Optical Coherence Tomography Evaluation of Peripapillary and Macular Structure Changes in Pre-perimetric Glaucoma, Early Perimetric Glaucoma, and Ocular Hypertension: A Systematic Review and Meta-Analysis

Background: This study aimed to assess the differences in the average and sectoral peripapillary retinal nerve fiber layer (pRNFL), macular ganglion cell plus inner plexiform layer (mGCIPL), and macular ganglion cell complex (mGCC) thickness using optical coherence tomography (OCT) in patients with pre-perimetric glaucoma (PPG) compared to those with early perimetric glaucoma (EG) and ocular hypertension (OHT). Methods: A comprehensive literature search of the PubMed database, the Cochrane Library, and Embase was performed from inception to March 2021. The weighted mean difference (WMD) with the 95% confidence interval (CI) was pooled for continuous outcomes. Results: Twenty-three cross-sectional studies comprising 2,574 eyes (1,101 PPG eyes, 1,233 EG eyes, and 240 OHT eyes) were included in the systematic review and meta-analysis. The pooled results demonstrated that the average pRNFL (WMD = 8.22, 95% CI = 6.32–10.12, P < 0.00001), mGCIPL (WMD = 4.83, 95% CI = 3.43–6.23, P < 0.00001), and mGCC (WMD = 7.19, 95% CI = 4.52–9.85, P < 0.00001) were significantly thinner in patients with EG than in those with PPG. The sectoral thickness of pRNFL, mGCIPL, and mGCC were also significantly lower in the EG eyes. In addition, the average pRNFL and mGCC were significantly thinner in the PPG eyes than those in the OHT eyes (pRNFL: WMD = −8.57, 95% CI = −9.88 to −7.27, P < 0.00001; mGCC: WMD = −3.23, 95% CI = −6.03 to −0.44, P = 0.02). Similarly, the sectoral pRNFL and mGCC were also significantly thinner in the PPG eyes than those in the OHT eyes. Conclusion: OCT-based measurements of peripapillary and macular structural alterations can be used to distinguish PPG from EG and OHT, which can help understand the pathophysiology of glaucoma at earlier stages. Studies that employ clock hour classification methods and longitudinal studies are needed to verify our findings. Systematic Review Registration: https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=239798 CRD42021239798


INTRODUCTION
Glaucoma is a group of neurodegenerative diseases that is characterized by the progressive loss of retinal ganglion cells (RGCs) and axons, followed by the irreversible visual field (VF) deterioration (1). Glaucoma is one of the leading causes of blindness, and ∼111.8 million people worldwide are expected to suffer from glaucomatous optic neuropathy through 2040. This imposes a huge burden on public health systems (2). Elevated intraocular pressure (IOP) is believed to be one of the major risk factors of glaucoma. Patients that have increased IOP with normal appearance of the optic disc can have about nine times the risk of developing glaucoma and are regarded as ocular hypertension (OHT) individuals or glaucoma suspects (3). Currently, reducing the IOP is the only effective method for glaucoma treatment (4,5). However, since glaucoma has an insidious onset and obscure symptoms especially at the earlier stages, such as pre-perimetric glaucoma (PPG) and early perimetric glaucoma (EG), patients are usually diagnosed at an advanced stage with severe VF defects (2,6). Thus, early detection of glaucomatous damage is crucial for hypotensive therapies to slow glaucoma progression and ameliorate the quality of life (7,8).
Several methods have been utilized for the diagnosis of glaucoma, and of these, standard automated perimetry-based VF examination is the gold standard for evaluating the severity of glaucomatous damage (9). Nevertheless, studies have shown that ganglion cell loss can precede VF defects in glaucoma (10)(11)(12)(13)(14)(15)(16)(17)(18) indicating that morphological changes occur earlier than functional damage. Since severe functional damage is closely related to the central region of the VF, it is difficult to solely rely on poor patient performance in standard automated perimetry (19). Moreover, the VF test is occasionally unreliable, which impairs its diagnostic efficacy. Therefore, more objective and reproducible methods are required for assessing the peripapillary and macular structure changes in glaucoma.
Thus, we conducted this systematic review and meta-analysis to facilitate a better understanding of glaucomatous progression form OHT without apparent ONH configuration changes to the pre-perimetric stage with structural deterioration, and from the pre-perimetric to the early perimetric stage with impaired VF in view of the peripapillary and macular structural alterations, and to enable ophthalmologists to discriminate PPG from EG, and PPG from OHT.

METHODS
This systematic review and meta-analysis adhered to the Preferred Reporting Items for Systematic Reviews and Metaanalysis (PRISMA) statement methodology and the Metaanalysis of Observational Studies in Epidemiology (MOOSE) guidelines (49,50). Three investigators (YT, TW, and YH) independently performed the literature search, data extraction, and quality assessment based on the same standard. The study was registered in PROSPERO (CRD42021239798).

Literature Search
We performed a comprehensive literature search of the PubMed database, Embase, and the Cochrane Library from inception to March 2021 using the following strategy with the combination of free text terms and Medical Subject Headings (

Inclusion and Exclusion Criteria
Eligible studies were included if they fulfilled the following criteria: (1) original article; (2) inclusion of PPG and (EG or OHT) with the same diagnostic standard: for PPG, patients had to demonstrate characteristic glaucomatous optic nerve damage (i.e., neuroretinal rim thinning, excavation, or notching) without a reproducible VF; for EG, in addition to the typical glaucomatous optic disc changes (i.e., neuroretinal rim thinning, excavation, or notching), the mean deviation (MD) of the VF defect had to exceed −6 dB based on the Hodapp-Anderson-Parrish VF severity grading scale (51); for OHT, patients with an elevated IOP > 21 mmHg but with normal optic disc appearance and VF were included; and (3) inclusion of at least one of the following quantitative indicators measured by OCT -pRNFL, mGCIPL, or mGCC thickness.

Data Extraction
The following details were extracted with regard to the studies: title, first author, publication year, study type, region, source of patients, number of patients and eyes, mean age of patients, female/male ratios, type of spectral domain OCT (SD-OCT) or time domain OCT (TD-OCT) devices, type of glaucoma, diagnostic standards, average and sectoral pRNFL, mGCIPL and mGCC thickness, scan area and protocol of the ONH and macular region, and MD of the VF. Disagreements were resolved through consensus after discussion among all authors.

Quality Assessment
The Agency for Healthcare Research and Quality (AHRQ) methodology checklist was used to evaluate the quality of the included cross-sectional studies.

Statistical Analyses
All statistical analyses were conducted using Review Manager V5.4.1 (Cochrane Collaboration, London, United Kingdom) and Stata V12.0 (StataCorp LLC, Texas, America). We employed the weighted mean difference (WMD) with the 95% confidence interval (CI) to pool the mean differences of OCT parameters between the PPG and EG groups, and between the PPG and OHT groups. A P-value of <0.05 was considered to be statistically significant. Heterogeneity was estimated using Cochrane's Q test and I 2 statistics. A fixed-effects model was used when I 2 < 50% (52); otherwise, a random-effects model (Der Simonian-Laird method) was used. We performed subgroup analyses according to the type of glaucoma, type of OCT device, and macular scan area. Subgroups with less than two included studies were excluded to prevent further discrepancy. "Leave-one-out" sensitivity analyses were performed to validate the stability and   (53,54). In some articles, the eight-quadrant classification method was used to display sectoral thickness while other studies employed the four-quadrant classification method. Since most of the studies used the four-quadrant classification (only one study used the eight-quadrant classification, which we included for the combined analysis regarding the pRNFL thickness), we transformed the eight-quadrant data to fourquadrant data to reduce heterogeneity using the following formula to combine the means and standard deviations of the two groups: x 1 and x 2 are the mean pRNFL or mGCIPL or mGCC thickness of the two adjacent sections in the eight-quadrant classification. σ 1 and σ 2 are the standard deviations of the two groups. N 1 and N 2 are number of the eyes in the two sections. x 12 and σ 12 are the combined mean and standard deviation (i.e., x 1 and σ 1 refer to parameters of the temporal superior quadrant, x 2 and σ 2 refer to parameters of the temporal inferior quadrant, and x 12 and σ 12 refer to combined parameters of the temporal superior and inferior quadrants).

