A Model of Progression to Help Identify Macular Damage Due to Glaucoma

Abstract

The central macula contains a thick donut shaped region of the ganglion cell layer (GCL) that surrounds the fovea. This region, which is about 12 degrees (3.5 mm) in diameter, is essential for everyday functions such as driving, reading, and face recognition. Here, we describe a model of progression of glaucomatous damage to this GCL donut. This model is based upon assumptions supported by the literature, and it predicts the patterns of glaucomatous damage to the GCL donut, as seen with optical coherence tomography (OCT). After describing the assumptions and predictions of this model, we test the model against data from our laboratory, as well as from the literature. Finally, three uses of the model are illustrated. One, it provides an aid to help clinicians focus on the essential central macula and to alert them to look for other, non-glaucomatous causes, when the GCL damage does not fit the pattern predicted by the model. Second, the patterns of progression predicted by the model suggest alternative end points for clinical trials. Finally, the model provides a heuristic for future research concerning the anatomic basis of glaucomatous damage.

The macular region of the retina represents only about 2% of the total area of the retina, but it contains over 30% of all the ganglion cells.¹ Essential visual functions, such as reading, driving, and face recognition, depend upon the integrity of the ganglion cell layer (GCL) in this region.^2–7 

Recent optical coherence topography (OCT) research has increased our understanding of macular involvement in glaucoma. There is now extensive OCT evidence that early damage affects the GCL of the macula^8–24 and that early progression of glaucoma affects this region as well.^19,25–33 Figure 1A shows the GCL thickness map of a healthy eye. As we move from the periphery toward the center of the macula, the GCL becomes thicker (dark red and white), except for the center of the fovea, where there are no ganglion cells. As a result, in healthy individuals, the topography of the GCL on the OCT resembles the shape of a donut. Although the appearance of this donut region varies in healthy eyes, as seen in Figure 1B, the donut shape is intact. On the other hand, the donuts in eyes with glaucoma typically have pieces missing, as illustrated by the GCL thickness maps from 15 eyes with glaucoma (Figs. 1C–E), which varied from early to advanced based upon the mean deviation (MD) of the 24-2 visual field as indicated on the figure and described in the figure legend. Supplementary Figure S1 contains the GCL thickness maps for all 54 healthy controls and all 62 glaucoma eyes used in the analysis below. 

Figure 1.

 

(A) Ganglion cell thickness (GCL) map for a healthy eye showing the size and sectors of the donut shaped central macular regions. (B–E) Five individuals with healthy eyes (B), and eyes with early (C), moderate (D), and advanced (E) glaucoma. The stage of the patient groups was based upon the mean deviation (MD) of the 24-2 visual field as explained in References 36 and 58.

The clinician is faced with the challenge of distinguishing patterns of GCL thickness due to glaucoma from variations present in healthy eyes, as well as from patterns due to other causes, such as retinal or neurological pathologies and/or artifacts. Here, we describe a model, which is based upon empirically supported assumptions, that predicts the pattern of GCL loss seen as glaucoma progresses. We supply support for this model by applying it to cross-sectional data, as well as longitudinal data. Finally, we discuss its potential use in the clinic, in clinical trials, as well as a heuristic for better understanding the progression of glaucoma. 

The Model of Donut Damage

The model of progression of damage to the donut shaped GCL of the central macula, called the “donut region” here, is based upon previous work.^{8,9,12,28,31,33–37} The schematics in the top row of Figure 2 summarizes the assumed anatomic relation between regions of the donut and the circumpapillary region (dashed black circle) of the disc. The axons of the ganglion cells within the green border in Figure 2A (top panel) enter the temporal quadrant (green arc) of the disc, whereas the ganglion cells within the red border in Figure 2B (top panel) enter the temporal half of the inferior quadrant of the disc. Further, we assume that the superior vulnerability zone (SVZ) and inferior vulnerability zone (IVZ), shown as red arcs, are the most likely regions to be abnormally thin in early glaucoma.^12,37–43 Although the exact location of these regions can vary among individuals due to anatomic differences, these variations will have relatively minor effects on the model's predictions.^44–47 Note that the superior (see Fig. 2A) and inferior retinal (see Fig. 2B) regions are shown separately to make it easier to see the anatomic differences between them, and as a reminder that early glaucoma can start in either one of these regions, as well as both regions, simultaneously.^9,10,12 

Figure 2.

