PERIMETRIC OUTCOMES OF MELBOURNE RAPID FIELDS IN PATIENTS WITH GLAUCOMA and DEMOGRAPHIC AND DIAGNOSTIC CHARACTERISTICS OF VIRTUAL REALITY VISUAL FIELD TESTING IN PRIMARY CARE GLAUCOMA SCREENING

Abstract

MANUSCRIPT 1 Purpose: The Humphrey Field Analyzer (HFA) is the most widely used perimeter in glaucoma, but its cost and low portability limit its use in resource-constrained settings. Melbourne Rapid Fields (MRF) is a low-cost, tablet-based perimeter system that enables at-home testing and testing in low-resource environments. Despite growing evidence that the MRF can obtain reliable estimates of the visual field, no study has synthesized its perimetric performance through meta-analysis and compared it with HFA. Methods: Following PROSPERO registration, we conducted a systematic review to compare MRF with HFA in patients with glaucoma. Searches were performed using controlled vocabulary related to perimetry assessment with these devices. Meta-analyses and narrative synthesis were conducted to summarize their performance. Results: Of 290 studies screened, 9 were included, aggregating data from 777 eyes. In pooled data comparing MRF to HFA, MRF had faster testing (-1.33 minutes), less-negative MD (+2.08 dB), and higher PSD (+0.77 dB). Global agreement with HFA was high (MD ICC = 0.94; MD R² = 0.85; PSD ICC = 0.85). Test-retest reliability was similar between devices (MRF ICC = 0.96; HFA ICC = 0.95). AUCs were comparable for MD (0.84 vs 0.85) but lower for PSD (0.81 vs 0.93). MRF had higher fixation losses, false positives, and false negatives. Conclusion: MRF had shorter test durations, high global agreement with HFA, excellent repeatability, and a bias toward less negative global results. MRF showed greater pointwise variability at certain thresholds and worse reliability measures. Despite these differences, MRF returned diagnostic information comparable to HFA. Précis: MRF had shorter test durations than HFA, high agreement with HFA global indices, modest bias in MD and PSD measurements, and comparable concordance and repeatability. MRF had poorer reliability indices, while maintaining good diagnostic performance. MANUSCRIPT 2 Purpose: To evaluate the false-positive proportion and rater reliability of virtual reality-based visual field (VR-VF) perimetry within a primary care-embedded glaucoma screening program. Design: Pre-specified sub-study of an ongoing comparative effectiveness glaucoma screening trial (Glaucoma and Retinopathy Screening Study [GRaSS]) (NCT06882356). Participants: Adults ≥40 years diagnosed with diabetes receiving routine care at Mass General Brigham community health centers. Methods: Within the parent GRaSS trial, patients with diabetes undergoing standard-of-care diabetic eye disease screening with an autonomous AI (AAI) were screened for glaucoma using an investigational AAI algorithm, intraocular pressure (IOP) measurement, and confirmatory VR-VF testing in AAI-flagged individuals or those with IOP ≥24 mmHg. This sub-study invited a predefined low-risk cohort (not AAI-flagged and bilateral IOP mmHg) to undergo VR-VF testing to characterize VR-VF performance in a normative group. Two masked glaucoma-trained clinicians dichotomized VR-VFs using predefined criteria; intra-rater reliability was assessed by random sequential repeat grading. Optic nerve morphology on fundus photographs, as adjudicated by a human expert, served as the reference standard. False-positive VR-VF was defined as abnormal VR-VF in participants with bilaterally normal optic nerves and IOP mmHg. Logistic regression evaluated the association between Hispanic ethnicity and false-positive VR-VF. The ZIP code-level sociodemographic context of sites and participants was characterized using the Social Deprivation Index (SDI) and American Community Survey data. Main Outcome Measures: Primary outcome: proportion of false-positive VR-VF among normal participants. Secondary outcome: inter- and intra-rater agreement for VR-VF interpretation. Exploratory outcomes: association between Hispanic ethnicity and false-positive VR-VF and sociodemographic characteristics of recruitment sites and participants. Results: Sociodemographic data were collected for 145 participants; 109 completed VR-VF testing (218 eyes). VR-VF had excellent inter-rater reliability (Cohen κ = 0.96) (95% CI: 0.91-0.99). Intra-rater reliability was high, with Cohen κ of 0.90 (95% CI: 0.83-0.96) and 0.91 (95% CI: 0.85-0.96). Among 81 normal participants, 44.4% (95% CI: 33.6%-55.2%) had false-positive VR-VF results. Hispanic ethnicity was associated with higher odds of false-positive VR-VF after adjustment for age, sex, and SDI (aOR, 3.37) (95% CI: 1.17-9.63). Participants predominantly resided in highly deprived areas. Conclusions: VR-VF demonstrated excellent inter- and intra-rater agreement but yielded a high false-positive rate in this primary care glaucoma screening context.