Canadian Glaucoma Study: Title and subTitle Break2. Risk Factors for the Progression of Open-angle Glaucoma FREE

Clinical trials have confirmed the importance of intraocular pressure (IOP) in the development¹ of open-angle glaucoma and its progression.² Furthermore, treatment to reduce IOP was shown to reduce the risk of development of glaucoma from 9.5% to 4.4% in the Ocular Hypertension Treatment Study,³ and to reduce progression of clinically manifest glaucoma from 27% to 12% in the Collaborative Normal Tension Glaucoma Study (CNTGS)⁴ and from 62% to 45% in the Early Manifest Glaucoma Trial (EMGT).⁵ However, within the periods of observation, there were patients whose glaucoma continued to progress despite the prescribed therapy for IOP lowering and those whose glaucoma remained stable despite receiving no treatment.^{4 - 5}

Clinical and scientific evidence suggests the existence of ocular and systemic factors in addition to, or even independent of, IOP that may be responsible for the development and progression of glaucoma. Despite several well-executed clinical trials, there remains little consensus on the identity of these factors. The CNTGS showed that women were 1.9 times more likely than men to have progression, and that self-reported migraine increased the progression risk by 2.5 times. The EMGT was unable to confirm the importance of either sex or migraine. On the other hand, the EMGT, but not the CNTGS, found age to be a significant factor for progression. The Ocular Hypertension Treatment Study found that age, but not sex or self-reported migraine, was a significant factor for the conversion to glaucoma. Differences in study populations, disease stage, study procedures, and whether patients were treated may explain these differences; however, the identification of risk factors besides IOP remains elusive.

The Canadian Glaucoma Study (CGS) is a multicenter interventional cohort study. Its primary objective was to determine baseline demographic and systemic risk factors, including susceptibility to peripheral vasospasm, and hematologic, coagulation, and immunopathological variables associated with the progression of visual field damage in open-angle glaucoma under an interventional protocol for IOP control. The protocol was established in this manner to minimize the influence of IOP maintenance and variability among participants such that other risk factors could be better elucidated. Its secondary objectives included establishing the relationship between structural and functional progression, determining the utility of the newer techniques of confocal scanning laser tomography and short-wavelength automated perimetry, and providing practical guidelines and paradigms for the follow-up of patients with glaucoma with respect to diagnostic tests. The last patient examination was in 2005. The purpose of this article is to describe the risk factors associated with glaucomatous visual field progression.

The study participants and procedures have been described elsewhere in detail.⁶ The following is a brief synopsis. The CGS is a multicenter Canadian study involving 5 hospital-based university departments and is registered with the ClinicalTrials.gov Protocol Registration System (identifier NCT00262626). It was approved by the research ethics committee of each participating center, and informed consent was obtained from each patient. After enrollment and documentation of baseline demographic, systemic, and ocular measures, patients were followed up with a standardized treatment protocol for IOP control and tested every 4 months with functional and structural tests. Patients with visual field progression were subjected to a further reduction in IOP.⁶

Patients were either newly or previously diagnosed as having open-angle glaucoma. Patients with pseudoexfoliation and pigmentary glaucoma were also enrolled if they satisfied the eligibility criteria. Inclusion criteria were (1) best-corrected visual acuity of 6/10 or better with the Early Treatment Diabetic Retinopathy Study chart; (2) photographically documented glaucomatous optic disc changes; (3) glaucomatous visual field changes including localized visual field defects, mean deviation better than −10 dB, and a positive glaucoma hemifield test; and (4) nonoccludable anterior chamber angles. Exclusion criteria were (1) significant nonglaucomatous ocular disease; (2) chronic nonglaucomatous ocular medication; (3) systemic diseases with known effects on the visual field; (4) greater than 6.00 diopters (D) (equivalent sphere) of myopia or hyperopia, or greater than 2.50 D astigmatism; and (5) previous incisional glaucoma surgery.