 

A schematic of a model of progression of the GCL donut. (A) Superior macula. The top panel illustrates the assumption that the axons entering the temporal quadrant of the disc (green arc) come from the ganglion cells within the green border. We further assume that the initial damage at the disc occurs in the superior vulnerability zone (SVZ; red arc). The middle panel illustrates that initial damage to the SVZ does not result in damage to the macular region or the GCL donut. However, we assume that damage progresses toward and into the temporal quadrant (red curved arrow). As damage progresses from the blue arcuate to the green arcuate to the yellow arcuate regions, damage in the central GCL donut progresses from temporal superior (TS) to superior (S) to nasal superior (NS), as shown in the lower panel. (B) Inferior macula. The top panel illustrates the assumption that the axons entering the temporal portion of the inferior quadrant of the disc come from the ganglion cells within the red border. We further assume that the initial damage at the disc occurs in the inferior vulnerability zone (IVZ; red arc). The middle panel illustrates that initial damage to the IVZ results in macular damage, including to the temporal inferior (TI) region of the donut. Further, we assume that damage progresses toward and into the temporal quadrant (red curved arrow). As damage progresses from the blue arcuate to the green arcuate to the yellow arcuate regions, damage in the central GCL donut progresses from TI to inferior (I) to nasal inferior (NI), as shown in the lower panel.

To predict the pattern of damage to the GCL donut, we assume that glaucomatous damage originates at the optic disc and appears initially in the IVZ and/or SVZ.^48–50 As the disease progresses, the damage deepens and widens, and the degree of donut involvement increases as damage extends into the temporal quadrant of the disc, as indicated by the dark red curved arrows in the middle panels.^{10,12,28,31,33,35,37,47,51,52} Note that we will not discuss damage that progresses toward the nasal quadrant,^37,53 as that does not affect the macula. The blue, green, and yellow arcuates in the middle row of Figure 2 represent progressive glaucomatous damage of the macula (thin black circle) over time, starting with blue and progressing to yellow. Note that the shape of these arcuate patterns was based upon the tracing of GCL axon bundles in Jansonius et al.,^54,55 as illustrated in figure 6 of Hood 2017.¹² 

Consider the superior retina first (see Fig. 2A, middle). In the earliest stages of glaucoma, the region within the red borders is typically affected first, and the superior macula is relatively unaffected.^{12,41,42,51,52} With further progression, as indicated by the curved dark red arrow, the damage at the disc enters the superior portion of the temporal quadrant and the blue arcuate region, then, over time, the green arcuate regions and finally the yellow arcuate regions are affected. This results in damage first to the temporal superior (TS) sector of the donut as shown by the blue arcuate, then to the TS + S (superior) sectors via the blue and green arcuates, and finally to all three donut sectors via the blue, green, and yellow arcuates. This is schematized by the red curved arrow in the bottom panel of Figure 2A. 

Now consider the inferior region (see Fig. 2B, middle). In most eyes with early damage to the IVZ, the temporal inferior (TI) and inferior (I) regions of the donut are already affected due to damage to the temporal region of IVZ,^{9,10,12,35,38,51–53,56} which has been called the macular vulnerability zone (MVZ).^8,9 With further progression, the damage at the disc enters the inferior portion of the temporal quadrant. This results in further damage via the yellow arcuate that now includes the nasal inferior (NI) sector. This is schematized by the red curved arrow in the bottom panel of Figure 2B. This pattern of progression in the inferior macula is illustrated in figures 3 and 4 of Shin et al.³¹ and figures 2 and 3 of Kim et al.,²⁸ whereas the anatomic relationship between damage in the inferior donut and the MVZ and IVZ can be seen in figure 3 of Lee et al.³⁵ 

Figure 3.

 

(A) GCL thickness map of an eye with early glaucoma. The donut shaped region is divided into six sectors: temporal superior (TS), superior (S), nasal superior (NS), nasal inferior (NI), inferior (I), and temporal inferior (TI). (B) A pie chart for the GCL thickness map in panel B showing the GCL thickness (um), the percentile based upon an age-matched group of healthy controls, and a color code of significance level (red: ≤1%, yellow: >1%, ≤5%, and green: >5%), for each sector. (C) Top: The first row of the table indicates the number of eyes within each superior donut sector with age-matched percentiles >5% (green cells) or ≤5% (orange cells). Middle: The prediction from the model in Figure 2. Lower: The same as the top table, but for the 54 healthy eyes. (D) Same as C for the inferior macula.

Figure 4.

 

One of the 10 eyes with early damage that is progressing that illustrates the agreement with the model. (A) The pie charts (left) and the GCL thickness maps (middle) are shown for time 1 and time 2 (31 months later). The schematic on the right illustrates the agreement with the model. (B) The thickness change map generated by subtracting the GCL thickness values at time 1 from those at time 2. Green and white indicate a decrease in thickness over time.

To summarize, we assume that initial glaucomatous damage includes the temporal half of the superior and/or inferior quadrants of the disc, the SVZ and IVZ. With time, the damage expands into the temporal quadrant of the disc. Based on the anatomic paths of the axons between the ganglion cell bodies in the macula and the optic disc, the model predicts a progression of thinning from temporal to nasal sectors of the central donut of ganglion cells as indicated in Figure 2. 

Based upon the model, we expect that with glaucoma, in both the superior and inferior half of the donut, the temporal sectors should have the highest rate of abnormal thinning, followed by the inferior/superior and last the nasal sectors. In addition, on average, the rate of abnormal thinning should be higher for the inferior sectors, compared to the corresponding superior sectors. 

Cross-Sectional Data and the Model

Figure 3 presents an application of the model to cross sectional data. The cross-sectional group consisted of 62 glaucoma or suspect eyes and 54 healthy eyes. Details can be found in the Supplementary Material and in previous publications.^36,57 The GCL thickness maps of all 116 eyes are shown in Supplementary Figures S1, S2, and the Supplementary Material contains details of the OCT protocols. The patient group included 32 eyes with a 24-2 MD better than -6 dB and 30 with MD worse than -6 dB. OCT cube scans centered on the fovea were obtained from all eyes as previously described.⁵⁷ The commercial software divides the donut shaped region of thick GCL around the fovea into 6 sectors, as seen in Figure 3A where it is superimposed on the GCL thickness maps of an eye with glaucoma. Figure 3B is a pie chart showing the GCL thickness (um) and percentile for each of the 6 sectors of the donut, the TS, S, nasal superior (NS), nasal inferior (NI), I, and TI. Age adjusted percentile levels are provided by the commercial software and color-coded such that green corresponds to thickness greater than fifth percentile, yellow corresponds to between the first and fifth percentiles, and red corresponds to values under the first percentile. 

The table in Figure 3C (upper panel) indicates, for each of the 3 superior sectors, the number of the 62 glaucomatous eyes with GCL thickness greater than the 5th percentile (see green in Fig. 3B), as well as the number with GCL thickness less than the 5th percentile (see yellow or red in Fig. 3B). The table in Figure 3D (upper panel) contains the same data for the 3 inferior donut sectors. As predicted by the model, the number of TS (TI) sectors below the 5th percentile (red or yellow) was greater than for sector S (I), which was greater than for NS (NI). Further, as discussed below, TI had the greatest number, 51 (82%) of the 62 eyes with GCL thickness below the 5th percentile, whereas NS had the fewest, 31 (50%). 

For the 54 healthy eyes (see Fig. 3, bottom tables), there was little difference in the number of yellow or red sectors across sections, and the small differences were not consistent with the model, suggesting intersubject variability follows a random pattern instead. 

Longitudinal Data

In addition, to evaluate the model, we compared the first and last scans of 10 eyes with early glaucoma (EG) at baseline, which were previously labeled as “definite progressors” (DPs).⁵⁸ These DP eyes showed clear evidence of progression on both their OCT and the VF measurements (24-2 and/or 10-2). In all 10 DP eyes, there was progression in the inferior half of the GCL donut consistent with the model. Figure 4 provides an example. The TI and I sectors in the pie charts (left column) changed as predicted from time 1 to time 2, which was about 31 months later. At time 1, the thinning was largely in the TI sector and this sector is yellow in the pie chart, whereas at time 2 that sector had become red on the pie chart and the I sector yellow. It is easier to see the actual pattern of change with the “change map” in the bottom panel.⁵⁸ This map is obtained by subtracting the GCL thickness map for time 1 from the GCL thickness map for time 2. In this pseudo-color map, white and green indicate thinning while red indicates thickening, which is typically due to measurement error. For the eye in Figure 4, in the interval between time 1 and time 2, both the TI and I sectors showed thinning on the change map. 

In 6 of the 10 DP eyes, the change over time was largely restricted to the inferior half of the donut, as seen on the change map in Figure 4B (red arrow), and in the bottom panels of Supplementary Figure S3A and S3B for all 6 eyes. In the remaining four eyes, there was progression in the superior donut sectors, as well as in the inferior sectors in the change maps (see Supplementary Fig. S4, lower row). All these changes in the 10 EG eyes were consistent with the model. 

Use as a Null Model for the Clinic

To help the clinician, the model can be considered a “null model” or “nominal/ideal model,” where a failure of the model suggests a failure of one or more of the assumptions, thus suggesting a non-glaucomatous pattern. Figure 5 contains 3 examples with patterns inconsistent with (i.e. rejecting) the model. In such cases, the clinician is advised to check the OCT b-scans, especially through the fovea (see the black lines in the left panels of Fig. 5), as well as 10-2 visual fields and fundus photographs/examinations, to rule out non-glaucomatous causes for this abnormality. 

Figure 5.

 

Three examples of eyes with patterns inconsistent with the model of glaucoma. (A) The GCL donut has a preserved region in the inferior temporal segment (black arrow, left panel) and clear thinning of the interior (I) and nasal interior (NI) sectors (red arrow, left panel). The b-scan (right panel) along the black line in the left panel showed clear damage to the receptors (red arrow). (B) The GCL donut (left panel) is distorted due to an epiretinal membrane (ERM), which can be seen on the b-scan in the right panel. (C) The GCL donut from one of the early glaucoma eyes in Supplementary Figure S2. There is a greater thinning in the nasal sectors of the donut (red arrows), than in the temporal sectors (black arrows). The horizontal b-scan (right panel) showed a thinning of the GCL and RNFL consistent with glaucoma.

The GCL donut in the first example (see Fig. 5A), has a preserved region in the TI segment (black arrow, left panel) and clear thinning of the I and NI sectors (red arrow, left panel). Whereas the 10-2 (not shown) confirmed the damage, the b-scan (see Fig. 5A, right panel) along the black line in Figure 5A (left panel) showed clear damage to the layer of photoreceptors (see the red arrow, right panel of Fig. 5A). 

The second example showed a distorted GCL donut (see Fig. 5B, left panel). The most likely cause in such cases is poor segmentation due to an epiretinal membrane (ERM). The b-scan in Figure 5B (right panel) shows an ERM (red arrow) that is clearly distorting the retina. 

The final example illustrates that glaucomatous damage can, but rarely does, deviate from the patterns predicted by the model. Figure 5C (left panel) shows a GCL thickness map from one of the EG eyes in Supplementary Figure S2. It has a greater thinning in the nasal sectors of the donut (red arrows), than in the temporal sectors (black arrows). This is a clear failure of the model. However, the 10-2 (not shown) had a subtle defect in the corresponding region, and the horizontal b- scan (Fig. 5C, right panel) although the fovea showed a thinning of the GCL and RNFL (red arrow in right panel), consistent with glaucoma. 

