Accuracy of pupil assessment for the detection of glaucoma: a systematic review and meta-analysis

Methodological quality of the review: Medium confidence

Author: Chang DS, Xu L, Boland MV, Friedman DS

Geographical coverage: India, Nigeria, United Kingdom (UK), United States of America (USA), Sweden, Germany, Japan and Australia

Sector: Glaucoma

Sub-sector: Relative afferent pupillary defect (RAPD) detection

Equity focus: None specified

Review type: Effectiveness review

Quantitative synthesis method: Meta-analysis

Qualitative synthesis method: Not applicable


Glaucoma is a specific disease of the optic nerve and is often more severe in one eye. When large enough, this asymmetry in disease severity can cause a relative afferent pupillary defect (RAPD). Better detection of RAPDs may be one way to identify persons with glaucoma.


To assess the accuracy of using pupillary light reflex (PLR) in detecting glaucoma.

Main findings:

A total of 30 studies met the inclusion criteria. Two studies were population-based studies, 20 were case-control and eight were case series of patients with glaucoma. Among the 30 studies, 10 did not include normal subjects as a control group and nine did not report accuracy of glaucoma detection.

Overall, 11 studies – including a total of 7271 participants – were included in the meta-analysis. Two studies were population-based studies, and the remaining were case-control studies. Geographical coverage of studies included India, Nigeria, UK, USA, Sweden, Germany, Japan and Australia. The pooled estimate corresponded to a sensitivity of 0.63 (95% confidence interval [CI], 0.43-0.80) with a specificity of 0.85 (95% CI, 0.77-0.97). After excluding two studies that used the swinging flashlight test (SFT), the sensitivity increased to 0.74 (95% CI, 0.59-0.85) with the specificity of 0.85 (95% CI, 0.77-0.90). Authors note that study designs and different pupil measurement techniques explained part of the heterogeneity between the studies.

Overall, testing procedures for pupil assessment of included studies were not standardised, and there was no consistent cut-off for each measurement.

In conclusion, authors note that the available evidence suggests that patients with glaucoma often have an abnormal pupil response to illumination. The measurement of pupil response to light provides an objective test of visual function, which identifies a substantial portion of those with glaucoma in some studies. However, there is insufficient evidence from population-based studies assessing pupil response under controlled conditions to support wider use of this approach. The authors also state that an RAPD is relatively non-specific and may be caused by a number of conditions other than glaucoma. Future studies investigating conditions and co-founders affecting pupil responses, and methods improving the accuracy of glaucoma detection, would be useful.


Authors searched Medline and Embase to identify potentially relevant studies. The search included combined controlled vocabulary search terms and text words for glaucoma, visual field defect, optic neuropathy, retinal nerve fibre layer thickness (study population) and RAPD, swinging flashlight test (SFT), pupillography, pupilometer, pupil perimeter and pupillary light reflex (PLR). All searches were last updated on June 22nd 2012, and were not limited by study design or by language. Authors reviewed the bibliographies of all selected articles and searched the Web of Science for articles that cited the selected articles to identify additional studies.

Two authors independently screened titles and abstracts for inclusion, retrieved potentially relevant studies and determined eligibility. Eligible studies were diagnostic accuracy studies or studies that assessed pupil response in populations including more than ten patients with glaucoma. Articles without original data or without English abstracts were excluded from the review.

Extracted data included study design, patients’ characteristics, study methodology, techniques in measuring pupil responses, definition of RAPDs and main results. The primary outcome of the review was to assess the accuracy of detecting glaucoma using RAPD. Secondary outcomes included comparing the magnitude of RAPDs in subjects with and without glaucoma, and assessing the association between visual field loss and the magnitude of RAPD. Methods used to extract data were not reported in the review. Methodological quality of included studies was assessed by one author using modified QUADAS tools that included 14 items.

Extracted data from the 11 studies included were used to create forest plots of sensitivity and specificity to depict study-specific estimates of sensitivity and specificity in receiver-operating characteristic space. Two studies were population-based studies and the remaining were case-control studies. Authors applied the mixed-effect bivariate summary receiver operating curve models to derive inferences about diagnostic test accuracy using STATA/SE. The bivariate models account for the correlation between sensitivity and specificity observed across studies, which captures the functional relationship between sensitivity and specificity because the threshold varies within each study. Authors also performed a pre-specified sensitivity analysis to determine whether performance improved after excluding studies that used the SFT because this approach is the least sophisticated of those studied.

Applicability/external validity:

Given that the majority of the studies included clinic patients with known or suspected diseases, authors note that the evidence is not applicable to routine screening and primary care settings, and the estimates of sensitivity and specificity may be over-estimates of the true effect.

Geographic focus

The majority of studies included in the review were conducted in high-income countries, and few in low- and middle-income countries. As discussed above, the authors note that due to high heterogeneity across included studies for most tests, it is not possible to conclude which test is the most accurate.

Summary of quality assessment:

Overall, there is medium confidence in the conclusions about the effects of this study. The search for evidence was partially comprehensive as authors did not contact authors/experts for potentially relevant studies. Although the searches included all languages, two studies were excluded because the abstract was in a different language. Additional limitations were identified in the methods used to extract data of included studies. Authors did not report whether this was conducted by two authors (independently), which may impact the reliability of the reported characteristics and results of the included studies.

Authors conducted a pooled analysis of two population-based studies and case-control studies, which may have not been appropriate due to the high heterogeneity across the studies for most tests. Excluding the two population-based studies from the analysis was sensible, however due to high heterogeneity across the case-control studies, authors could not conclude which test was the most accurate. Therefore, it may have been most appropriate to conduct a narrative synthesis of the findings. Nevertheless, authors did not draw strong conclusions based on the included studies and appropriately acknowledged and described the heterogeneity and limitations of the included studies. Authors also noted limitations of the review including the lack of an agreed reference standard for the diagnosis of glaucoma and a standard study protocol, and the exclusion of abstracts written in languages other than English.


Chang DS, Xu L, Boland MV, Friedman DS. Accuracy of pupil assessment for the detection of glaucoma: a systematic review and meta-analysis. Ophthalmology. 2013 Nov;120(11):2217-25. Source