Trends are smoothed by locally weighted least-squares regression.
Of note, 291 of 607 FEs (47.9%) were initially treated along with the study eye at baseline (time 0). Error bars indicate 95% CIs for the estimated probability of FE treatment.
eFigure 1. Flowchart of treatment of the fellow eye in CIGTS participants.
eFigure 2. Scatterplots of fellow eye (FE) vs study eye (SE) slopes over time for mean deviation (MD) and intraocular pressure (IOP).
eFigure 3. Interaction plot showing the effect of mean deviation (MD) over time on the hazard of fellow eye treatment in the subset of eyes that were not initially treated at baseline (n = 316).
Customize your JAMA Network experience by selecting one or more topics from the list below.
Niziol LM, Gillespie BW, Musch DC. Association of Fellow Eye With Study Eye Disease Trajectories and Need for Fellow Eye Treatment in Collaborative Initial Glaucoma Treatment Study (CIGTS) Participants. JAMA Ophthalmol. 2018;136(10):1149–1156. doi:10.1001/jamaophthalmol.2018.3274
What is the association of disease trajectory and need for treatment between the 2 eyes of a patient with glaucoma?
In this post hoc analysis of 607 participants of the Collaborative Initial Glaucoma Treatment Study, 414 participants (68.2%) had bilateral treatment for glaucoma within 7 years after initial treatment, with a correlation observed between eyes on mean deviation trajectories. Characteristics associated with an increased hazard of fellow eye treatment included older age, hypertension, higher intraocular pressure, large cup-disc ratio, and worse mean deviation.
These findings suggest that most patients with unilateral glaucoma will develop bilateral glaucoma and therefore should be closely monitored, especially if disease progression is noted in the initially diagnosed eye.
Clinicians would benefit from knowing the association of glaucomatous change in 2 eyes of the same patient.
To estimate the time between initial glaucoma treatment of the study eye (SE) and the need for treatment of the fellow eye (FE) and to ascertain the concordance between rates of progression in SEs and FEs.
Design, Setting, and Participants
In this post hoc analysis of participants from the Collaborative Initial Glaucoma Treatment Study, 607 participants with newly diagnosed open-angle glaucoma in 1 or both eyes were followed up for up to 11 years. An SE was designated at baseline and randomized to medical or surgical treatment. By protocol, FEs were treated when eligible or at physician discretion. Survival analysis methods were used to estimate the probability of FE treatment over time and to test baseline and time-dependent predictors of treatment. Disease trajectory was calculated with linear regression as the patient eye–specific slopes of mean deviation (MD) and intraocular pressure (IOP) over time, and correlations between SE and FE trajectories were calculated. Data were collected from October 1993 to December 2004. Data were analyzed from September 2012 to May 2018.
Main Outcomes and Measures
Time to FE treatment and slopes over time of MD and IOP in SEs and FEs.
Of the 607 included patients, 334 (55.0%) were male and 337 (55.5%) were white, and the mean (SD) age was 58.0 (10.9) years. A total of 291 FEs (47.9%) were treated at baseline, 123 (20.3%) were eventually treated, and 193 (31.8%) never received treatment. The probability of FE treatment for open-angle glaucoma was 0.57 at 1 year and 0.68 at 7 years after randomization. Characteristics significantly associated with an increased hazard of FE treatment included older age (hazard ratio [HR], 1.33; 95% CI, 1.08-1.64; P = .007), hypertension (HR, 1.76; 95% CI, 1.16-2.67; P = .008), higher IOP (HR, 1.24; 95% CI, 1.20-1.29; P < .001), large cup-disc ratio (HR, 1.40; 95% CI, 1.23-1.58; P < .001), and worse MD. Correlations in MD slopes between SEs and FEs initially, eventually, and never treated were 0.73, 0.71, and 0.34, respectively. Comparable correlations in IOP slopes were 0.57, 0.24, and 0, respectively.
Conclusions and Relevance
Among patients with newly diagnosed open-angle glaucoma, almost half had initial treatment of both eyes. After 7 years, approximately two-thirds of patients had bilateral treatment. Of the variables predictive of FE treatment, modifiable factors included hypertension and IOP. Slopes of MD were similar between SEs and treated FEs. This implies that SE change is a harbinger of FE change and therefore warrants close surveillance.
The binocular nature of open-angle glaucoma (OAG) has been documented, with a population-based study of Latino individuals1 reporting that roughly half of those identified had the condition in both eyes. Chen and coauthors found that glaucoma may develop at different times2 and progress at different rates between eyes of an individual.3,4 To our knowledge, comparative information is lacking on timing of initial treatment and trajectory of disease between eyes of an individual with newly diagnosed OAG.
To address this knowledge gap, we assessed fellow eye (FE) information on patients whose study eye (SE) was treated for newly diagnosed OAG in the Collaborative Initial Glaucoma Treatment Study (CIGTS). Our aim was to investigate the incidence and timing of glaucoma treatment in FEs and predictors of FE treatment and to compare disease trajectories of FEs with SEs. When a patient is diagnosed with OAG that warrants treatment in only 1 eye, knowing characteristics of the initially untreated FE that are associated with progression and eventual treatment and knowing whether FE trajectory of disease is related to that of the initially diagnosed eye would be useful in follow-up of such eyes. This knowledge could prompt earlier FE intervention to reduce the risk of visual impairment due to glaucoma.
The CIGTS enrolled 607 participants with newly diagnosed OAG in 1 or both eyes at 14 clinical centers in the United States from 1993 to 1997. Participants were randomized to topical medications or trabeculectomy to treat their glaucoma and were followed up every 6 months through 2004. Follow-up visits included Humphrey 24-2 full-threshold visual field (VF) testing, Goldmann applanation tonometry, and best-corrected Early Treatment of Diabetic Retinopathy Study visual acuity testing to collect mean deviation (MD), intraocular pressure (IOP), and visual acuity, respectively. Although an SE was designated at baseline for bilaterally eligible participants, FEs were closely monitored and treated when protocol requirements for treatment eligibility were met or at the discretion of the treating physician. When both eyes were eligible, choice of SE was at the discretion of the treating ophthalmologist. The SE was almost always the worse eye on at least 1 clinical measure related to glaucoma (ie, IOP, MD, and cup-disc ratios; 603 of 607 eyes [99.3%]). Data for this study were analyzed between September 2012 and May 2018. This study was approved by the Investigative Review Board at the University of Michigan, and written informed consent was obtained from all participants.
Eligibility of an SE required 1 of 3 criteria to be met: (1) a qualifying IOP of 20 mm Hg or greater, a Humphrey 24-2 VF with 3 or more contiguous points on the probability plot at the less than 2.0% level, a glaucoma hemifield test outside normal limits, and optic disc change compatible with glaucoma; (2) a qualifying IOP of 20 to 26 mm Hg, a Humphrey 24-2 VF result with 2 or more contiguous points in the same hemifield on the probability plot at the less than 2.0% level, and glaucomatous optic disc damage; or (3) a qualifying IOP of 27 mm Hg or greater with glaucomatous optic disc damage. Fellow eyes meeting one of these eligibility requirements at baseline were treated, with allowance for physician discretion, along with the SE as randomized. If an FE did not meet eligibility at baseline, the treating ophthalmologist had the option to treat or not treat. When the decision was to treat the ineligible FE, the treatment given was to match the SE treatment. If the ophthalmologist elected not to treat the ineligible FE, the FE could subsequently be treated after it met one of the treatment criteria during follow-up or at the physician’s discretion. Although requirements for FE treatment were well defined in the protocol, physician discretion or patient preference sometimes altered the choice and/or timing of treatment.
