Latanoprost for open-angle glaucoma (UKGTS): a randomised, multicentre, placebo-controlled trial

Background Treatments for open-angle glaucoma aim to prevent vision loss through lowering of intraocular pressure, but to our knowledge no placebo-controlled trials have assessed visual function preservation, and the observation periods of previous (unmasked) trials have typically been at least 5 years. We assessed vision preservation in patients given latanoprost compared with those given placebo. Methods In this randomised, triple-masked, placebo-controlled trial, we enrolled patients with newly diagnosed open-angle glaucoma at ten UK centres (tertiary referral centres, teaching hospitals, and district general hospitals). Eligible patients were randomly allocated (1:1) with a website-generated randomisation schedule, stratiﬁ ed by centre and with a permuted block design, to receive either latanoprost 0·005% (intervention group) or placebo (control group) eye drops. Drops were administered from identical bottles, once a day, to both eyes. The primary outcome was time to visual ﬁ eld deterioration within 24 months. Analyses were done in all individuals with follow-up data. The Data and Safety Monitoring Committee (DSMC) recommended stopping the trial on Jan 6, 2011 (last patient visit July, 2011), after an interim analysis, and suggested a change in primary outcome from the diﬀ erence in proportions of patients with incident progression between groups to time to visual ﬁ eld deterioration within 24 months. This trial is registered, number ISRCTN96423140. Findings We enrolled 516 individuals between Dec 1, 2006, and March 16, 2010. Baseline mean intraocular pressure was 19·6 mm Hg (SD 4·6) in 258 patients in the latanoprost group and 20·1 mm Hg (4·8) in 258 controls. At 24 months, mean reduction in intraocular pressure was 3·8 mm Hg (4·0) in 231 patients assessed in the latanoprost group and 0·9 mm Hg (3·8) in 230 patients assessed in the placebo group. Visual ﬁ eld preservation was signiﬁ cantly longer in the latanoprost group than in the placebo group: adjusted hazard ratio (HR) 0·44 (95% CI 0·28–0·69; p=0·0003). We noted 18 serious adverse events, none attributable to the study drug. Interpretation This is the ﬁ rst randomised placebo-controlled trial to show preservation of the visual ﬁ eld with an intraocular-pressure-lowering drug in patients with open-angle glaucoma. The study design enabled signiﬁ cant diﬀ erences in vision to be assessed in a relatively short observation period. placebo-controlled trial to assess the beneﬁ t of topical medical treatment (eye drops) for reduction of loss of vision in patients with open-angle glaucoma. Our ﬁ ndings provide strong evidence for the vision-preserving beneﬁ t of lowering of intraocular pressure, supporting evidence from previous randomised trials that were not masked or placebo-controlled. The study also provides evidence of the vision-preserving beneﬁ ts of topical prostaglandin analogues. The trial design meant a fairly short observation period was needed to show treatment eﬀ ects on vision, with the diﬀ erence between treatment groups evident at just 12 months compared with typical observation periods of roughly 5 years in previous trials. The short trial duration will have a major beneﬁ cial eﬀ ect on development and assessment of new treatments, increasing the likelihood of these treatments being made available for patient beneﬁ t.


Introduction
Open-angle glaucoma is a chronic progressive optic neuropathy that causes a loss of vision, predominantly aff ecting the mid-peripheral visual fi eld at fi rst and later damaging central vision as the disease progresses; although increasingly evidence suggests unrecognised damage to central vision early in the disease course. Glaucomatous vision loss is associated with restricted mobility, 1 falls and motor vehicle accidents, 2 and is the leading cause of irreversible blindness worldwide and the second major cause for blind registration in the UK. 3, 4 The prevalence of open-angle glaucoma increases exponentially with age and the level of intraocular pressure. 5 Drugs to reduce intraocular pressure have been used for decades to slow or halt progressive vision loss in patients with open-angle glaucoma, yet there has been no placebo-controlled trial to assess vision preservation with this treatment. For most patients, the fi rst-choice drug is a prostaglandin analogue, 6 although a recent Cochrane review of medical interventions for glaucoma reported no published evidence for a protective eff ect on vision. 7 The United Kingdom Glaucoma Treatment Study (UKGTS) is the fi rst placebo-controlled trial to assess the eff ect of intraocular-lowering treatment on vision preservation.
