Summary

• One trial (N=616) of frail elderly persons found no difference between vision screening, including components for glaucoma, versus no screening on vision outcomes (mean logarithm of the minimum angle of resolution [logMAR] distance visual acuity scores 0.27 vs. 0.25, p=0.32, and mean logMAR near visual acuity scores −0.01 vs. −0.03, p=0.26) and vision-related quality of life (National Eye Institute Visual Function Questionnaire-25 [NEI-VFQ-25] mean composite scores 84.3 vs. 86.4, p=0.49) after 1 year.

Evidence

The prior screening CER included no trials comparing screening with no screening.² We identified one good-quality trial (n=616) conducted in Australia comparing vision screening by an optometrist with no screening that included components relevant for diagnosis of glaucoma (IOP, direct ophthalmoscopy, and visual field) as well as other visual testing (visual acuity, contrast sensitivity, and slit lamp examination; Appendix B Table 1).¹³² In the screened group, interventions for screen-positive persons included referral to an ophthalmologist or public hospital eye clinic and/or an occupational therapist (for home modifications, mobility training, or a cane); those in the control group received no vision assessment or intervention. The mean age was 81 years and 68 percent were female; race and ethnicity were not reported. Thirty-one percent of participants needed help with activities of daily living at baseline and 52 percent were taking more than four medications. At baseline, 46 percent of participants had experienced a fall in the past year. At baseline, mean visual acuity was 0.22 logMAR (Snellen 20/30), the mean NEI-VFQ-25 score was 85.5 (scale 0 to 100, higher is better), 63 percent had cataracts, 39 percent had undergone cataract surgery, and 98 percent wore glasses. Fourteen percent of patients had glaucoma at baseline and 50 percent self-reported vision as “good.” In addition to appropriate randomization, the trial blinded outcome assessors and data analysts and attrition was low (11% screening arm and 16% control arm; Appendix B Table 2). Nearly half (48.7%) of the patients in the screening arm were judged to need treatment, though only 5.5 percent of patients judged to need treatment were referred for glaucoma management. Other interventions were new glasses (29.8%), referral for cataract surgery (4.9%), referral for age-related macular degeneration (AMD) (1.6%), and referral to an occupational therapist (7.7%).

At 1 year, there were no differences in vision parameters or vision-related quality of life. Mean distance visual acuity was 0.27 vs. 0.25 logMAR (p=0.32), mean near visual acuity −0.01 vs. −0.03 logMAR (p=0.26), and NEI-VFQ-25 mean composite scores were 84.3 vs. 86.4 (p=0.49). Nearly three-quarters of patients in the control group reported having seen an eye care professional in the 12 months prior to study, which could have attenuated potential benefits of screening.

Summary

• One trial (n=616) found screening associated with an increased risk for falls versus no screening (incidence rate ratio 1.57, 95% CI 1.20 to 2.05, and risk of one or more falls 65% vs. 50%, RR 1.31, 95% CI 1.13 to 1.50); screening was associated with increased risk for fractures that was not statistically significant (RR 1.74, 95% CI 0.97 to 3.11).

Evidence

No trial in the prior screening CER compared harms of screening with no screening.² A previously-described trial¹³² of vision screening (including components for identification of glaucoma) versus no screening in frail elderly reported risk of falls and fracture (Appendix B Table 1).⁷² In the trial, 46 percent of patients had fallen in the past year. Although the trial hypothesized that screening would reduce the risk of falls, screening was associated with increased incidence of falls (758 vs. 516 falls, incidence rate ratio 1.57, 95% CI 1.20 to 2.05), risk of one or more falls (65% vs. 50%, RR 1.31, 95% CI 1.13 to 1.50), and risk of two or more falls (38% vs. 31%, RR 1.24, 95% CI 0.99 to 1.54) versus no screening. Screening was also associated with increased risk of fracture, though the difference was just above the threshold for statistical significance (10% vs. 5.7%, RR 1.74, 95% CI 0.97 to 3.11, p=0.06).

Summary

• Retinal nerve fiber layer thickness on spectral domain-OCT was associated with a pooled sensitivity of 0.79 (95% CI 0.75 to 0.83) and specificity of 0.92 (95% CI 0.87 to 0.96) for distinguishing between glaucomatous eyes and controls, based on 15 studies (N=4,242); the pooled AUROC curve was 0.90 (95% CI 0.86 to 0.93), based on 16 studies (N=4,060).
• Ganglion cell complex thickness on spectral domain-OCT was associated with a pooled sensitivity of 0.74 (95% CI 0.68 to 0.80) and specificity of 0.91 (95% CI 0.80 to 0.96) for distinguishing between glaucomatous eyes and controls, based on nine studies (N=1,522); the pooled AUROC curve was 0.88 (95% CI 0.84 to 0.92), based on six studies (N=765).
• Tonometry was associated with a pooled sensitivity of 0.48 (95% CI 0.31 to 0.66) and specificity of 0.94 (95% CI 0.90 to 0.96), based on 13 studies (N=32,892).
• The Humphrey Visual Field Analyzer was associated with a pooled sensitivity of 0.87 (95% CI 0.69 to 0.95) and specificity of 0.82 (95% CI 0.66 to 0.92) for distinguishing between glaucomatous eyes and controls, based on six studies (N=11,244).
• Evidence on diagnostic accuracy was limited for other screening tests: cup-to-disc ratio on spectral domain-OCT, swept source-OCT, optic disc photography, ophthalmoscopy/biomicroscopy/stereoscopy, pachymetry, and afferent papillary defect.
• One pilot study (n=56) and a followup study (n=256) found a telemedicine screening intervention performed in a primary care setting had variable sensitivity but high specificity for identifying persons with glaucoma compared with a face-to-face evaluation by an ophthalmologist.

Evidence

The prior screening CER² included a systematic review¹⁴⁵ and 83 additional studies on the diagnostic accuracy of tests for glaucoma. Since the prior screening CER, several diagnostic tests have been superseded by newer technologies and are not included in this review. For example, for imaging the optic nerve and retinal structures, OCT has superseded Heidelberg retina tomography and scanning laser polarimetry; for evaluating visual field loss, the Humphrey Field Analyzer has superseded frequency doubling technology. In addition, the prior screening CER included case-control studies, which were excluded from this review, and had an emphasis on comparative diagnostic accuracy, which was not the focus of this review. The prior screening CER concluded that it was unclear whether any one test or combination of tests was suitable for glaucoma screening in the general population, due to the lack of a definitive diagnostic reference standard for glaucoma and heterogeneity in the design and conduct of the studies.

This review includes 53 diagnostic accuracy studies (sample sizes 46 to 8623, N=65,464 in 59 publications.) (Table 1; Appendix B Tables 3–4)^{16,53–56,59–61,64,65,67–71,73–77,79–81,84–86,88,89,93,95–97,99–102,104–112,116,117,120,122,125,128,130,131,133,135–138,144} The largest groups of studies evaluated spectral domain-OCT (k=29, N=11,434) and tonometry (k=17, N=49,742), followed by visual fields (k=10, N=11,633), ophthalmoscopy/biomicroscopy/stereopscopy (k=3, N=17,519), optic disc photography (k=4, N=3,133), pachymetry (k=2, N=6,129), telemedicine (k=2, N=308), and afferent pupillary defect (k=1, N=107). Most studies evaluated more than one test of diagnostic accuracy. Forty-six studies included a single eye per participant in the analysis, and six studies^{75,93,101,112,116,122} allowed two eyes per participant (the number of eyes analyzed was unclear in one study¹⁰⁵). In most studies, the reference standard was based on findings related to ophthalmic structure (e.g., appearance of optic disc) as well as function (e.g., visual fields), though exact criteria differed (Appendix B Table 3).

Table 1

Diagnostic Accuracy, Pooled Analyses.

Mean age ranged from 38.2 years to 82.2 years (median 58 years). The proportion of females enrolled ranged from 13.3 to 72.3 percent (median 55%) in studies that reported gender. Twelve studies reported race/ethnicity. Two studies restricted enrollment to Asian persons,^130,131 one study restricted enrollment to Latino persons,⁸⁰ and one study restricted enrollment to White persons.⁸¹ In the other studies, the proportion of White participants ranged from 17 to 99 percent. In two of the studies, the majority of participants (61% and 62%) were Black.^86,110 Studies were conducted in Western Europe (N=16), the U.S. (N=13), and Asia (N=18); two studies were conducted in Turkey and one study each was conducted in Hungary, Australia, New Zealand, and Croatia. The prevalence of glaucoma ranged from 1.1⁹³ to 73.6¹⁰⁰ percent. Seven studies were rated good quality^{56,59,73,80,109,111,122} and the remainder were rated fair quality (Appendix B Table 5). Methodological limitations in the fair-quality studies included lack of blinding and uncertain interval between index and reference tests.

Tests of Ophthalmic Structure

The diagnosis of glaucoma is typically made by using tests of both ophthalmic structure and function together. Tests of eye structure include OCT, optic disc photography, and clinical examination with an ophthalmoscope or slit-lamp (biomicroscopy). Thirty studies (N=11,618) evaluated OCT (Appendix B Tables 3–4). Four studies were rated good quality (N=2,575)^56,59,73,122 and 27 were rated fair quality (N=8,859)^{53–55,64,67,71,75,76,79,81,93,96,97,99–101,104–106,116,120,125,128,131,136,144} (Appendix B Table 5).

Optical Coherence Tomography

There are three types of OCT: time domain, spectral domain, and swept source. Time domain-OCT represents the earliest technology and became commercially available in 1996.¹⁴⁶ Time domain-OCTs have a movable reference light and can produce 400 axial scans of the eye per second. Time domain-OCT was not included in this review as it has been superseded by spectral domain-OCT, which entered the market in 2006, uses a fixed reference light, and can produce 50,000 axial scans per second, resulting in images with greater resolution.¹⁴⁶ Swept source is the latest OCT technology and is even faster than spectral domain-OCT (200,000 or more axial scans per second), but is not yet in widespread use. The primary parameters used on OCT are the thickness of the retinal nerve fiber layer and the ganglion cell complex.

The prior screening CER included 48 studies of OCT. Based on average retinal nerve fiber layer estimates on OCT, sensitivity ranged from 24 to 96 percent and specificity ranged from 66 to 100 percent. Many studies (k=34) in the prior screening CER used time domain-OCT and are not included in this review. Two studies of spectral domain-OCT were carried forward from the prior USPSTF report (k=2, n=283),^55,131 and we identified 27 new studies (N=14,199). Twenty-nine studies (N=14,482) evaluated spectral domain-OCT^{53–56,59,64,67,71,73,75,76,79,81,93,96,97,99–101,104,106,116,120,122,125,128,136,144} and three studies (n=120, 145, and not reported) assessed swept source-OCT.^104–106

Retinal Nerve Fiber Layer Thickness

Retinal nerve fiber layer thickness on spectral domain-OCT was associated with a pooled sensitivity of 0.79 (95% CI 0.75 to 0.83) and specificity of 0.92 (95% CI 0.87 to 0.96) for diagnosing eyes with glaucoma versus no glaucoma (healthy eyes, glaucoma suspect, and/or ocular hypertension), based on 15 studies (N=4,242)^{53,54,56,59,64,71,73,76,81,99,104,106,128,131,136} (Table 2, Figures 2–3). Pooled estimates were similar when the analysis was limited to studies in which the control group was healthy eyes (9 studies, N=2,404, sensitivity 0.81, 95% CI 0.74 to 0.86 and specificity 0.96, 95% CI 0.89 to 0.99).^{53,54,56,73,81,99,104,106,131} Pooled estimates were also similar when the analysis was limited to studies that measured retinal nerve fiber layer based on the mean overall thickness as opposed to mean inferior,⁸¹ mean outer/inferior,^71,104 or mean temporal/inferior retinal nerve fiber layer thickness^53,54 (12 studies, N=3,819, sensitivity 0.79, 95% CI 0.74 to 0.84 and specificity 0.90, 95% CI 0.85 to 0.93),^{56,59,64,71,73,76,99,104,106,128,131,136} and when results were limited to the 12 fair-quality studies (N=1,880, sensitivity 0.80, 95% CI 0.74 to 0.85; specificity 0.94, 95% CI 0.88 to 0.97).^{53,54,64,71,76,81,99,104,106,128,131,136} In three good-quality studies (N=2,400),^56,59,73 sensitivity ranged from 0.69 to 0.81 and specificity ranged from 0.79 to 0.94. One study (n=129) also reported accuracy of retinal nerve fiber layer thickness for diagnosing ocular hypertension versus healthy eyes (sensitivity 0.08, 95% CI 0.005 to 0.63; specificity 1.00, 95% CI 0.96 to 1.00).⁸¹

Table 2

Sensitivity and Specificity, Spectral Domain-OCT Retinal Nerve Fiber Layer Thickness.

Figure 2

Glaucoma vs. Control, Spectral Domain-OCT Sensitivity and Specificity for Retinal Nerve Fiber Layer Thickness. Abbreviations: CI = confidence interval; df = degrees of freedom; OCT = optical coherence tomography.

Figure 3

Glaucoma vs. Control, Spectral Domain-OCT Retinal Nerve Fiber Layer Thickness. Abbreviations: HSROC = hierarchical summary receiver operating characteristic; OCT = optical coherence tomography.

Five studies on diagnostic accuracy of retinal nerve fiber layer thickness on spectral domain-OCT were not pooled because they were based on the inter-eye retinal nerve fiber layer thickness asymmetry⁷⁹ or because they evaluated more than one eye per participant.^{75,93,101,122} Details of these studies are shown in Appendix B Tables 3 and 4.

Retinal nerve fiber layer on spectral domain-OCT was associated with high discrimination for distinguishing glaucomatous eyes from non-glaucoma, with a pooled AUROC curve of 0.90 (95% CI 0.86 to 0.93, I²=96%), based on 16 studies (N=4,060)^{53–56,59,64,71,76,96,97,99,100,104,106,120,128} (Figure 4). Discrimination was similar for the two good-quality studies (N=1,944, pooled AUROC 0.87, 95% CI 0.80 to 0.94, I²=86%)^56,59 and 14 fair-quality studies (N=2,116, pooled AUROC 0.90, 95% CI 0.86 to 0.94, I²=97%).^{53–55,64,71,76,97,99,100,104,106,120,128} All studies reported an AUROC greater than or equal to 0.83, with the exception of two studies that reported an AUROC of 0.78.^54,71 Results were similar in a sensitivity analysis restricted to studies that utilized overall mean retinal nerve fiber layer thickness (12 studies, N=3,634, AUROC 0.92, 95% CI 0.89 to 0.94, I²=80%) and in an analysis stratified according to whether the non-glaucoma group was healthy eyes (10 studies, N=2,262, AUROC 0.92, 95% CI 0.89 to 0.94, I²=84%),^{53–56,96,99,100,104,106,120} glaucoma suspects (4 studies, N=496, AUROC, 0.90, 95% CI 0.86 to 0.94, I²=51%),^64,76,96,99 or ocular hypertension (3 studies, N=319, AUROC 0.80, 95% CI 0.71 to 0.89, I²=87%)^54,71,96 (Figure 5).

