Inclusion and Exclusion Criteria

Study patients had previously diagnosed low-pressure glaucoma that fulfilled the following eligibility criteria: all known untreated IOP ≤21 mm Hg; open iridocorneal angles; at least 2 reproducible visual fields with glaucomatous defects in 1 or both eyes on automated perimetry (Humphrey Field Analyzer; Carl Zeiss Meditec, Inc, Dublin, California, USA), with the location of the field defect being consistent with the photographic appearance of the optic nerve head; and age ≥30 years. To determine eligibility based on IOP, all patients receiving IOP-lowering treatment underwent a 4-week washout without therapy. Baseline IOP (measured with a calibrated Goldmann applanation tonometer) had to be ≤21 mm Hg in both eyes with <5 mm Hg difference between the eyes on an office diurnal curve (8:00 AM, 10:00 AM, 12:00 PM, 4:00 PM) assessed prior to randomization.

Ocular exclusion criteria included the following: a history of IOP >21 mm Hg in the patient record, best-corrected visual acuity worse than 20/40 in either eye, a history of angle closure or an occludable angle by gonioscopy, prior glaucoma incisional surgery, inflammatory eye disease, prior ocular trauma, diabetic retinopathy or other diseases capable of causing visual field loss or optic nerve deterioration, extensive glaucomatous visual field damage with a mean deviation worse than −16 decibels (dB), or a clinically determined threat to central fixation in either eye. Systemic exclusion criteria included a resting pulse <50 beats/minute; severe or uncontrolled cardiovascular, renal, or pulmonary disease that would preclude safe administration of a topical beta-adrenergic antagonist; and a prior myocardial infarction or stroke. Continuation of systemic medications that could affect IOP was allowed as long as the doses remained constant throughout the trial.

Randomization, Treatment, and Masking

Patients were randomly assigned to receive monotherapy with either brimonidine tartrate 0.2% containing benzalkonium chloride (Alphagan; Allergan, Inc, Irvine, California, USA) or timolol maleate 0.5% containing benzalkonium chloride (Timoptic; Merck & Co Inc, West Point, Pennsylvania, USA) twice daily in both eyes, including the morning before each visit. To allow for higher patient attrition in the brimonidine group attributable to an expected rate of adverse events of approximately 20%, randomization and delivery of medications (provided by Allergan, Inc, Irvine, California, USA) to the sites were stratified in blocks of 7 (4 to brimonidine and 3 to timolol). The randomization list was maintained and masked study medications were provided in new 10-mL white bottles labeled with the assigned randomization number directly to the clinical centers by an independent pharmacy (Fountain Valley Cancer Center Pharmacy, Fountain Valley, California, USA). Ocular treatment other than the study medication was not permitted. Investigators, patients, and the visual field reading and coordinating centers were all masked to patient assignment.

Endpoints requiring discontinuation from the study included: treated IOP >21 mm Hg that was repeated within 1 month, safety concern as judged by the treating physician, symptomatic ocular allergic adverse events (hyperemia, pruritus, stinging, and/or conjunctival folliculosis) requiring medication cessation, retinal events that could alter visual acuity or visual field (eg, age-related macular degeneration), the occurrence of systemic (eg, respiratory or cardiovascular) adverse events that prevented the administration of topical timolol, nonocular intolerable events associated with topical brimonidine (eg, xerostomia, fatigue, drowsiness), or if the patient moved or declined continued participation. Collection of data from discontinued patients ceased at their final study visit. Data up to this point were included in the analysis, but discontinued patients were no longer followed as part of the study.

Study Visits

Patients were examined at 1 and 4 months after initiation of treatment. Subsequent visits were at 4-month (±2 weeks) intervals. Pre- and post-randomization morning visits recorded the following: ocular and systemic history, blood pressure, pulse, corrected visual acuity, IOP, slit-lamp examination, and optic disc evaluation for cup-to-disc ratio and the presence or absence of disc hemorrhage. Gonioscopy and stereoscopic optic disc photographs were performed annually. Full-threshold standard achromatic perimetry (Humphrey 24-2) visual field was performed at 4-month intervals according to protocol guidelines.