Previous trials to assess intraocular-lowering treatment on vision preservation in open-angle glaucoma have compared medical treatment with no treatment, 8 combined medical and laser therapy with no treatment (the Early Manifest Glaucoma Trial 9 [EMGT] in 2002), one medical treatment with another, 10,11 medical, laser, or surgical intraocular pressure reduction with no treatment, 12 medical with laser and surgical pressure reduction, 13 initial medical with surgical pressure reduction, 14,15 and diff erent laser and surgical sequences in addition to medical treatment (the Advanced Glaucoma Intervention Study 16  angle glaucoma to assess the vision-preserving effi cacy of one drug in a placebo-controlled trial. The observation period for trials of visual fi eld preservation in patients with open-angle glaucoma has typically been 5 years or longer, 9,12,15,16 with the shortest trial lasting 30 months. 11 Long trial duration hinders the assessment of new interventions to prevent vision loss and increases the cost of drug development, which, in turn, reduces the likelihood of new treatments being made available for patient benefi t. Therefore, the design of UKGTS incorporated approaches for the measurement of outcomes that have potential to shorten study design, including repeated tests on visits (clustering) at the beginning and end of the observation period and imaging the retinal nerve fi bre layer and optic nerve head. Both the clustering and inclusion of imaging can increase the precision of estimates of the rate (speed) of change. 17,18 We aimed to assess the eff ect of the prostaglandin analogue latanoprost on the visual fi eld preservation of patients with open-angle glaucoma in a comparatively short trial.

Study design and participants
The UKGTS is a randomised, multicentre, triple-masked, parallel-group, placebo-controlled trial, undertaken in ten participating centres (tertiary referral centres, teaching hospitals, and district general hospitals) throughout the UK. The study design and baseline characteristics of UKGTS participants have been published previously. 19,20 We consecutively identifi ed participants with newly diagnosed, untreated open-angle glaucoma. Our eligibility criteria were closely modelled on those from the EMGT 21 to allow comparison and meta-analysis. Open-angle glaucoma was defi ned as the presence of glaucomatous visual fi eld defects in at least one eye with corresponding damage to the optic nerve head (see supplemental procedures in the appendix) and an open iridocorneal drainage angle on gonioscopy. We allowed patients with primary open-angle glaucoma and pseudoexfoliation glaucoma, but not pigment dispersion glaucoma. Exclusion criteria included advanced glaucoma (visual fi eld mean deviation worse than -10 dB in the better eye or -16 dB in the worse eye), mean baseline intraocular pressure of 30 mm Hg or higher, Snellen visual acuity worse than 6/12, and poor image quality (>40 μm mean pixel height standard deviation) with the Heidelberg retina tomograph (Heidelberg Engineering, Heidelberg, Germany). After the Moorfi elds Eye Hospital Reading Centre confi rmed eligibility and the patient gave informed consent, the Moorfi elds Clinical Trials Unit assigned a study identifi cation number.
The study was undertaken in accordance with good clinical practice guidelines and adhered to the Declaration of Helsinki. The trial was approved by the Moorfi elds and Whittington Research Ethics Committee on June 1, 2006 (reference 09/H0721/56). All patients provided written informed consent before screening investigations. An independent Data and Safety Monitoring Committee (DSMC) was appointed by the trial steering committee. The trial manager monitored adverse events, which were reported immediately to the operational DSMC at Moorfi elds Eye Hospital. Serious adverse events were reported to the Medicines and Healthcare Products Regulatory Agency.

Randomisation and masking
We randomly allocated participants (1:1) to either latanoprost 0·005% or latanoprost vehicle eye drops (placebo) alone once a day in both eyes. Patients were enrolled by clinicians at each site; once eligibility was confi rmed by the reading centre, the Moorfi elds Clinical Trials Unit assigned patients the next available study identifi cation number. We did randomisation in permuted blocks of varying sizes (block sizes range from 4 to 10), stratifi ed by participating centre. The randomisation schedule, drawn up by the research and development statisticians at Moorfi elds Eye Hospital on a randomisation website, was sent to the Pharmaceutical Manufacturing Unit, which labelled the bottles with the participant study identifi cation number only. Latanoprost and placebo eye drops were provided in identical-looking bottles. Participants, clinicians, and assessors of the primary outcome were masked to the treatment allocation and clinicians were encouraged not to tell the participants their intraocular pressure measurements.

