INTRODUCTION
Primary open‐angle glaucoma (POAG) is a chronic neurodegenerative eye disease and the leading cause of irreversible blindness worldwide (Quigley & Broman, 2006). Among the various forms of glaucoma, POAG is the most prevalent, affecting approximately 1%–3% of middle‐aged to elderly individuals (Springelkamp et al., 2017). Due to ageing populations and the starkly increasing prevalence of (high) myopia, both important risk factors for POAG, the prevalence of POAG is projected to increase in the coming decades (Tham et al., 2014).
It is estimated that by 2040 there will be 80 million individuals/patients worldwide with POAG (Tham et al., 2014). Since there are no curative treatments, minimizing progression is the only way to prevent visual field (VF) loss in POAG patients. Therefore, accurate monitoring of progression is imperative.
Currently, optical coherence tomography (OCT) is one of the most common modalities for monitoring structural progression in POAG. However, in cases of advanced POAG, OCT is no longer able to detect significant changes in the retinal nerve fibre layer (RNFL), ganglion cell complex layer (GCC) and ganglion cell inner plexiform layer (GC‐IPL), also known as the floor effect (Chan & Miller, 2007). Glaucomatous structural and functional damage tends to be regionally diverse, and OCT parameters become increasingly uninformative as more and more of the retina reaches this floor effect.
It has been hypothesized that OCT angiography (OCTA), a novel imaging modality that captures vascular parameters, might be more sensitive for progression in intermediate to late‐stage POAG patients compared to RNFL or GCL thickness. Previous studies have shown that POAG is associated with vascular changes in the retinal microcirculation and that these changes become more prominent as POAG progresses (Rao et al., 2020). It is unclear if these vascular changes occur parallel to or before VF progression.
Nevertheless, several cross‐sectional studies have reported an added value of OCTA to existing diagnostic modalities in POAG care (Andrade De Jesus et al., 2020; Bekkers et al., 2020). On top of that, it is a relatively fast, objective and non‐invasive imaging technique. OCTA might therefore be an excellent adjunct to existing diagnostic modalities in POAG care.
However, longitudinal research is scarce, non‐early (i.e. intermediate and advanced) POAG patients are underrepresented in the existing literature, and it remains unclear which OCTA markers are most relevant for the monitoring or prediction of POAG progression (Triolo & Rabiolo, 2020; Van Melkebeke et al., 2018). To address these gaps in knowledge, we performed a systematic review of the literature and multiple meta‐analyses to assess the potential clinical usefulness of OCTA in intermediate and advanced‐stage POAG and identify the most important parameters to monitor this subgroup.
METHODS
This research adheres to the Preferred Items for Systematic Reviews and Meta‐analyses (PRISMA) (Page et al., 2021) and the Meta‐analysis Of Observational Studies in Epidemiology (MOOSE) guidelines (Brooke et al., 2021). The study protocol was registered in the PROSPERO prospective register of systematic reviews on the 6th of March 2025, under registration number CRD420251005304. PROSPERO is an international database for registering systematic reviews, established to prevent duplicates, increase transparency and guarantee systematic review quality.
Search strategy
A literature search was conducted on the 12th of August 2025 in Medline, Embase, Web of Science and the Cochrane database of systematic reviews. We only included studies with an available abstract, with human participants and studies in which a full text was available. The full search strategy can be found in the Supplementary Methods.
Reference lists of included studies were screened for additional relevant studies.
Study selection
First, two reviewers (KK and VAV) independently screened titles and abstracts for eligible articles, in which the relation between OCTA parameters and POAG was assessed in a longitudinal study design, without any other restricting selection criteria. Second, the full text of the included studies was reviewed. Studies following the progression of POAG over time using an OCTA were included.
Exclusion criteria were: (1) inclusion of secondary or angle‐closure glaucoma, (2) no use of OCTA, (3) cross‐sectional designs or case reports, (4) no intermediate or advanced POAG at baseline in the study population, or if POAG stage was not categorized; a mean visual field mean deviation (VF MD) higher than −6 dB at baseline, (5) reviews or editorials, (6) not published in a peer‐reviewed journal, (7) unavailable full‐text and (8) POAG outcomes compared to other ocular diseases, instead of healthy controls, (9) intervention trials. Finally, the results of both reviewers were compared, and discrepancies were discussed. If the two researchers could not reach a consensus, the broader co‐author group was consulted to reach a consensus.
