Performance of perimetric glaucoma staging systems and their preference patterns among the Indian eye care practitioners

Abstract

Context:

Various systems are available to stage glaucomatous field defects into different severity grades. Still, the preference for such systems among Indian eye care practitioners (ECPs) and the consensus between them are not explored.

Purpose:

To compare the clinical performance of perimetric glaucoma staging systems and to assess their preference patterns among Indian ECPs.

Setting and Design:

Retro-prospective cross-sectional study.

Methods:

Three investigators stratified 400 perimetry reports using four systems: Hodapp Anderson Parrish (HAP), enhanced Glaucoma Severity Staging (eGSS), Advanced Glaucoma Intervention Study (AGIS), and Collaborative Initial Glaucoma Treatment Study (CIGTS). Their performance was compared in terms of agreement and association. Following this, 30 ECPs from across the country segregated a subset of reports (n = 20) into different stages based on their system of choice, and their preferences/perceptions about the available perimetry staging systems were enquired and documented.

Statistical Analysis Used:

Descriptive statistics, Kruskal–Wallis, Cohen’s kappa, and Spearman’s correlation.

Results:

For the consolidated severity stages, substantial agreements (K ranging from 0.61 to 0.75) were found between all the systems, except for HAP–CIGTS (K = 0.58) and AGIS–CIGTS (K = 0.51). All ECPs were aware of perimetry staging systems, and a high preference was reported for HAP (97%) due to ease of implementation and interpretation.

Conclusion:

Among the four staging systems, the highest agreement was found between HAP and eGSS. A majority of ECPs proposed the need for a uniform classification system across the practices for clinical comparisons and patient education.

Perimetry is an integral part of glaucoma management to quantify the associated visual field defects (VFDs) as well as to monitor the progression.[1,2] Various systems are available to stage glaucomatous field defects into different severity grades.[3,4,5] Currently, there is sparse information regarding the preference for such systems among eye care practitioners (ECPs) and the consensus between their clinical outcomes. Hence, this study aimed to compare the clinical performance of perimetric glaucoma staging systems and to assess the pattern of preference and perception toward the existing systems among Indian ECPs.

Methods

Study participants

This cross-sectional retro-prospective study considered a set of visual field (VF) reports obtained using the Swedish Interactive Thresholding Algorithm (SITA) standard 24-2 protocol of the Humphrey Field Analyzer (HFA; Carl Zeiss Meditec, Inc., Dublin, CA, USA). These reports were extracted by scrutinizing the perimetry database of the consecutive glaucoma patients who visited the outpatient service of the glaucoma department at a tertiary eye care center based in Chennai, India. Glaucoma was defined based on the examination of definite structural alterations according to the International Society of Geographic and Epidemiologic Ophthalmology (ISGEO) criteria.[2] The first 400 reliable and repeatable reports that met the inclusion criteria were collected and categorized by a single grader according to the Hodapp Anderson Parrish (HAP) system. This first-stage categorization (normal, mild, moderate, severe) was performed to ensure an identical proportion of reports in normal as well as different levels of defect severity (n = 100 in each). The grader considered the overall extent of VFD using both mean deviation (MD) values in decibels (dB) and the number of defective VF locations and their proximity to fixation as flagged by the pattern deviation (PD) probability plot of HFA.[6] All the reports included were within acceptable limits of reliability (<33% fixation losses and false negatives and <15% false-positive responses). Any reports exhibiting substantial non-glaucomatous disease-related VFDs resulting from either retinal diseases or lesions, and/or any patient with a clinical history of long-term intake of medications that can have a potential effect on VF sensitivity, and/or any patient with an incomplete medical record were not included.

Performance of the perimetric glaucoma staging systems

Three trained graders categorized the full set of HFA reports (de-identified) in consensus based on four functional classification systems – HAP,[6,7] enhanced Glaucoma Severity Staging (eGSS),[7,8] Advanced Glaucoma Intervention Study (AGIS),[7,9] and Collaborative Initial Glaucoma Treatment Study (CIGTS).[7,10] Along with the randomly organized reports, supportive materials [Supplementary File 1] were also handed over to the graders as a reference to different grading criteria. Graders segregated each report into normal and abnormal/defective, and the latter was further subcategorized into different staging levels according to the applied system.

Since each of the four systems has its set criteria of subdividing each stage into substages for fine categorization, they end up with different numbers of stages. Hence, to make the outcomes of the systems comparable, consolidated stages (CSs) were also arrived at. Normal (CS1) was represented by HAP normal, eGSS stage 0 and borderline, AGIS normal, and CIGTS normal; mild (CS2) by HAP early, eGSS stage 1, AGIS mild, and CIGTS mild; moderate (CS3) by HAP moderate, eGSS stage 2 and stage 3, AGIS moderate, and CIGTS moderate; and finally, severe (CS4) by HAP severe, eGSS stage 4 and stage 5, AGIS severe, and end-stage, and CIGTS severe and end-stage. Graders were masked to the identity of the patients. Apart from the categorical staging levels, global indices such as MD, pattern standard deviation (PSD), both in dB, and visual field index (VFI in %) of each report were also exported for further analysis.

Preference patterns for the glaucoma staging systems among Indian ECPs

This phase included 30 ECPs (15 ophthalmologists and 15 optometrists) from across the country with a minimum of 2 years of clinical experience in glaucoma evaluation, diagnosis, and/or management. After obtaining consent for participation, an online survey was conducted that included 20 HFA reports, a subset (n = 20) drawn from the main set (n = 400). The reports were de-identified and randomly arranged in a Google form with adequate clinical data, and ECPs were instructed to segregate the reports into different stages based on their system of choice. Following this, a set of questions [Supplementary File 2] on ECPs’ preference/perception about the available perimetry staging systems were administered. The questions aimed at knowing the awareness of ECPs about existing VF staging systems, their choice for a specific system during routine practice for clinical judgment as well as for patient education purposes, and then the list of VF factors/indices they typically consider to draw clinical decisions. All these responses were obtained and recorded for further analysis.

Statistical analysis

All the data were entered in Microsoft Excel, and statistical analysis was performed using Statistical Package for Social Sciences (SPSS; version 25). Tests for normality (Kolmogorov–Smirnov) were conducted, and suitable statistical methods were applied. The distribution (frequency) of VF reports into different stages was compared across the classification systems. Then, the agreement between the classification systems was estimated based on gross-stage (GS) classification (normal vs. abnormal/defective) as well as for the consolidated ones (normal, mild, moderate, severe). These inter-system and inter-grade agreements were assessed using the kappa statistic; based on Cohen’s kappa coefficient (κ), the agreement strength was flagged as none (κ = 0.0–0.20), minimal (κ = 0.21–0.39), weak (κ = 0.40–0.59), moderate (κ = 0.60–0.79), strong/substantial (κ = 0.80–0.90), and almost perfect (κ above 0.90).[11] Descriptive statistics of the global indices such as MD, PSD, and VFI were compared across all staging systems (Kruskal–Wallis). Finally, Spearman’s rank correlation was used to assess the association between the predicted outcomes of each system. From the online survey responses, first, the inter-practitioner agreement analysis was run for the whole sample of ECPs and also by segregating them into two subgroups based on their years of clinical experience. The preference and perception of ECPs regarding the staging systems were appraised using frequency and percentages.

Results

Clinical performance of perimetric glaucoma staging systems

A total of 400 reports (n = 203 right eye reports, 51%) were included in the final analysis. The percentages of VF flagged as normal with eGSS, AGIS, and CIGTS were 21% (n = 85), 25% (n = 99), and 16% (n = 63), respectively [Table 1]. While considering CSs, the AGIS system staged the highest percentage of reports (45%) under the category of mild, the eGSS system flagged 38% of reports under the moderate category, while CIGTS identified 35% of the reports under the severe category [Fig. 1].

Table 1.

Details of visual field reports categorized as normal and abnormal using the HAP, eGSS, AGIS, and CIGTS classification systems

Systems	Stages	eGSS		AGIS		CIGTS		Total
		Normal	Abnormal	Normal	Abnormal	Normal	Abnormal
HAP	Normal	74	26	84	17	58	42	100
	Abnormal	11	289	15	284	5	295	300
eGSS	Normal			78	8	61	24	85
	Abnormal			21	293	2	313	315
AGIS	Normal					59	40	99
	Abnormal					4	297	301
Total		85	315	99	301	63	337	400

AGIS=Advanced Glaucoma Intervention Study, CIGTS=Collaborative Initial Glaucoma Treatment Study, eGSS=enhanced Glaucoma Severity Staging, HAP=Hodapp Anderson Parrish, VFD=visual field defect

Figure 1.