Outcomes
After the funding decision and before trial initiation, the primary endpoint was changed from the one used in the EMGT-at least three test locations showing signifi cant deterioration at the p<0·05 level in three consecutive 30-2 visual fi elds-to at least three visual fi eld locations worse than baseline at the 5% levels in two consecutive reliable visual fi elds and at least three visual fi eld locations worse than baseline at the 5% levels in the two subsequent consecutive reliable visual fi elds; the locations identifi ed in the fi rst and second pair were not required to be identical. This change was made because the visual fi eld test in the UKGTS was smaller in extent , and therefore fewer locations were tested than in the EMGT. Time to progression was defi ned as time from baseline to the fourth visual fi eld that confi rmed progression. The primary endpoint was assessed on the day of each visit and then verifi ed by the Reading Centre.
After the publication of the National Institute for Health and Care Excellence Guideline Diagnosis and Management of Chronic Open Angle Glaucoma and Ocular Hypertension, 6 the independent DSMC requested an (unplanned) interim analysis on Jan 6, 2011, based on time-to-event. The Steering Committee therefore formally adopted time-to-event for the primary outcome. On the basis of the interim analysis, the DSMC requested an early termination of the trial and patients were scheduled for exit visits over the subsequent 3 months. We give survival data of diff erences in time from baseline to the event of confi rmed deterioration between treatment groups. Additional endpoints were intraocular pressure higher than 35 mm Hg on two successive occasions (safety endpoint) and visual acuity reduction to less than 6/18 (non-glaucomatous vision change endpoint). We also report the original primary endpoint-proportion of patients with deterioration at 24 months in each group. Secondary outcomes included visual fi eld deterioration rate (speed), measurements from imaging of the retinal nerve fi bre layer and optic nerve head, and scores from patient-related outcome measures questionnaires at 24 months, or at the time of a trial endpoint. 19 Because visual fi eld deterioration is more likely at higher intraocular pressures, patients leaving the trial after an endpoint would cause the mean intraocular pressure of those remaining to reduce over time. Therefore, missing data (from patients having reached an endpoint or being lost to follow-up) were imputed with the last observation carried forward (LOCF) method.

Procedures
We did visual fi eld testing, intraocular pressure measurement, and imaging at 11 scheduled visits over 24 months, with clustering of the tests at baseline, 18 months, and 24 months; 16 visual fi elds test were scheduled over 24 months. We measured vision function through testing of the visual fi eld. The visual fi eld test estimates retinal sensitivity, measured in dB, at 54 locations across the central 24 degrees, one eye at a time; 52 of which are analysed in the instrument software for change over time. A summary measure of vision function is the average loss at all locations: the mean deviation. Criteria for incident damage are based on identifi cation of visual fi eld locations that are repeatedly worse than baseline; these criteria are sensitive and identify change before it is noticed by the patient. Visual fi eld testing was done with the Humphrey Field Analyser Mark II (or II-i) with the Swedish interactive threshold algorithm standard 24-2 programme (Carl Zeiss Meditec, Dublin, CA, USA). We measured visual fi eld deterioration with the Humphrey Field Analyzer II-i Guided Progression Analysis software (version 5.1.1). Two independent glaucoma-fellowship-trained ophthalmologists (Nick Strouthidis [Moorfi elds Eye Hospital] and Paul Foster [UCL]) confi rmed that the deterioration was consistent with open-angle glaucoma on review of the visual fi elds and a fundus photograph taken at the time that the patient reached the endpoint. For eye-related variables, the values for the eye with worse baseline mean deviations were taken. Detail of other procedures is given in the appendix.

Statistical analysis
At the start of the trial, we decided the sample size based on our outcome at 18 months. However, because recruitment was slower than we expected, we recalculated the sample size in Oct 8, 2008, for an outcome at 24 months. We further revised the sample size in June 3, 2009, because of a greater than anticipated attrition rate. We established the fi nal sample size (516 participants) to have 90% power with a two-sided error (α=0·05) to detect a diff erence between 24% and 11% in incident visual fi eld deterioration in 24 months' follow-up, allowing for a 25% loss to follow-up over the study period.
We analysed all patients in the treatment group to which they were randomly assigned with all available data up to the point of withdrawal. If no data were available for any post-baseline assessment, we excluded patients from the analysis. We censored data from patients undergoing any ocular surgery at the last visit before surgery and included these patients in the loss to follow-up numbers.
We used survival analysis to compare the diff erences in time from baseline to the event of confi rmed deterioration between treatment groups. We estimated treatment eff ect as hazard ratios (HR) with 95% CIs computed by a Cox proportional hazards model that included terms, prespecifi ed in the statistical analysis plan, for treatment, baseline covariates (age at baseline, race [white or non-white], sex, baseline intraocular pressure, baseline mean deviation, blood pressure, refractive error, axial length, and central corneal thickness), and study centre. For patients who had no post-baseline data, we developed a regression model with the remaining data and used  because the trial was terminated early 1 did not meet eligibility criteria 231 analysed, including participants with <21 months' follow-up 32 did not attend all follow-up 7 completed 18 months' follow-up before trial extension 31 had incomplete follow-up because the trial was terminated early 1 VF endpoint suspected to be non-glaucomatous 1 did not meet eligibility criteria 258 allocated to latanoprost 28 did not attend any post-baseline study visits, including 3 reaching the IOP endpoint 27 did not attend any post-baseline study visits, including 2 who did not meet trial eligibility criteria this to estimate times to progression. The Cox model was re-run with the imputed values for the patients with no post-baseline data and estimates compared with the available case model. We investigated reasons for loss to follow-up with logistical regression of covariates (including treatment group, intraocular pressure, baseline mean deviation, age, sex, and ethnic origin) on an indicator of loss. Summary measurements for continuous, normally distributed, outcome values were diff erences in means and corresponding 95% CIs. For non-parametric equivalents, we used non-normally distributed outcome measures. For categorical variables, we used a Pearson's χ² or Fisher's exact test. We used survival analysis and linear regression to assess glaucoma severity as a risk for visual fi eld deterioration. We used linear regression to assess imaging outcomes for evidence of deterioration before a confi rmed visual fi eld endpoint in patients with baseline mean deviation lower than -10 dB (appendix). All statistical tests used a two-sided p value of 0·05. We did statistical analyses with Stata (version 12). This trial is registered, number ISRCTN96423140.

Role of the funding source
The funders had no input into the design, conduct, data collection, analysis, result interpretation, or reporting of the study. Some of the authors are employed by the sponsor organisation (Moorfi elds Eye Hospital). The organisation itself had no input. The corresponding author has full access to all data and had fi nal responsibility to submit for publication. Data are n (%) or mean (SD). BP=blood pressure. *Data are missing for participants who did not attend the fi rst post-allocation visit. and, therefore, provided no outcome data to analyse. Four patients were found to not fulfi l eligibility criteria after randomisation, but the two with post-baseline data were included in the analysis. We analysed data for the primary trial outcome for 461 patients (fi gure 1). Table 1 shows baseline characteristics of participants and table 2 shows eye-related parameters. Treatment groups were mostly similar; arteriosclerosis was more common in patients given latanoprost than in controls (table 1). Mean intraocular pressure at baseline seemed slightly higher in the placebo group than in the latanoprost group (table 1). Within   latanoprost group and 230 in the placebo group). 59 (25·6%; 95% CI 20·1-31·8) patients in the placebo group reached the deterioration endpoint at 24 months compared with 35 (15·2%; 10·8-20·4) in the latanoprost group (p=0·006). In one additional participant, an endpoint was deemed potentially to be due to a nonglaucomatous process. The visual fi eld deterioration endpoint was reached in both eyes of ten participants. Based on the 94 patients with visual fi eld deterioration events during the 24 month observation period, time to fi rst deterioration was signifi cantly longer in the latanoprost group than in the placebo group (adjusted HR 0·44, 95% CI 0·28-0·69; p=0·0003; fi gure 2). Treatment diff erences were evident at 18 months (0·43; 0·26-0·71; p=0·001), and 12 months (0·47; 0·23-0·95; p=0·035). 193 (75%) participants in the placebo group and 210 (81%) in the latanoprost group were taking trial drops and under observation or had reached an endpoint at 12 months. In the model with imputation, the primary endpoint HR was 0·43 (95% CI 0·26-0·69; p=0·0005). At the fi rst visit after treatment allocation, mean intraocular pressure reduction from baseline was 5·0 mm Hg (SD 3·6) in the latanoprost group and 1·4 mm Hg (3·1) in the placebo group. At 24 months, the mean intraocular pressure reduction from baseline was 4·0 mm Hg (3·4) in the latanoprost group and 1·3 mm Hg (3·6) in the placebo group (fi gure 3A). After we applied the LOCF adjustment, the intraocular pressure reduction at 24 months from baseline was 3·8 mm Hg (4·0) in the latanoprost group and 0·9 mm Hg (3·8) in the placebo group (fi gure 3B).
