Edited by: Valerie Purvin, Midwest Eye Institute, United States
Reviewed by: A. V. Rukmini, Duke-NUS Medical School, Singapore; Essam Mohamed Elmatbouly Saber, Benha University, Egypt
This article was submitted to Neuro-Ophthalmology, a section of the journal Frontiers in Neurology
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Computerized pupillary light reflex assessment devices (CPLRADs) may serve as an effective screening tool for glaucomatous optic neuropathy, since they can dynamically detect abnormal pupillary responses from a novel sequence of light stimuli and functionally-shaped stimuli. The aim of this study was to systematically evaluate the current state of advanced CPLRADs and accuracy of application in detecting glaucoma. An electronic literature search of PubMed, MEDLINE, and Embase from database inception to December 2019 was performed. Studies that reported data on the use of computer-aided pupillometry with monocular and/or binocular monitoring in glaucoma patients were included. Two review authors independently conducted the study selection and extracted study data. A total of twenty-five studies were included in this review; eight studies with a total of 829 subjects were included in this meta-analysis. Data were pooled using a random-effect model, since the significant heterogeneity (
Glaucoma comprises a heterogeneous group of diseases characterized by progressive loss of retinal ganglion cells (RGCs) and their optic nerve axons, discernable by cupping of the optic nerve head, with associated visual-field damage or even blindness (
The pupillary light reflex (PLR) is driven by photoreceptors [i.e., rods, cones, and intrinsically photosensitive retinal ganglion cells (ipRGCs)] that control pupil dilation or constriction in response to light that falls on the retina. Consequently, how much light enters the eye can be modified, aiding the adaptation to various levels of darkness and light (
In routine clinics, although visual field (VF) testing by standard automated perimetry (SAP) is efficient in detecting functional changes, the test remains subjective and time-consuming (
CPLRADs are able to dynamically measure the entire waveform of the pupillary response under a novel sequence of controlled color light stimuli and functionally-shaped stimuli, possibly detecting and evaluating glaucomatous optic neuropathy (
A diagram of the computerized pupillary light reflex assessment devices. CPLRADs contain three main parts: an illumination system (infrared illumination, alternating color illumination, background illumination, functionally-shaped stimuli program), an image capturing system (infrared digital camera, monitoring camera), and a computerized image analysis system (capturing the following pupil parameters: baseline pupillary size, maximum contraction velocity, maximum dilation velocity, amplitude (ratio), time to max contraction, time to maximal dilation).
In this study, we reviewed state-of-the-art CPLRAD techniques for measuring and quantifying the PLRs in glaucoma patients. Furthermore, since the stimulus presented can be varied in color, shape, intensity, duration, and size, we assessed the accuracy of using those techniques for direct clinical decision making for glaucoma diagnosis and evaluation.
We performed a systematic review according to the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) statement. The protocol was developed at the start of our investigation. Ethical approval and patient consent were not required, because all data analyses were performed with previously published studies.
Systematic literature research using the PubMed, MEDLINE, and Embase databases, as well as references of the included studies, was performed from database inception through a search end date of December 10, 2019. For PubMed, Embase, and MEDLINE, we used both controlled vocabulary and text words for synonymous terminology within titles and abstracts in the development of search strategies. The search strategy contained all possible combinations of terms: chromatic pupillometry, automated pupillography, pupillometer, pupillary light reflex, PLR, relative afferent pupillary defect, RAPD, pupil, glaucoma, visual field defect, retinal nerve fiber layer thickness, RNFL thickness, retinal ganglion cell loss (
Articles reporting the use of CPLRADs to examine abnormal PLRs/RAPDs in glaucoma were eligible if they met all of the following criteria: (1) recruited participants with diagnosed glaucomatous optic neuropathy; (2) used samples of participants aged 18 years or over; (3) utilized an acceptable reference standard; and (4) reported sufficient results to calculate metrics for the diagnostic accuracy of the technique used.
An article was excluded if: (1) testing was performed on infants/toddlers; (2) the number of participants with glaucoma was <10; or (3) the study was not published in English or was a review study, letter, or conference paper.
Two reviewers independently extracted the data of selected articles using a piloted data extraction form. Disagreements between individual judgments were resolved by discussions with a senior reviewer (XJQ). The extracted information included the title, authors, publication year, study design, characteristics of each study population (including the sample size and mean age), equipment for measuring pupillary responses, binocular or monocular study category, stimulus color, stimulus pattern, and main outcomes (
The characteristics of studies included in the meta-analysis.
Pillai et al. ( |
Case-control | G56.9 | RAPiDo | B | U | c | U | Significant | Sensitivity: 89% |
India | study | H35.21 | MD, CDR asy | Specificity: 91.7% | |||||
G130, H43 | ( |
AUC: 0.94 | |||||||
Prospective | Median MD | Amp-FF-W: Sensitivity: | |||||||
Rao et al. ( |
Cross-sectional | G61 | RAPDx | B | 1–2 min | w/b/g/r/y | Peripheral/Full | asy = 6.7 dB | 5–45%; |
India | study | H46 | field | (range: 0.1–27.2 | Specificity: 95%; | ||||
G47, H42 | dB) | AUC: 0.60–0.82 | |||||||
Cross-sectional | Standard setting: Sensitivity: | ||||||||
Waisbourd et | study | 66.2 ± 13.6 | RAPDx | B | 1–2 min | w | Full-field | MD asy > 5 dB; | 93.3%;Specificity: |
al. ( |
G60 | CDR asy≥0.20 | 41.2%; AUC: 0.84 for | ||||||
OH/GS21 | detecting MD asymmetry | ||||||||
Cross-sectional | Mean MD asy: | Full field (ConMaxVelLatR) | |||||||
Tatham et al. ( |
study | G69.1 | RAPDx | B/M | U | w/b/g/r/y | Peripheral/Full | 2.2 ± 3.1dB; | Sensitivity: 53%; |
USA | G66, H50 | H51.3 | field | Significant MD | Specificity: 80%; AUC: 0.75 | ||||
Prospective | asy ( |
(green stimulus) | |||||||
Full model: | |||||||||
Chang et al. ( |
Case-control | G67 ± 11 | RAPDx | B/M | 1 min | w/b/g/r/y | Peripheral/Full | MD asy < −5 dB; | Sensitivity 84%; |
USA | study | H60 ± 10 | field | Specificity 80%; | |||||
G148, H71 | AUROC: 0.90 | ||||||||
Kalaboukhova | Case-control | G65 ± 10 | A custom- | B | 15 s | w | Peripheral/Full | Mean MD: | PARm (a cut-off point of |
et al. ( |
study | H63 ± 8 | built | field | 6.3 dB | 1.16) Sensitivity | |||
Sweden | G30, H30 | pupillometer | (range: 0.31– | 86.7%; Specificity | |||||
18.80 dB) | 90%; AUROC 0.923 | ||||||||
Case-control | A | Paracentral/ | Peripheral stimulus pattern: | ||||||
Wride et al. ( |
study | G59.6 ± 16.7 | PupilmetrixTM | M | short | w | Bjerrum/ | U | Sensitivity 93.1%; |
UK | G29, H30 | H69.9 ± 13.6 | PLR60 | period | Peripheral | Specificity 76.7%; | |||
AUROC 0.907 | |||||||||
Kalaboukhova | Case-control | G69 ± 7 | A custom- | B | 15 s | w | Peripheral/Full | Minimum MD | PARm (a cut-off point |
et al. ( |
study | H59 ± 14 | built | field | asy: 1.6 dB | of 1.16): Sensitivity | |||
Sweden | G17, H15 | pupillometer | 82.4; Specificity | ||||||
86.7; AUROC 0.929 |
Each of the full text articles was independently reviewed by two reviewers (LGS, DZ) and scored with the Quality Assessment of Diagnostic Accuracy Studies 2 scores (QUADAS-2) tool. The methodologic quality of the included studies was evaluated with the QUADAS-2 tool. The risk of bias was assessed in four domains (i.e., patient selection, index test, reference standard, and flow and timing). The applicability concerns were assessed in three domains (i.e., patient selection, index test, reference standard). Each domain was assessed by indicating a “low,” “high,” or “unclear” rating.
We used Stata software version 15.0 (StataCorp, College Station, TX) to assess meta analysis. Extracted data were synthesized by creating forest plots of sensitivity and specificity. Heterogeneity was evaluated by Cochrane's
The initial literature search identified 5,003 citations (1,562 articles in Embase, 1,877 articles in PubMed and 1,564 articles in Medline).
Preferred reporting items for systematic reviews and meta-analyses (PRISMA) flow diagram for study inclusion.
In summary of
For the shape pattern stimulus, 15 studies used full-field illumination paradigms (
Among our 25 included studies, a control group with normal subjects was not designed in six studies and the accuracy of diagnosing glaucoma by pupillary assessment tests was not stated in the other 11 studies. Thus, our final meta-analysis incorporated eight studies with 829 subjects (
Forest plot showing a summary of sensitivity and sensibility for studies included in the meta-analysis.
In the subgroup analysis, we divided the eight studies into two groups according to the different light stimulus colors (
We also performed a sensitivity analysis on the synthetic sensitivity and specificity.
Sensitivity analysis of the included studies.
Performance of quality assessment of diagnostic accuracy studies 2 (QUADAS-2) evaluation.
Computer-aided PLR assessment devices are drawing increasing attention from both the vision science and the medical equipment engineering domains (
The RAPD is recognized as a characteristic finding of glaucoma with various techniques. A previous systematic review and meta-analysis performed by Chang et al. reported data showing that pupillography has a sensitivity of 75% with a specificity of 85% in clinical-based studies by evaluating various methods of RAPD detection including the earlier version of automated pupillography (
More importantly, the current state of CPLRADs is more complex than that of the previous earlier version of the automated pupillograph (
Some of the included studies had problems regarding the risk of bias according to the QUADAS-2 quality assessment. For patient selection, participants in all eight meta-analysis studies were enrolled on the basis of different inclusion criteria. Five studies enrolled patients with glaucoma of any cause (
The effect of glaucoma severity on the diagnostic ability of CPLRAD parameters and inter-eye asymmetry in MD have a significant effect on the AUCs (
Dark adaptation may be associated with the results of PLR capture to properly evaluate these studies, it is necessary to determine if all the participants underwent proper dark adaptation. In each of the studies, with different periods of time, participants were dark-adapted before light exposure. Eleven of 25 studies allowed 1–2 min for dark adaptation (
Simpler monochromatic CPLRADs might be more suitable for practical glaucoma screening. Based on various color stimuli, in our subgroup analysis, the sensitivity in the CLG was 0.74 (95% CI: 0.60–0.88) and the specificity was 0.87 (95% CI: 0.79–0.95). In the MLG, the sensitivity was 0.91 (95% CI: 0.86–0.96) and specificity was 0.75 (95% CI: 0.58–0.93). In addition to considering diagnostic accuracy, the MLG has other advantages over the CLG because the former can be done quickly, is easily reproducible, has a relatively low cost, and avoids more exposure to light radiation.
Using chromatic CPLRADs, however, provides a deeper understanding of rod, cone, and intrinsically-photoreceptive retinal ganglion cells (ipRGCs) contributing to the PLR. The PLR driven by ipRGCs and classical photoreceptors is impaired in glaucoma (
Proper model selection for a larger number of CPLRAD parameters can improve the diagnostic accuracy in glaucoma. Chang et al. demonstrated that using logistic regression including asymmetry in pupillary contraction latency, velocity, amplitude, and age may increase the sensitivity and specificity of pupillography in glaucoma (
This study has certain limitations: (1) the search strategy was limited to only those articles written in English; (2) none of our 25 studies, considered how to control the cognitive load and emotional factors that possibly alter both pupil size; and (3) in some of the studies, the glaucoma subjects were notably older than the control subjects. For instance, Rao et al. reported an average age of 61 years for glaucoma patients and 46 years for control subjects (
A future challenge is to identify the optimum combination of the large numbers of features generated by CPLRADs. As larger sample sizes are available, novel techniques, such as deep learning and image processing can be used to provide better diagnostic ability. It is important to identify that some features perform better on left/right eye stimulation. Furthermore, the application of CPLRADs does not provide any insight into the potential to measure structural damage, such as injury to the optic nerve head (ONH). The development of novel stimuli and assessment may enable calculation of the specific damage to the retina and ONH. If the CPLRADs were to be applied in different settings, such as the community, it might be that their accuracy for glaucoma diagnosis would be lower. Since many factors affect the PLR, further tests are needed to identify the cause of an abnormal PLR.
In conclusion, our results revealed that the diagnostic abilities of even the best CPLRD parameters are only moderate in glaucoma. The diagnostic abilities of the CPLRAD measurements were significantly influenced by the inter-eye asymmetry and within-eye asymmetry in case of glaucomatous damage. Further research on the mechanism of ipRGC in glaucoma should be deeply explored by chromatic pupillography to investigate other factors, such as sleep qualities in glaucoma patients.
All datasets presented in this study are included in the article/
LS designed the study and drafted the manuscript. DZ acquired the data and undertook the statistical analysis and interpretation. XQ drafted the manuscript. AL undertook the statistical analysis. YW and CZ revised the manuscript and supervised the study. All authors contributed to the article and approved the submitted version.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
The Supplementary Material for this article can be found online at: