Secondary Logo

Journal Logo

New Glaucoma Insights: Review Article

Home-Based Perimetry for Glaucoma: Where Are We Now?

Daka, Qëndresë MD, PhD*,†; Mustafa, Rona MD*; Neziri, Burim MD, PhD*,†; Virgili, Gianni MD, PhD‡,§; Azuara-Blanco, Augusto MD, PhD

Author Information
doi: 10.1097/IJG.0000000000002022

Abstract

Glaucoma can permanently damage vision if left untreated and is accounted among the leading causes of irreversible blindness worldwide.1,2 The disease is projected to affect 112 million by 2040, with the number of affected individuals remaining much higher than the number of diagnosed patients.3–7 Besides late diagnosis, delayed follow-up due to insufficient capacity within hospital eye services is also responsible for severe vision loss from glaucoma.7,8

Increased glaucoma case load will be challenging for health care systems with the demand expected to exceed the supply due to shortages of ophthalmologists and greater life expectancy of glaucoma patients.9,10 Visual field (VF) assessment, currently done in ophthalmology units, is a cornerstone of glaucoma diagnosis and the most important test to monitor disease progression and treatment efficacy.11–14 A minimum of 6 VF tests over the first 2 years after diagnosis has been recommended to identify the rate of disease progression and risk for visual impairment. Frequent testing is required in patients with higher risk of severe functional loss, helps reduce the noise effect due to variability, and facilitates early detection of change.15,16 A downside of increased VF testing frequency is the location of the equipment at specialized services (ie, eye clinics) and the limited clinic capacity, which causes backlogs especially now in the COVID-19 period, which has disrupted in-clinic perimetry.17,18 Furthermore, visits to eye clinics are time consuming, expensive, and inconvenient for patients and their careers.19 Therefore, frequency of VF testing may be done substantially less often than recommended contributing to visual impairment of individuals.15

New remote models of health care such as telehealth or virtual clinics have accelerated as reasonable options to overcome insufficient clinical capacity and improve patient outcomes.8,10 These strategies reduce clinic visits and have been proposed for glaucoma detection and monitoring.10,19 Portable perimeters designed for home use appear as a potential tool for remote home monitoring of VF in order to reduce the number of clinic visits.20 They have a theoretical advantage of providing an abundance of VF data, which may help the ophthalmologist to confirm disease in glaucoma suspects and to possibly assess disease progression.10,15,16 A computer simulation study has shown that weekly home monitoring improved early detection of rapid VF loss in glaucoma even when the compliance was moderate and test variability increased.16 In addition, home perimeters may help improve patient engagement with their disease and potentially improve adherence with treatment and monitoring recommendations.19,20

Technology has made possible development of innovative portable perimeters that use tablets, personal computers, or head-mounted displays (HMD). Such devices may be used for assessing VF loss without clinical visits.19 However, what is still uncertain is their performance and ability to detect and quantify the severity of VF damage, which is an initial step of technology evaluation. For these reasons, we completed a systematic review assessing the diagnostic accuracy, repeatability, usability, and acceptability of the different home-based perimeters for glaucoma diagnosis and detection of disease progression during monitoring.

METHODS

The review was performed following the format of Preferred Reporting Items for Systematic Reviews and Meta-Analyses Diagnostic Test Accuracy Studies (PRISMA-DTA) statement checklist.21 The full protocol was registered in PROSPERO International Prospective Register of Systematic Reviews and can be retrieved from: https://www.crd.york.ac.uk/prospero (registration no. CRD42020226167).

This systemic review was conducted in accordance with the PICO (problem or population, intervention, comparison, and outcome)framework.22

Studies evaluating devices by direct head-to-head verification with standard automated perimetry (SAP), regardless of publication status, time, location, examiner, or setting were included. Studies describing the development of the test or reporting outcomes that were not relevant to evaluate diagnostic accuracy or usability (eg, comparing retinal sensitivities at different locations) were not considered. We also excluded studies that enrolled participants with significant visual loss associated with other eye diseases. Editorials without primary research data, as well as studies and abstracts published in non-English language were excluded.

Publications were included if: (a) design of the study was cross-sectional (glaucoma diagnosis) or longitudinal (glaucoma progression); (b) included adult participants of both sexes; (c) participants have undergone VF testing with any portable perimeter designed for home use and any SAP test as comparator; (d) assessed test sensitivity and specificity for glaucoma diagnosis (primary outcome) or detection of disease progression, test-retest repeatability at one time point, proportion of patients able to use the test and reasons for failure, or acceptability of the test (secondary outcomes).

The literature searches included search of the electronic databases: PUBMED, SCOPUS, and clinicaltrials.gov. Last search date was May 2021. We also used handsearch of reference lists of relevant publications. A comprehensive search strategy was designed to retrieve the highest number of relevant publications with the assistance of a medical librarian. For details regarding the search strategy see Table, Supplemental Digital Content 1 (https://links.lww.com/IJG/A603).

Studies identified were screened initially based on titles and abstract, and retrieved in full text for final assessment of the eligibility criteria for inclusion. Study authors were contacted when available information was not sufficient. The process was conducted independently by 2 authors (Q.D. and B.N.), whereas a third author (A.A.-B.) was involved to resolve discrepancies.

A standardized data extraction form to extract trial data was predesigned. Two review authors (Q.D. and R.M.) extracted data from included studies independently and inconsistencies were resolved by involvement of a third author (A.A.-B.). The following data were recorded: study characteristics (author, publication year, country, funding); study aim; study design (type, time, setting); target condition (recruitment, definition, examinations, diagnosis); participant characteristics (selection criteria, age, gender); interventions (index test and reference standard characteristics, test conduction, test interpretation); and relevant outcomes. When available, whole-field analysis rather than quadrant, zone, or point analysis data were extracted for diagnostic accuracy outcomes.

Risk of bias of studies that reported test diagnostic accuracy data was evaluated with QUADAS-2 tool that comprises 4 domains: patient selection, index test, reference standard, and flow and timing.23 For more transparent and consistent evaluation, we developed a QUADAS-2 guidance (Table, Supplemental Digital Content 2, https://links.lww.com/IJG/A604). The quality of studies was evaluated independently by 2 review authors (Q.D. and A.A.-B.) and final assessment was decided by consensus and discussion with a third author (G.V.). Study authors were contacted to clarify study details. Studies that did not report diagnostic test accuracy data were not evaluated with QUADAS-2 tool.

All data considered in the review were reported in a narrative and tabular format as extracted from primary studies. Outcome measures were reported as used in the primary studies. Due to the novelty of these techniques, all threshold levels, cutoffs, and subgroup analysis for diagnostic accuracy are presented. For meta-analysis, we aimed to pool whole-field test sensitivity and specificity for detection of glaucoma (as defined in primary studies). However, due to sparse and heterogeneous data a meta-analysis was not conducted, and accuracy data were presented graphically in forest plots, which were obtained using Review Manager (RevMan) [Computer program]. Version5.3. Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014.

RESULTS

A total of 2340 records were identified initially. Following screening, 2305 were excluded due to duplicity, irrelevance, or study design, whereas 35 were retrieved in full text for further evaluation of eligibility. Finally, 18 studies were found to be eligible (Fig. 1). List of the excluded studies in the last step together with the first reason noted for exclusion is provided in Table (Supplemental Digital Content 3, https://links.lww.com/IJG/A605).

F1
FIGURE 1:
PRISMA flow chart of study selection.

Characteristics of the Included Studies

Descriptions of characteristics of the included studies are outlined in Table 1. Their publication dates ranged between 2005 and 2021. The majority were cross-sectional clinical-based studies and were conducted in USA,29,37,40,41 India,27,30 UK,26,31 Australia,33,36 Singapore,24 Greece,34 Nepal,35 Denmark,38 and Philippines.39 Two home-based longitudinal studies were conducted in UK25 and Australia28, while one was a multicenter clinical-based study conducted in UK and India.32 Participants were recruited from clinics and had detailed ophthalmological examinations, except in 5 studies24,32,34,40,41 where examinations were not reported. Three studies25,28,32 included patients with glaucoma or ocular hypertension, 11 studies24,26,27,29–31,33,36–38,41 included glaucoma patients and healthy controls, 3 studies35,39,40 also included patients with other diagnosis, and 1 study34 did not specify participant diagnosis.