In participants reaching a visual fi eld endpoint, the mean change in mean deviation was -1·6 dB (IQR -0·6 to -2·6 dB; fi gure 4). Mean change in visual acuity (decimal) at last visit from baseline was +0·01 for eyes with visual fi eld progression and -0·02 for eyes without visual fi eld progression (Wilcoxon rank-sum test for diff erence p=0·3). Change from baseline visual acuity was -0·01 in the placebo group and -0·02 in the latanoprost group (Wilcoxon rank-sum test for diff erence p=0·9).
27 (10%) patients in the latanoprost group and 28 (11%) participants in the placebo group did not attend any post-baseline appointments. We explored the following associations for patients lost to follow-up (those with no post-baseline visits): age, sex, baseline intraocular pressure, baseline visual fi eld loss, ethnic origin, and treatment group. There were no associations at the p<0·05 level.
A further 32 (12%) patients in the latanoprost group and 39 (15%) in the placebo group had shorter than 21 months' follow-up because of failure to attend follow-ups or they had an intraocular pressure endpoint (fi gure 1). Of those  (table 3,  table 4). Of the 3750 scheduled appointments before planned trial exit or loss to follow-up, 63 (2%) were missed; 32 by patients in the placebo group and 31 by those in the latanoprost group. 79 participants had incomplete followup from early trial termination or had completed 18 months' follow-up before the extension of the observation period (fi gure 1, appendix). An intraocular pressure safety endpoint was reached in six participants, two of whom simultaneously reached a visual fi eld endpoint. No patient reached a visual acuity reduction endpoint. 192 adverse events (99 in patients allocated to placebo and 93 in patients allocated to latanoprost) were reported in 98 participants (50 participants in the latanoprost group and 48 in the placebo group; table 3). 13 of 153 mild adverse events and one of 21 moderate adverse events were thought to be possibly related to study drugs (four in the placebo group, ten in the latanoprost group). 18 serious AEs were reported (nine in the placebo group and nine in the latanoprost group), none attributed to the study drug. Seven patients withdrew because of a possible adverse reaction to study drug (table 4): local allergy (in two patients), drop intolerance (in three), diarrhoea (in one), and asthma (in one).

Discussion
Systematic reviews of treatment for glaucoma have reported no placebo-controlled trials in patients with openangle glaucoma with visual function as the outcome (panel). 9,23 To our knowledge, no evidence exists for a protective eff ect on vision for the most frequently prescribed class of drugs to lower intraocular pressureprostaglandin analogues. 7 To our knowledge, UKGTS is the fi rst placebo-controlled trial to show visual fi eld preservation through lowering of intraocular pressure with topical drugs in patients with open-angle glaucoma and to show the visual-fi eld-preserving eff ect of   latanoprost, the drug most frequently used in high-income countries. Previous studies, such as the EMGT 9 and the Collaborative Normal Tension Glaucoma Study, 12 have assessed the outcome of combined treatments to lower intraocular pressure. Our placebo-controlled trial is of clinical relevance in view of the results of a recent trial by Krupin and colleagues suggesting very diff erent visionpreserving action of two drugs with similar intraocularpressure-lowering effi cacy. 11 The eff ect size in the UKGTS was large, with an adjusted HR of 0·44 (95% CI 0·28-0·69), associated with initial intraocular pressure-lowering of 5·0 mm Hg (SD 3·6) in the latanoprost group and 1·4 mm Hg (3·1) in the placebo group. The initial intraocular pressure reduction from baseline (5·0 mm Hg of 19·6 mm Hg; 26%) with latanoprost in our trial was lower than the 31% (peak) and 28% (trough) reduction from baseline reported in a meta-analysis by van der Valk and colleagues. 