Study screening and selection were conducted using Covidence (Covidence systematic review software, 2025, Veritas Health Innovation, Melbourne, Australia).
Data collection and risk of bias assessment
From each of the included studies we extracted the name of the authors, title, year of publication, study design, sample size, type of OCTA device and software, baseline characteristics of the study population, OCTA parameters and layers, measure of POAG progression, the effect estimate for the association between the OCTA parameters and POAG progression after adjustment for confounding factors with corresponding 95% confidence interval (CI) (or standard error if the 95% CI was not reported) and p‐values. The risk of bias was assessed using the QUADAS‐2 tool for diagnostic studies (Whiting et al., 2011).
Data synthesis and analysis
If multiple studies were published from the same research group, we evaluated if there was any overlap in the analysed patient population. If no overlap existed, both studies were included. If any overlap was found, we included the most recent study.
We recalculated the 95% CI if this was not reported, using the reported effect estimate in combination with standard error and/or p‐value. A meta‐analysis was performed for comparisons for which at least three or more studies had reported results using the same exposure (i.e. the OCTA parameter being investigated, such as baseline vessel density [VD], rate of VD change, foveal avascular zone area [FAZ], etc.) and outcome in the same retinal layer (e.g. superficial retinal plexus, deep retinal plexus or choroid). The heterogeneity of the meta‐analyses was measured by calculating I2.
Heterogeneity is a diverse concept describing the variability or differences between included studies. Typically, at least three types of heterogeneity are recognized: clinical heterogeneity (e.g. difference in study populations or interventions), methodological heterogeneity (e.g. differences in study design such as how outcomes were measured) and statistical heterogeneity (variability of the actually observed effects). The first two types of heterogeneity cannot be easily quantified and were assessed qualitatively; the latter can be assessed with several metrics such as the aforementioned I2.
Results from both a random‐ and fixed‐effect model were presented. In case of low statistical heterogeneity (I2 < 0.5), the fixed‐effects model was considered the primary outcome; and if statistical heterogeneity was high (I2 ≥ 0.5), the random‐effects model was considered the primary outcome. Funnel plots were generated to evaluate potential publication bias; for a guide to interpreting funnel plots we recommend Brush et al.
(2024). Briefly, funnel plots display the study effect (e.g. beta or odds ratio) on the horizontal axis and a measure of precision (e.g. standard error or sample size) on the vertical axis. If studies with non‐significant or negative results are less likely to be published, this would lead to an asymmetrical funnel shape.
A p‐value < 0.05 was considered statistically significant. Statistical analyses were performed in R for Windows version 4.2.3 (R Foundation for Statistical Computing, Vienna, Austria), using the meta package version 8.0–2.
RESULTS
A total of 4727 records were retrieved, of which 1808 were duplicates, leaving 2919 studies to be screened (Table S1). After screening and full‐text assessment, 18 studies (Chen et al., 2024; Jo et al., 2021; Kim et al., 2019; Kwon et al., 2019; Lee et al., 2022; Lee et al., 2023; Lee et al., 2024; Lee, Park, & Park, 2021; Lee, Shin, et al., 2021; Li et al., 2022; Lin et al., 2019; Ninomiya et al., 2023; Shang et al., 2023; Shin et al., 2021; Shoji et al., 2017; Song et al., 2024; Wang et al., 2022; Yospon & Rojananuangnit, 2023) were included in our review, of which 7 (Chen et al., 2024, Lee, Park, & Park, 2021, Lee et al., 2022, Lee et al., 2023, Shin et al., 2021, Wang et al., 2022) were eligible for meta‐analysis (Figure 1) (Page et al., 2021). All included studies were published from 2017 onwards, with 10 studies (48%) published in the last 2 years (Figure S1).
All studies examined the association of OCTA parameters with VF progression, and two studies additionally examined the association of OCTA with conventional OCT parameters (Kim et al., 2019; Li et al., 2022).
Study characteristics
A total of 3186 participants were analysed within the 18 included studies with a mean follow‐up of 50.6 ± 34.3 months. The included studies were typically either prospective observational studies or retrospective chart reviews of glaucoma specialist clinics or university hospitals. Most studies enrolled all consecutive patients that met the inclusion criteria and consented to participation.
Almost all included studies were conducted in East‐Asian countries (e.g. China, Korea or Japan), and although rarely explicitly described, the vast majority of included participants will have been of East‐Asian ancestry. We observed high heterogeneity in the device and software used for OCTA acquisition (Table 1).