Outcome Measures

The main outcome measure was visual field progression in 1 eye as determined by pointwise linear regression. Visual field progression analysis was performed using Progressor software (Medisoft, Inc, Leeds, UK). Visual field analyses were performed by an independent reading center (Devers Eye Institute, Legacy Health System, Portland, Oregon, USA) masked to the treatment assignment. Linear regression of the sensitivity (in dB) was performed at each test location to obtain the rate of change at that location, based on all fields up to and including the last examination. Default Progressor criteria were used to define a significant negative slope (at least −1.0 dB/year for inner points and −2.0 dB/year for edge points) at the P < 5% level. Edge points for the Humphrey 24-2 field included the 2 outer nasal locations, 1 above and 1 below the horizontal. Criteria for visual field progression required confirmation at the next 2 examinations of a significant negative slope at the same 3 or more test locations. For eyes not reaching a progression endpoint, we used the time to last follow-up visit (either the final scheduled visit or the censored one). Progressing visual field locations were not required to be contiguous. Because of these criteria, progression could not be determined until 5 visual field tests had been collected (the 16-month visit). Consequently, for the purposes of this manuscript, only those eyes with enough visual field tests to determine a progression outcome (yes/no) were included in the analysis (1 at baseline plus 4 tests following randomization).

We examined clinical characteristics that predicted the development of visual field progression. The following demographic and ocular parameters were investigated: age at baseline, sex, family history of glaucoma, central corneal thickness (CCT), cup-to-disc ratio (estimated during slit-lamp examination), refractive error spherical equivalent (SE), and lens status. Pre-randomization data collected and investigated for risk assessment were: IOP mean, peak, and fluctuation during the diurnal curve; baseline pulse; systolic (SBP) and diastolic blood pressure (DBP); presence of systemic comorbidities (migraine, Raynaud phenomenon, systemic hypertension, and diabetes mellitus); and use of systemic medications (categorized as systemic antihypertensives, systemic beta-blockers, and antidiabetic agents). Post-randomization variables consisted of series length (ie, follow-up time), detection of at least 1 optic disc hemorrhage on stereophotographs any time during follow-up, and mean, peak, and fluctuation of IOP and blood pressure during follow-up. For numerical variables, the mean was calculated by averaging all values recorded during the follow-up period. Fluctuation was defined as the standard deviation (SD) of all measurements in the same interval. Mean ocular perfusion pressure (MOPP) was estimated by the equation MOPP = 2/3 × [DBP + 1/3 × (SBP – DBP)] – IOP. Data on systemic comorbidities and medications were obtained from participants’ self-reports.

Statistical Analysis

Descriptive statistics are presented with frequency tables and graphs, whereas estimates of center and dispersion are described as mean and SD, respectively. Cox proportional hazards model was used to investigate the risk of progression (progression: yes or no, based on the progression criteria described above) based on follow-up time (progression endpoint or data censoring for progressing and nonprogressing eyes, respectively). Variables with P < .10 in the univariate model were entered in a multivariate model. Generalized estimating equations were used to control for inter-eye relationships. Since one of the assumptions of regression analyses is that the predictors should not be strongly and significantly correlated with one another (ie, no colinearity), if pairs of predictor variables had moderate to strong significant correlation, the variable with the most significant P value in the univariate analysis was entered in the multivariate model. Cox proportional hazards multivariate model was performed using a backward elimination approach based on likelihood ratios. Variables in the saturated multivariate model with P > .10 were subsequently removed and variables were entered in the model if P < .05. Alpha level was set at 5% (2-sided) and computerized statistical analyses were performed using SPSS for Windows (version 16.1; IBM SPSS Statistics Inc, Armonk, New York, USA).