Stacked bar graph illustrating the number of visual field reports (%) categorized into four consolidated stages (normal, mild, moderate, severe) on the basis of four perimetry staging systems

For the GS classification, linear weighted kappa analysis showed a substantial degree of agreement between eGSS and AGIS systems and a moderate degree of agreement between the remaining pairs. Meanwhile, for CSs, weak degrees of agreement were found between HAP versus CIGTS and AGIS versus CIGTS, whereas the rest of the pairs showed moderate degrees of agreement [Table 2]. All pairs of classification systems showed a strong positive correlation for classifying the VF reports into CSs (Spearman rank, r = 0.75–0.87, P < 0.001). In the agreement analysis within the CSs the normal stages showed moderate agreement (κ = 0.73), severe stages showed a weak agreement (κ = 0.58), whereas the mild and moderate stages showed minimal agreement (κ = 0.32 and 0.26 respectively) across the four classification systems.

Table 2.

Inter-system agreement analysis using kappa while making gross stage categorization (normal/abnormal) and consolidated staging (CS 1–4)

Classification systems	κ coefficients		P
	GSs	CSs
HAP and eGSS	0.73	0.75	<0.001
HAP and AGIS	0.79	0.67
HAP and CIGTS	0.63	0.58
eGSS and AGIS	0.80	0.61
eGSS and CIGTS	0.78	0.68
AGIS and CIGTS	0.66	0.51

κ coefficients=Weighted kappa coefficients, AGIS=Advanced Glaucoma Intervention Study, CIGTS=Collaborative Initial Glaucoma Treatment Study, GSs= Gross-stages, CSs=Consolidated stages, eGSS=Enhanced Glaucoma Severity Staging, HAP=Hodapp Anderson Parrish

Comparison of the distribution of global indices (MD, PSD, and VFI) across CSs for all four systems revealed a statistically significant difference (P < 0.001) in MD (dB) for CS1 [Fig. 2a-c, respectively]. Meanwhile, the differences in the other two indices (PSD and VFI) were found to be statistically insignificant (P = 0.59 and 0.15, respectively) across the four classification systems. Similarly, for the remaining CSs (CS2–CS4), all three indices were found to be significantly different (P < 0.001, Kruskal–Wallis/post hoc) across the four classification systems.

Figure 2.

Box plot displaying the distribution of global indices (a: mean deviation, b: pattern standard deviation, and c: visual field index) across the consolidated stages (normal, mild, moderate, severe) for all four visual field classification systems

Preference patterns for perimetric staging systems among Indian ECPs

A total of 30 ECPs participated in the study, out of which 15 were ophthalmologists and 15 were optometrists. The years of clinical experience of ECPs varied from 2 to 30 years in glaucoma practice.

All ECPs (100%) were aware of the systems available for grading glaucomatous VFDs. Regarding the choice of classification systems, 15 (50%) opted for/preferred HAP, 10 (34%) opted for eGSS, four (13%) opted for AGIS, and one (3%) opted for CIGTS. VF reports graded with the HAP system showed a weak agreement (Fleiss κ = 0.57) across all ECPs. For further analysis, ECPs were divided into two groups with equal numbers in each based on their years of clinical experience (<10 or ≥10 years). In group 1 (<10 years), six (40%) of them preferred the HAP system, whereas in group 2, nine (60%) ECPs preferred the HAP system. Fleiss kappa analysis revealed a weak agreement (κ = 0.53) in group 1, whereas group 2 showed a moderate agreement (κ = 0.62) while grading the VF report using the HAP system.

ECPs reported a variety of factors that they relied upon while interpreting HFA reports. The most frequently reported ones were the global indices which included MD, PSD, and VFI, followed by reliability indices which include, fixation loss, false positive, and false negative. Twenty-nine (97%) participants responded positively to the need for a standard functional classification system. Sixteen (53%) ECPs reported that they used staging systems for educating their patients regarding their VF status, disease severity, and progression.

Discussion

The current study assessed the clinical performance of perimetric glaucoma staging systems (HAP, eGSS, AGIS, and CIGTS) and their preference patterns among Indian ECPs. ECPs included both ophthalmologists and optometrists. Optometrists were involved as they are the primary ECPs who are most often the first line of contact for glaucoma patients, especially in the secondary and tertiary eye care levels or placed within an independent practice dealing with glaucoma patients. These optometrists conduct comprehensive ophthalmic evaluations including VF investigation and report interpretation. They are also involved in enhancing awareness about glaucoma, community screening, and also monitoring disease progression. Since these active roles make optometrists an integral part of glaucoma evaluation and management, we decided to include an equal proportion of optometrists and ophthalmologists (involved in glaucoma evaluation and diagnosis with a minimum of 2 years of experience). ECPs were included based on a convenient sampling, and we attempted to maintain a relatively equal proportion of ECPs from various parts of the country and did not pool in information from a single region.

The results of the study revealed moderate degrees of inter-system agreement, except for CIGTS versus HAP and AGIS. A majority of Indian ECPs who participated in our survey agreed with HAP for both clinical staging, monitoring, and patient education. ECPs agreed upon the need for a standardized system for clinical comparisons across practices. When compared with the other three systems, CIGTS had a higher percentage of reports flagged as abnormal/defective, suggesting that the system overrates the presence of glaucomatous functional defects; this could be because of its sole dependence on total deviation (TD) probability plot, which does not adjust for generalized VF depression caused by ocular opacities and pupillary miosis.[12] Hirasawa et al.[13] stated that the eGSS grades were not consistent with the CIGTS system; however, in this study, the two systems had a strong positive correlation in both GSs and CSs. The eGSS system graded the reports more severely compared to AGIS and had a strong positive correlation and a moderate agreement for grading reports in CSs, which was on a par with a previous study by Ng et al.[3] For GSs, they showed a substantial agreement; however, the previous study showed a moderate agreement. A study published recently showed none, moderate, and no agreement for HAP versus eGSS, HAP versus AGIS, and, AGIS versus eGSS systems, respectively.[14] The current study, on the other hand, revealed a moderate agreement for all three pairs while grading the reports in CSs. Discrepancies in agreement level within CSs across the four systems [Fig. 1 and Table 3] might have occurred because of differences in the number of categories for classifying VFDs in each of the systems [Supplementary File 1], thereby suggesting the need for a standardized classification system.

Table 3.

Sample VF reports illustrating the level of agreement/disagreement for the consolidated stages across the four classification systems

Classification systems	Report 1	Report 2	Report 3	Report 4
	MD: -1.12	MD: -1.43	MD: -2.72	MD: -18.31
	PSD: 1.43	PSD: 1.71	PSD: 2.25	PSD: 13.42
	VFI: 99	VFI: 97	VFI: 96	VFI: 53
Systems
HAP	Normal	Mild	Moderate	Severe
eGSS	Normal	Normal	Normal	Severe
AGIS	Normal	Normal	Normal	Moderate
CIGTS	Normal	Mild	Mild	Severe

AGIS=Advanced Glaucoma Intervention Study, CIGTS=Collaborative Initial Glaucoma Treatment Study, eGSS=enhanced Glaucoma Severity Staging, HAP=Hodapp Anderson Parrish, MD=Mean deviation (dB), PSD=Pattern standard deviation (dB), VFI=visual field index (%)

As there is no clinical standard for staging VFD severity, we relied upon the three global indices (MD, PSD, and VFI) to characterize and compare the categorical classification systems. The differences in these indices noticed in this study might have occurred because of varying criteria followed by each of the systems [Supplementary File 1]. On the contrary, a previous study found that there was no significant difference in these indices across GSs of VFDs between the AGIS and eGSS systems.[3] The three investigators exhibiting excellent agreement for all the classification systems implied reproducibility of these grading systems in our study. They reported that since they were aware of the HAP criteria, grading reports with this system was very easy, followed by eGSS, wherein initially, they found little difficulty but gradually managed well. AGIS and CIGTS were quite complex and time-consuming; however, after learning effect, the investigators were able to pick up the pace. The AGIS system has a website for classifying VFDs, but it requires entering all the values one by one on the website and was reported to not help the investigators in saving time.[15]

Our study is the first to report statistical information on the views of ECPs on the available functional classification systems of glaucoma. A similar finding, that is, HAP is the most widely used classification system in a clinical setting, was obtained in our study, which can be because of the easily memorable factors considered while grading.[7] The second most preferred system in this study, eGSS, is reported to be the best choice for VFD classification system in both clinics as well as in research as it uses only two global indices (MD and PSD) to categorize VF with the help of a nomogram;[3] however, in situations when this nomogram is not available, it can make categorizing of VFDs impossible. AGIS and CIGTS were the least preferred systems in our study. AGIS was originally designed particularly for intervention purposes; hence, only pure glaucomatous VFDs can be graded and it is not intended for the defective points that arise on the TD plot because of ocular opacities and pupillary miosis. Both AGIS and CIGTS are time-consuming, which is not acceptable in the busy schedules of ECPs.[7,9,10]

The global and reliability indices were the most dependent factors with a few others like the pattern of the defect, PD plot, and glaucoma hemifield test (GHT) for interpreting the HFA VF report by ECPs. Majority of ECPs responded positively to the need for a standardized classification system. A similar statement was made by Ng et al.,[3] wherein they mentioned the need for one common VFD system. A moderate agreement for ECPs in group 2 for grading VF reports can be attributed to the application of the HAP system for a longer duration in their practice, which, in turn, might have made them thorough with the system criteria. Grading the report with just one of the criteria of the system is subjected to errors in staging, which might be the case for group 1. Thorough knowledge of the classification systems, at least one in detail, is necessary and should be included as a part of the educational curriculum.

The lack of a standard functional classification system hindered determination of validity of the other systems, which, if present, would have also aided to decide on the best treatment modality and monitoring the progression of glaucoma.[4] The functional findings of classification were not correlated with the structural characteristics, which can provide further insight into monitoring and understanding the progression of glaucoma. A standardized classification system (ideally five stages) can be applied in routine clinical practice, which will help in a rapid yet accurate staging of functional defects, thus ensuring repeatability within and between ECPs and in parallel suitable for patient education. A quickly accessible software program can be designed and developed perhaps with artificial Intelligence (AI) to help in predicting the presence of the disease, and categorizing the same according to severity level might have substantial clinical value.

Conclusion

Though a considerable agreement was shown between the systems, each of the systems led to different levels of severity staging depending on their categorization criteria. Each of the systems was shown to have its distinct strengths and limitations. Among Indian ECPs, HAP seemed to be the most preferred classification system for clinical interpretation as well as for patient education.

Ethical approval

The study adhered to the tenets of the Declaration of Helsinki, and the experimental measures were reviewed and accepted by the Institutional Review Board and the Medical Ethics Committee of Vision Research Foundation, Chennai, India.

Financial support and sponsorship:

Nil.

Conflicts of interest:

There are no conflicts of interest.

Supplementary File 1: Supportive Material for Grading Glaucomatous Visual Field Defects

1. Hodapp Anderson Parish (HAP)

   Visual field (VF) categories – normal, early, moderate, and severe.

   • Minimum criteria for diagnosing acquired glaucoma damage:

     • a Glaucoma Hemifield Test outside the normal limits on at least two fields; OR

     • a cluster of three or more non-edge points in a location typical for glaucoma, all of which are depressed on the pattern deviation plot at a P < 5% level and one of which is depressed at a P < 1% level on two consecutive fields; OR

     • a corrected pattern standard deviation that occurs in less than 5% of normal fields on two consecutive fields.

   • If the VF report does not qualify at least one of the minimum criteria – normal VF.

2. Brusini’s enhanced Glaucoma Staging Severity (eGSS) system.

   VF categories – stage 0, borderline, stage 1, stage 2, stage 3, stage 4, and stage 5; graded based on the nomogram given below [Fig. 1 ^{(1.1MB, tif)} ] with the help of mean deviation (MD) and pattern standard deviation (PSD) values.