In participants bilaterally eligible at baseline, both SEs and FEs were started the randomized treatment and proceeded separately through the intervention sequence. If initial treatment failed to control IOP VF loss for a given eye, argon laser trabeculoplasty was administered. Subsequent inability to control IOP or VF loss led to treatment crossover (ie, trabeculectomy for those randomized to medication and medication for those randomized to trabeculectomy). Fellow eyes that became eligible during follow-up were started the first step of randomized treatment, not the current treatment of the SE. Further details on the CIGTS protocol have been described elsewhere.5
Patient demographic characteristics and baseline FE characteristics were summarized with means and standard deviations for continuous variables and with frequencies and percentages for categorical variables and were stratified by whether FEs received initial treatment at baseline, were eventually treated during follow-up when eligible, or were never treated. Tests for differences between initially, eventually, and never treated FEs were performed with analysis of variance and χ2 tests. Linear regression was used to estimate disease trajectory, defined as the slope of MD or IOP over time. Trajectories of FEs and SEs were calculated for each participant with at least 5 years of follow-up. Trajectories were compared between SEs and FEs with descriptive statistics, correlations (r), and scatterplots stratified by FE treatment categories. In FEs that were eventually treated, locally weighted least-squares regression6 was used to create smoothed trends of MD and IOP over time, before and after treatment. Kaplan-Meier analysis was used to estimate the cumulative probability of FE treatment over time. Cox regression was used to evaluate factors associated with time to FE treatment in initially untreated eyes. Factors investigated included patient characteristics (ie, age at randomization, sex, race, education, marital status, glaucoma family history, smoking status, and presence of diabetes, hypertension, or other vascular disease), baseline clinical characteristics of the FE and SE (ie, randomized treatment, glaucoma diagnosis, spherical equivalent, and disc hemorrhage), and time-dependent clinical characteristics of FEs and SEs (ie, MD, IOP, visual acuity, and vertical cup-disc ratio). Model results were adjusted for center effects. Model selection was performed using the method of best subsets with the score statistic as the selection criterion.7 The proportional hazards assumption was assessed by testing covariate by time interactions; when significant, interaction terms were included in models. Results are presented as hazard ratios (HRs) and 95% CIs. A P value less than .05 was considered statistically significant, and all P values were 2-tailed. This investigation was a post hoc analysis of data from the CIGTS. SAS version 9.4 (SAS Institute) was used for all statistical analysis.
Of the 607 included patients, 334 (55.0%) were male and 337 (55.5%) were white, and the mean (SD) age was 58.0 (10.9) years. A total of 291 of 607 FEs (47.9%) were initially treated along with the SE at baseline (Figure 1). Of the 316 initially untreated FEs, 193 (61.1%) never received treatment and 123 (38.9%) were eventually treated for OAG. Of the 414 FEs that were treated (either initially or eventually), 224 (54.1%) were given medications, 21 (5.1%) received argon laser trabeculoplasty, and 169 (40.8%) received surgery. Although criteria for treatment of the FE were suggested, treatment was at the physician’s discretion (96 of 291 initially treated FEs [33.0%] were treated by physician discretion). Further, the randomized treatment was mostly but not always the initial treatment performed on FEs (31 of 414 treated FEs [7.5%] received nonrandomized treatment) (eFigure 1 in the Supplement).
Table 1 displays comparisons between FEs initially, eventually, and never treated on demographic and baseline clinical factors. Fellow eyes that were initially treated had significantly worse baseline clinical measures than FEs that were eventually or never treated over follow-up, including having lower baseline MD (mean [SD], −3.6 [3.4] dB vs −2.5 [2.3] dB vs −1.5 [1.7] db, respectively; P < .001), higher IOP (26.4 [5.0] mm Hg vs 23.2 [3.2] mm Hg vs 21.3 [3.5] mm Hg; P < .001), and larger vertical cup-disc ratio (0.64 [0.16] vs 0.57 [0.16] vs 0.51 [0.18]; P < .001). Study eye measures of MD and vertical cup-disc ratio showed similar patterns of decreasing clinical severity among participants whose SEs were initially, eventually, and never treated. Black patients (n = 231) were more likely than white patients and patients of other races (n = 376) to have their FEs initially treated (126 of 231 [54.5%] vs 165 of 376 [43.9%]), equally likely to have their FEs eventually treated (46 [19.9%] vs 77 [20.5%]), and less likely to have their FEs never treated (59 [25.5%] vs 134 [35.6%]) (P = .02). Participants randomized to medication were more likely to have their FE initially treated at baseline than participants randomized to surgery (176 of 307 [57.3%] vs 115 of 300 [38.3%], respectively; P < .001); over the follow-up period, participants randomized to medication were less likely to have their FEs eventually treated than participants randomized to surgery (38 of 307 [12.4%] vs 85 of 300 [28.3%]). Overall, there was no bias in FE treatment by randomization assignment (214 of 307 FEs [69.7%] treated from participants randomized to medication and 200 of 300 FEs [66.7%] treated from participants randomized to surgery; P = .42).