Literature Search
A total of 334 studies were retrieved in our screening, of which 85 duplicates were removed and 208 articles were excluded by titles and abstracts. We further excluded 17 studies with unextractable data and 1 study that used a different diagnostic standard. Finally, 23 studies were integrated into the qualitative and quantitative analyses . The flow diagram of literature search is shown in Figure 1.

Characteristics of Included Studies
According to our eligibility criteria, 23 cross-sectional studies comprising 2,574 eyes (1,101 PPG eyes, 1,233 EG eyes, and 240 OHT eyes) were included in the systematic review and the meta-analysis. Detailed characteristics of the included studies are summarized in Table 1. The AHRQ checklist scores of all the included cross-sectional studies were not <5, demonstrating that the studies were of good quality. Details are presented in Table 2.

Subgroup Analyses
The stratified subgroup analysis according to the type of glaucoma (Table 3) (Table 4), the average and quadrant pRNFL thickness were significantly lower in the EG eyes than in the PPG eyes regardless of the kind of OCT that was used except for the nasal quadrant of pRNFL in the Cirrus SD-OCT subgroup (WMD = 6.38, 95% CI = −1.02 to 13.78, P = 0.09).

Peripapillary RNFL Thickness
Six studies evaluated average pRNFL thickness in patients with PPG and OHT with I 2 < 50%; thus, fixed-effects model was used.

Publication Bias
No significant publication bias was shown according to the results of Egger's and Begg's tests (P > 0.05, Table 6), and no obvious asymmetry or correlation between study and effect size was observed in the funnel plot in terms of pRNFL and mGCC thickness (Supplementary Figures 2-5). However, slight asymmetry was noted in the funnel plot of average mGCIPL thickness (Supplementary Figure 6A), but not in those of superior and inferior mGCIPL thickness (Supplementary Figures 6B,C).

Sensitivity Analyses
There was no obvious change in the results after "leave-oneout" sensitivity analyses, indicating the reproducibility and stability of our results (Supplementary Figures 7-11). However, the sensitivity analysis of average mGCIPL thickness in patients with PPG and EG indicated that the study by Kim et al. (40) contributed mostly to the heterogeneity ( Table 7). After excluding this study, heterogeneity was largely reduced (I 2 = 34%, Supplementary Figure 12) and the funnel plot became symmetrical ( Table 6 and Supplementary Figure 13, P-value of Egger's test increased from 0.195 to 0.745). In addition, the sensitivity analysis of average mGCC thickness in PPG eyes and EG eyes demonstrated that the study by Aydogan et al. (39) was the major source of the heterogeneity ( Table 8).
After excluding this study, the I 2 decreased from 83% to 64% (Supplementary Figure 14). In the sensitivity analysis of average mGCC thickness in the PPG eyes and OHT eyes, the study by Aydogan et al. (39) was also shown to introduce the heterogeneity mostly ( Table 9). After excluding this study, no heterogeneity was noted (I 2 = 0%, Supplementary Figure 15).

DISCUSSION
In the present study, we first pooled the average and sectoral pRNFL, mGCIPL, and mGCC thickness in patients with PPG and EG. Our results demonstrated that the average and sectoral pRNFL, mGCIPL, and mGCC were significantly thinner in the EG eyes than in the PPG eyes. These findings were consistent across several studies (26, 29, 34, 38-41, 43, 44), whereas the results of eight studies were not significantly different in terms of the average or sectoral pRNFL thickness (27,28,30,31,(35)(36)(37)42); two studies reported there was no significant difference in the EG eyes compared to the PPG eyes concerning mGCC thickness (28,36), and one study did not demonstrate a significant reduction in the EG eyes regarding the average and superior mGCIPL thickness (33). Currently, the exact biomechanisms of glaucomatous neurodegeneration remain poorly understood (1,55). Nevertheless, continuous and progressive glaucomatous damage may lead to configuration changes in retinal ganglion cell dendrites, soma, and axons (56,57), causing the attenuation of thickness in corresponding residing sites, inner plexiform layer, ganglion cell layer and RNFL (20). Our pooled results of pRNFL, mGCIPL, and mGCC thickness suggested that more serious structural damage occurred in EG than in PPG.
To explore the source of heterogeneity across the included studies, we performed subgroup analyses according to the different types of glaucoma. Our findings did not show a significant decrease in the superior quadrant of pRNFL (P = 0.08), and the average, as well as the inferior mGCC thickness in the EG eyes when patients with open-angle glaucoma were enrolled (average: P = 0.09; inferior quadrant: P = 0.16). This was probably mainly due to the relatively small sample size in the subgroup of OAG regarding the pRNFL and the mGCC thickness (N = 2), because of which heterogeneity could not be excluded. Another reason was that no quality assurance step was taken in the study of Cennamo et al. (36). Since only a single OCT examination was performed to evaluate each parameter by one experienced ophthalmologist, ensuring the reproducibility and reliability of the examination results was difficult. Based on the anatomy of retina, ∼30-50% RGCs are centered within the 4.5-mm-circle region of the fovea (44, 58-60); thus, using a relatively small scan area of 3 × 3 mm may not allow the differentiation of EG from PPG or suspected glaucoma (41,61). For this reason, we also performed a subgroup analysis of the macular scan area. Interestingly, our pooled data showed that when the 6 × 6 mm scan protocol was used, there was a significant decrease in mGCC thickness in the EG eyes compared with the PPG eyes, whereas there was no difference when the 7 × 7 mm scan protocol was used. One reason for the different diagnostic performances between these two scan protocols may be the decreased signal-to-noise ratio. Although the enlarged scan area could cover the region with the most abundant RGCs, the concomitant decrease in signal-to-noise ratio and increase in test-retest variability (62) may have undervalued the assessment of GCC thickness measured by SD-OCT. Another explanation was the strict inclusion criteria by Arintawati et al. (28). In this study, investigators only accepted subjects with all the typical glaucomatous changes observed in fundus photographs to avoid false-positive cases. Consequently, patients with higher severity may have been enrolled; thus, showing a minimal difference between the PPG and EG groups. Unfortunately, because of the small sample size (N = 3), we were unable to include other scan areas and protocols, introducing difficulties in the overall evaluation of the diagnostic values regarding different macular scan protocols. In addition, sensitivity analysis demonstrated that the study of Aydogan et al. (39) contributed mostly to the heterogeneity of average mGCC thickness. The main reason was that the age was not well-matched among PPG and EG groups (P < 0.001), which may induce potential bias since glaucoma is an age-related optic neuropathy (55). However, after excluding this study, the heterogeneity decreased (I 2 decreased from 83 to 64%).
Recently, in addition to OCT, several studies have focused on the macular microvasculature changes in PPG and EG via OCT angiography, revealing the progression patterns of glaucoma with respect to microvascular dysregulation (41,42,44,63,64). Although macular vessel density (VD) was reported to significantly decrease both in PPG and EG, one study demonstrated that mGCC thickness, unlike macular VD, could serve as a tool to discriminate PPG from EG (41). The study also showed that the percentage loss of mGCC thickness was significantly higher than that of macular VD both in PG and EG. Another study reported that both inferotemporal and superotemporal pRNFL thickness were significantly decreased in EG eyes compared to PPG eyes whereas only the inferotemporal sector of the radial peripapillary capillary VD experienced a significant decrease (42). Considering previous OCT studies that suggest that structural deterioration usually occurs before functional loss (10)(11)(12)(13)(14)(15)(16)(17)(18), these findings consolidate the results of our study, which indicate that the OCT  evaluation of macular structure changes could help to clarify the pathophysiological mechanisms of glaucoma and differentiate PPG from EG.
In a subgroup analysis of pRNFL thickness according to the type of SD-OCT, the pooled results were generally consistent with the combined pooled data except for the nasal quadrant in the Cirrus SD-OCT subgroup. The main reason was that Kim et al. (34) utilized the optic disc Cube 200 × 200 scanning protocol, whereas the other two studies focused on a 3.46-mmdiameter circle region of the ONH (31,35). Although all types of SD-OCT could detect characteristic glaucomatous damage patterns of pRNFL thickness, different algorithms, software, and parameters may have induced subtle differences in diagnostic performance (65). However, our pooled results demonstrate the important diagnostic value of SD-OCT in evaluating the severity of the glaucomatous damage of the ONH.  Apart from this, studies have shown that individuals with OHT are at higher risk of developing glaucoma than others (3,4,66); however, there is no optimal IOP cut-off that possesses both reasonable sensitivity and specificity (3), which may delay the early diagnosis until patients are found to have apparent optic disc structural deterioration. Although the widely used cut-off of 21 mmHg has high sensitivity, it can decrease the specificity to 44% (67). In addition, spectrum bias usually occurs when studies inappropriately include the control groups without any suspicious symptoms of the disease, which can impair the diagnostic efficacy when clinically non-relevant individuals are enrolled (68)(69)(70). Thus, we also pooled the average and sectoral pRNFL and mGCC thickness in patients with PPG and OHT.
The pooled results revealed that the pRNFL and mGCC were significantly thinner in the PPG eyes than those in the OHT eyes, which were consistent with several investigations (45,47,48). However, two studies reported that there was no difference in patients with PPG and OHT in terms of the average mGCC thickness (39,43); one study demonstrated that the average pRNFL thickness could not serve as a valued indicator for differential diagnosis (43), and another study showed that no difference was noted in the PPG eyes and OHT eyes regarding the superior mGCC thickness (46).
Similar to the investigation of exploring the heterogeneity across different OCT parameters for the differential diagnosis of PPG and EG, we also performed subgroup analysis of average pRNFL thickness according to the type of OCT. However, no difference was noted in the pooled results with either SD-OCT or TD-OCT. Compared to the traditional TD-OCT, SD-OCT is the latest generation of OCT with ultra-high scanning speed and retinal image resolutions, and is reported to have higher diagnostic abilities in terms of sectoral pRNFL and macular thickness. However, both SD-OCT and TD-OCT showed comparable reproducibility regarding the mean pRNFL (71). This was consistent with our finding. In the sensitivity analysis, the study of Aydogan et al. (39) also contributed mostly to the heterogeneity of mGCC thickness in patients with PPG and OHT. This may be due to usage of the different macular scanning protocol.
Despite the strengths of our study, some of its limitations should be considered. First, mild asymmetry was shown in the funnel plot of average mGCIPL thickness (Supplementary Figure 6A), suggesting potential publication bias. To elucidate the source of the bias, we performed a "leave-one-out" sensitivity analysis ( Table 7). The results showed that the study by Kim et al. (40) contributed mostly to the heterogeneity in average mGCIPL thickness, wherein some problems of automated segmentation software occurred, although measures were taken to minimize the consequence of the segmentation error. Therefore, after excluding this study, low heterogeneity (I 2 = 34%) was noted (Supplementary Figure 12), and the funnel plot became symmetrical (Supplementary Figure 13). Second, regarding the subgroup analysis of the macular scan area, due to the relatively small sample size, we did not include other scan protocols apart from the 6 × 6 mm and 7 × 7 mm scan protocols. Further, the small sample size may have also introduced heterogeneity in the analysis of mGCC thickness, although sensitivity analyses proved the reliability of our results. Further studies should be included to comprehensively evaluate the influence of scan area and protocols on the assessment of macular structure changes in PPG, EG, and OHT. Moreover, the heterogeneity was high regarding many of our findings, indicating that the results should be cautiously interpreted.

CONCLUSION
The OCT-based assessment of peripapillary and macular structural changes could be potentially utilized to discriminate PPG from EG and OHT. This facilitates a better understanding of the pathophysiology of glaucoma and provides references for ophthalmologists to manage individuals suspected to have glaucoma and glaucoma patients according to the extent of severity in a non-invasive way. Further studies that employ clock hour classification methods that can monitor the configuration alterations in a narrower range and longitudinal studies are needed to verify our findings.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.

AUTHOR CONTRIBUTIONS
YT, TW, XZ, and BJ: conceptualization and design. YT, TW, and YH: literature search, data extraction, quality assessment, and data analysis. YT, TW, XZ, YH, and BJ: manuscript writing and editing. BJ: supervision. All authors approved the final version of the manuscript.