Figure 4

AUROC, Spectral Domain-OCT Retinal Nerve Fiber Layer Thickness by Comparison. Abbreviations: AUROC = area under the receiver operating characteristic curve; CI = confidence interval; ES = estimate; OCT = optical coherence tomography.

Figure 5

AUROC Curves, Spectral Domain-OCT Retinal Nerve Fiber Layer Thickness. Abbreviations: AUROC = area under the receiver operating characteristic curve; CI = confidence interval; ES = estimate; OCT = optical coherence tomography.

Three studies (N=364) of retinal nerve fiber layer reported a pooled AUROC of 0.76 (95% 0.63 to 0.90, I²=83%) for discrimination of glaucoma suspect from healthy eyes.^55,96,97 One study (n=122) of retinal nerve fiber layer reported an AUROC of 0.64 (95% CI 0.54 to 0.75) for discrimination of ocular hypertension from healthy eyes.⁹⁶ Four other studies reported discrimination of retinal nerve fiber layer but were not pooled due to inadequate data (N=1,335)^{73,125,136,144} or because they enrolled more than one eye in some participants (N=659).^67,75,116 Details are shown in Appendix B Tables 3 and 4.

Ganglion Cell Complex

The ganglion cell complex is composed of three thickness areas which can be imaged using OCT: the retinal nerve fiber layer, the inner plexiform layer, and the ganglion cell layer. Ganglion cell complex on spectral domain-OCT was associated with a pooled sensitivity of 0.74 (95% CI 0.68 to 0.80) and pooled specificity of 0.91 (95% CI 0.80 to 0.96) for identifying individuals with glaucoma, based on nine studies (N=1,522)^{53,54,71,73,76,81,104,106,136} (Table 3; Figures 6 and 7). Estimates were similar when three studies^71,76,136 in which persons with ocular hypertension or glaucoma suspects were excluded from the analysis (6 studies, N=1,145, sensitivity 0.76, 95% CI 0.66 to 0.83 and specificity 0.92, 95% CI 0.86 to 0.96). Estimates were also similar when studies that reported only the ganglion cell layer or inner plexiform layer^{71,76,104,106} were excluded (5 studies, N=998, pooled sensitivity 0.73, 95% CI 0.60 to 0.83 and pooled specificity 0.95, 95% CI 0.87 to 0.98). One good-quality study (n=456) reported sensitivity of 0.62 (95% CI 0.41 to 0.80) and specificity of 0.93 (95% CI 0.91 to 0.96);⁷³ in eight fair-quality studies (N=542) pooled sensitivity was 0.75 (95% CI 0.68 to 0.81) and pooled specificity was 0.91 (95% CI 0.78 to 0.97).^{53,54,71,76,81,104,106,136}

Table 3

Sensitivity and Specificity, Spectral Domain-OCT Ganglion Cell Complex Thickness.

Figure 6

Glaucoma vs. Control, Spectral Domain-OCT Sensitivity and Specificity for Ganglion Cell Complex Thickness. Abbreviations: CI = confidence interval; df = degrees of freedom; OCT = optical coherence tomography.

Figure 7

Glaucoma vs. Control, Spectral Domain-OCT Ganglion Cell Complex Thickness. Abbreviations: HSROC = hierarchical summary receiver operating characteristic; OCT = optical coherence tomography.

Six fair-quality studies found ganglion cell complex, ganglion cell layer, and inner plexiform layer associated with high discrimination for distinguishing glaucoma from non-glaucoma (N=765, AUROC 0.88, 95% CI 0.84 to 0.92, I²=68%)^{53,54,75,96,104,106} (Figure 8). Results were similar when the non-glaucoma groups were stratified as healthy eyes (5 studies, N=564, AUROC 0.87, 95% CI 0.82 to 0.92, I²=70%),^{53,54,96,104,106} glaucoma-suspect eyes (2 studies, N=354, AUROC 0.84, 95% CI 0.69 to 1.00, I²=92%),^75,96 or eyes with ocular hypertension (2 studies, N=224, AUROC 0.76, 95% CI 0.70 to 0.82, I²=0%)^54,96 (Figure 9). Results were also similar when the analysis was restricted to two studies that assessed the ganglion cell complex (N=211, AUROC 0.87, 95% CI 0.73 to 1.00, I²=89%).^53,54 Five studies could not be pooled due to inadequate data (e.g., reported sensitivity, specificity and/or AUROC without confidence intervals, only reported odds ratios),^{71,122,125,136} or did not report standard AUROC.⁷³ Four studies were not pooled because they enrolled more than one eye in some participants.^67,75,93,101 Details are provided in Appendix B Tables 3–4.

Figure 8

Ganglion Cell Analysis.

Figure 9

Ganglion Cell Analysis by Control Group.

Cup-to-Disc Ratio

One study (N=286) found the cup-to-disc ratio on spectral domain-OCT associated with sensitivity of 0.84 (95% CI 0.77 to 0.89) and specificity of 0.72 (95% CI 0.60 to 0.81) for identifying persons with glaucoma versus healthy eyes. The cup-to-disc ratio threshold for a positive test was not specified.⁸¹

Three studies (n=1,870) found the spectral domain-OCT vertical cup-to-disc ratio associated with an AUROC that ranged from 0.74 to 0.94^100,101,144 (Table 4). These studies were not pooled because they enrolled more than one eye in some participants^101,144 and one did not report SD.¹⁴⁴

Table 4

Glaucoma vs. Control, Spectral Domain-OCT Cup-to-Disc Ratio.

Swept Source–OCT

Swept source-OCT utilizes a longer wavelength than spectral domain-OCT to visualize deeper structures and is faster than spectral domain-OCT.

Two studies (reported in 3 publications, N=266) assessed the diagnostic accuracy of swept-source OCT using retinal nerve fiber layer thickness.^104–106 One study found wide-field retinal nerve fiber layer thickness map associated with sensitivity of 0.95 (95% CI 0.90 to 0.98) and specificity of 0.89 (95% CI 0.75 to 0.97) for distinguishing between participants with glaucoma and participants with healthy eyes.¹⁰⁶ The other study reported an AUROC for distinguishing persons with glaucoma from those with healthy eyes of 0.85 (95% CI 0.78 to 0.92) for the retinal nerve fiber layer outer/inferior sector and 0.83 (95% CI 0.75 to 0.90) for the outer/temporal sector of the ganglion cell inner plexiform layer.¹⁰⁴ Another article (n=not reported; 184 eyes)¹⁰⁵ reported an AUROC of 0.85 (95% CI 0.78 to 0.91) for discriminating between early perimetric glaucoma and healthy eyes of 0.85 (95% CI 0.78 to 0.91) for the retinal nerve fiber layer thickness and 0.87 (95% CI 0.80 to 0.92) for the ganglion cell inner plexiform layer (inferior temporal).

Optic Disc Photography

Four studies (N=3,133) reported diagnostic accuracy of cup-to-disc ratio on optic disc photography, separate from OCT.^64,74,86,107 One study (n=2,631) screened participants with indirect ophthalmoscopy as well as disc photographs to assess the optic disc¹⁰⁷ (Table 5).

Table 5

Glaucoma vs. Control, Optic Disc Photography Cup-to-Disc Ratio.

Two studies reported similar discrimination of cup-to-disc ratio on optic disc photography, with AUROCs of 0.85 (95% CI 0.74 to 0.96) and 0.81 (95% CI 0.74 to 0.92).^64,74 In one of these studies, sensitivity was 0.64 (95% CI 0.45 to 0.81) and specificity was 0.73 (95% CI 0.45 to 0.92) for distinguishing persons with glaucoma from glaucoma suspects; the cup-to-disc ratio threshold was not reported.⁶⁴ Two studies did not report discrimination; in one study sensitivity was 0.18 (95% CI 0.09 to 0.31) and specificity was 0.67 (95% CI 0.62 to 0.71) based on a cup-to-disc ratio threshold of 0.4.¹⁰⁷ In the other study, sensitivity was 0.71 (95% CI 0.54 to 0.85) and specificity was 0.49 (95% CI 0.44 to 0.55) for distinguishing between glaucoma and nonglaucoma, based on a cup-to-disc ratio of 0.65 for average-sized or large discs and 0.5 for small discs.⁸⁶

Ophthalmoscopy, Biomicroscopy, and Stereoscopy

Five studies reported accuracy of cup-to-disc ratio on ophthalmoscopy, biomicroscopy, and stereoscopy (N=17,519)^{80,84,107,130,135} (Table 6). Studies were not pooled because the methods used to determine cup-to-disc ratio as well as the cutoffs to define a positive screen varied. Although specificity was high in all studies, sensitivity varied widely (range 0.18 to 0.92).

Table 6

Ophthalmoscopy/Biomicroscopy/Stereoscopy Cup-to-Disc Ratio.

Tests of Ophthalmic Function

Tests of optic nerve function include measurements of IOP through tonometry and visual field assessment.

Visual Fields

The Humphrey Field Analyzer has superseded frequency doubling technology as standard of care for the assessment of visual fields. Although the Humphrey Field Analyzer was often used as part of the reference standard for the diagnosis of glaucoma, 10 studies (N=11,633) reported diagnostic accuracy of the Humphrey Field Analyzer against a reference standard.^{64,74,80,88,89,95,108,112,117,122} In these studies, Humphrey Field Analyzer methods varied: five studies used the Swedish Interactive Threshold Algorithm-Standard 24-2,^{64,74,80,108,122} two use the 76-point 30 degree suprathreshold,^88,89 and one study each used the Full Field 120 Protocol,⁹⁵ the 630 Armaly Full Field Test,¹¹² and the Central 30-2 (Goldmann III stimulus).¹¹⁷

The Humphrey Field Analyzer was associated with pooled sensitivity of 0.87 (95% CI 0.69 to 0.95) and pooled specificity of 0.82 (95% CI 0.66 to 0.92), based on six studies (N=11,244)^{64,80,88,95,108,117} (Figures 10 and 11). There were too few studies of specific Humphrey Field Analyzer methods to conduct a meaningful analysis stratified by method (Table 7). One good-quality study (N=6,082)⁸⁰ reported a sensitivity of 0.88 (95% CI 0.83 to 0.92) and specificity of 0.64 (95% CI 0.63 to 0.65) using SITA-Standard 24-2.

Figure 10

Glaucoma vs. Control, Humphrey Field Analyzer Visual Field. Abbreviation: HSROC = hierarchical summary receiver operating characteristic.

Figure 11

Glaucoma vs. Control, Visual Field Sensitivity and Specificity. Abbreviations: CI = confidence interval; df = degrees of freedom.

Table 7

Glaucoma vs. Control, Humphrey Field Analyzer Sensitivity and Specificity.

The mean deviation on the SITA-Standard 24-2 was associated with a pooled AUROC (0.83, 95% CI 0.70 to 0.97, I²=88%), based on three studies (N=288).^64,74,108 (Figure 12) and the pattern standard deviation was associated with an AUROC of 0.87 (95% CI 0.76 to 0.99), based on two studies (N=242)^74,108 (Figure 13). One other study found the 76-point 30 degree suprathreshold associated with an AUROC of 0.87 (CI not reported).⁸⁹

Figure 12

Glaucoma vs. Control, AUROC Humphrey Field Analyzer Visual Field Mean Deviation. Abbreviations: AUROC = area under the receiver operating characteristic curve; CI = confidence interval; ES = estimate.

Figure 13

Glaucoma vs. Control, AUROC Humphrey Field Analyzer Visual Field Pattern Standard Deviation. Abbreviations: AUROC = area under the receiver operating characteristic curve; CI = confidence interval; ES = estimate.

One study (n=175; 280 eyes) found the Humphrey Field Analyzer, Swedish Interactive Threshold associated with sensitivity of 0.70 (95% CI 0.63 to 0.77) and specificity of 0.95 (95% CI 0.89 to 0.98).¹²² Another study (n=104; 182 eyes) found the Humphrey Field Analyzer, Armaly full field test associated with a sensitivity of 0.64 (95% CI 0.56 to 0.72) and specificity of 0.64 (95% CI 0.48 to 0.78).¹¹² Because these studies included more than one eye of some participants, they were not included in pooled analysis.

Tests of IOP Measurement

Tonometry

Seventeen studies (n=49,742) evaluated the accuracy of tonometry for identifying glaucoma.^{60,65,68,70,73,77,80,84,86,89,93,102,107,117,133,135,138} Tonometry was associated with a pooled sensitivity of 0.48 (95% CI 0.31 to 0.66) and pooled specificity of 0.94 (95% CI 0.90 to 0.96) for diagnosing glaucoma from non-glaucomatous or healthy eyes, based on 13 studies (N=32,892)^{65,68,73,77,80,84,86,102,107,117,133,135,138} (Figures 14 and 15; Table 8). The IOP cutoff was 21 to 22 mm Hg in all studies except for two, which used cutoffs of 22.6⁸⁴ and 25 mm Hg.⁹³ Results were similar when one study¹³⁸ that compared diagnostic accuracy for probable glaucoma with not probable glaucoma was excluded from the analysis (12 studies, N=28,726, pooled sensitivity 0.47, 95% CI 0.29 to 0.66 and pooled specificity 0.94, 95% CI 0.90 to 0.97). When stratified by tonometry method, sensitivity was higher for Goldmann tonometry (4 studies, N=11,690; sensitivity 0.66, 95% CI 0.36 to 0.87)^65,77,80,102 than for other methods (9 studies, N=21,202, sensitivity 0.39, 95% CI 0.22 to 0.58) (Table 8). However, the sensitivity estimate for Goldmann tonometry was imprecise. Specificity was similar regardless of tonometry technique. Results were also similar when the analysis was limited to fair-quality studies (11 studies, N=26,305, pooled sensitivity 0.54, 95% CI 0.34 to 0.72 and specificity 0.94, 95% CI 0.89 to 0.97).^{65,68,77,84,86,95,102,107,117,135,138} Only two studies were rated good quality^73,80 (sensitivity 0.24, 95% CI 0.19 to 0.30 and 0.19, 95% CI 0.07 to 0.39 and specificity 0.97, 95% CI 0.97 to 0.97 and 0.89, 95% CI 0.86 to 0.92). One study that included more than one eye per individual (n=3,039, eyes=6,060) reported a sensitivity of 0.07 (95% CI 0.01 to 0.19) and specificity of 0.99 (95% CI 0.99 to 0.99) for glaucoma versus non-glaucoma, based on an IOP threshold of >25 mm Hg using a rebound tonometer.⁹³ Two studies (N=418) found tonometry associated with low sensitivity (0.01, 95% CI 0.00 to 0.05 and 0.27, 95% CI 0.20 to 0.36) and high specificity (0.98, 95% CI 0.94 to 1.00 and 0.81, 95% CI 0.73 to 0.88) for distinguishing glaucoma suspects versus healthy controls.^86,117

Table 8

Tonometry Sensitivity and Specificity Glaucoma vs. Control.

Figure 14

Glaucoma vs. Control Tonometry. Abbreviation: HSROC = hierarchical summary receiver operating characteristic.

Figure 15

Glaucoma vs. Control Tonometry Sensitivity and Specificity. Abbreviations: CI = confidence interval; df = degrees of freedom.

In three studies (N=4,684), discrimination of tonometry based on the AUROC ranged from 0.66 to 0.78.^60,77,89 All three studies used Goldmann applanation tonometry (Figures 16 and 17).

Figure 16

Glaucoma vs. Control, Goldman Applanation Tonometry. Abbreviations: HSROC = hierarchical summary receiver operating characteristic.

Figure 17

Glaucoma vs. Control, Other Tonometry Techniques. Abbreviations: HSROC = hierarchical summary receiver operating characteristic.

One study (N=6,310) that evaluated intereye IOP asymmetry on tonometry⁷⁰ was not pooled; details are shown in Appendix B Tables 3 and 4.

Other

Summary

• One cross-sectional study (n=145) that was not in the prior CER found a questionnaire associated with low sensitivity (0.20, 95% CI 0.03 to 0.56) but high specificity (0.96, 95% CI 0.91 to 0.99) for identifying persons with glaucoma.

Evidence

One fair-quality, cross-sectional study (n=145) not in the prior screening CER² reported the diagnostic accuracy of a weighted screening questionnaire for identifying persons with glaucoma (Appendix B Tables 6–8).¹¹⁷ In the instrument, the highest weights were assigned for taking steroid medication and having a previous glaucoma diagnosis; less highly weighted risk factors were previous eye injury or stroke, age, race, prior eye surgery, high blood pressure, being nearsighted, and family history of diabetes or glaucoma. Two out of ten participants with glaucoma were correctly identified as having glaucoma based on the questionnaire alone (sensitivity 0.20, 95% CI 0.03 to 0.56), and 116 out of 121 correctly identified as not having glaucoma (specificity 0.96, 95% CI 0.91 to 0.99). The study was conducted in the U.S., but applicability to screening was likely limited because previous glaucoma diagnosis was one of the most heavily weighted risk factors.

Summary

• Treatment was associated with greater reduction in IOP compared with placebo or no treatment (16 trials, N=3,706, mean difference −3.14 mm Hg, 95% CI −4.19 to −2.08, I²=95%); there was an interaction between drug class and effects of medical therapy on IOP (p for interaction <0.0005), though estimates favored treatment for all drug classes.
• Treatment with topical therapy decreased risk of glaucoma progression compared with placebo or no treatment at 24 to 120 months (7 trials, N=3,771, RR 0.68; 95% CI 0.49 to 0.96, I²=53%; absolute risk difference [ARD] −4.8%, 95% CI −8.5 to −1.0).
• Evidence on effects of medical therapy on quality of life was very limited, with one trial (n=461) reporting no differences between latanoprost and placebo in general or vision-related quality of life measured using the EuroQol-5D (EQ-5D) (1.7 vs. 1.7, p=0.98), Short Form Health Survey-36 (SF-36) (4.8 vs. 5.0, p=0.94), Glaucoma Quality of Life-15 (GQL-15) (2.7 vs. 3.2 p=0.66), or Glaucoma Activity Limitation-9 (GAL-9) (3.0 vs. 3.2 p=0.87) scales at 24 months.

Evidence

The prior treatment CER³ primarily focused on head-to-head comparisons of glaucoma treatment, but included a systematic review that found medical (topical) therapy for ocular hypertension associated with reduced risk of onset of visual field defects compared with placebo or no treatment (10 trials, N=3,648, odds ratio [OR] 0.62, 95% CI 0.47 to 0.81).¹⁴⁷ For this review, we included 13 placebo-controlled trials and four trials of medical therapy compared with no treatment^{21,123,126,143} in patients with OAG or ocular hypertension. Nine trials^{21,78,87,92,94,114,126,127,143} were in the systematic review utilized in the prior treatment CER and we identified eight additional trials,^{35,62,63,121,123,124,134,142} including the U.K. Glaucoma Treatment Study³⁵ (UKGTS), which reported effects on quality of life and visual acuity in addition to IOP and visual field progression (Appendix B Table 9).

Across trials, sample sizes ranged from 20 to 1,636 participants (N=4,665). Mean age ranged from 55 to 74 years and 34 to 75 percent of participants were female. In 10 trials that reported race/ethnicity, the proportion of White participants ranged from 68 to 100 percent. Two trials enrolled patients with untreated, newly diagnosed OAG (excluding advanced disease and pigment dispersion),^35,63 three trials enrolled mixed populations with OAG or ocular hypertension (proportion with OAG 40%, 76%, and not reported),^62,124,142 and 12 trials enrolled patients with ocular hypertension (elevated IOP but normal visual fields; often also normal optic discs). OAG or ocular hypertension was diagnosed using a variety of tests, including perimetry, tonometry, gonioscopy, and visualization of the optic nerve by ophthalmoscopic examination and/or imaging; only the UKGTS³⁵ utilized OCT (primarily time domain-OCT) as part of the diagnostic evaluation. Mean baseline IOP ranged from 19.6 to 27.3 mm Hg; mean baseline IOP was <22 mm Hg in the two trials^35,63 of patients with early untreated OAG and ≥22 mm Hg in the other trials. Treatment was a beta-blocker in 10 trials (timorol in 6 trials,^{78,87,121,126,143} levobunolol in 2 trials,^62,123 and betaxolol in 2 trials^92,121) a carbonic anhydrase inhibitor in five trials (dorzolamide in 4 trials,^{63,114,115,124,142} and brinzolamide in 1 trial¹²⁴), a prostaglandin analogue (latanaprost) in one trial,³⁵ and an alpha agonist (brimonidine) in one trial.¹³⁴ One trial did not evaluate a specific drug but allowed various topical therapies, with a target IOP ≤24 mm Hg or ≥20 percent IOP reduction.²¹ The duration of followup ranged from 1.5 months^63,121 to 120 months,⁸⁷ with followup >1 year in 10 trials. One study was multinational,^114,115 and the others were conducted in the U.S.,^{21,62,78,94,121,124,127,134,142} U.K.,^35,92,143 Sweden,^63,87 Italy,¹²³ and Canada.¹²⁶

In 12 trials,^{21,35,62,63,78,87,92,114,115,121,124,126,142} randomization and analysis was per individual (2 of which enrolled only one eye^63,92); one trial¹²⁷ randomized by individual but reported a per-eye analysis; and three trials^94,134,143 randomized one eye in each individual (with the other eye serving as the control). One trial¹²³ reported a per-eye analysis, but randomization by eye or individual was unclear. Four trials were rated good quality,^{35,92,115,127} and 12 were rated fair quality^{21,62,78,87,94,121,123,124,126,134,142,143} (Appendix B Table 10). Methodological limitations in the fair-quality trials included unclear reporting of randomization, allocation concealment, and blinding methods, and high attrition in some studies.

Intraocular Pressure

Mean IOP was the most commonly reported outcome, reported in all but two trials.^{21,35,62,63,87,92,94,115,121,123,126,127,134,142,143} Overall, treatment was associated with greater reduction in IOP compared with placebo or no treatment (16 studies, N=3,706, mean difference −3.14 mm Hg, 95% CI −4.19 to −2.08, I²=95%) (Figure 18). Statistical heterogeneity was substantial, though inconsistency was in the magnitude of effect but not the direction of effect, with all trials reporting effects on IOP that favored treatment (range −0.70 to −7.00 mm Hg). A funnel plot did not indicate small sample effects (Egger’s test p=0.16) (Figure 19), but results were difficult to interpret due to statistical heterogeneity. There was an interaction between drug class and effects of medical therapy on IOP (p for interaction <0.0005), though estimates favored treatment for all drug classes. Beta blockers were associated with a pooled mean difference of −3.75 mm Hg (95% CI −5.43 to −2.06; I²=92%), based on nine trials (N=455); prostaglandin analogues with a mean difference of −2.70 mm Hg (95% CI −3.34 to −2.06), based on one trial (n=516); alpha agonists with a mean difference of −2.30 mm Hg (95% CI −3.52 to −1.08), based on one trial (n=30); and carbonic anhydrase inhibitors with a mean difference of −1.20 mm Hg (95% CI −2.30 to −0.61), based on four trials (N=1,635). One trial (n=1,636) found treatment to target using various medications associated with a mean difference of −4.60 mm Hg (95% CI −4.85 to −4.35). Estimates also consistently favored medical therapy, with differences ranging from −2 to −4 mm Hg, when analyses were stratified according to ocular hypertension, untreated OAG, or mixed status at baseline (Figure 20), baseline IOP (<20 mm Hg vs. ≥20 mm Hg) (Figure 21), and quality (fair vs. good) (Figure 22), or duration (<1 year vs. ≥1 year) (Figure 23), though statistically significant interactions were present (Table 9). In the OHTS, effects of medication compared with placebo on IOP were almost identical in Black persons and persons of other races.¹⁴⁸

Figure 18

Medical Treatment vs. Placebo/No Treatment on IOP, by Drug Class. Abbreviations: CI = confidence interval; IOP = intraocular pressure; NR = not reported; OAG = open angle glaucoma; OHT = ocular hypertension; SD = standard deviation.

Figure 19

IOP Funnel Plot. Abbreviations: IOP = intraocular pressure; s.e. = standard error.

Figure 20

Medical Treatment vs. Placebo/No Treatment on IOP, by Population. Abbreviations: CI = confidence interval; IOP = intraocular pressure; NR = not reported; OAG = open angle glaucoma; OHT = ocular hypertension; SD = standard deviation.

Figure 21

Medical Treatment vs. Placebo/No Treatment on IOP, by Baseline IOP. Abbreviations: BLIOP = baseline intraocular pressure; CI = confidence interval; IOP = intraocular pressure; mm Hg = millimeters mercury; NR = not reported; OAG = open angle glaucoma; (more...)

Figure 22

Medical Treatment vs. Placebo/No Treatment on IOP, by Quality. Abbreviations: CI = confidence interval; IOP = intraocular pressure; NR = not reported; OAG = open angle glaucoma; OHT = ocular hypertension; SD = standard deviation.

Figure 23

Medical Treatment vs. Placebo/No Treatment on IOP, by Duration. Abbreviations: CI = confidence interval; IOP = intraocular pressure; NR = not reported; OAG = open angle glaucoma; OHT = ocular hypertension; SD = standard deviation.

Table 9

Medical Treatment vs. Placebo/No Treatment, Pooled Analyses.

Progression of Glaucomatous Changes

Nine trials reported effects of topical medical therapies compared with placebo or no treatment on risk of glaucoma progression (Appendix B Table 9).^{21,35,78,87,92,94,114,115,126,143} Glaucoma progression was defined as progression of visual field defects,^35,78 progression to a glaucoma diagnosis (among those with ocular hypertension, based on visual field defects or optic disc changes),^87,92,126 or progression of visual field defects or optic disc changes.^21,114,115 Definitions and measurement methods for progression of visual field loss varied, but were based on the development of focal or reproducible visual field defects or by the development of any reproducible visual field defect (Appendix B Table 9). No trial reported the proportion of patients with overall visual field loss exceeding a minimum clinically important threshold such as a change in mean deviation of >3 to 5 decibels (dB, a measure of light intensity when testing visual fields).¹⁴⁹

Treatment with topical therapy decreased risk of glaucoma progression compared with placebo or no treatment (seven trials, N=3,771, RR 0.68; 95% CI 0.49 to 0.96, I²=53%; ARD −4.8%, 95% CI −8.5% to −1.0%; Figure 24) at 24 to 120 months. Estimates were similar in the new UKGTS trial,¹⁵⁰ which evaluated patients with untreated OAG (n=461, RR 0.59, 95% CI 0.41 to 0.86), and trials included in the prior treatment CER of patients with ocular hypertension (6 trials, N=3,310, RR 0.71, 95% CI 0.46 to 1.08; I²=57%) (Figure 25). Results were also similar in fair-quality studies (4 trials, N=1,978, RR 0.57, 95% CI 0.33 to 1.00, I²=44%) and good-quality studies (3 trials, N=1,793, RR 0.76, 95% CI 0.52 to 1.30, I²=15%; Figure 26). There was no interaction between medication type (p for interaction=0.30) or study quality (p for interaction=0.36) and effects on risk of glaucoma progression. Two trials^94,143 (ns=34 and 62) were not included in the meta-analysis because they randomized one eye in each individual and reported a per-eye analysis, but both found treatment associated with decreased risk of progression (RR 0.63, 95% CI 0.17 to 2.39 and RR 0.38, 95% CI 0.13 to 1.16). An analysis restricted to progression of visual field defects (excluding optic disc changes as a criterion for progression) produced similar results (6 trials, N=3,679, RR 0.73, 95% CI 0.53 to 1.05, I²=25%); Figure 27).^{21,35,78,92,114,126} In the UKGTS, latanaprost was associated with a small and non-statistically significant difference in overall visual field loss, measured by the mean deviation (−0.23 vs. 0.14 dB, p=0.07); there was no difference in visual acuity (−0.01 vs. −0.02 logMAR, p=0.9).³⁵ In the OHTS, there was no interaction between race and risk of progression from ocular hypertension to OAG (hazard ratio [HR] 0.50, 95% CI 0.28 to 0.90 for Black patients, and HR 0.36, 95% CI 0.23 to 0.57 for other races; p for interaction=0.40).¹⁴⁸

Figure 24

Medical Treatment vs. Placebo/No Treatment on Progression to Glaucoma. Abbreviations: CI = confidence interval; OAG = open angle glaucoma; OHT = ocular hypertension.

Figure 25

Medical Treatment vs. Placebo/No Treatment on Progression to Glaucoma, by Population. Abbreviation: CI = confidence interval.

Figure 26

Medical Treatment vs. Placebo/No Treatment on Progression to Glaucoma, by Quality. Abbreviations: CI = confidence interval; OAG = open angle glaucoma; OHT = ocular hypertension.

Figure 27

Medical Treatment vs. Placebo/No Treatment on Progression of Visual Field Defects. Abbreviations: CI = confidence interval; OAG = open angle glaucoma; OHT = ocular hypertension.

Summary

• There were no significant differences in risk of serious adverse events (3 trials, N=3,140, RR 1.14, 95% CI 0.60 to 1.99; I²=32%), withdrawal due to adverse events (5 trials, N=648, RR 2.40, 95% CI 0.71 to 19.32; I²=0%), or any adverse event (2 trials, N=1,538, RR 1.56, 95% CI 0.59 to 4.03; I²=82%).
• Two trials found treatment associated with increased risk of ocular adverse events (primarily itching, irritation, tearing, dryness, or taste issues) compared with placebo (RR 1.21, 95% CI 1.10 to 1.33 in a trial of various treatments and RR 3.52, 95% CI 2.46 to 5.02 in a trial of dorzolamide).

Evidence

Eight trials (in 9 publications) of medical treatments compared with placebo or no treatment reported harms (Appendix B Table 9).^{21,35,62,78,114,115,124,127,142} There were no statistically significant differences in risk of serious adverse events (three trials, N=3,140, RR 1.14, 95% CI 0.60 to 1.99; I²=32%; Figure 28),^{21,35,114,115} withdrawal due to adverse events (five trials, N=648, RR 2.40, 95% CI 0.71 to 19.32; I²=0%; Figure 29),^{35,62,78,127,142} or any adverse event (two trials, N=1,538, RR 1.56, 95% CI 0.59 to 4.03; I²=82%).^35,114,115 However, estimates were imprecise and the estimate for any adverse event was based on two trials, with substantial statistical heterogeneity (RR 1.04; 95% CI 0.73 to 1.47 in a trial of latanaprost³⁵ and RR 2.30, 95% CI 1.69 to 3.12 in a trial of dorzolamide^114,115). Two trials found treatment associated with increased risk of ocular adverse events compared with placebo (RR 1.21, 95% CI 1.10 to 1.33 in a trial of various treatments^114,115 and RR 3.52, 95% CI 2.46 to 5.02²¹ in a trial of dorzolamide). The most common ocular adverse events were localized itching, irritation, tearing, dryness, or taste issues. Because of extreme statistical heterogeneity (I²=94%), the pooled estimate was unreliable and not reported. The OHTS found no interaction between race and likelihood of experiencing one or more serious adverse events (p for interaction=0.16) or any adverse event (p for interaction=0.58) with medical treatment compared with placebo.¹⁴⁸

Figure 28

Medical Treatment vs. Placebo/No Treatment on Serious Adverse Effects. Abbreviations: CI = confidence interval; OAG = open angle glaucoma; OHT = ocular hypertension.

Figure 29

Medical Treatment vs. Placebo/No Treatment on IOP Withdrawals Due to Adverse Effects. Abbreviations: CI = confidence interval; IOP = intraocular pressure; OAG = open angle glaucoma; OHT = ocular hypertension.

Summary

• Three trials (N=1,875) found netarsudil to be noninferior to timolol for IOP lowering at 3 to 12 months.
• A pooled analysis of two trials (N=985) found netarsudil and latanaprost to be associated with similar effects on IOP (mean diference 0.3 mm Hg) and likelihood of IOP ≤18 mm Hg 57.4% vs. 65.5%) at 12 months.
• One trial (N=413) found latanoprostene bunod 0.024 or 0.040 percent associated with slightly greater effects on IOP (difference 1.2 mm) compared with latanoprost at short-term (1 month) followup.
• Two trials (N=840) found latanoprostene bunod associated with slightly greater IOP reduction compared with timolol at 3 months (difference −1.0 to −1.3 mm Hg).
• A pooled analysis of two trials (N=840) found latanoprostene bunod associated with increased likelihood of IOP ≤18 (20.2% vs. 11.2%; p=0.001) and IOP reduction ≥25 percent (22.9% vs. 19.0%; p<0.001) at 3 months.

Evidence

Eight trials and two meta-analyses evaluated the effects of latanaprostene bunod or netarsudil compared with an older glaucoma medication (Appendix B Table 11).^{57,58,66,91,98,113,129,139–141} The Rho Kinase Elevated IOP Treatment (ROCKET) trials compared netarsudil with timolol (ROCKET-1 and ROCKET-2¹²⁹ and ROCKET-4⁹⁸ trials); the VOYAGER study compared different doses of latanaprostene bunod with latanaprost;¹⁴⁰ the LUNAR¹¹³ and APOLLO^139,141 trials (acronyms not defined) compared latanaprostene bunod with timolol; and the MERCURY-1^58,66 and MERCURY-2⁵⁷ trials (acronym not defined) compared netarsudil with latanaprost.

Sample sizes ranged from 411 to 985 (N=4,113). The mean age ranged from 61 to 66 years and 50 to 68 percent of participants were female. All trials enrolled mixed populations of patients with OAG or ocular hypertension. The proportion of patients with OAG in the ROCKET trials ranged from 62 to 68 percent and in the MERCURY-1 and MERCURY-2 trials ranged from 72 to 77 percent; the three trials of latanaprostene bunod did not report the proportion of patients with OAG. Mean baseline IOP ranged from 20.7 to 26.7 mm Hg. The duration of followup was 3 months in all trials except for three, which had 1-¹⁴⁰ or 12-month followup.^66,91 All trials were multinational except for the MERCURY trials, which were conducted in the U.S. Three trials (LUNAR, APOLLO, and MERCURY-2)^57,113,141 were rated good quality and five trials were rated fair quality.^{66,91,98,129,140} Methodological limitations in the fair-quality trials included unclear reporting of randomization, allocation concealment, and blinding of outcome assessors; the ROCKET and MERCURY-1 trials also had high and differential attrition (Appendix B Table 12).^91,98,129 All of the trials focused on the outcome of IOP.

The ROCKET trials (N=1,875) found netarsudil to be noninferior to timolol in mean IOP reduction at 3 months (3 trials) and 12 months (1 trial).⁹¹ The MERCURY-1 trial (N=480) found netarsudil and latanaprost associated with similar reduction in IOP (mean difference 0.3 mm Hg for netarsudil vs. latanaprost) and likelihood of achieving IOP ≤18 mm Hg RR (57.4% vs. 65.5%, RR 0.73; 95% CI 0.61 to 0.88) at 12 months.⁶⁶

The short-term (1 month) VOYAGER trial (n=413) found latanaprostene bunod 0.024% and 0.04% associated with greater reduction in IOP compared with latanaprost (mean differences 1.23 mm Hg, 95% CI 0.37 to 2.10; 0.04% and 1.16 mm Hg, 95% CI 0.29 to 2.03, respectively) and increased likelihood of IOP ≤18 (68% vs. 64% vs. 47%, p<0.05 for both latanaprostene bunod doses vs. latanaprost).¹⁴⁰ Two trials (LUNAR and APOLLO, N=840) found latanaprostene bunod associated with greater IOP reductions than timolol.^113,141 In a pre-planned pooled analysis, mean differences in IOP ranged from −1.0 to −1.3 mm Hg at 3 months (p<0.001).¹³⁹ Latanaprostene was also associated with a higher likelihood of IOP ≤18 at week 2, week 6, and 3-month timepoints (20.2% vs. 11.2%; p=0.001) and IOP reduction ≥25 percent at all timepoints (2.9% vs. 19.0%; p<0.001).

Summary

• Netarsudil was associated with increased risk of ocular adverse events (3 trials, N=1,875, RRs ranged from 1.69 to 2.07), withdrawal due to adverse events (3 trials, N=1,875, RRs ranged from 4.73 to 38.20), and any adverse event (1 trial, n=708, 80% vs. 60%, RR 1.33, 95% CI 1.20 to 1.47) compared with timolol.
• Netarsudil also associated with increased risk of any adverse event (1 trial, n=480, RR 1.45, 95% CI 1.27 to 1.66), ocular adverse events (1 trial, n=480, RR 1.76, 95% CI 1.50 to 2.07) and withdrawal due to adverse events (1 trial, n=480, RR 12.82, 95% CI 4.71 to 34.85) compared with latanaprost at 12 months.
• One trial (n=413) found latanaprostene bunod and latanaprost associated with similar likelihood of any adverse events and withdrawals due to adverse events.
• Two trials (N=840) found latanaprostene bunod associated with increased risk of ocular adverse events compared with timolol (pooled RR 1.72; 95% CI 1.22 to 2.42); estimates for withdrawal due to adverse events were imprecise.

Evidence

Eight head-to-head trials and two meta-analyses reported adverse events associated with newly approved glaucoma medications compared with older medications (Appendix B Table 11).^{57,58,66,91,98,113,129,139–141} In the ROCKET trials (k=3, N=1,875), netarsudil was associated with increased risk of ocular adverse events compared with timolol.^91,98,129 The most common ocular adverse events were conjunctival redness or hemorrhage, corneal deposits (cornea verticillata, typically asymptomatic and without effects on vision), blurry vision, tearing, and itching. The proportion of patients with ocular adverse events ranged from 73 to 88 percent with netarsudil and from 41 to 50 percent with timolol; RRs ranged from 1.51 to 2.07 at 3 to 12 months (ARDs ranged from 26% to 38%). Netarsudil was also associated with increased likelihood of withdrawal due to adverse events (RRs ranged from 4.73 to 38.20; ARDs ranged from 8% to 34%), though estimates were imprecise. One trial found netarsudil associated with increased likelihood of any adverse event compared with timolol (80% vs. 60%, RR 1.33, 95% CI 1.20 to 1.47).⁹⁸

Netarsudil was also associated with increased risk of any adverse event (RR 1.45, 95% CI 1.27 to 1.66) and ocular adverse events (RR 1.76, 95% CI 1.50 to 2.07) at 12 months compared with latanaprost, based on one trial (n=480).⁶⁶ Netarsudil was also associated with increased risk of withdrawal due to adverse events compared with latanaprost at 3 months (2 trials, N=986, RR 7.40, 95% CI 2.94 to 18.65)⁵⁷ and 12 months (one trial, n=480, RR 12.82, 95% CI 4.71 to 34.85).⁶⁶

One short-term (1 month) trial (n=413) found no differences between latanaprostene bunod and latanaprost in risk of any adverse event or withdrawal due to adverse events.¹⁴⁰ Two trials (N=840) found latanaprostene bunod associated with increased risk of ocular adverse events compared with timolol (pooled RR 1.72, 95% CI 1.22 to 2.42).¹³⁹ Estimates for withdrawal due to adverse events were imprecise (RR 1.96; 95% CI 0.22 to 17.40 and RR 0.48, 95% CI 0.12 to 1.88).

Summary

• The large (n=718), good-quality Laser in Glaucoma and Ocular Hypertension (LiGHT) trial reported SLT and medical therapy were associated with similar effects on IOP, visual acuity, visual field, general quality of life, and glaucoma-specific utility, symptoms, and quality of life at 3 years.
• Three smaller, fair-quality trials found SLT and medical therapy similar for IOP at 4 to 12 months and 5 years; the trials did not evaluate other ocular and health outcomes.

Evidence

The prior treatment CER³ included two trials^103,151 (N=220) comparing SLT with medical therapy. One trial¹⁰³ was carried forward for this update but the other¹⁵¹ was ineligible because a high proportion of patients had capsular glaucoma and use of an outdated intervention (argon laser). Three additional trials (in four publications; N=925) not included in the prior treatment CER of SLT compared with a topical prostaglandin analogue were added for this update (Appendix B Table 13).^{37,82,118,119}

The largest study was the LiGHT trial, which enrolled 718 participants with OAG or ocular hypertension and visual acuity ~20/120 or better.^37,82 In the medical therapy arm of LiGHT, patients received stepped therapy with prostaglandins as initial therapy, followed by beta blockers, topical carbonic anhydrase inhibitors, or alpha agonists. Patients were excluded if they had prior surgery or were currently on or had prior exposure to glaucoma medical therapy. Forty-five percent of patients were female and mean age was 63 years. Seventy percent of patients were White, 20 percent Black, and 7.1 percent Asian. Mean baseline IOP was 24.5 mm Hg; the majority of randomized patients were diagnosed with OAG (77.3%) compared with ocular hypertension (22.7%). Of those with OAG, disease severity was most commonly assessed as mild (88.6%), followed by moderate (20.1%) and severe (10.4%).

The sample sizes in three smaller trials (including the trial in the prior treatment CER)¹⁰³ ranged from 32 to 167 participants; 48 percent to 55 percent were female.^103,118,119 Mean ages ranged from 52 years to 66 years and mean baseline IOP ranged from 22.8 mm Hg to 29.3 mm Hg. The proportion of patients with OAG ranged from 43 to 59 percent and the proportion with ocular hypertension ranged from 41 to 57 percent. All studies excluded patients with prior laser or glaucoma surgery. In two trials,^103,118 patients were randomized to 360° SLT; in the other trial,¹¹⁹ patients were randomized to 90°, 180°, or 360° SLT. The duration of followup ranged from 4 months to 5 years. Three trials were conducted in the United Kingdom (U.K.)^{37,82,118,119} and one in Hong Kong.¹⁰³

One RCT^37,82 was rated good quality; the remaining three were rated fair quality (Appendix B Table 14).^103,118,119 Methodological limitations in the fair-quality trials included unclear reporting of randomization, allocation concealment, and blinding methods.

Summary

• The LiGHT trial reported similar adverse and serious adverse event rates between SLT and medical therapy; evidence on harms from three smaller trials was limited.

Evidence

The LiGHT trial found no differences between SLT and medical therapy in patients experiencing any adverse event (73.3% vs. 71.8%) or any serious adverse event (18.0% vs. 18.8%) (Appendix B Table 13).^37,82 SLT and medical therapy were associated with similar risk of ocular adverse events such as ocular irritation, retinal hemorrhage, or floaters (52% vs. 61%) and serious ocular adverse events such as trauma, central retinal artery occlusion, or choroidal neovascularization (2.2% vs. 1.7%). In the SLT group, 34.4 percent of patients experienced SLT-related transient adverse events, including discomfort, transient blurred vision, transient photophobia, and hyperemia. Evidence on harms of SLT compared with medical therapy from other trials was limited, due to suboptimal reporting and imprecision.^103,118,119