There were 2 baseline visits separated by 7 to 10 days. At the first visit, demographic, ocular, and medical histories were recorded followed by a full eye examination, standard automated perimetry (SAP) with the 30-2 full threshold program of the Humphrey Field Analyzer (Carl Zeiss Meditec, Dublin, California), the short-wavelength automated perimetry 30-2 program of the Humphrey Field Analyzer, and confocal scanning laser tomography with the Heidelberg Retina Tomograph (Heidelberg Engineering GmbH, Heidelberg, Germany). At the second baseline visit, SAP and short-wavelength automated perimetry were repeated. Both baseline SAP visual fields had to meet the eligibility criteria. Finger blood flow measurements were made with laser Doppler flowmetry using a protocol to determine susceptibility to peripheral vasospasm.⁷ Patients were defined as vasospastic if the ratio of maximum (after exposure to heat) to minimum (after exposure to cold) blood flow exceeded 7.⁷ Blood samples were obtained at baseline, and tests for clinical chemistry, general hematology, coagulation, and immunopathology were conducted. A full list of the blood tests is given in the previous CGS publication.⁶ Finally, stereo optic disc photography was performed.

Follow-up visits occurred at 4-month intervals. Eye examination, SAP, short-wavelength automated perimetry, and confocal scanning laser tomography were conducted at every visit (3 examinations annually), whereas finger blood flow measurements and stereo optic disc photography were conducted at 32- and 28-month intervals, respectively (3 times over 5 years).

Newly diagnosed patients were required to have a 30% or greater reduction in IOP, the result of which was defined as the target IOP. Previously diagnosed patients had a physician-defined target IOP based on their history and previous rate of change of the visual field and optic disc. If visual field progression was confirmed, patients were required to have an additional 20% or greater reduction from the previous target IOP, the result of which was the new target IOP. A stepwise treatment protocol⁶ from topical monotherapy, adjunct topical therapy, argon laser trabeculoplasty, and/or systemic carbonic anhydrase inhibitors and trabeculectomy was used to achieve the target IOP. Only IOP measurements obtained after attaining target IOP were included in the analysis.

Glaucomatous progression was defined on the basis of visual field change determined with SAP using the glaucoma change probability analyses.⁸ Progression was suspected when 8 or more locations in the total deviation change probability map, with 4 or more clustered locations in a single hemifield, were flagged. A first confirmation examination was then performed to verify the change within 7 to 10 days. Progression was confirmed when there was an overlap of 4 or more locations, with at least 2 locations clustered within a single hemifield. If progression was not confirmed at this examination, a second confirmation examination was conducted within 7 to 10 days. Hence, 2 of 3 examinations had to confirm visual field progression. The visual field examinations were reviewed by the Visual Field Reading Committee, and its decision was relayed to the referring center or physician.

The original sample size estimate was based on the hypothesis that patients with peripheral vasospasm have a lower progression rate compared with nonvasospastic patients.⁶ The estimated sample size of 220 patients was based on the following assumptions: (1) a progression rate of 20% in the vasospastic group and 35% in the nonvasospastic group; (2) an equal number of vasospastic and nonvasospastic patients; (3) a 2-sided α of .05; (4) a β of .20, and (5) a cumulative 5-year attrition rate of 35%. The analysis was based on the study eye, which was chosen by a random selection technique at the beginning of the study if both eyes were eligible.

The end point occurred at the first visual field progression and, in patients who had progression, only data up to that point were used. In the newly diagnosed patients, IOP measurements obtained before the target IOP was attained were not included in the analyses. To explore the potential effect of differences in exposure variables leading to bias in outcomes in patients completing and not completing 5 years of follow-up, we compared the difference in all baseline variables in these 2 groups of patients. In addition, we compared IOP in the follow-up, both in a repeated-measures analysis of variance and at 1, 2, 3, and 4 years of follow-up in patients completing and not completing 5 years of follow-up.

Risk factors for progression were explored with Kaplan-Meier survival analyses with the log-rank test for the univariate analyses. We also explored whether progression rates were different among centers. Because IOP was the only time-dependent variable, it was analyzed as a covariate in the survival analysis. The mean and standard deviation of IOP in the follow-up were also explored. Variables were entered into the Cox proportional hazards model only if P < .10 in a forward stepwise analysis and if the hazards were judged to be proportional by examining the negative-log plots of the survivor functions. All 2-way interaction terms were evaluated in the model and were included if the −2 × log likelihood score indicated a better model fit. Hazard ratios were derived from the final model, which included only the significant terms. These analyses were performed with a commercial software package (SAS, version 9.0; SAS Institute Inc, Cary, North Carolina).

The study population and baseline characteristics have been described in a previous publication.⁶ Briefly, the CGS enrolled 258 patients (131 men and 127 women) with glaucoma. The median (interquartile range [IQR]) age at enrollment was 65.0 (55.3-72.0) years. There were 23 patients (8.9%) with pseudoexfoliation glaucoma and no patients with pigmentary glaucoma. Forty-six patients (17.8%) were newly diagnosed and 212 (82.2%) had been previously diagnosed. At baseline, patients had early to moderate visual field damage with a median (IQR) mean deviation of the 2 baseline visual fields of −4.04 (−5.82 to −2.40) dB. There were no statistically significant sex differences at baseline in any ocular measures studied, except that men had a higher median untreated IOP compared with women (26.0 and 25.0 mm Hg, respectively).⁶ Of the systemic baseline measures, women had a 3-fold higher prevalence of thyroid disease and a 2-fold higher prevalence of migraine, while there were almost 2.5 times more diabetic men than women.⁶ The enrollment period spanned 5 years; however, 50% and 80% of the enrollment occurred within 2.0 and 3.4 years, respectively. In newly diagnosed patients, target IOP was obtained with topical medications usually within 3 office visits, 1 to 3 weeks apart.

The median (IQR) follow-up was 5.3 (3.7-7.0) years, with 167 patients (64.7%) completing 5 or more years and 67 (26.0%) completing 7 or more years of follow-up. Of the 91 patients (35.3%) not completing 5 years of follow-up, 26 (29%) were lost to follow-up, 18 (20%) withdrew because of poor health, 14 (15%) withdrew without stating a reason, 12 (13%) found the tests too difficult or time consuming, 11 (12%) relocated, 5 (5%) died, 4 (4%) were no longer able to travel to the study site, and 1 (1%) withdrew because of pregnancy. The cumulative overall progression rate determined by survival analysis at 2, 3, 4, and 5 years was 11.1% (95% CI, 7.7%-16.0%), 17.9% (95% CI, 13.4%-23.7%), 22.0% (95% CI, 17.0%-28.2%), and 30.7% (95% CI, 24.7%-37.6%), respectively. Seventy-one patients reached the end point. Of these, visual field progression was confirmed in 45 patients (63%) at the first confirmation examination and 26 patients (37%) at the second confirmation examination. There was no difference in the survival times of newly and previously diagnosed patients (P = .42).

Descriptive baseline statistics of all variables explored in this study in the no-progression and progression groups are shown in Table 1. Univariate survival analyses disclosed 7 variables that were significantly associated with progression. Of the demographic variables, only female sex and higher baseline age were associated with progression. Abnormalities in results of 2 baseline blood tests, namely anticardiolipin antibody (ACA) and red blood cell distribution width, were significant. Finally, of the ocular variables, higher mean deviation (better visual field) and higher mean and lower standard deviation of IOP in the follow-up (before confirmed progression and treatment intervention) were related to progression. The median (IQR) standard deviation and range of IOP were 2.07 (1.5-2.8) mm Hg and 6.0 (4.0-9.0) mm Hg, respectively.

