparing initial treatment randomized to medications or surgery. Ophthalmology . 2001; 108: 1943–1954. [CrossRef] [PubMed]
Lichter PR Mills RP CIGTS Study Group. Quality of life study: determination of progression. In: DR Anderson Drance SM eds. Encounters in Glaucoma Research 3: How to Ascertain Progression and Outcome . Amsterdam: Kugler Publications; 1996: 149–163.
Gordon MO Beiser MS Brandt JD The Ocular Hypertension Treatment Study: baseline factors that predict the onset of primary open-angle glaucoma. Arch Ophthalmol . 2002; 120: 714–720. [CrossRef] [PubMed]
Ocular Hypertension Treatment Study Group; European Glaucoma Prevention Study Group Gordon MO Torri V Miglior S Validated prediction model for the development of primary open-angle glaucoma in individuals with ocular hypertension. Ophthalmology . 2007; 114: 10–19. [CrossRef] [PubMed]
Medeiros FA Weinreb RN Sample PA Validation of a predictive model to estimate the risk of conversion from ocular hypertension to glaucoma. Arch Ophthalmol . 2005; 123: 1351–1360. [CrossRef] [PubMed]
Investigators AGIS The advanced glaucoma intervention study (AGIS): 12. Baseline risk factors for sustained loss of visual field and visual acuity in patients with advanced glaucoma. Am J Ophthalmol . 2002; 134: 499–512. [CrossRef] [PubMed]
Ernest PJ Schouten JS Beckers HJ An evidence-based review of prognostic factors for glaucomatous visual field progression. Ophthalmology . 2013; 120: 512–519. [CrossRef] [PubMed]
Drance S Anderson D Schulzer M Risk factors for progression of visual field abnormalities in normal-tension glaucoma. Am J Ophthalmol . 2001; 131: 699–708. [CrossRef] [PubMed]
De Moraes CG Liebmann JM Greenfield DS Risk factors for visual field progression in the low-pressure glaucoma treatment study. Am J Ophthalmol . 2012; 154: 702–711. [CrossRef] [PubMed]
Heijl A Leske MC Bengtsson B Early Manifest Glaucoma Trial Group. Measuring visual field progression in the Early Manifest Glaucoma Trial. Acta Ophthalmol Scand . 2003; 81: 286–293. [CrossRef] [PubMed]
Kass MA Heuer DK Higginbotham EJ The Ocular Hypertension Treatment Study: a randomized trial determines that topical ocular hypotensive medication delays or prevents the onset of primary open-angle glaucoma. Arch Ophthalmol . 2002; 120: 701–713. [CrossRef] [PubMed]
The Advanced Glaucoma Intervention Study (AGIS): 7. The relationship between control of intraocular pressure and visual field deterioration. The AGIS Investigators. Am J Ophthalmol . 2000; 130: 429–440. [CrossRef] [PubMed]
De Moraes CG Sehi M Greenfield DS Chung YS Ritch R Liebmann JM. A validated risk calculator to assess risk and rate of visual field progression in treated glaucoma patients. Invest Ophthalmol Vis Sci . 2012; 53: 2702–2707. [CrossRef] [PubMed]
Figure 1
 
Frequency distribution curve for the faster component of visual field decay for the main study group shows a bimodal distribution. If a cutoff value of 36%/y is taken to be the separation point of the two subdivisions, the rate of decay (n = 222, mean [±SD]) for the rapid progressor group was 65.0%/y (±14.2%/y) and for the slow progressor group (n = 545) was 6.0 ± 7.8%/y. The rapid progressor group consisted of 29.0% of all eyes.
Figure 1
 
Frequency distribution curve for the faster component of visual field decay for the main study group shows a bimodal distribution. If a cutoff value of 36%/y is taken to be the separation point of the two subdivisions, the rate of decay (n = 222, mean [±SD]) for the rapid progressor group was 65.0%/y (±14.2%/y) and for the slow progressor group (n = 545) was 6.0 ± 7.8%/y. The rapid progressor group consisted of 29.0% of all eyes.
Figure 2
 
The histograms for visual field MD and VFI rate of change, separately by the rapid and slow progressor groups of eyes.
Figure 2
 
The histograms for visual field MD and VFI rate of change, separately by the rapid and slow progressor groups of eyes.
Figure 3
 
Example of a patient from the rapid progressor group, showing the raw threshold data and the gray scale of the patient's initial and last visual fields. Below are the rate of decay data for the fast component and slow component of this patient's eye analyzed using the pointwise exponential regression method, along with an example of the change of the visual field MD and VFI data for the same time period.
Figure 3
 
Example of a patient from the rapid progressor group, showing the raw threshold data and the gray scale of the patient's initial and last visual fields. Below are the rate of decay data for the fast component and slow component of this patient's eye analyzed using the pointwise exponential regression method, along with an example of the change of the visual field MD and VFI data for the same time period.
Figure 4
 
An example of a patient from the slow progressor group, showing the raw threshold data and the gray scale of the patient's initial and last visual fields. Below are the rate of decay data for the fast component and slow component of this patient's eye analyzed with the pointwise exponential regression method, along with plots of the visual field MD and VFI data for the same time period. When the fitted regression line of MD and VFI had a positive slope (improvement in visual field), the fitted line was adjusted as a flat line using the average of the first three values.
Figure 4
 
An example of a patient from the slow progressor group, showing the raw threshold data and the gray scale of the patient's initial and last visual fields. Below are the rate of decay data for the fast component and slow component of this patient's eye analyzed with the pointwise exponential regression method, along with plots of the visual field MD and VFI data for the same time period. When the fitted regression line of MD and VFI had a positive slope (improvement in visual field), the fitted line was adjusted as a flat line using the average of the first three values.
Figure 5
 