The within-eye slopes of MD over time showed a strong positive linear correlation between SEs and FEs in the initially treated group (r = 0.73; P < .001) and eventually treated group (r = 0.71; P < .001), whereas FEs that were never treated showed slopes of MD that were less strongly correlated with those from the SEs (r = 0.34; P < .001) (eFigure 2 in the Supplement). The slope of IOP over time between SEs and FEs showed a similar trend, albeit with correlations of decreased magnitude (initially treated FE: r = 0.57; P < .001; eventually treated FE: r = 0.24, P = .01; never treated FE: r = 0; P = .98). For categories of SE slopes of MD and IOP, descriptive statistics of corresponding FE slopes are displayed in Table 2 by FE treatment categories. For FEs that were initially and eventually treated, slopes of MD followed the pattern on average as the slopes from SEs. In FEs that were never treated, slopes of MD were on average close to 0 dB per year no matter the severity of the corresponding SE slopes. Similar results were seen for IOP slopes between SEs and FEs, although the strength of the association in FEs that were eventually treated was more modest.
The smoothed trend of MD over time before and after treatment in FEs that were eventually treated in relation to FEs that were initially and never treated is shown in Figure 2A and B, respectively. Initially treated FEs presented with the worst MD, which remained stable over time (mean, −3.5 dB). Never treated FEs showed relatively stable MD over time (mean, −1.1 dB). In FEs eventually treated for glaucoma, MD began to decline in the 2 years before treatment, and once treated, these FEs showed stable MD over the remaining follow-up at the level seen in initially treated FEs. For IOP (Figure 2C and D), initially and never treated FEs again showed relatively stable levels over time, with never treated FEs having higher IOP (mean, 20 mm Hg) than initially treated FEs (mean, 17 mm Hg). In FEs eventually treated for glaucoma, IOP showed an increase in the 2 years before treatment and stabilized after treatment at a level slightly below that of initially treated FEs (mean, 16 mm Hg).
Figure 3 displays Kaplan-Meier curves showing probabilities of FE treatment for glaucoma by time since randomization for both the entire sample (n = 607) and FEs not initially treated (n = 316). In all 607 FEs, the probability of being treated at baseline was 0.48. The cumulative probability of FE treatment at 1, 3, 5, and 7 years after randomization was 0.57, 0.61, 0.65, and 0.68, respectively. In the 316 initially untreated FEs, the probability of FE treatment at 1, 3, 5, and 7 years was 0.17, 0.25, 0.33, and 0.39, respectively.
Patient characteristics significantly associated with time to treatment of initially untreated FEs included older age and hypertension. For every 10-year increase in age, the hazard of FE treatment increased by 33% (HR, 1.33; 95% CI, 1.08-1.64; P = .007). Hypertension was associated with a 76% increase in the hazard of FE treatment (HR, 1.76; 95% CI, 1.16-2.67; P = .008). Time-varying FE characteristics that were significantly associated with treatment included higher IOP and cup-disc ratio. For every 1–mm Hg increase in IOP, the hazard of treatment increased by 24% (HR, 1.24; 95% CI, 1.20-1.29; P < .001); every 0.1 increase in vertical cup-disc ratio was associated with an increase of 40% in hazard of treatment (HR, 1.40; 95% CI, 1.23-1.58; P < .001). For these variables, the proportional hazards assumption was met.
The association of MD with the hazard of treatment in initially untreated FEs changed over time from randomization. Specifically, initially worse MD was associated with an increased hazard of FE treatment. This hazard further increased over time (interaction P = .04). For example, compared with an MD of 0 dB at baseline, a 1-dB worse MD was associated with a 12%, 18%, 24%, and 31% increased hazard of FE treatment at 1, 3, 5, and 7 years after randomization, respectively. A 3-dB worse MD was associated with a 42%, 65%, 92%, and 123% increased hazard of FE treatment at 1, 3, 5, and 7 years after randomization, respectively (eFigure 3 in the Supplement).
Although clinical measures of the SE were tested for associations with time to FE treatment, measures of the FE better predicted this outcome than the corresponding SE measures. Among the 14 clinical centers, differences in the hazard of treatment were found (P < .001). Three centers showed a significantly increased hazard of treatment (HRs ranged from 1.89 [95% CI, 1.01-3.54] to 6.18 [95% CI, 2.32-16.46] compared with the average center effect), and 1 showed a decreased hazard of treatment (HR, 0.32 [95% CI, 0.17-0.63]). Randomized treatment was not found to have a significant association with time to eventual FE treatment in the adjusted model (HR, 0.69; 95% CI, 0.44-1.08; P = .11).
Although ophthalmologists participating in the CIGTS selected an SE for randomization and glaucoma treatment, FE treatment could start at baseline or follow-up according to protocol requirements or by physician discretion. Nearly half of FEs began treatment at baseline, and 68% of FEs received treatment by 7 years. While OAG usually affects both eyes, it may present in each eye at different times. In the Early Manifest Glaucoma Trial,8 24% of participants had both eyes diagnosed with OAG at baseline, which is half of the 48% we found in participants of the CIGTS. This difference is likely because the Early Manifest Glaucoma Trial excluded people with advanced VF defects at diagnosis,8 whereas CIGTS did not exclude based on VF findings. If those with more advanced SE OAG are more likely to develop FE OAG, such a difference would result. The 48% of patients we identified with newly diagnosed bilateral OAG is similar to results found by Varma et al1 in the Los Angeles Latino Eye Study, which reported that 47% of participants had evidence of bilateral OAG in their study of prevalent OAG among Latino individuals. The study participants were older and had more severe MD compared with participants in the CIGTS, which would imply more bilateral disease than we observed. However, in the CIGTS, treatment of borderline FEs by physician discretion at baseline may have inflated the number with bilateral treatment.
We found a strong positive linear correlation between SEs and FEs in the slope of MD over time both for the initially treated group (r = 0.73; P < .001) and eventually treated group (r = 0.71; P < .001). For IOP, correlations were of decreased magnitude (initially treated FE: r = 0.57; P < .001; eventually treated FE: r = 0.24; P = .01). Chen and Park2 reported a correlation of 0.29 (P = .04) for VF progression between SEs and FEs in 48 patients observed with initially unilateral VF loss from OAG. In a subsequent study of 152 patients undergoing treatment for bilateral OAG, Chen4 reported a between-eye correlation for VF progression of 0.35 (P < .001).
We identified 316 initially untreated FEs, among which 33% were treated by 5 years. Chen and Park2 identified VF progression in only 3 FEs among 48 patients with initially unilateral VF loss from OAG. Their Kaplan-Meier estimate of FE progression at 5 years was 7.2%, which is much lower than the 33% with FE treatment we report. This difference relates to the more stringent criteria imposed by Chen and Park,2 who required VF progression in the FE, whereas our risk relates to initiation of treatment for a number of possible reasons, including increased IOP or worsening VF.
The baseline predictors of subsequent FE treatment for the 316 FEs not treated at baseline included clinical measures common to glaucoma assessment—higher IOP, worse MD, and larger vertical cup-disc ratio. The associations of these factors with FE treatment for glaucoma are consistent with previous studies addressing risk factors for glaucoma,9-14 as is older age. The association of a modifiable factor, hypertension, with FE treatment for glaucoma lacks consistency in prior reports. Quigley et al9 did not find hypertension to be associated with development of glaucomatous VF loss among participants with ocular hypertension, whereas results from the Blue Mountains Eye Study15 showed a 50% increased risk of OAG among people with hypertension. A caveat to our hypertension finding is our lack of blood pressure measurements, which prevented assessment of the reported association of low diastolic perfusion pressure (the difference between diastolic blood pressure and IOP) with risk of OAG.10,16,17
Among factors we found were not associated with FE treatment for glaucoma were 2 that have been found to be associated with an increased risk of glaucoma in prior studies: family history of glaucoma18,19 and black race.20 An immediate family history of glaucoma was reported by 201 participants in the CIGTS and 231 reported their race as black, so we had a robust sample to evaluate these factors. Family history of glaucoma lacked predictive significance for FE treatment in univariate and multivariable analyses. Although we found that black patients were more likely to have their FEs initially treated than white patients and patients of other races (55% vs 44%, respectively) and less likely to have their FEs never treated than white patients and patients of other races (26% vs 36%), we did not find a significant association of race with time to FE treatment in initially untreated FEs. Black patients were found to have a significantly higher prevalence of hypertension than white patients and patients of other races (110 of 231 [47.6%] vs 115 of 376 [30.6%], respectively) and significantly larger vertical cup-disc ratio measurements (mean [SD], 0.61 [0.16] vs 0.57 [0.19]). After adjusting for hypertension and vertical cup-disc ratio, race no longer showed a significant association with time to FE treatment. A similar result was found in the Ocular Hypertension Treatment Study,11 where the effect of race on onset of primary OAG was significant in univariate analysis but did not show an independent association with the onset of primary OAG after adjustment for vertical cup-disc ratio and central corneal thickness.
Data collected on participants of the CIGTS for investigating time to FE treatment provided several strengths. Participants were newly diagnosed with OAG, followed up with every 6 months, and examined by certified staff according to a detailed protocol, and data on both eyes were collected. The longitudinal nature of our study permitted us to evaluate associations that took into account intervisit dependencies in variables, such as IOP and MD. However, our study had limitations. The time of day at which IOP measurements occurred was not standardized. Our FE treatment information cannot be claimed to reflect a diagnosis of OAG, as ophthalmologists in the CIGTS were permitted discretion in when to initiate treatment of the FE. This discretion could mean some FEs were treated that had ocular hypertension and some were not treated that had early OAG.
Among the characteristics that were significantly associated with an increased risk of FE treatment, 2 modifiable factors were identified: hypertension and higher IOP. The substantial incidence of treatment for the FE of patients in whom the treating physician decided only 1 eye needed treatment initially, and the synchrony between trajectories of clinical change in initially treated and untreated eyes, support the need for close follow-up of such patients. Attention to not only IOP but also hypertension may reduce the risk of FE treatment.
Accepted for Publication: May 25, 2018.
Corresponding Author: David C. Musch, PhD, MPH, Department of Ophthalmology and Visual Sciences, Medical School, University of Michigan, 1000 Wall St, Ann Arbor, MI 48105-1912 (firstname.lastname@example.org).
Published Online: August 9, 2018. doi:10.1001/jamaophthalmol.2018.3274
Author Contributions: Dr Musch had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Gillespie, Musch.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Niziol, Musch.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Niziol, Gillespie.
Obtained funding: Gillespie, Musch.
Study supervision: Musch.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Musch is a member of independent data and safety monitoring boards for Glaukos Corp, InnFocus, and Ivantis. No other disclosures were reported.
Funding/Support: This study was supported by research grant R21 EY020912 from the National Eye Institute and by a departmental grant from Research to Prevent Blindness. Dr Musch is a recipient of the Lew R. Wasserman Merit Award from Research to Prevent Blindness.
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Meeting Presentation: This paper was presented in part at the 2013 annual meeting of the Association for Research in Vision and Ophthalmology; May 7, 2013; Seattle, Washington.
Create a personal account or sign in to: