Key Questions and Analytic Framework

The scope and Key Questions were developed by the Evidence-based Practice Center (EPC) investigators, USPSTF members, and Agency for Healthcare Research and Quality (AHRQ) Medical Officers using the methods developed by the USPSTF.⁴⁹ The analytic framework and Key Questions that guided the review are shown in Figure 1. In the Key Questions, “OAG” refers to POAG patients and glaucoma suspects. Eleven Key Questions were developed for this review:

Figure 1

Analytic Framework and Key Questions: Glaucoma. *Includes open-angle glaucoma suspects. Note: Subpopulations of interest include those defined by age, sex, race/ethnicity, and setting (e.g., rural or urban), etc.

Search Strategies

A research librarian searched the Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews, and MEDLINE (January 2011 to February 9, 2021), for relevant studies and systematic reviews. The search relied primarily on the previous systematic review for the USPSTF to identify potentially relevant studies published before 2011 (we reassessed all articles included in that systematic review using the eligibility criteria). Search strategies are available in Appendix A1. To supplement electronic searches, we reviewed reference lists of relevant articles. Ongoing surveillance was conducted to identify major studies published since February 2021 that may affect the conclusions or understanding of the evidence and the related USPSTF recommendation. The last surveillance was conducted on January 21, 2022, and identified no studies affecting review conclusions. One retrospective observational study⁵⁰ comparing glaucoma screening to no screening was identified during surveillance but was not eligible for inclusion due to observational design and serious methodological limitations (control group was non-participants/non-responders and study did not control for potential confounders).

Study Selection

At least two reviewers independently evaluated each study to determine eligibility. We selected studies on the basis of inclusion and exclusion criteria developed for each Key Question (Appendix A2).

Articles were selected for full-text review if they were about OAG or glaucoma suspect in adults 40 years of age or older, were relevant to a Key Question, and met the pre-defined inclusion criteria. We excluded studies of patients with narrow-angle glaucoma, secondary OAG (including exfoliation glaucoma), or advanced glaucoma (e.g., with severely impaired vision). We restricted inclusion to English-language articles and excluded studies published only as abstracts. Studies of non-human subjects were also excluded, and studies had to report original data.

For screening, we included studies on a complete eye examination (as defined in the studies), various components of a complete eye examination (ophthalmoscopy, perimetry, tonometry, pachymetry, evaluation for afferent pupillary defect), and imaging tests (optic disc photography, optical coherence testing [OCT], and fundus photography). We excluded screening tests that are considered outdated or no longer used, such as the water drinking test, the Heidelberg Retina Tomograph, scanning laser polarimetry, and older OCT technology (time-domain OCT). For treatment, we included first line medical treatments (defined as prostaglandin analogues, beta-blockers, alpha2 agonists, and carbonic anhydrase inhibitors), SLT, and newly FDA-approved medical treatments (latanoprostene bunod and netarsudil). We excluded studies of combination treatment and trabeculectomy, which are not considered first line therapy for ocular hypertension or early glaucoma, and outdated therapies (e.g., argon laser trabeculoplasty). The comparison for screening was no screening and the main comparison for treatment was placebo or no treatment. We also included head-to-head trials that compared latanoprostene bunod or netarsudil with first-line medical therapies. For screening, referral, and treatment, outcomes were IOP, visual field loss, visual acuity, optic nerve damage, visual impairment (defined as visual acuity <20/70 or <20/100), quality of life, function, and harms (e.g., eye irritation, corneal abrasion, infection, anterior synechiae, cataracts), reported at least 4 weeks after initiating the intervention. We included randomized controlled trials (RCTs) of screening and treatment and cohort and cross-sectional studies on screening test diagnostic accuracy. We excluded diagnostic accuracy studies that used a case-control design, due to potential spectrum bias.⁵¹ Telemedicine studies of screening were included if they were conducted in primary care settings. Studies on imaging test diagnostic accuracy that utilized artificial intelligence to analyze images were included if they evaluated a clinical cohort (e.g., did not analyze images in a databank), did not use a case-control design, reported validation testing, and utilized algorithms available for widespread use. Studies on screening accuracy were not restricted by clinical setting, although results from primary care settings were highlighted if available. This report utilized primary studies and systematic reviews were used to identify potentially eligible studies. In accordance with USPSTF methods, studies rated poor quality (see below) were excluded.

The selection of literature is summarized in the literature flow diagram (Appendix A3). Appendix A4 lists the included studies, and Appendix A5 lists the excluded studies with reasons for exclusion.

Data Abstraction and Quality Rating

For studies meeting inclusion criteria, we created data abstraction forms to summarize characteristics of study populations, interventions, comparators, outcomes, study designs, settings, and methods. One investigator conducted data abstraction, which was reviewed for completeness and accuracy by another team member.

Predefined criteria were used to assess the quality of individual studies by using criteria developed by the USPSTF. Studies were rated as “good,” “fair,” or “poor” per USPSTF criteria, depending on the seriousness of methodological shortcomings (Appendix A6).⁴⁹ For each study, quality assessment was performed by two team members. Disagreements were resolved by consensus.

Data Synthesis and Analysis

We performed a random effects meta-analysis using the profile likelihood model to summarize the effects of first-line medical treatments versus placebo or no treatment on likelihood of glaucoma progression (based on progression of visual field loss, with or without optic nerve changes), serious adverse events, and withdrawal due to adverse events and mean IOP. Glaucoma progression, serious adverse events, and withdrawal due to adverse events were evaluated as dichotomous outcomes using the relative risk. IOP was evaluated as a continuous outcome using the mean difference. For mean IOP, adjusted differences were utilized when reported; otherwise, the difference in followup IOP was utilized when available, followed by the difference in change from baseline. Further, differences based on per-individual data were used when available. For trials that randomized each individual to a treatment but reported a per-eye analysis (i.e., two eyes per individual), the mean IOP was averaged between the two eyes and the standard deviation (SD) for the mean IOP was calculated by assuming a correlation of 0.5 between an individual’s eyes. For trials in which one eye in each individual was randomized to the medical treatment and the other eye received the control treatment, the mean difference based on the within-subject comparison was utilized. If the SD for the within-subject mean difference was not reported, it was calculated based on the reported SD for each treatment group, again, assuming a correlation of 0.5. When the SD for the followup IOP was not reported, it was imputed using the average coefficient of variation from other included trials. Comparable interventions within the same study were combined in the primary analysis, so each study was represented once in a meta-analysis in order to avoid overweighting. Analyses were stratified by the type of medical treatment (alpha agonist, prostaglandin analogue, beta-blocker, carbonic anhydrase inhibitor, or mixed) and prespecified study-level subgroup analyses were conducted on the following factors: glaucoma status (OAG, ocular hypertension, or mixed); quality (good or fair); mean IOP (<20 vs. ≥20 mm Hg), and duration of followup (<1 year vs. ≥1 year). For glaucoma progression, a sensitivity analysis restricted to trials that defined progression based on visual field loss (excluding optic nerve changes) was conducted. Statistical heterogeneity was assessed using the Cochran Q-test and I² statistic.⁵² When at least 10 studies were available for meta-analysis, we tested for small sample effects using graphical (funnel plot) and statistical (Egger’s test) methods. All meta-analyses were conducted using Stata 14.2 or Stata/SE 16.1 (Statacorp, College Station, Texas).

For diagnostic accuracy, a bivariate logistic random effects model was used to summarize sensitivity and specificity of screening tests simultaneously for identifying glaucomatous eyes from those without glaucoma (healthy eyes, glaucoma suspect, or ocular hypertension). A bivariate model was used to account for the correlation between sensitivity and specificity, to produce summary values for sensitivity and specificity with corresponding 95 percent confidence intervals (CIs). For the bivariate model, at least four studies were needed to pool. Meta-analysis was restricted to studies that used one eye per individual; studies that used both eyes were not included in the meta-analysis because they did not report the correlation between eyes and inclusion would result in overweighting. When studies reported a range of testing cutoffs, data based on the most commonly used cutoff (e.g., IOP >21 mm Hg) or closest to it were utilized. For one study⁵³ that reported sensitivities across multiple specificities without reporting a cutoff, the sensitivity and specificity pair with the fewest misclassifications were utilized. Meta-analysis using a random effects Dersimonian-Laird model was also performed to summarize the area under the receiver operating characteristic (AUROC) curve as reported in individual studies. Stratified analyses were conducted based on the type of control (healthy eye, glaucoma suspect, or ocular hypertension) and study quality (good or fair). Sensitivity analyses were conducted on factors related to specific imaging tests: for retinal nerve fiber layer on spectral domain-OCT, sensitivity analysis was restricted to studies that measured retinal nerve fiber layer based on the mean thickness; for ganglion cell complex on spectral domain-OCT, sensitivity analysis was restricted to studies that utilized measures of the retinal nerve fiber layer, inner plexiform layer, and the ganglion cell layer; and for studies of tonometry, sensitivity analysis was restricted to studies that measured IOP using Goldmann tonometry. Statistical heterogeneity was assessed using the I²; however, this value is often high and difficult to interpret in diagnostic accuracy studies because it is dependent on sample size and is a univariate measure that does not account for variability in sensitivity or specificity estimates due to threshold effects.

For all Key Questions, the overall strength of evidence was determined using the approach described in the USPSTF Procedure Manual.⁴⁹ The strength of evidence was rated “high”, “moderate”, “low” or “insufficient” based on study quality, consistency of results between studies, precision of estimates, study limitations, and risk of reporting bias.⁴⁹ Additionally, the applicability of the findings to U.S. primary care populations and settings was assessed. Discrepancies were resolved through consensus discussion.

USPSTF and AHRQ Involvement

The authors worked with USPSTF liaisons at key points throughout the review process to develop and refine the analytic framework and Key Questions and to resolve issues around scope for the final evidence synthesis.

AHRQ staff provided oversight for the project, coordinated systematic review, reviewed the draft report, and assisted in an external review of the draft evidence synthesis.

Expert Review and Public Comment

The draft research plan was posted for public comment from February 13, 2020 to March 11, 2010. The comments were reviewed and no changes to the scope or Key Questions were required, though some edits were made for clarity. The eligibility criteria table (Appendix A2) was revised to clarify included and excluded diagnostic tests for glaucoma; tests that are no longer used were excluded. A final research plan was posted on the USPSTF’s Web site on June 11, 2020.

A draft version of this report was reviewed by content experts and Federal partner representatives (Appendix A7), and edits were made for clarity. In addition, the draft was posted for public comment from October 26, 2021 to November 22, 2021. The comments were reviewed and minor edits were made for clarity. However, no changes to the studies or findings were required.