The receiver operating characteristic curve for discriminating rapid progressors from slow progressors with the multivariable logistic regression model. The predictive value of the multivariable logistic regression model after cross-validation was 0.72 (±0.035). This can be interpreted as the probability of correct classification as either a rapid progressor or a slow progressor using the multivariable logistic regression model developed in the current study.
Figure 5
 
The receiver operating characteristic curve for discriminating rapid progressors from slow progressors with the multivariable logistic regression model. The predictive value of the multivariable logistic regression model after cross-validation was 0.72 (±0.035). This can be interpreted as the probability of correct classification as either a rapid progressor or a slow progressor using the multivariable logistic regression model developed in the current study.
Table 1
 
Characteristics of the Study Groups
Table 1
 
Characteristics of the Study Groups
AGIS UCLA
Eyes, n 389 378
Patients, n 309 257
Sex, n (%)
 Male 185 (47.6%) 170 (45.0%)
 Female 204 (52.4%) 208 (55.0%)
Race, n (%)
 Caucasian 174 (44.7%) 289 (76.4%)
 African American 211 (54.3%) 24 (6.4%)
 Hispanic 4 (1.0%) 9 (2.4%)
 Other 0 (0.0%) 56 (14.8%)
Age, y 64.7 ± 9.6 65.0 ± 11.7
Baseline visual acuity 0.8 ± 0.3 0.8 ± 0.2
Baseline IOP, mm Hg 15.3 ± 5.0 15.4 ± 5.2
Baseline MD,* dB −10.8 ± 5.5 −5.3 ± 5.4
Baseline number of medications, n 2.8 ± 0.9 1.6 ± 1.2
Diabetic medication, n (%) 70 (18.0%) 39 (10.3%)
Hypertension medication, n (%) 197 (50.6%) 99 (26.2%)
Baseline vertical cup-to-disc ratio 0.8 ± 0.2 0.7 ± 0.2
Follow-up years, y 8.1 ± 1.1 9.2 ± 2.8
Number of visual fields 15.7 ± 3.0 13.7 ± 5.8
Table 2
 
Summary Statistics for Rates of Change With Visual Field Mean Deviation and Visual Field Index
Table 2
 
Summary Statistics for Rates of Change With Visual Field Mean Deviation and Visual Field Index
Rapid Progressors Slow Progressors P Value
MD rate of change
 Mean −0.68 −0.03 <0.0001
 SD 0.61 0.01
 Range −3.75 to 0.43 −1.6 to 1.38
VFI rate of change
 Mean −2.76 −0.15 <0.0001
 SD 2.52 0.02
 Range −13.92 to 1.12 −4.38 to 4.59
Table 3
 
Comparison of Baseline Characteristics of the Study Sample According to Rates of Decay (Rapid Progressors Defined as Eyes With a Rate of Decay Equal to or Exceeding 36%/y)
Table 3
 
Comparison of Baseline Characteristics of the Study Sample According to Rates of Decay (Rapid Progressors Defined as Eyes With a Rate of Decay Equal to or Exceeding 36%/y)
Rapid Progressors Slow Progressors P Value
N % N %
Total 222 28.9 545 71.1
Sex
 Male 112 50.5 243 44.6 0.1623
 Female 110 49.5 302 55.4
Race
 Caucasian 120 54.1 343 62.9 0.0022
 African American 88 39.6 147 27.0
 Hispanic 2 0.9 11 2.0
 Other 12 5.4 44 8.1
Age at first visit
 Mean 65.93 64.11 0.0253
 SD 9.86 11.02
 Range 38.38–88.69 7.60–89.16
Baseline visual acuity
 Mean 0.81 0.83 0.4608
 SD 0.23 0.27
 Range 0.1–1.33 0.05–3.5
Baseline IOP
 Mean 14.88 15.54 0.1161
 SD 5.45 4.93
 Range 1–45 2–34
Baseline MD
 Mean −11.35 −6.74 <0.001
 SD 5.50 5.85
 Range −25.36 to 0.12 −23.67 to 2.29
Baseline number of medications
 Mean 2.44 2.11 0.0005
 SD 1.18 1.24
 Range 0–4 0–5
Diabetic medication
 Yes 40 18.0 69 12.7% 0.0698
 No 182 82.0 476 87.3%
Hypertension medication
 Yes 97 43.7 199 36.5 0.0766
 No 125 56.3 346 63.5
Baseline vertical cup-to-disc ratio
 Mean 0.81 0.73 <0.001
 SD 0.13 0.18
 Range 0.2–1 0–1
Table 4
 
Results of Multivariable Logistic Regression for Prediction of Visual Field Progression Using Both the Advanced Glaucoma Intervention Study and University of California-Los Angeles Databases
Table 4
 
Results of Multivariable Logistic Regression for Prediction of Visual Field Progression Using Both the Advanced Glaucoma Intervention Study and University of California-Los Angeles Databases
P Value Odds Ratio 95% CI for Odds Ratio
Lower Upper
Sex, reference, female 0.071 1.371 0.973 1.933
Race, reference, Caucasian 0.619 1.108 0.738 1.657
Age, per decade 0.019 1.236 1.037 1.479
Visual acuity, per unit change 0.072 1.943 0.933 3.977
Baseline IOP, per 1 mm Hg 0.626 0.991 0.957 1.026
Baseline MD,* per dB <0.001 1.112 1.074 1.152
Baseline number of medications 0.985 0.998 0.856 1.165
Diabetes medication, yes 0.168 1.394 0.864 2.228
Hypertension medication, yes 0.656 0.920 0.637 1.324
Baseline vertical cup-to-disc ratio, per 0.1-unit change 0.001 1.226 1.089 1.389
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×