Five variables (abnormal ACA level, higher baseline age, higher mean IOP and lower standard deviation of IOP in the follow-up, and female sex) qualified for entry in the Cox proportional hazards model. Four of these were independently predictive of visual field progression (Table 2). None of the 2-way interaction terms was significant. Abnormal baseline ACA level yielded a hazard ratio of 3.86, indicating that patients with a positive ACA test result were almost 4 times as likely as those with a negative result to have progression.While highly predictive and statistically significant (eFigure 1), only 11 (5.4%) of the 204 patients tested had an abnormal ACA level. Higher baseline age was also highly significant, with every year of age adding a 4% increased independent risk of progression(Table 2).

Mean IOP in the follow-up was highly significantly associated with progression, with a hazard ratio of 1.19, indicating that every 1 mm Hg added a 19% increased risk of progression(Table 2). Most patients had a mean IOP in the follow-up of 14 to 19 mm Hg (Figure 1). The 33.3rd and 66.7th percentiles of the distributions were 15.5 and 17.0 mm Hg, respectively. The survival curves for the whole group divided into equal thirds based on mean IOP in the follow-up are shown in Figure 2. Finally, women were almost twice as likely as men to have progression (Table 2, Figure 3).

There were no differences between patients who did and did not complete 5 years of follow-up in the 8 continuous baseline variables (shown in Table 1; P > .05, group ttest or Mann-Whitney test). Of the 28 categorical variables shown in Table 1, patients who did not complete 5 years of follow-up had a higher proportion of individuals with abnormal random glucose level (P = .02, Fisher exact test), red blood cell count (P = .04), and hematocrit (P = .03). None of the other 25 variables was significantly different between these groups of patients. Intraocular pressure in the follow-up was not significantly different in patients completing and not completing 5 years of follow-up in both the repeated-measures analysis (P = .91) or mean values at 1, 2, 3, and 4 years of follow-up (P > .28, group ttest), suggesting that there was no bias with respect to IOP exposure in these 2 groups of patients.

Although patients with peripheral vasospasm tended to have a lower progression rate compared with nonvasospastic ones, this difference failed to reach statistical significance in a univariate analysis (P = .14). There were significantly more vasospastic women than vasospastic men (52.1% vs 33.6%; P = .004, χ²test). The tendency of vasospastic patients to have a lower progression rate persisted when the data were stratified by sex. Because men had a lower progression rate than women, there was a significant difference between vasospastic men, who had the lowest progression rate, and nonvasospastic women, who had the highest progression rate (eFigure 2) (P = .03).

The primary objective of the 2 major randomized glaucoma clinical trials with nontreated arms, namely the CNTGS⁴ and the EMGT,⁵ was to determine whether IOP lowering had a beneficial effect on the disease course. The CNTGS derived participants from clinical samples of patients with glaucoma who had statistically normal IOP, whereas the EMGT drew newly diagnosed patients predominately from a large population screening program and excluded patients only if their mean IOP was greater than 30 mm Hg (both eyes) or greater than 35 mm Hg (at least 1 eye). Despite the different study designs and populations, both studies showed that IOP lowering had a beneficial effect across the IOP range in glaucoma. Although these studies and others,^{9 - 11} either cross-sectional or longitudinal, have been used to identify non–IOP-dependent factors related to progression, they were not intended to answer these questions and may have been limited by issues related to post hoc analyses. The CGS was primarily designed to address whether, in an interventional protocol that minimized IOP variation among patients (hence theoretically reducing the potential of IOP as a confounding variable), peripheral vasospasm and a variety of blood tests that provide an insight into the general vascular health of patients were predictive of visual field progression in glaucoma. We did not analyze any data after the first end point. The effect of treatment interventions on progression will be the subject of a future report.