TABLE 1 - Main Characteristics of Included Studies
References Aim Study Design Condition Population Intervention Outcomes
Ding et al,24 Singapore Empirically derive a recommended test stimulus and evaluate its feasibility Prospective Cross-sectional Clinic No data on recruitment and examinations GON: established disease Healthy: no known VF defects 66 GON; 30 healthy Phase 1: 29 GON/0 healthy Phase 2: 17 GON/0 healthy Phase 3: 57 GON/30 healthy Included: ≥21 y; BCVA≥6/12; reliable 24-2 HVF within 1 mo Excluded: preperimetric disease; 10-2 and unreliable HVF Age (mean y): GON: 67.1±11.1 Healthy: 57.6±12.4 Gender (F%: GON/Healthy): 36.4/80 Reference standard: SAP HFA; SITA-Standard 24-2 Time: mean 6.92±1.12 min GON/5.19±0.697 min controls Clinic Index test: VFF; entopic; same as 24-2 Time: mean GON: 3.60±1.85 min; controls:1.11±0.486 min Phase 1: establish stimulus: variable size; 30 Hz; BW Phase 2: establish frequency: variable Hz, BW; 2×2 Phase 3: test of recommended 2×2 stimulus; 30 Hz, BW Clinic Diagnostic accuracy Test-retest repeatability
Jones et al,25 UK FFS, IGA, CoO Assess accuracy and adherence in a pilot sample Prospective Longitudinal Cohort 6 mo Clinic & home Recruited first 20 responders. Established diagnoses. Assessed by a glaucoma optometrist: history, VA, and VF 18 POAG/NTG; 1 ACG; 1 SEG Included: BCVA>0.5 in the better eye Excluded: surgery or laser within 6 mo before participation Age (median, y): 71 (62–78) Sex (F, %): 50 Reference standard: SAP HFA 3; SITA Fast 24-2 Time: median 3.9 min (quartiles: 3.3–4.6) Clinic: 4 per eye; 2 baseline & 2 after Index test: Eyecatcher modified; ZEST threshold 24 points of the 24-2 grid; below sensitivity set to12.6 dB Time: median 4.5 min (quartiles: 3.9-5.2) Home: 1 test per eye/month; recorded; remind send 10 with and 10 no practice; support available Accuracy Usability Acceptability
Jones et al,26 UK FFS Examine the feasibility of applying in a busy GON clinic as a triage test to identify high-risk and false-positive Prospective Case-control Clinic Recruited opportunistically from GON clinic. Full visual assessment by local clinical team 77: 44 POAG; 5 NTG; 2 PAC; 2 ACG; 1 PG; 8 OHT, 3 complex; 3 OHT/borderline GON; 10 nil abnormal (new referrals) One eye randomly selected Included: capacity to provide written informed consent Excluded: no attempt to select/filter Age (median, y): 70 (59–77) Sex (F, %): 43 Reference standard: SAP HFA; SITA Fast 24-2 Time: median 3.5 min (3.3-4.1) Clinic: data extracted from patients’ medical records Index test: Eyecatcher V2.0; updated; eye tracker 6 dB more intense stimulus than the expected threshold 22 locations (11of the 24-2) Time: median: 2.5 min (2.4-2.7) Clinic: twice tested Diagnostic accuracy Test-retest repeatability Usability Acceptability
Pradhan et al,27 India SRDII Compare the reliability indices; threshold sensitivity and subjective experience Prospective Cross-sectional Clinic Recruited: patients and suspects from GON clinic; staff or those who came in routine exam Exam: history, BCVA, VF, IOP, fundus, biomicroscopy GON: ONH, RNFL, VF changes Suspects: suspicious ONH Healthy: IOP≤21 mm Hg, normal 54 initially; 42 analyzed (84 eyes) 21 GON; 14 suspect; 49 normal Included: ≥18 y; BCVA≥20/40; URE within±5 Dsph and ±3 Dcyl. Excluded: trauma history or inflammation; intraocular surgery within 6 mo; presence of any retinal or neurological disease Age (mean, y): 46.4±17.8 (19-83) Sex (F, %): 50 Reference standard: SAP HFA II; full threshold 24-2 Clinic: unreliable if FP or FN >33% Index test: GV; threshold 24-2; Goldman size III; measures a minimum threshold of 8 dB Clinic: blind spot located initially Time: median: 11 min (10-11.5) Usability Acceptability
Prea et al,28 Australia ORIA Examine the short-term uptake, compliance, and performance of home monitoring Prospective Longitudinal Cohort 6 mo Clinic & home Recruited during clinic review. SG: any form of GON controlled by medication and/or surgery; no signs of progression on exam, VF, and structural scan. 101 (186 eyes) initially: 9 normal; 86 POAG; 66 other GON; 25 GS Included: OHT or SG; VA≥20/40; able to understand English instructions; at least 2 HFA 24-2 SITA before trial Excluded: ocular surgery or changed medications in the preceding 6 mo Age (median, y): 66.5 (21-89) Sex (F, %): 32 Reference standard: SAP HFA; SITA standard 24-2 Time: the average 6.4±1.2 min Clinic:1 baseline; 1 after Index test: MRF app; threshold; 66-point radial test grid Time: the average 5.2±1.2 min Home: 12:1 test per eye; weekly interval; reminder sent 78 (127 eyes) ≥2 tests; 14 data only for 1 eye Accuracy Usability Acceptability
Razeghinejad et al,29 USA Olleyes Inc. Investigate the performance Prospective Cross-sectional Clinic Recruited: patients-GON clinic; hospital staff and volunteers. Exam: history, VA, IOP, fundus, biomicroscopy, and gonioscopy. GON: controlled OAG Healthy: normal retina, ON, and HVF; IOP<21 mm Hg. 55 initially; 51 analyzed (102 eyes) 26 OAG; 25 healthy subjects Included: reproducible abnormal SAP Excluded: conditions or medications affecting vision or reaction time; URE >±5.00 Dsph and >±2.00 Dcyl; BCVA<20/30; closed angle; ocular surgery within 6 mo (except cataract/kerato-refractive) Age (mean, y): GON:66.04 (23-86); healthy: 53.96 (30-79) Sex (F, %: GON/healthy): 50/68 Reference standard: SAP HFA; SITA 24-2 Time: 5.62 min GON/ 4.77 min controls Clinic: FP, FL, and FN ≤15% Index test: VisuALL; eye-tracking; T threshold Central 50 locations of 24 degrees in a 6 degrees grid pattern that straddles the horizontal and vertical midlines Time:9.28 min GON/6.13 min controls Clinic Diagnostic Accuracy Usability Acceptability
Mees et al,30 USA Federal grant Investigate the ability to identify glaucoma Prospective Case-Control Clinic in India Recruited patients and family members from clinic. Exam: OCT, HFA, IOP, fundus, biomicroscopy. Classified by one physician initially; two others masked to the first final. 227 Initially;199 Tested 157 analyze: 62 GON/95 controls Included: VA≥20/40; reliable HFA GON: previous diagnosis; Controls: family; no retinal disease or GON Excluded: 28—either unreliable HFA or could not complete study; 42—ambiguous diagnosis after tests Age (median):GON:54.2/Control:49 Sex (F, %: GON/Healthy): 46/47 Reference standard: SAP HFA 2; SITA standard 24-2 Clinic: unreliable if FP or FN>33% Index test: CFA; suprathreshold; 30 degrees; 18 dB 54 positions as the HFA 24-2SITA Standard Time: average of 3:29 min Clinic Diagnostic accuracy Acceptability
Jones et al,31 UK FFS Present and compare concordance of the results, speed, completion rates, and satisfaction Prospective Case-control Clinic Recruited from clinics. Exams: VF, VA, CS, HFA-GHT. GON “outside normal limits” Healthy “within normal limits”; MD: a measure of GON severity 18 (36 eyes): 12GON; 6 Healthy Included: GON: POAG in both eyes (4+ years). Attempt to recruit sample with a wide range of disease severity Excluded: GON: any other ocular disease (except cataract surgery) Age (median, y): POAG 70/controls 76 Reference standard: SAP HFA; SITA standard 24-2 Time: 6.9 m (6.5, 7.5) GON/4.8 min (4.6, 5.2) controls Clinic: repeated if “unreliable” Index test: Eyecatcher; eye tracker; suprathreshold Goldman III ~ to 10 dB in HFA; adapted 24-2 grid Time: mean 5.1 min (4.6, 5.5) GON/3.4 (3.2, 3.5) controls Clinic Usability Acceptability
Prea et al,32 Australia CET, JGRF, Glance Optical PTY To establish the medium-term repeatability Prospective Longitudinal Cohort 6 mo Clinic UK and India Recruited on their normal review dates. Previously diagnosed. GS:IOP<21and OHT:IOP>21, normal VF and ON; Treated SG=ON and/or VF change 60 Included: 6 GS; 24 POAG;18 JOAG; 3 OHT; 9 Other GON Included: GS, OHT, treated SG Excluded: condition or drug affecting vision; change of GON treatment during study; inability to understand English; VA<6/12; unreliable HFA Age (mean, y): 57.9 (15-89) Sex (F, %) 38 Reference standard: SAP HFA 24-2; SITA fast(25) or SITA Standard(35) Time: SITA-fast 4.3±0.2; SITA-standard 6.2±0.1 min Clinic: 4 times Index test: MRF; threshold; 66-point radial test grid Time:4.6±0.1 min Clinic: 4 times (baseline, 2,4,6 mo) Baseline: HFA first; other visits: randomized One eye with lower MD on HFA Accuracy Usability Acceptability
Schulz et al,33 Australia Biogen Idec, NHMRCA Independent validation to determine performance for glaucoma detection and follow-up Prospective Cross-sectional Clinic Recruited from GON clinic. GON: ON, NRR and NRFL loss (VF loss and IOP not a criteria); glaucoma specialist confirmed. Manifest GON: reproducible VF defects on HFA. 85 Included: 25 controls; 60 4OAG One eye; (worst in manifest GON) Included: GON: stable IOP and VF, HFA SITA 24-2 at least 3 times; new to MRF. Controls: new to both tests Excluded: VA<6/12; cataract or high RE; pathology likely to affect VF Age (mean, y): control: 53.8±19.9 GON: 64.0±10.4 Sex (F, %: GON/healthy):58.2/60 Reference standard: SAP HFA; 24-2 SITA Time: 5.44 min±1.1 Clinic: in last 3 mo; only reliable used Index test: MRF; thresholding; ZEST; 30 dB range Bayesian protocol; modified 24-2 grid or radial Time: 4.16 min±1.8 Clinic: 2 times (baseline; after 4-6 mo) Diagnostic Accuracy Usability Acceptability
Tsapakis et al,34 Greece not declared Present and evaluate the reliability in order to assess the possibility of glaucoma screening through the Internet Prospective Cross-sectional Clinic Recruited on GON clinic. No data on examinations. 10:20 eyes ×52 points=1040 points compared point by point Included: as appeared randomly and consecutively at the glaucoma department within hours Age (mean, y): 67.9 y (47-81) Reference standard: SAP HFA; SITA standard central 30-2 (76 points); inner 24 degrees (52 points) compared Clinic Index test: TVFST; eye tracker; suprathreshold; 24 degrees; 52 points; virtual photometer; 3 levels: −4 dB;−8 dB; −12 dB Time: 2-3 min Clinic Diagnostic Accuracy Usability Acceptability
Johnson et al,35 USA RPB M and S Technologies Evaluate and compare the performance of the screening procedure for clinic based visual field testing. Prospective Cross-sectional Clinic in Nepal Recruited attendees of the clinic; able to provide correct age. Exam by glaucoma fellow: IOP, BCVA, slit-lamp gonioscopy, ON, macula, RNFL, any conventional. 206 (411 eyes) Included: 210 healthy;183 GON;18 DR 373 analyzed One eye—if no light perception Included: healthy: IOP<21 mm Hg, normal ON, RNFL and retina, RE; DR: HbA1c and retina appearance; GON: IOP≥21 mmHg, RNFL and ON loss Excluded: URE >4 Dsph, 3 Dcyl; VA≤20/60; neurologic, ocular, systemic, or treatment affecting VF Age (mean, y): normal: 42.42 (18-78) GON: 54.7 (18-82); DR: 54.3 (45-64) Reference standard: SAP HFA; 24-2 SITA standard Clinic: eyes:198 controls; 160 GON; 15 DR Index test: VFE; suprathreshold; Goldman V; 16 dB Central 30 degrees; 96 locations; 24 per quadrant Time: average 3.18 min (SD=16.88 s) Clinic Diagnostic accuracy
Kong et al,36 Australia AMPTMF Determine the correlation between the outcomes obtained in a clinical setting Prospective Cross-sectional Clinic Recruited in GON research unit. Exam: ONH and retina, OCT, HFA, gonioscopy, ON photograph 90 included: 51 POAG; 22 PAC 5 mixed GON;12 healthy One eye randomly selected Included: VA≥6/12; reliable HFA. Excluded: retinal or corneal disease; requirement of an English interpreter; intraocular surgery within 6 mo Age (mean, y): 69.5 y (18-91) Reference standard: SAP HFA; 24-2 SITA standard Time: 6.3±0.1 min Clinic: before MRF; within 3 mo Index test: MRF; full field threshold; fast Time: average 5.7±0.1 min Radial pattern; 66 locations, 30 dB range Clinic: practice before; 2 tests on same day (4 retested within 4 wk) Test-retest repeatability
Lowry et al,37 USA no funding Determined the correlation between the oucomes, including analysis of ROC curves for glaucoma detection Prospective Case-control Clinic Recruited a convenience sample from clinics. GON: NRR, ON, RNFL exam; VF defects on at least 2 HVF. GON severity: MD-HPA criteria. Controls: family and friends; other patients based on fundus and VA. Right eye; left if criteria not meet 93 included: 30 healthy; GON: 35 mild;16 moderate; 12 severe Included: BCVA≥20/60; no other ocular or neurologic causes of VF loss; able to reliably perform HVF and POP within the first two attempts for both Excluded: / Age (mean, y): GON:64.8±14.6 Controls: 63.8±103 Sex (F, % GON/controls): 51/53 Reference standard: SAP HFA; 24-2 SITA standard clinic: within 3 mo with POP Index test: POP; suprathreshold; 24 degrees horizontally and 20 degrees vertically; 3 levels of stimuli:−16.7 dB, −21.7 dB, −26.7 dB Time: <5 min Clinic: unreliable repeated up to 3 times Diagnostic Accuracy Usability
Olsen et al,38 Denmark Synoptik Foundation, FFS Develop and evaluate a protocol for screening for glaucoma using the DMCO Prospective Retrospective Exploratory analysis Clinic Recruited attendees of the clinic. Random selection and grouping by glaucoma specialist based on full eye exam including: ON & HFA. GON: typical VF defects; Controls: OHT or other risk factors for GON with normal VF. 123 invited; 100 accepted; 97 52 GON; 9 borderline; 36 control Repeated tests; both eyes Included: ≥18 y; HFA familiarity; VA≥0.5 Excluded: tunnel vision; pregnancy dense cataract; conditions preventing from sitting correctly; myopia >6 D Age (mean, y): 66 (33-88) Sex (F, %): 55 Reference standard: GSS (SAP HFA; 30-2 SITA fast) Clinic: phase 1:within 6 mo before Phase 2: new HFA; same day Index test: DMCO; 42 point of central 24 degrees; 3 versions: standard (BW); basic (appear/disappear); advanced: (light gray) Time: standard median 2.5 min GON/1.26 min controls Clinic: phase1: DMCO version evaluated: 91 3-basic (31); 3-standard (60); 3-advanced (91) Phase 2: determine work distance: 26 Phase 3: identify best algorithm Diagnostic accuracy Acceptability
Santos and Morabe,39 Philippines not declared Determine the reliability in detecting VF loss among ophthalmology patients Prospective Case-control Clinic Out-patients requiring VF exam for their ophthalmic condition. Exam: BCVA and biomicroscopy. 122 Followed; 77 included (137) Prior diagnosis: 4 other; 24PAC; 2SG; 61PACG;10POAG;36GS Included: ≥18 y; able to understand and follow commands; good HFA fixation; BCVA near≥6/60 Excluded: unable to finish both tests Age (mean, y): 58 (18-82) Sex (F, %): 68 Reference standard: SAP HFA; 30-2 SITA Time: mean 7.50 min (SD+0.08) Clinic Index test: VFE; Goldmann size V; 16 dB intensity Time: mean 3.21 min (SD+0.01s) Clinic: 20 min after HFA Diagnostic accuracy
Wroblewski et al,40 USA NEI, DEI Proof of concept study exploring the practicality, reliability, and validity. Prospective Cross-sectional Clinic Recruited in GON and neurophthalmolgy clinic. No data on examinations. 22 POAG;17 normal; 1PG; 6GS 3 LTG; 3 NAG;1 CACG;1 ADOA; 1 PPh; 2 OND; 2 MS; 2 CS; 3 Tu Included: all patients undergoing treatment regardless of age, sex, ethnicity, or health status; previous experience with HFA II Excluded: <18 y Reference standard: SAP HFAII Time: SITA-Standard 6.1±1.0 min Clinic:51 SITA Standard &28 SITA Fast Index test: VirtualEye; Eyetracker; full threshold 24-2 32 degrees horizontal and 24 degrees vertical; MAN or VG, 40 dB rang Time: average MAN 10.6±3.3 min, VG 9.4±2.1 min Clinic: eye with larger VF defects in HFA; 2 times Usability Acceptability
Ianchulev et al,41 USA Compare the sensitivity and specificity to detect VF defects Prospective Case-series Clinic No data on recruitment and examinations. 33 patients (58 eyes) HFA defects:10 without; 28 mild/moderate; 20 severe Included: GON or GS; CDR>0.5; BCVA≥20/200; Excluded: unreliable HVF;VA<20/100 Age (mean, y): mean 57.5 Sex (F, %): 55 Reference standard: SAP HFA Clinic: within 6 mo prior Index test: Peristat; threshold; 4 levels (16.7-26.7 dB); 24 degrees horizontally and 20 degrees vertically Clinic Diagnostic accuracy Usability
ACG indicates angle closure glaucoma; ADOA, autosomal dominant optic atrophy; AMPTMF, AMP Tomorrow Marker Foundation; BCVA, best-corrected visual aquity; BNRC, Brazilian National Research Council; BW, Black on White; CACG, chronic angle closure glaucoma; CDR, cup disc ratio; CET, Cambridge Eye Trust; CF, C3 visual field analyzer; CNT, correlated neighborhood thresholding; CoO, College of Optometrist; CS, contrast sensitivity; CS, cortical stroke; Dcyl, Diopter cylindrical; DEI, Doheny Eye Institute; DMCO, Damato Multifixation Campimetry Online; DR, diabetic retinopathy; Dsph, Diopter spherical; F, female; FL, fixation loss; FFS, fight for sight; FP, false positive; GHT, glaucoma hemifield test; GON, glaucoma optic neuropathy; GS, glaucoma suspects; GV, gear vision; HFA, Humphrey field analyzer; HPA, Hodapp-Parrish-Anderson; HVF, Humphrey visual field; IGA, International Glaucoma Association; IOP, intraocular pressure; JGRF, Jukes Glaucoma Research Fund; JOAG, juvenile open-angle glaucoma; LTG, low-tension glaucoma; M, male; MAN, manual; MRF, Melbourne rapid fields; MS, multiple sclerosis; NAG, narrow angle glaucoma; NEI, National Eye Institute; NHMRCA, National Health and Medical Research Council of Australia; NRR, neuro retinal rim; NTG, normal-tension glaucoma; OAG, open angle glaucoma; OCT, optical coherence tomography; OHT, ocular hypertension; ON, optic nerve; OND, optic nerve drusen; ONH, optic nerve head; ORIA, Ophthalmic Research Institute of Australia; PAC, primary angle closure; PD, pattern deviation; PG, pigmentary glaucoma; POAG, primary open-angle glaucoma; POP, Peristat online perimetry; PPh, pseudophakia; RE, refractive error; RNFL, retinal nerve fiber layer; RPB, research to prevent blindness; SAP, standard automated perimetry; SEG, secondary glaucoma; SG, stable glaucoma; SITA, Swedish Interactive Threshold Algorithm; SRDII, Samsung Research and Development Institute India; TD, total deviation; TVFST, telemedicine visual field screening test; URE, uncorrected refractive error; VA, visual acuity; VFE, visual fields easy; VFF, visual field fast; VG, visual grasp.

Eleven portable perimeters designed for home use were compared with SAP Humphrey Field Analyzer (HFA) (Carl-Zeiss Meditec, Dublin, CA) and one also with glaucoma staging system.38

Quality assessments of 11 studies reporting diagnostic accuracy data are reported in Table 2. Based on our QUADAS-2 guidance, studies were considered to have a “high” risk of bias with regard to patient selection as they were either selected in a predetermined, nonrandomized manner; or recruited from a selected population24,26,29,30,33–35,37,38 or excluded patients with unreliable test results24,30,37,38,41 or unable to finish tests.39 With respect to conduction and interpretation of index test and reference standard, only 2 studies were considered to have a “low” risk of bias24,35; 2 studies were rated as “high” risk of bias for index test as the used threshold was not prespecified and 2 studies had “low” risk of bias for reference standard26,38; 7 studies were rated as “unclear” for both domains.29,30,33,34,37,39,41

TABLE 2 - Results of the Quality Assessment for 11 Diagnostic Accuracy Studies (QUADAS-2) Tool
Risk of Bias Applicability Concerns
References Patient Selection Index Test Reference Standard Flow and Timing Patient Selection Index Test Reference Standard
Ding et al24 High Low Low Low NA NA NA
Jones et al25 NA NA NA NA NA NA NA
Jones et al26 High High Low Low NA NA NA
Pradhan et al27 NA NA NA NA NA NA NA
Prea et al28 NA NA NA NA NA NA NA
Razeghinejad et al29 High Unclear Unclear Unclear NA NA NA
Mees et al30 High Unclear Unclear High NA NA NA
Jones et al31 NA NA NA NA NA NA NA
Prea et al32 NA NA NA NA NA NA NA
Schulz et al33 High Unclear Unclear Low NA NA NA
Tsapakis et al34 High Unclear Unclear Low NA NA NA
Johnson et al35 High Low Low Low NA NA NA
Kong et al36 NA NA NA NA NA NA NA
Lowry et al37 High Unclear Unclear Low NA NA NA
Olsen et al38 High High Low Low NA NA NA
Santos and Morabe39 High Unclear Unclear Low NA NA NA
Wroblewski et al40 NA NA NA NA NA NA NA
Ianchulev et al41 High Unclear Unclear Low NA NA NA
NA indicates not applicable.

Flow and timing domain was assigned with “low” risk of bias in most studies24,26,33–35,37–39,41 except in 1 study where it was assigned with “high” risk of bias because <80% of participants were included in the analysis30 and in 1 study where it was rated as “unclear.”29

Summary of Performance Data From the Included Studies

Devices were tested in patients with different disease severity using distinct test patterns and different reference standards; therefore, comparison of their results need to be interpreted with extreme caution.

Diagnostic accuracy for glaucoma was assessed in 11 studies24,26,29,30,33–35,37–39,41 (Table 3), whereas accuracy for detection of glaucoma progression was assessed in 3 studies.25,28,32 Diagnostic accuracy data distinguishing mild/moderate diseased patients from healthy controls and avoiding unit-of-analysis issues (eg, on a per-patient rather than per-VF point basis) were available only for 4 studies, presented graphically in Figure 2. For graphical presentation, “best algorithm” (the one with the highest discriminatory ability-highest area under the curve value)38 or “medium contrast sensitivity threshold” as described by the authors37 were considered.

TABLE 3 - Results of 11 Studies that Provided Data on Diagnostic Accuracy of Devices for Glaucoma Diagnosis
References Diagnostic Accuracy
Ding et al24 VFF correspondence of scotoma detection with HFA, whole field analysis PD &TD (50 GON; 30 Controls): sensitivity 91.2%; specificity 97%
Jones et al26 Mean hit rate in Eyecatcher compared with MD in HFA, whole field analysis (24 MD <−6 dB; 22 MD >−2 dB): AUROC=0.97 (95% CI: 0.94–0.99) FP identification (0.7 arbitrary cut-off): sensitivity: 73%; MD<−6 dB (0.7 arbitrary cut-off): specificity: 100%
Razeghinejad et al29 VisuALL mean sensitivity vs. HFA, whole field analysis (52 GON; 50 Controls): ROC VisuALL (0.98) vs. ROC HFA (0.93)
Mees et al30 Identical miss rate for each stimulus point between CFA and HFA: AUROC (MD <−6 dB)=0.78±0.05 SD; AUROC (MD >−6 dB)=0.87±0.04 SD
Schulz et al33 Comparison of MRF and HFA global indices (60 GON; 20 age-matched controls): AUROC (95% CI) MD: 0.85 (0.77–0.93) HFA; 0.84 (0.76–0.93) MRF; P<0.0001 AUROC (95% CI) PSD: 0.93 (0.88–0.98) HFA; 0.81 (0.76–0.95) MRF; P<0.0001 AUROC (95% CI) VFI: 0.88 (0.80–0.95) HFA; 0.85 (0.79–0.96) MRF; P<0.0001 Subset of 36 manifest GON (HFA reference): AUROC (MRF): 0.89 MD, 0.85 PSD; 0.88 MRF
Tsapakis et al34 Binary output of visual field points in TVFST compare with visual field points in HFA (1040 points): High threshold (−4dB), generalized Youden index cut-off point 28: AUROC=0.762 (0.73–0.79); sensitivity=0.637(0.592–0.680); specificity=0.735 (0.696–0.771) Medium threshold (−8 dB), generalized Youden index cut-off point 25: AUROC=0.782 (0.75–0.81); sensitivity=0.790 (0.755–0.822); specificity=0.646 (0.599–0.690) Low threshold (−12 dB) generalized Youden index cut-off point 16: AUROC=0.837 (0.81–086); sensitivity=0.942 (0.939–0.936); specificity=0.497 (0.503–0.509)
Johnson et al35 The percentage of missed points in VFE plotted as a function of the percentage of points worse than the normal 5% probability level in HFA (160 GON; 198 controls): AUROC: controls vs. all glaucoma 0.687 (0.64–0.75); controls vs. moderate/advanced 0.784 (0.73–0.84)
Lowry et al37 Comparison of adjacent test points missed in POP and HVA, whole field (63 mild/worse GON; 30 Controls): −16.7 dB; cut-off: 1-2 missed point: AUROC=0.81 (95% CI, 0.71–0.90); sensitivity 54%; specificity 100% −21.7 dB; cut-off: 1 missed point: AUROC=0.77 (95% CI, 0.67–0.87); sensitivity 59%; specificity 93% −26.7 dB; cut-off: 1 missed point: AUROC=0.77 (95% CI, 0.67–0.87); sensitivity 54%; specificity 100% Comparison of adjacent test points missed in POP and HVA, whole field (28 moderate/worse GON; 65 mild GON/controls): −16.7 dB; cut-off: 4 missed point: AUROC=0.87 (95% CI, 0.80–0.95); sensitivity 86%; specificity 85% −21.7 dB; cut-off: 3 missed point: AUROC=0.85 (95% CI, 0.77–0.94); sensitivity 71%; specificity 94% −26.7 dB; cut-off: 2 missed point: AUROC=0.85 (95% CI, 0.76–0.93); sensitivity 71%; specificity 92% No significant difference between the 3 AUROC for mild/worse GON vs. controls (P=0.36); or moderate/worse GON vs. combined controls/mild GON (P=0.75)
Olsen et al38 Comparison of the number of points missed on a DMCO test with the GSS categories (53 GSS0; 17 GSS; 14 GSS2; 11 GSS3; 11 GSS4): Standard 1+2 algorithm (sensitivity; specificity): GSS0 (2%; 98%); GSS1 (12%; 88%); GSS2 (71%; 29%); GSS3 (100%; 0%); GSS4 (100%; 0%) Standard 1+2 algorithm; GSS0 (53) vs. all other GSS (53); cut-off 4.5: AUC 0.90; sensitivity 64.2%; specificity 98.1%
Santos and Morabe39 Clinical interpretation of VF defects in VFE vs. VF defect in HFA (74 Defects, 63 No defects): Sensitivity 91%; Specificity 100%
Ianchulev et al41 Clinical interpretation of VF defects in Peristat and HFA, quadrant analysis (232 quadrants, all scotoma levels): Reviewer 1: sensitivity 83%, specificity 96%; reviewer 2: sensitivity 80%, specificity 94%; reviewer 3: sensitivity 83%, specificity 96% Clinical interpretation of VF defects in Peristat and HFA, quadrant analysis ( 196 quadrants, moderate-severe defects only): Reviewer 1: sensitivity 84%, specificity 97%; reviewer 2: sensitivity 86%, specificity 94%; reviewer 3: sensitivity 84%, specificity 96%
AUROC indicates area under the receiver operating characteristic curve; CFA, C3 visual field analyzer; DMCO, Damato Multifixation Campimetry Online; FP, false positive; GON, glaucoma optic neuropathy; GSS, glaucoma scoring system; HFA, Humphrey Field Analyzer; MD, mean deviation; MRF, Melbourne rapid fields; PD, pattern deviation; POP, Peristat online perimetry; PSD, pattern standard deviation; ROC, receiver operating characteristics; TD, total deviation; TVFST, telemedicine visual field screening test; VFE, visual fields easy; VFF, visual field fast; VFI, visual fields index.

F2
FIGURE 2:
Forest plots of whole-field sensitivity and specificity for detection of glaucoma patients (as defined in primary studies) from controls from 4 studies that avoided substantial unit of analysis error. CI indicates confidence interval; FN, false negative; FP, false positive; TN, true negative; TP, true positive.

Five studies provided data on area under the receiver-operating characteristic curve (AUROC)26,29,30,33,35; 3 provided data on point30,34 or quadrant analysis41; and 1 did not provide data on the number of patients included on the analysis26 and therefore were not presented graphically in forest plots.

Sensitivity and specificity data for glaucoma diagnosis were available for: Eyecatcher, Visual Field Fast (VFF), Telemedicine Visual Field Screening Test (TVFST), Visual Fields Easy (VFE), Damato Multifixation Campimetry Online (DMCO), Peristat Online Perimetry (POP), and an older version of Peristat. In general, depending on devices and used thresholds, sensitivity ranged from 54% in detecting mild to 91% in detecting moderate/severe glaucoma patients, while specificity ranged between 50% and 100% for any form of glaucoma. Sensitivity ranged between 59% and 91%, whereas specificity increased to 93% to 100% in the 4 studies that avoided substantial unit of analysis error.

AUROC curves were provided for: VisuALL, C3 fields analyzer (CFA), Eyecatcher, Melbourne Rapid Fields (MRF), TVFST, VFE, POP, and DMCO. Depending on devices and used thresholds, AUROC ranged from 0.69 in detecting mild glaucoma to 0.97 in detecting moderate/severe glaucoma.

Twelve studies reported usability25–29,31–34,37,40,41 and acceptability,25–34,38,40 while test-retest repeatability at one time point was reported in 3 studies.24,26,36 We describe below characteristics of the different technologies included in this review.

Tablet-based Perimeters

VFF (Leonard Yip, Apple AppStore) is a free application for iPad (Apple, Cupertino, CA) that uses fine noise-field stimulus flickering at high frequency. It covers the same VF area as the 24-2 HFA and has the possibility to adjust the frequency, size, and color of stimulus.

VFF testing was performed in a dark room with an iPad set to auto-brightness and mounted in a mobile stand. Room brightness and screen luminance were measured with a photometer. One eye was tested first, with patient resting on chinrest and fellow eye occluded, 38 cm from the screen. Fixation was switched to the 4 screen corners to test quadrants of the VF, while position of the fixation dot was kept always at central eye level. Scotomas were perceived as abnormality of flicker color or frequency and identified using touch screen for automated quantification.

VFF sensitivity and specificity for detecting moderate/severe glaucoma, with recommended 2×2stimulus-30 Hz, black/white, were 91.2% and 97%, respectively. Testing was tolerated well and completed in 1.1 to 3.6 minutes. Repeatability of whole-field scotoma area in glaucoma patients revealed an intraclass correlation coefficient (ICC) of 0.96.24

Eyecatcher (City, University of London, UK) is a free test delivered on a tablet. It tests one eye at time, with the fellow eye occluded, ~50 cm from the screen.

The modified version tested for home-based perimetry used a ZEST thresholding algorithm, which required fixation of central cross during the test and a button press for response. In this version, stimuli were presented in a 10 cd/m2 white background and individuals received visual feedback of the stimulus location. Measurable values ranged from 12.6 to 48 dB (sensitivities <12.6 dB could not be measured).

Testing was performed in a dark room. Participants were asked to do a test every month for 6 months. Although a fixation tracker can potentially be adapted, in this report there was no control over fixation, distance, or ambient lighting; however, testing was recorded by a tablet camera. Median time duration of this test was 4.5 minutes, without any variations among 6 sessions, and was accepted by volunteer glaucoma patients who were willing and able to comply with a monthly regimen. Adherence to a 6-month regime was 98.3% with 1 patient quitting after 4 sessions due to chronic symptoms of vertigo that were also experienced with HFA. Concordance between 6 Eycatcher tests measured at home and four HFA tests done in clinic was good (r=0.94, P<0.001) with 95% coefficient of repeatability of ±3.4 dB. MD values deviated >3 dB from the median in 9% of tests; however, majority of the anomalous test could be identified by machine learning techniques that were applied to tablet camera recording the test (AUROC=0.78). Variations of the ambient illumination had no effect on VF measurements.25

Two previous eye-tracking and head-tracking versions of Eyecatcher, intended for detection of clinically meaningful defects, used a suprathreshold perimetry (Goldman III-fixed intensity stimuli presented against a white 31 cd/m2 background) adapted for a 24-2 grid (±15 degrees horizontal and ±9 degrees vertical). Eye-tracker monitored fixation and viewing distance to determine the position of stimulus relative to point of fixation and maintain constant stimulus size irrespective of viewing distance, thus removing the need for fixation crosses and chin rests. The device analyzed eye-movement responses to determine whether the user saw the stimulus, whereas luminance calibration was done by a photometer. With this version, participants were required to follow the dot that appeared randomly on the tablet screen. Average test duration time was 5.1 minutes for the version that used a stimuli equivalent to 10 dB31 and 2.5 minutes for the updated version that used a 6 dB more intense stimuli than the expected threshold.26 Eyecatcher AUROC curve was 0.97.26 The device was rated more enjoyable, easier to perform, less tiring, and less hard to concentrate on than SAP (all P<0.001) and there was no difference in task comprehension, which was high for both tests.26,31 Perception of device between new referrals and follow-up patients was similar.26 Completion rate was >90%26,31 with the main reasons for failure being eye-tracking problems26,31 and software errors.26 Mean hit rate 95% coefficient of repeatability was 0.19 (19% of the test dynamic range).26

VFE (George Kong Softwares, Melbourne, Australia) is a free downloadable application for iPad (Apple, Cupertino, CA) that uses a suprathreshold strategy (Goldman V-16 dB fixed intensity stimuli presented on 10 cd/m2 backround luminance) to test 96 locations in central 30 degrees. It generates a graphic output of targets seen and not seen as well as reliability indices.

With VFE one eye was tested at time, with the fellow eye occluded, at ~33 cm from the screen. Fixation was shifted to a red point in 4 tested quadrants and the screen was pressed each time the target was detected. Average test time was 3.19 minutes. iPad calibration was done with a photometer/radiometer.

Reported AUROC for all glaucoma’s was 0.687, and 0.784 AUROC for moderate/severe glaucoma,35 with 91% sensitivity and 100% specificity in subjects requiring VF examination as a part of evaluation for the ophthalmological condition.39

MRF (GLANCE Optical Pty Ltd, Melbourne, Australia), a threshold version of VFE, is an application available for iPad (Apple, Cupertino, CA). It uses a 66-point radial grid (34 degrees horizontal and 25 degrees vertical) with spot size increasing in the periphery (Goldman III centrally and IV at 17 degrees eccentricity on 5 cd/m2 background luminance). It has 8 possible threshold levels and a 3-step paradigm in a range of 30 dB. MRF guides individuals through the test by voice prompts that are available in multiple languages and generates a report with reliability indices, sensitivity values, total deviation map, the pattern deviation (PD) map, and VF gray scale. A space arm with forehead rest can be used for head and distance control; otherwise, subjects can place an elbow of the arm contralateral to the tested eye at the edge of the iPAD for distance control, ~33 cm from the screen.

MRF testing is done in dim light while screen brightness is automatically adjusted to maximal hardware brightness by the application. At each test one eye was tested first, with the fellow eye occluded. Fixation was changed four times to test peripheral locations and individuals were instructed to touchscreen when the stimulus was seen (Bluetooth keybord can be used). iPad calibration was done with a photometer/radiometer whereas fixation monitoring was implemented with a blind-spot monitor. The average time duration of the MRF test was between 4.2 and 5.7 minutes among glaucoma patients with different disease severities.28,32,33,36

In a study evaluating MRF to detect disease progression, weekly testing was conducted. Patients were willing and able to perform MRF home-based perimetry at weekly intervals with 88% of participants completing more than 1 examination and 69% completing all 6 examinations. Of 69%, 72% returned home examinations weekly on 1 reminder and 87% on 3 weekly reminders. IT reasons and interest (26% each) were the main reasons for lack of uptake, whereas IT reasons and life demands (20% each) for lack of compliance.28 Comparing results with the reference standard no significant difference was observed (all P>0.05) for MD, pattern standard deviation (PSD) versus PD, and visual fields index (VFI) versus visual capacity among participants who completed at least 2 MRF tests at-home and 2 in-clinic HFA. There was high concordance for MD (95% limits of agreement, −6.2 to 8.8 dB) in the Bland-Altman analysis and high correlation of average MD between in-clinic and at-home outcomes (R=0.85).28

MRF compliance rate with clinical visits at baseline, second, fourth, and sixth month was 100%, 87%, 70%, and 97%, respectively. Concordance between the corrected MD of the MRF and SITA-standard was substantial with an average coefficient of accuracy (Ca) of 0.83 and moderate between MRF and SITA-fast with a Ca of 0.77. Concordance was moderate to substantial for all test times with Ca ranging from 0.67 to 0.82 and ICC 0.71 to 0.81 between MRF and SITA-fast and Ca from 0.79 to 0.83 and ICC 0.81 to 0.90 between MRF and SITA-standard with a negative bias (−1.2 to −3.4 dB) for MRF.32

AUROC curves were: 0.84 MD, 0.81 PSD, and 0.85 VFI for all glaucoma’s and improved to 0.89 MD, 0.85 PSD, and 0.88 VFI for manifest glaucoma.33 MRF was found easy to use by 99% of participants with only one patient having difficulties with the touchscreen. Holding the iPad steady and using the touch screen presented challenges which were overcome by using an iPad stand.33

MRF repeatability assessment revealed an overall ICC of 0.93 MD and 0.89 PD, suggesting a minimal learning effect. ICC was lower (0.73 MD and 0.74 PD) among patients with mild defects in HFA. Bland-Altman analysis showed a small bias (0.1 to 0.5 dB) between tests 1 and 2 that was constant across the range of field loss.36

Web-based Perimeters

Online TVFST is a software based on Microsoft as well as Google’s Android platform that allows the use of virtual reality glasses with a 6-inch Android smartphone. It uses a supra-threshold strategy to test the central 24 degrees (52 points) of VF at 3 uncalibrated intensity levels (−4, −8, −12 dB). Stimuli are available in white/gray on a black background or black/gray on a white background and variable presentation rate. TVFST output is binary with reliability indices given. The software analyzes and validities test results and allows combination of the results from 2 or more tests into a single test.

The TVFST version for tablet/computer took 2 to 3 minutes for testing 1 eye, with the fellow eye occluded. Distance was determined by looking at the central fixation point and moving toward display until the blind spot dot disappeared. Subjects were instructed to click mouse each time the stimulus was seen on screen. Testing was performed in a dark room, with screen brightness adjusted to 50% of the maximum. A web camera was used as a photometer to detect ambient illumination, eye tracking to set the central target’s location accordingly, and blind spot method to monitor fixation and head position as well as to adjust stimuli position and size.

TVFST reported accuracy range was: 0.76 to 0.84 area under the curve, 64% to 94% sensitivity, and 74% to 50% specificity depending on used thresholds.34 It was completed, well tolerated, and found simple and easy to use by all participants.34

POP (Keep Your Sight Foundation, San Francisco, CA) is a free web-based test for any computer monitor larger than 17 inches. It uses a supra-threshold strategy to test VF 24 degrees from fixation horizontally and 20 degrees vertically using 3 levels of standardized thresholds (−16.7, −21.7,−26.7 dB). POP generates reliability indices and a grayscale VF display.

To standardize testing and reduce variability across computer monitors a predefined strategy was used. One eye was tested, with fellow eye occluded, at ~50 cm. Distance was determined by fixating at a central white circle and adjusted until a blinking indicator in the periphery disappeared. Testing was performed in a dark room with fixation at a central point and the subjects were instructed to push the keyboard spacebar each time they saw the stimulus. The blind spot stimulus was used for head position and fixation. Testing was guided by investigators and took around 5 minutes.

POP diagnostic evaluation revealed a range of AUROC curves from 0.81 to 0.77, sensitivity 54% to 59%, and specificity 93% to 10% for mild glaucoma increasing to 0.87 to 0.85 AUROC curves, 86% to 71% sensitivity, and 85% to 94% specificity for moderate/severe glaucoma, depending on contrast sensitivity thresholds and criteria for labelling a case positive.37

An older version of the Peristat test revealed 80% to 83% sensitivity for all glaucoma’s that increased to 84% to 86% when cases with mild defects were excluded, while specificity reminded between 94% and 97%.41 POP completion rate was 95%, when unreliable tests were repeated up to 3 times by participants.37 Peristat older version was successfully used by participants without any assistance although 40% were computer illiterate.41

DMCO is a free web-based test, accessed at http://www.testvision.org, that examines 42 points of the central 24 degrees. It does not generate reliability indices whereas results are presented as a map of the points seen and not seen.

Three versions of DMCO have been tested: STANDARD with black-on-white stimulus that requires mouse movement at the stimulus and click on the appeared target (smiley icon) in order for the next stimulus to be presented, BASIC where a circle with a number that appears and disappears changes to a flower with 4 numbered petals on the mouse point and the participant has to select the same number that appeared on the circle to confirm the participant was looking at a fixation target, and ADVANCED that is similar to STANDARD but with light-gray stimuli initially presented, and with missed points retested using darker stimuli. One eye at a time was tested, with fellow eye occluded, at a distance determined by focusing on a fixation target and adjusting until a large flashing stimulus disappeared into the normal blind spot. Testing was performed in a room with normal lighting.

Different DMCO algorithms reported AUROC that ranged from 0.79 to 0.90 compared with doctors’ diagnosis. DMCO STANDARD 1+2 had a 0.90 AUROC curve, with 64.2% sensitivity and 98% specificity for detecting all glaucoma stages, with sensitivity highly dependent on glaucoma severity. Fatigue was noted during the first phase of testing, as the number of participants in the three tests declined. On the second phase, 84% accepted invitation.38

HMD-based Perimeters

The virtual reality CFA (Remidio; Glen Allen, VA) allows 3-dimensional assessment of the 30 degrees VF. It uses a suprathreshold strategy, 24-2 pattern, to test 54 points with a contrast of ~18 dB. CFA testing time was around 3.5 minutes and reliability indices are provided.

CFA calibration was done with a lux meter (background brightness 4 cd/m2, stimuli brightness 60 cd/m2), whereas patients were instructed to maintain gaze at a central fixation point and respond by a handheld clicker when a stimulus was seen. AUROC curve of CFA was 0.87 for severe glaucoma and 0.78 for mild/moderate glaucoma.30 It was preferred by 93% of participants compared with HFA, whereas 60% declared that they would use it at home if available. CFA was found easier to use and more comfortable (P<0.001) than HFA.30

Gear vision is an application developed for a commercially available HMD (Samsung GearVR) and a compatible smartphone (Samsung S8) that can measure a minimum threshold of 8 dB based on the maximum brightness of S8 display. It determines the blind spot before the test and provides reliability parameters. Eyes are tested individually without the need for occlusion. Head movements and periodic breaks are allowed if required. For VF sensitivity mapping, both threshold and suprathreshold strategies can be used.

The median time duration of the 24-2 full-threshold strategy, using Goldman size III stimulus tested at 54 points, was 11 minutes. GV study participants (including glaucoma patients, suspects and healthy controls) experienced less body and eye discomfort (11%, 14%) compared with HFA (48%, 60%) and preferred GV (68%). There were 13 unreliable GV tests compared with 6 HFA tests, while 7 exclusions were due to failure to follow instructions, battery discharge or falling asleep.27

VisuALL (Olleyes Inc., NJ) combines an HMD, a laptop/phone/tablet, and a bluetooth connected handpiece for response. It tests 50 locations of the central 24 degrees VF in a 6-degree grid pattern with a background brightness of 3 cd/m2 and stimuli brightness of 3 to 120 cd/m2. Eyes are tested individually without the need for occlusion. HMD can be adjusted for interpupillary distance correction. The participant’s head can move during the test.

Fixation is checked by an eye tracker before the stimulus is shown and adjusted. Control of testing environment luminance and reports are similar to the HFA. Both threshold and suprathreshold strategies can be used.

VisuALL testing time, tested with the thresholding algorithm T, was 9.28 minutes among glaucoma patients and 6.13 minutes among healthy controls. ROC curve of VisuALL was greater compared with HFA (0.98 vs. 0.93) for mild/moderate glaucoma. VisuALL was well received with 1 participant not able to take the VisuALL test and 1 the HFA.28

VirtualEye (BioFormatix; San Diego, CA) is an HMD test that combines microdisplays and real-time eye tracker for fixation. It can test 30 degrees of VF using a full-threshold strategy of a 24-2 grid (stimulus size III, sensitivity range of 40 dB), but different strategies can be implemented. The device allows adjustments to facial structures and testing of each eye or both eyes per session. Testing is operated through a Windows computer and provides reliability indices, retinal sensitivity values, color-coded display, and automatic interpretation of results into normal, suspect, preperimetric, mild, moderate, or severe glaucoma.

Manual model that requires steady fixation and mouse click response on visible stimuli took around 10.6 minutes. An alternative “visual grasp” mode that requires first fixation at target and then at the direction of stimulus with response determined from the analysis of eye tracker data took around 9.4 minutes. VirtualEye was preferred over the HFA due to improved comfort and ease of use and there was no clear preference for the test mode that was completed 80% in manual and 61% in visual grasp mode. Reasons for unsuccessful tests included: difficulties in instructions, inability to adjust the visor or self-reported double vision and/or fatigue, software malfunctions, error leading to loss of data, hardware failure, and excessive fixation loss, false positive, or false negative.40

DISCUSSION

Home-based perimetry is a promising development that has a potential to facilitate and improve glaucoma care by remote assessment. The approach is potentially inexpensive compared to clinic-based testing, requiring only the low-cost device and proper instructions for patients, and, in theory, can be performed unsupervised in any environment. In theory home-based perimetry could potentially be used for frequent testing in patients at risk of disease progression, and facilitate glaucoma care for people having barriers to access eye clinics such as individuals with limited mobility, elderly, or those from rural area. Home-based perimetry can potentially make clinical trials more efficient by detecting early disease progression.24–41

We propose that a diagnostic accuracy assessment is an important initial step to explore performance of new technologies. Novel tests for home monitoring are typically evaluated first in a clinic or laboratory setting and an initial exploration of their ability to detect VF loss is conducted. For this purpose, a diagnostic study can be designed with the understanding that a novel technology should be able to confirm that a patient has VF loss (ie, sensitivity) and that a normal person does not have any VF loss (ie, specificity). In addition, ideally, the technology would be able to quantify the severity of VF damage in people with glaucoma. Once this initial evaluation is successfully completed the next challenge is to apply the technology to a home environment and assess the potential role of home monitoring to detect disease progression.

However, at the moment relevant evidence for home monitoring is insufficient. Most studies identified in our review were conducted in the clinic or laboratory with a highly selected population and thus are not generalizable to real life situations. Technical challenges such as control of screen distance, need for calibration, and data transferring process need to be resolved. Adoption of technology by elderly patients or among those with limited technological experience, and lack of adherence with frequent testing requirements are also important hurdles.42 Potential clinical use of some of the tests identified in our review is questionable as they either do not appear accurate enough or they are designed for uncalibrated suprathreshold perimetry. Threshold measurements required to detect changes have additional challenges, including higher variability than SAP, compared with simpler suprathreshold testing that are typically designed for diagnostic purposes. Other issues are the practical tradeoffs of free web-based tests versus technologies that require dedicated equipment.

Our review highlights that current literature consists mainly of feasibility or pilot studies, intended only to proof the concept. Early research has positive signals regarding the feasibility and acceptability among patients. However, the studies reporting diagnostic accuracy data have high risk of bias as have used a highly selected populations and results cannot be generalized to the general glaucoma population. An additional observation from our study is the lack of standardization in how perimetric results are reported.

Home-based perimeters still have a promising role as reflected in the increasing number of technologies. They may not replace SAP, but their addition, in theory, may offer a reasonable value that may be beneficial in some circumstances, including the provision of eye care during pandemics.25,31

More definite studies designed to address real life performance of home-based perimeters and the value of information captured are required as most devices were tested in experimental, highly controlled conditions. Reviewed tests were different from SAP in many ways including: test spots and stimulus characteristics, background luminance, screen size and curvature, threshold determination strategy, viewing distance and luminance control, fixation requirement, nontested eye occludens, analysis of scotomas, reliability indices, disease severity, and follow-up. Primary studies had major differences in various aspects of study design and conduct making comparison of devices impossible. Performance of the tested devices is likely to be overestimated due to inclusion of the highly selected populations. In most studies, investigators were not masked to conduction and interpretation of tests and clinical data.

A limitation of our review could be the exclusion of non-English studies or novel devices that did not enter into academic literature or did not fulfil criteria of our protocol, such as a low-cost HMD using smartphone and frequency doubling technology (FDT), “imo” (CREWT medical systems, Tokyo, Japan), the Advanced Vision Analyzer (Elisar; New City, NY), the PalmScan VF2000 Visual Field Analyzer (Micro Medical Devices; Calabasas, CA), and Vivid Vision Perimetry (Vivid Vision, San Francisco, CA).

Future studies may elucidate the performance and value of home-based perimeters. It will also be important to evaluate the potential concerns of individuals, sustainability, economic utility, as well as satisfaction and treatment adherence.25,26,42

REFERENCES

1. Bourne RR, Taylor HR, Flaxman SR, et al. Number of people blind or visually impaired by glaucoma worldwide and in world regions 1990-2010: a meta-analysis. PLoS One. 2016;11:e0162229.
2. Daka Q, Trkulja V. Efficacy and tolerability of mono-compound topical treatments for reduction of intraocular pressure in patients with primary open angle glaucoma or ocular hypertension: an overview of reviews. Croat Med J. 2014;55:468–480.
3. Tham YC, Li X, Wong TY, et al. Global prevalence of glaucoma and projections of glaucoma burden through 2040: a systematic review and meta-analysis. Ophthalmology. 2014;121:2081–2090.
4. Heijl A, Bengtsson B, Oskarsdottir SE. Prevalence and severity of undetected manifest glaucoma: results from the early manifest glaucoma trial screening. Ophthalmology. 2013;120:1541–1545.
5. Thapa SS, Paudyal I, Khanal S, et al. A population-based survey of the prevalence and types of glaucoma in Nepal: The Bhaktapur Glaucoma Study. Ophthalmology. 2012;119:759–764.
6. Budenz DL, Barton K, Whiteside-de Vos J, et al. Prevalence of glaucoma in an urban West African population: the Tema Eye Survey. JAMA Ophthalmol. 2013;131:651–658.
7. Jayaram H, Strouthidis NG, Gazzard G. The COVID-19 pandemic will redefine the future delivery of glaucoma care. Eye (Lond). 2020;34:1203–1205.
8. Foot B, MacEwen C. Surveillance of sight loss due to delay in ophthalmic treatment or review: frequency, cause and outcome. Eye (Lond). 2017;31:771–775.
9. Wong SH, Tsai JC. Telehealth and screening strategies in the diagnosis and management of glaucoma. J Clin Med. 2021;10:3452.
10. Aboobakar IF, Friedman DS. Home monitoring for glaucoma: current applications and future directions. Semin Ophthalmol. 2021;36:310–314.
11. European Glaucoma Society. Terminology and Guidelines for Glaucoma, 5th ed. Savona: Publicomm; 2020.
12. Hu R, Racette L, Chen KS, et al. Functional assessment of glaucoma: uncovering progression. Surv Ophthalmol. 2020;65:639–661.
13. Quigley HA. Glaucoma. Lancet. 2011;377:1367–1377.
14. Gardiner SK, Crabb DP. Frequency of testing for detecting visual field progression. Br J Ophthalmol. 2002;86:560–564.
15. Chauhan BC, Garway-Heath DF, Goñi FJ, et al. Practical recommendations for measuring rates of visual field change in glaucoma. Br J Ophthalmol. 2008;92:569–573.
16. Anderson AJ, Bedggood PA, George Kong YX, et al. Can home monitoring allow earlier detection of rapid visual field progression in glaucoma? Ophthalmology. 2017;124:1735–1742.
17. Broadway DC, Tibbenham K. Tackling the NHS glaucoma clinic backlog issue. Eye (Lond). 2019;33:1715–1721.
18. Faes L, Bachmann LM, Sim DA. Home monitoring as a useful extension of modern tele-ophthalmology. Eye (Lond). 2020;34:1950–1953.
19. Che Hamzah J, Daka Q, Azuara-Blanco A. Home monitoring for glaucoma. Eye (Lond). 2020;34:155–160.
20. Boote T, Tatham AJ. Home monitoring in glaucoma. Eur Ophth. 2018;12:102–104.
21. McInnes MDF, Moher D, Thombs BD, et al. Preferred reporting items for a systematic review and meta-analysis of diagnostic test accuracy studies: the PRISMA-DTA statement. JAMA. 2018;319:388–396.
22. Richardson WS, Wilson MC, Nishikawa J, et al. The well-built clinical question: a key to evidence-based decisions. ACP J Club. 1995;123:A12–A13.
23. Whiting PF, Rutjes AW, Westwood ME, et al. QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. Ann Intern Med. 2011;155:529–536.
24. Ding J, Tecson IC, Ang BCH, et al. The performance of iPad-based noise-field perimeter versus Humphrey Field Analyser in detecting glaucomatous visual field loss. Eye (Lond). 2021;20:1–12.
25. Jones PR, Campbell P, Callaghan T, et al. Glaucoma home monitoring using a tablet-based visual field test (Eyecatcher): an assessment of accuracy and adherence over 6 months. Am J Ophthalmol. 2021;223:42–52.
26. Jones PR, Lindfield D, Crabb DP. Using an open-source tablet perimeter (Eyecatcher) as a rapid triage measure for glaucoma clinic waiting areas. Br J Ophthalmol. 2021;105:681–686.
27. Pradhan ZS, Sircar T, Agrawal H, et al. Comparison of the performance of a novel, smartphone-based, head-mounted perimeter (GearVision) with the humphrey field analyzer. J Glaucoma. 2021;30:e146–e152.
28. Prea SM, Kong GYX, Guymer RH, et al. Uptake, persistence, and performance of weekly home monitoring of visual field in a large cohort of patients with glaucoma. Am J Ophthalmol. 2021;223:286–295.
29. Razeghinejad R, Gonzalez-Garcia A, Myers JS, et al. Preliminary report on a novel virtual reality perimeter compared with standard automated perimetry. J Glaucoma. 2021;30:17–23.
30. Mees L, Upadhyaya S, Kumar P, et al. Validation of a head-mounted virtual reality visual field screening device. J Glaucoma. 2020;29:86–91.
31. Jones PR, Smith ND, Bi W, et al. Portable perimetry using eye-tracking on a tablet computer—a feasibility assessment. Transl Vis Sci Technol. 2019;8:17.
32. Prea SM, Kong YXG, Mehta A, et al. Six-month longitudinal comparison of a portable tablet perimeter with the humphrey field analyzer. Am J Ophthalmol. 2018;190:9–16.
33. Schulz AM, Graham EC, You Y, et al. Performance of iPad-based threshold perimetry in glaucoma and controls. Clin Exp Ophthalmol. 2018;46:346–355.
34. Tsapakis S, Papaconstantinou D, Diagourtas A, et al. Home-based visual field test for glaucoma screening comparison with Humphrey perimeter. Clin Ophthalmol. 2018;12:2597–2606.
35. Johnson CA, Thapa S, George Kong YX, et al. Performance of an iPad application to detect moderate and advanced visual field loss in Nepal. Am J Ophthalmol. 2017;182:147–154.
36. Kong YX, He M, Crowston JG, et al. A comparison of perimetric results from a tablet perimeter and humphrey field analyzer in glaucoma patients. Transl Vis Sci Technol. 2016;5:2.
37. Lowry EA, Hou J, Hennein L, et al. Comparison of peristat online perimetry with the Humphrey perimetry in a clinic-based setting. Transl Vis Sci Technol. 2016;5:4.
38. Olsen AS, Alberti M, Serup L, et al. Glaucoma detection with damato multifixation campimetry online. Eye (Lond). 2016;30:731–739.
39. Santos AS, Morabe ES. VisualFields easy: an iPad application as a simple tool for detecting visual field defects. Philipp J Ophthalmol. 2016;41:22–26.
40. Wroblewski D, Francis BA, Sadun A, et al. Testing of visual field with virtual reality goggles in manual and visual grasp modes. Biomed Res Int. 2014;2014:206082.
41. Ianchulev T, Pham P, Makarov V, et al. Peristat: a computer-based perimetry self-test for cost-effective population screening of glaucoma. Curr Eye Res. 2005;30:1–6.
42. Jones L, Callaghan T, Campbell P, et al. Acceptability of a home-based visual field test (Eyecatcher) for glaucoma home monitoring: a qualitative study of patients’ views and experiences. BMJ Open. 2021;11:e043130.
Keywords:

glaucoma; home-based perimeters; diagnostic accuracy

Copyright © 2022 Wolters Kluwer Health, Inc. All rights reserved.