However, the Cirrus HD (OCT 6000, Carl Zeiss Meditec, Dublin), DRI OCT Triton (Triton, Topcon healthcare, Tokyo) and the Avanti Rt vue (Optovue RTVUE 100 xr Avanti, Optovue, Fremont) were the most frequently used in respectively five (28%), four (22%) and four (22%) studies. As for VD measurement software, out of 18 studies, 10 (55%) used a version of the Angiovue software (Optvue, Fremont) and 5 (28%) used Imagenet 6 (Topcon healthcare, Tokyo). Spectralis angiography software (Heidelberg Engineering, Heidelberg), ImageJ (1.46r, National Institute of Heath, Bethesda) and Angioplex (Carl Zeiss Meditec, Dublin) were all used once (6%).
OCTA Parameters
A variety of OCTA parameters were used in the included studies (Table 2): Peripapillary VD (N = 7), parafoveal VD (N = 3), FAZ parameters (N = 3) and microvascular dropout (MvD; N = 4). Lesser common parameters were fractal dimension, vessel perfusion density and flux index, and sectorial analyses (N < 3 studies).
Vessel density (meta‐analyses)
Five of the seven studies reporting on the relation between peripapillary VD at baseline and VF progression were eligible for meta‐analysis. A lower baseline peripapillary VD increased the risk of future VF progression, with a hazard ratio (HR) (95% CI) of 1.05 (1.02, 1.07) per year for each percentage decrease in baseline VD (Figure 2a). Notably, the studies that found a significantly elevated risk were all published by the Lee et al. research group, although their results were based on different cohorts.
Furthermore, a lower baseline peripapillary VD was associated with an increased rate of deterioration in visual field mean deviation (VF MD) during follow‐up, with a mean difference (95% CI) of −0.02 (−0.03, −0.01) dB faster deterioration per year for each percentage decrease in baseline VD (Figure 2b).
Three studies reporting the relation between parafoveal VD at baseline and VF progression were eligible for meta‐analysis. Baseline parafoveal VD did not affect the risk of VF progression (HR [95% CI]: 0.99 [0.94, 1.06] for each percentage decrease in baseline VD per year; Figure 2c).
Additionally, a negative annual peripapillary VD change was significantly associated with a higher risk of VF progression (HR [95% CI]: 2.03 [1.30, 3.18] for each percentage decrease in VD per year; Figure 2d).
Lastly, sectorial analyses (i.e. nasal, superior, inferior or temporal) of baseline macular VD furthermore showed no significant association with VF progression (Wang et al., 2022).
Foveal avascular zone
Unfortunately, there was a high amount of heterogeneity among studies investigating the FAZ and it was therefore not reasonably possible to perform a meta‐analysis. Li et al., 2022 investigated FAZ change over time compared to VF progression during the same period, whereas Ninomiya et al. (2023) investigated baseline FAZ compared to future VF progression.
Additionally Ninomiya et al. (2023) investigated the FAZ of the superficial plexus whereas Wang et al. (2022) investigated the FAZ of the deep plexus (Table 2), which are most likely correlated but not necessarily equivalent.
Ninomiya et al. (2023) did find a significantly higher reduction in VF MD during follow‐up in patients with a larger FAZ area, reduced FAZ circularity and an increased FAZ perimeter in the inferior hemifield at baseline, with respectively a beta (95% CI) of −0.10 (−0.03, −0.17) μm2, 0.06 (0.00, 0.13) μm and − 0.12 (−0.05, −0.19) μm (Ninomiya et al., 2023). In other words, a larger and less circular (i.e. more disrupted) FAZ predicts faster future VF MD deterioration of the inferior hemifield.
However, they did not find this association in the superior hemifield. None of the other studies investigating FAZ parameters and VF progression could replicate this finding (Li et al., 2022; Wang et al., 2022) (Table 2).
Sectorial analyses found that while the FAZ area, perimeter and circularity were not associated with a change in superior VF MD (Table 2), the inferior and inferocentral VF MD showed a significant association (Ninomiya et al., 2023). A smaller FAZ area, for instance, was associated with a higher VF MD with a beta of −0.197 μm2 (p < 0.001) (Table 2).
Microvascular dropout
Four studies reported the relationship between the presence of peripapillary choroidal MvD at baseline and VF progression. However, similar to the FAZ findings, a meta‐analysis was not possible due to heterogeneity in study design. The definition of MvD was either a qualitative assessment (i.e. the presence or absence of any MvD) (Jo et al., 2021; Kwon et al., 2019), or a quantitative assessment of the area (Kim et al., 2019) or angular circumference of MvD (Lee, Park, & Park, 2021).
Additionally, a variety of either the superficial plexus, deep plexus or choroid was examined. Nevertheless, the baseline presence of MvD was significantly associated with a higher risk of VF progression in 2 studies (Table 2). The rate of MvD area change and MvD angular circumference were also significantly associated with VF progression in one study, with a HR (95% CI) of 1.02 (1.01, 1.03) per mm2 increase per year and 1.52 (1.31, 1.75) per degree increase per year respectively (Lee et al., 2022) (Table 2).
Furthermore, Lin et al. (2019) found a significant association between the presence of a deep peripapillary MvD and the rate of OCT RNFL thickness loss (beta −0.85 μm, p = 0.041) (Lin et al., 2019). In line with this, Lee et al.
(2023) found a significant association between the presence of a superficial peripapillary MvD at baseline and VF progression, with a HR (95% CI) of 1.97 (1.11, 3.56) per year (Lee et al., 2023) (Table 2).
Other parameters
Yospon and Rojananuangnit (2023) investigated the relation of peripapillary vessel perfusion density and flux index with VF MD, and found a significant correlation of r = 0.63 (p < 0.001) and r = 0.42 (p < 0.001), respectively (Yospon & Rojananuangnit, 2023) (Table 2). Wang et al. (2022) investigated the association of baseline fractal dimension of the macular superficial and deep plexus with VF progression, but found a non‐significant HR (95%) of 1.00 (0.71, 1.41) and 0.80 (0.51, 1.24), respectively (Wang et al., 2022) (Table 2).
Risk of bias
Risk of bias was evaluated using the QUADAS‐2. Overall, the methodological quality of the included studies was moderate to good (Table S2). Funnel plots of the meta‐analysed studies showed no signs of publication bias (Figure S2).
DISCUSSION
In this study, we systematically reviewed and meta‐analysed the existing literature on the clinical usefulness of OCTA for the monitoring of progression in intermediate to advanced POAG patients. A reduction in peripapillary VD increased the risk of VF progression. Peripapillary VD loss during follow‐up was also significantly associated with a decrease in VF MD.
On the contrary, parafoveal VD at baseline was not associated with VF progression. Although changes in FAZ parameters and MvD appear to be associated with VF progression, the scarce literature on these parameters withheld us from performing a meta‐analysis. Similarly, vessel perfusion density and flux index might be promising predictors; however, the available literature is still limited.
In advanced POAG patients, OCT is known to reach a ‘floor effect’ as it is no longer able to identify further loss of RNFL.4 A systematic review by Migeul et al., supporting our analyses, suggested the use of OCTA to measure progression longitudinally (Miguel et al., 2022). Peripapillary VD seems a promising monitoring tool and predictor of VF progression in intermediate and advanced POAG. Studies on preperimetric and early‐stage POAG found a faster rate of optic nerve head VD (i.e. peripapillary VD and parapapillary VD) loss to be associated with visual field loss (Tansuebchueasai et al., 2024), and a lower baseline peripapillary VD to be associated with higher odds of VF progression (Park et al., 2019).
Our study shows similar findings in advanced POAG, suggesting that OCTA could be a helpful prognostic and monitoring tool in POAG care. Future studies will have to investigate the diagnostic accuracy (i.e. specificity, sensitivity and area under the receiver‐operator curve), optimal OCTA imaging frequency and thresholds for the rate of VD loss that warrant more careful monitoring or treatment.
Parafoveal VD in our meta‐analysis, on the other hand, showed no significant increase in risk of VF progression. Although research on macular vascular parameters such as parafoveal VD is limited, Hou et al. (2020) found that macular VD, a combination of parafoveal and perifoveal VD, decreases more rapidly than macular GCC in POAG eyes (Hou et al., 2020).
However, they only measured the rate of VD and GCC loss during follow‐up in POAG eyes, not their association with VF progression, nor the predictive capacity for future VF progression. It is also unclear how many of those POAG eyes had already reached the floor effect and would therefore naturally have a low to non‐existent loss of GCC thickness. The normalized rate of macular VD loss might be higher than that of ganglion cell complex (GCC) loss in a randomly selected POAG eye, but based on our meta‐analysis there is currently no evidence that components of macular VD, such as parafoveal VD, can be used to predict future visual field loss.