The CGS identified 4 statistically significant variables that were independently predictive of progression, namely positive ACA test result, higher baseline age, higher mean follow-up IOP (before confirmed progression and treatment intervention), and female sex. Equally important, patients with diabetes, hypertension, cardiovascular disease, migraine, and pseudoexfoliation glaucoma did not have a higher progression rate;however, the CGS had limited statistical power to elucidate factors with low prevalence. We were also not able to demonstrate that a variety of baseline hematologic and coagulation indexes influenced progression in treated glaucoma.

In the CGS, the distribution of mean IOP had a relatively low dispersion, with the central 50% of the values falling between 15 and 18 mm Hg and the central 75% between 14 and 19 mm Hg. We excluded IOP measurements once progression had occurred, since treatment intervention would significantly affect the mean and standard deviation of IOP in the follow-up. In the untreated arm of the EMGT, every 1–mm Hg decrease in IOP equated to around a 10% decrease in risk of progression.² Similarly, the risk of conversion to open-angle glaucoma from ocular hypertension was around 10% for every 1–mm Hg increase in IOP.¹ Despite our best attempts to reduce the confounding role of IOP, mean IOP before the first end point was a powerful predictor of progression, with approximately a 20% increase in risk of progression for every 1–mm Hg increase in mean IOP. The CGS is therefore in agreement with every major clinical trial in glaucoma over the past decade in confirming the potent effect of IOP on the progression of glaucoma across the spectrum of IOP.^{3 - 5 ,12} However, a substantial number of patients with mean follow-up IOP in the lowest tertile had progression, and there was no clear-cut IOP level below which progression did not occur. Equally important, while the CGS was not designed to address this issue, a notable proportion of untreated patients do not show progression by study criteria^{4 - 5} or have very slow progression,¹³ perhaps without a significant effect on quality of life. Hence, while we and others have shown the potency of IOP in dictating glaucoma progression and the notion that every 1–mm Hg decrease translates into a significant risk reduction, the risk reduction probably cannot be 100%. After exhausting all other options, achieving low target IOP by surgery should probably be counterbalanced with the potential of serious surgical complications, such as hypotony, leaky blebs, and endophthalmitis.

In the CGS, 43% of the patients were classified as vasospastic, while, in a previous study using the same methods and classifications,⁷ 65% of patients with normal-tension glaucoma and 26% of controls were categorized as vasospastic. In a cross-sectional study,⁹ it was postulated that vasospastic patients had a more IOP-dependent disease, suggesting that a higher IOP would exacerbate the neuropathy because of an inability of the optic nerve head circulation to regulate blood flow in the face of lower ocular perfusion pressure. The inference from this work was that vasospastic patients would respond better to IOP reduction. The CGS was designed to address this hypothesis, and, on the whole, we were unable to find convincing evidence to support it. There was a tendency for vasospastic patients to have a lower progression rate than nonvasospastic patients; more specifically, when subdivided by sex, vasospastic men had a significantly better outcome than did nonvasospastic women. However, despite the significantly higher prevalence of vasospasm in women, the potentially protective effects of vasospasm in treated glaucoma were likely mitigated by the higher progression rate in women.

Female sex was a predictor of progression, with almost twice as many women as men showing progression. This finding corroborates those of the CNTGS but not the EMGT. The reasons for these discrepancies are not obvious, but they may have to do with genetic and environmental differences between the populations, as well as sample selections, study procedures, and analysis methods that may or may not have disclosed differential effects of treatment between men and women. Postmenopausal women may be susceptible to hormonal factors that predispose them to a higher risk of progression. Evidence from the Rotterdam Eye Study¹⁴ and the Blue Mountains Eye Study¹⁵ indicate that early menopause was significantly associated with a higher prevalence of glaucoma, suggesting that endogenous estrogen may have a protective effect against glaucoma in women. Aging has different effects on retrobulbar circulation in men and postmenopausal women not receiving hormone therapy,¹⁶ whereas estrogen replacement decreases ophthalmic artery resistance^{17 - 18} and plasma viscosity.¹⁸ The relevance of these findings for glaucoma progression is unclear. If lack of estrogen were a strong factor in glaucoma, then presumably the prevalence of glaucoma in postmenopausal women compared with age-matched men would be higher; however, population-based prevalence studies do not consistently support this hypothesis.^{19 - 25}

Patients with an abnormal ACA level were almost 4 times as likely as those with a normal ACA level to have progression. Although the hazard ratio for this variable was remarkably large and highly statistically significant, only a small minority of the patients in the CGS had a positive ACA test result and substantially more patients with a negative ACA result than those with a positive one had progression. Anticardiolipin is one of the antiphospholipid antibodies found to have elevated levels in patients with acquired prothrombotic syndromes.²⁶ The ACA levels can also be elevated in miscarriage, systemic lupus erythematosus, ischemic stroke, and myocardial infarction.²⁷ While there is interest in the role of ACA in rheumatology, immunology, obstetrics, and cardiology, there is considerable debate as to whether elevated ACA levels are the cause or the effect of a variety of clinical disorders.²⁸ There are 2 published reports on ACA levels and glaucoma; however, they reported contrary findings^{29 - 30} and, because of their cross-sectional design, could not address the issue of causality. Since an abnormal ACA level at baseline was highly predictive of progression in the CGS, this topic merits further investigation, specifically in phenotyping patients with positive ACA results and determining the ocular and systemic factors relevant for glaucoma progression.

There are some limitations of the CGS. Thirty-five percent of the patients did not complete 5 years of follow-up. We were unable to find systematic evidence that those who completed at least 5 years of follow-up had a different exposure to potential risk factors than those who did not. Using 2 different analyses, we determined that the IOP risk among these 2 groups was not significantly different. However, 3 of the 36 variables (shown in Table 1) were significantly different in that the group not completing 5 years of follow-up contained a higher proportion of patients with abnormal glucose level, red blood cell count, and hematocrit. These findings may indicate a subgroup of patients with biochemical or hematologic disorders not completing the follow-up. These variables were not identified as significant risk factors for progression. Nonetheless, we cannot rule out the potential of other noninvestigated measures having a differential influence in patients completing and not completing the 5-year follow-up.

Progression end points based on visual field criteria have been subject to considerable discussion.^{31 - 34} Because of visual field variability and the absence of a reference standard for visual field progression, end point criteria have to be chosen with adequate specificity to minimize false-positive end points and, at the same time, maintain good sensitivity. Although criteria similar to those used by the CGS have good performance characteristics,^{33 ,35} we acknowledge the inherent limitation of perimetric end points. The CGS contained a mixture of newly and previously diagnosed patients. In the latter group, we were unable to monitor the number of treatment interventions before reaching the target IOP in the study. All observations in the CGS occurred after the target IOP was reached. While we cannot rule out the possibility that newly and previously diagnosed patients may have been exposed to different risks for progression, the progression rates in these 2 groups of patients were not different.

In summary, using a prospective interventional protocol, the CGS identified abnormal ACA level, higher mean IOP in the follow-up, higher baseline age, and female sex as significant independent risk factors for visual field progression in glaucoma.

Correspondence:Balwantray C. Chauhan, PhD, Department of Ophthalmology and Visual Sciences, Dalhousie University, Second Floor, Centennial Bldg, Queen Elizabeth II Health Sciences Centre, Halifax, NS, Canada B3H 2Y9 (bal@dal.ca).

Submitted for Publication:November 5, 2007; final revision received March 26, 2008; accepted March 30, 2008.

Financial Disclosure:None reported.

Funding/Support:This study was supported by the E. A. Baker Foundation of the Canadian National Institutes for the Blind (1994-2005), by the Glaucoma Research Society of Canada (2003-2005), and by unrestricted grants from Allergan Canada (2003-2005), Merck Frosst Canada (2003-2005), and Pfizer Canada (2003-2005).

This article was corrected online for error in data on 8/11/2008, prior to publication of the correction in print.