3. Advanced Glaucoma Intervention Study (AGIS) system.

   VF categories – based on a 20-point scale, Normal (0), mild (1–5), moderate (6–11), severe (12–17), end-stage (18–20). Scoring is done with the help of the AGIS VF Test Scoring reference plot, corresponding to the total deviation (TD) numeric plot [Fig. 2 ^{(1.1MB, tif)} ].

   Nature of defect:

   1. From the six test locations in the nasal region:

      • three or more clusters of depressed adjacent points (may or may not cross the midline) constitute a nasal defect and

      • two or fewer depressed test locations either above or below the horizontal midline including the other three test locations on the opposite of the horizontal midline with no depression constitute a nasal step.

   2. A hemifield with a cluster of three or more depressed sites (more than one cluster can be present in the hemifield) constitutes a hemifield defect.

4. Collaborative Initial Glaucoma Treatment Study (CIGTS) system.

   VF categories – based on a 20-point scale, normal (0), mild (1–5), moderate (6–11), severe (12–17), end-stage (18–20). It takes into consideration the TD probability plot, where each of the 52 points is scored individually [Fig. 3 ^{(879.1KB, tif)} ], the sum of which has a maximum value of 208. This value is then divided by 10.4 to yield scores in the 0–20 range, which are then graded in the same manner as the AGIS system.

   Defective point = P < 5% on the TD plot and has at least two neighboring points with P < 0.05 in the same vertical hemifield.

   Weightage is given based on the following factors:

   1. The minimum depth of the defect at a given point (P < 5%, <2%, <1%, <0.5% is given a weight of 1, 2, 3, and 4, respectively)

   2. Two most defective neighboring points, for example,

      • A point without two neighboring points all depressed to at least P < 0.05 is given a weight of zero.

      • A point at P < 0.01 with only two neighboring points of defect, both at P < 0.05, would receive a weight of 1.

Figure 1

Nomogram for grading visual field with the eGSS system

Figure 2

AGIS visual field test scoring reference plot

Figure 3

Example of visual field classification with the CIGTS system

Table 1.

Classification of VF defects with the HAP system

	Early	Moderate	Severe
MD	<-6 dB	<-12 dB	>-12 dB
Points depressed below 5% and 1% on the PD plot	<18 and <10	<37 and <20	>37 and >20
Sensitivity of central 5°	All points with at least 15 dB	No points with 0 dB	At least one point with 0 dB
No. of hemifields with <15 dB sensitivity in central 5°	-	Only one	Both

Table 2.

Scoring for AGIS system

Type of defect	No. of depressed points	dB	Score
Nasal step/nasal defect	1–6	Not specific	+1
Nasal defect	4–6	≤12	+1
Hemifield defect (points in a cluster)	2	One point depressed ≥12 dB	+1
	3–5	Not specific	+1
	6–12	Not specific	+2
	13–20	Not specific	+3
	>020	Not specific	+4
Deep defect (each hemifield)	Half or more of the adjacent defective locations in a hemifield	≥28 ≥24	+5 +4
		≥20	+3
		≥16	+2
		≥12	+1

Supplementary File 2: Survey

Title: Performance of perimetric glaucoma staging systems and their preference patterns among Indian eye care practitioners

Name of the eye care practitioner _________________________

Age (in years) ____________________________________________

Gender

1. Male

2. Female

3. Other.

Highest educational qualification ________________________

Years of professional experience ________________________

Practice in glaucoma since ______________________________

1. Are you aware of the classification systems used for staging visual field defects in glaucoma?

   1. Yes

   2. No.

2. Which classification system do you rely on while evaluating the stage of the visual field defect?

   1. Hodapp Anderson Parrish (HAP)

   2. Enhanced Glaucoma Severity Staging (eGSS)

   3. Advanced Glaucoma Intervention Study (AGIS)

   4. Collaborative Initial Glaucoma Treatment Study (CIGTS)

   5. Other.

3. What are the factors considered by you while interpretation of HFA field report? _________________________

4. Do you think it is necessary to have a standardized classification system for staging visual field defects?

   1. Yes

   2. No

   3. Other.

5. Do you use staging systems while educating your patient on visual field loss?

   1. Yes (specify below)

   2. No

   3. Other.