The value of FAZ parameters is unclear as the literature on the longitudinal relation of the FAZ and POAG is both limited and inconsistent. In particular, there is almost no overlap of the method of FAZ assessment nor the layer examined. Two studies in our review found a significant association while two studies found no association between FAZ parameters and VF progression.
A recent cross‐sectional study on early and intermediate POAG patients also found no differences in FAZ parameters between patients and controls (Verticchio Vercellin et al., 2024). In our study, Ninomiya et al. (2023) found that changes in inferior hemifield FAZ area, circularity and perimeter were correlated with a faster reduction of VF MD during 90‐month follow‐up (Ninomiya et al., 2023).
Similarly, Nishida et al. (2023) previously found that patients with a significant increase in FAZ area had faster rates of GCL thinning and VF MD loss (Nishida et al., 2023). The underlying mechanism behind this association is unclear.
However, Miguel et al. (2022) and Wang et al. (2022) found no significant association between FAZ parameters and VF progression (Li et al., 2022; Wang et al., 2022).
Due to the limited and heterogeneous literature, a meta‐analysis was not feasible.
Lastly, the presence of MvD at baseline was associated with a higher risk of VF progression in three studies out of four studies. Similarly, meta‐analysis was not possible due to heterogeneous study design; again, in particular due to differences in MvD definition and retinal layers examined. Nonetheless, previously Latif et al.
(2024) found a larger MvD area to be associated with a higher rate of VF MD loss during follow‐up (Latif et al., 2024). We similarly found a significant association between change in MvD and VF progression (HR [95% CI] 1.02 [1.01, 1.03] per mm2 increase per year). Even though literature is limited, MvD can be a relevant parameter for monitoring POAG.
Strengths and limitations
To our knowledge, this is the first systematic review and meta‐analysis to examine the literature on the association between OCTA parameters and POAG progression, specifically in intermediate and advanced POAG patients. Previous studies have reviewed the literature in earlier stages of POAG (Miguel et al., 2019). However, these results could not be readily applied to late‐stage (i.e. advanced) POAG patients.
Moreover, where possible we meta‐analysed the previously published results for a quantifiable and concise summary of the available evidence. On top of that, most included studies were of high quality, ensuring the reliability of our results.
Some limitations should also be considered. Firstly, as OCTA is a relatively novel imaging modality, we could only include a limited number of studies in our review. The limited available literature was also considerably heterogeneous in study design.
Consequently, only a subset of the included studies could be meta‐analysed. Moreover, a mean follow‐up of 50.6 ± 34.3 months may be too short to detect VF progression in intermediate and advanced POAG. Secondly, there is a lack of diagnostic accuracy metrics currently reported in the literature.
If OCTA is to be implemented as a diagnostic tool, diagnostic metrics (e.g. sensitivity, specificity, area under the receiver‐operator curve or Harrel's C‐statistic) would be required. Unfortunately, none of the included studies reported such metrics. Thirdly, almost all included studies were conducted in East‐Asian countries (e.g.
China, Korea and Japan) and will have included predominantly patients of East‐Asian ancestry. Our findings might not be fully generalizable to patients of European or African ancestries. Lastly, five studies in our review were published by the Lee et al.
(2024) research group (Lee et al., 2022; Lee et al., 2023; Lee et al., 2024; Lee, Park, & Park, 2021; Lee, Shin, et al., 2021). To minimize bias from repeated use of the same study population, we excluded studies by Lee et al. (2023) with overlapping or partially overlapping populations.
Our meta‐analysis and systematic review was limited by heterogeneity in study population, in OCT device, OCTA protocols and definition of VF progression. Future research should focus on longitudinal studies with OCTA with defined device and image protocol methods to estimate the OCTA parameters to reduce heterogeneity. On top of that, studying different devices and imaging protocols on the same study population can be useful for comparison.
Moreover, a consensus on the definition of VF progression is needed. This will make future longitudinal research more comparable and make clinical assessment of POAG progression more objective.
CONCLUSION
OCTA has the potential to be a valuable tool in monitoring progression of intermediate and advanced POAG. The meta‐analyses showed that a reduction in peripapillary VD in particular increases the risk of VF progression. Parafoveal VD showed no association with VF progression.
Changes in the FAZ and MvD are also reported to be potential predictors, but evidence is limited and heterogeneous. More longitudinal evidence, examining advanced POAG patients and analysing diagnostic outcomes (i.e. sensitivity and specificity) is needed to enable the next step towards clinical implementation.
Supporting information