24 However, in their meta-analysis, both the baseline peak (25·5 mm Hg) and trough (24·3 mm Hg) levels were higher than was the baseline pressure in the UKGTS (19·6 mm Hg). The intraocular pressure reduction at 24 months (3·8 mm Hg) is in line with data obtained for cohorts with lower baseline pressure-eg, 3·2 mm Hg from 17·6 mm Hg 25 and 3·9 mm Hg from 18·8 mm Hg. 26 The reduction in intraocular pressure lessened a little over the 24 months; in the LOCF analysis, the intraocularpressure reduction compared with baseline was 3·8 mm Hg in the latanoprost group and 0·9 mm Hg in the placebo group (fi gure 3B). The pattern of intraocularpressure reduction, with a strong initial response that diminished after 6 months and then stabilised, has been reported previously. 25 The pressure reduction in the placebo group (fi gure 3B)-which was greater at the fi rst post-allocation visit (1·4 mm Hg) than at the visit at 24 months (0·9 mm Hg)-might be a regression-to-themean eff ect due to clinician behaviour (with the desire to recruit participants into a trial, a clinician might be more likely to accept borderline clinical fi ndings as glaucoma when the intraocular pressure reading is high and, after randomisation and start of drops, might have a greater expectation of lower intraocular pressure at the fi rst postallocation visit), or might be a true hypotensive eff ect of latanoprost vehicle eye drops or a true placebo eff ect.
Only two previous trials have been published of medical (non-incisional) treatment for open-angle glaucoma with an untreated control group and vision function as an outcome-the study by Holmin and colleagues 8 and the EMGT. 9 Neither study was placebocontrolled or masked, although both had an objective (visual fi eld) outcome. The study by Holmin and colleagues 8 was very small (16 patients enrolled). Previous trials of open-angle glaucoma with visual fi eld deterioration as an outcome have used widely diff ering criteria to identify deterioration and have studied diff ering clinical populations than each other, 9,12,15,16 whereas we chose the recruitment criteria and main outcome of the UKGTS to be similar to those of the EMGT to allow comparison and meta-analysis of results. The outcome criterion was established from the glaucoma change probability maps, which are based on the pattern deviation, limiting the eff ect of homogeneous reduction in diff erential light sensitivity seen with cataract. 9 The observation period of the trial was short (24 months) and little change in lens opacity would be expected; as shown by the stability of the visual acuity over the duration of this study.
From our estimation of incident progression from the survival curves (fi gure 2), a visual fi eld endpoint was reached by 24 months in 34% of participants in the untreated group of the UKGTS and 25% in the untreated group of the EMGT, versus in 20% of participants in the   10 with vision function as a primary outcome that was not already identifi ed in the Cochrane reviews. We assessed trial quality on the basis of four metrics: allocation concealment, performance bias, detection bias, and attrition bias.

Interpretation
To our knowledge, the United Kingdom Glaucoma Treatment Study (UKGTS) is the fi rst randomised, triple-masked, placebo-controlled trial to assess the benefi t of topical medical treatment (eye drops) for reduction of loss of vision in patients with open-angle glaucoma. Our fi ndings provide strong evidence for the vision-preserving benefi t of lowering of intraocular pressure, supporting evidence from previous randomised trials that were not masked or placebo-controlled. The study also provides evidence of the vision-preserving benefi ts of topical prostaglandin analogues. The trial design meant a fairly short observation period was needed to show treatment eff ects on vision, with the diff erence between treatment groups evident at just 12 months compared with typical observation periods of roughly 5 years in previous trials. The short trial duration will have a major benefi cial eff ect on development and assessment of new treatments, increasing the likelihood of these treatments being made available for patient benefi t.
treated group of the UKGTS and 11% in the treated group of the EMGT. Since baseline intraocular pressure (19·9 mm Hg in UKGTS, 20·6 mm Hg in EMGT) and participant age (66 years in UKGTS, 68 years in EMGT) were similar, 20,21 the higher incident progression in the UKGTS might have been caused by the more sensitive criterion to identify progression. The relative sensitivity of the endpoint criteria can be judged from the median change in visual fi eld mean deviation between baseline examination and the time at which deterioration was confi rmed: -1·6 dB for the UKGTS criterion (fi gure 4) and -1·9 dB for the EMGT criterion. 27 Incident progression was reduced by 41% in the treated group of UKGTS, compared with a 58% reduction in the EMGT; this fi nding is consistent with the 16% smaller intraocular pressure-lowering treatment eff ect in the UKGTS (-3·8 mm Hg vs -4·5 mm Hg in the EMGT). Strengths of the UKGTS trial (and EMGT trial) are that newly diagnosed, previously untreated participants were studied, so that the results are free of infl uence from previous therapeutic interventions and the natural history of open-angle glaucoma can be studied in the untreated patients. Because consecutive new patients were assessed and recruited for the UKGTS trial, the population studied was representative of the target population considered for treatment and on which the trial results will be applied.
The UKGTS has the shortest observation period of any glaucoma trial with a vision function outcome, [9][10][11][12][13]15,16 with a statistically signifi cant diff erence between treatment groups evident at only 12 months, as well as at the planned 24 month analysis timepoint. However, because the signifi cance of the diff erence in proportions surviving in the EMGT was p=0·004 after 48 months' observation, 9 the diff erence between treatment groups in the EMGT was probably signifi cant over a much shorter observation period than was reported. The short observation period needed to observe a treatment eff ect can bring many benefi ts, speeding up novel drug development with consequent cost reduction and thereby increasing the likelihood of bringing new drugs to patients. The short observation period was achieved by adopting a sensitive change-from-baseline event criterion to identify visual fi eld deterioration, frequent visual fi eld tests, and a suffi ciently large sample size. Also, the UKGTS design included additional features that might further reduce study duration and sample size. We aimed to improve the accuracy and precision of the rate (speed) of progression by including the clustering of tests at the beginning and end of the observation period 17 and quantitative imaging of the retinal nerve fi bre layer and optic nerve head; 18 the analysis of these secondary outcomes will be reported elsewhere.
Because two-thirds of patients in the placebo group had no detectable deterioration within 24 months, and the small amount of change needed for an endpoint occurred in just one eye of almost 90% of participants with data, the questions arises as to whether some patients with glaucoma, particularly those with less visual fi eld loss at baseline and those at lower risk of progression, 28,29 could be monitored without treatment for a period. Investigators have previously suggested initial careful observation without treatment in patients with open-angle glaucoma 9 and normal tension glaucoma. 30 Because deterioration patterns are diffi cult to predict, 31 an observation period would be advantageous to identify patients who might not need treatment, thereby avoiding the unnecessary burden of treatment for such patients. The rate of progression of visual fi eld loss might change over extended observation periods, and so regular reassessments of the rate of progression would be advisable.
A potential limitation of the UKGTS was the relatively high loss to (or incomplete) follow-up, which can reduce the generalisability of the results. However, individuals lost to follow-up were generally similar to those remaining in the trial. In addition, participant ascer tainment was not population-based, but based on referral from primary care (community optometry practice); however, the participant characteristics in the UKGTS 20 were strikingly similar to those in the EMGT, in which ascertainment was largely based on population screening. 21 UKGTS participants were predominantly white (about 90%), which might also reduce gen eralisability of our fi ndings to some subpopulations in the UK or other countries who might exhibit greater susceptibility to vision loss. 32,33 Furthermore, individuals with advanced open-angle glaucoma were excluded so that the results of the trial only apply to patients with mild to moderate glaucoma. Subsequent publications will address the eff ect of treatment with respect to baseline intraocular pressure, glaucoma stage, participant age, and other risk factors.

Contributors
DFG-H, DPC, and CB conceived of and designed the trial and RPW guided the set-up and conduct of the trial. DFG-H was the chief investigator for the trial and had oversight of the trial throughout. GL and AS were the trial physicians and did the the day-to-day running of the study. FA was the trial coordinator. ETW was the chief technician. TGZ chaired the steering committee. GL, AS, NA, AA-B, RRB, DCB, IAC, JPD, SGF, KRM, AIM, AN, and PGS recruited patients and reported corresponding data. KS assisted with data preparation and analysis; CB, KP, RAR, and WX were the trial statisticians and did the fi nal statistical analyses of the results. DFG-H led the interpretation of the data and writing of the manuscript, assisted by TAH. All authors had input into the data interpretation and preparation of the fi nal report for publication.