Implications for Clinical Trials Involving Progression

The model also provides suggestions for ways to improve the detection of progression. The most common method today for measuring GCL progression is based upon global (g-) GCL thickness, which is the average of the GCL thickness within the donut. However, there is evidence that metrics that focus on change in sectors of the GCL region have better sensitivity and specificity than g-GCL thickness.^{28,30,59–61} For example, based upon the model, it is not surprising that change in the GCL thickness of the TI region has a better sensitivity and specificity for detecting progression than does change in g-GCL thickness.^30,59,61 Further, we have data suggesting that, as expected from the model, the sector of the donut with the optimal sensitivity for detecting progression will depend upon the stage of GCL loss (Gomide, Tsamis, Sun, et al., “Assessing the progression of glaucoma in the central region of ganglion cells using OCT.” ARVO 2023). For example, in earlier stages of glaucoma, measures of the GCL thickness of the inferior (TI and I) sectors are more likely to detect progression, whereas in more advanced glaucoma, the GCL thickness of the nasal (NS and NI) sectors are more likely to perform better because the other sectors have reached a floor by then. Thus, it may be possible to improve the sensitivity to progression by identifying a donut sector as a region of interest to be followed based upon a baseline OCT. 

Model Limitations and Heuristic for Future Work

The model is not meant to be a complete model of progression. It is meant to be a simplified model with clear assumptions for the purposes described above. A complete model would need to take into consideration that glaucomatous damage can be diffuse, as well as local, and that progression is not simply an expansion and deepening of existing defects but can involve local variations of existing defects.^{10,28,31,34,37} On the other hand, the model does need to be tested with a larger, real-world dataset, which may also include eyes with other diseases and glaucoma. Except for the examples in Figures 5A and 5B, the data presented here were from studies in which the patients were screened to exclude other pathologies, including high myopia. The model needs testing for eyes with high myopia. In general, real-world data from the clinic will show a richer variety of GCL donut patterns, and the diagnostic value of these patterns remains to be determined in this population. Finally, all the data in this study came from one commercial OCT instrument. The GCL thickness maps of OCT manufacturers differ in dimensions of the central macular region, in the way in which that region is subdivided, the scan protocols used, and whether they measure the GCL or GCL plus inner plexiform layer. Although these details may vary, we predict the model will generalize to other instruments as the underlying anatomy will dominate over variations in scanning. 

Summary

The model, as it stands, has two major virtues for clinical work. First, it focuses on the central macular region. There is now considerable evidence that macular damage occurs early in the glaucomatous process.^8–24 However, many glaucoma specialists^62–65 still depend largely upon OCT disc scans analyzed with circumpapillary retinal nerve fiber layer (cpRNFL) quadrant pie charts to define glaucoma. This approach has poor sensitivity and, most importantly, misses eyes with clear macular damage.^{24,27,66–68} On the other hand, the model of the central macular GCL donut puts the focus on the most important region to preserve, the central donut of ganglion cells. Further, in its current form, it should prove useful as a null model in the clinic. In particular, a failure of the model suggests a failure of one or more of the assumptions and alerts the clinician to examine other OCT, fundus/disc photographs, and visual field information for the cause, which is likely to be non-glaucomatous. 

Acknowledgments

Supported in part by the Jane and David Walentas Glaucoma Research Fund, Columbia University Department of Ophthalmology; an unrestricted grant to the Department of Ophthalmology from Research to Prevent Blindness, Inc., New York, NY, USA; and National Institutes of Health (Bethesda, MD, USA) Grants EY-02115 (D.C.H.), EY-032182 (E.T.) and EY-025253 (C.G.D.M.). 

Disclosure: D.C. Hood, Topcon, Inc. (F, C), Heidelberg Engineering (F, C), Novartis, Inc. (C); S. La Bruna, None; A. Leshno, None; G.A. Gomide, None; M.J. Kim, None; G.A. Cioffi, None; J.M. Liebmann, None; C.G. De Moraes, Carl Zeiss Meditec, Inc. (C, R); Topcon, Inc. (C); Heidelberg Engineering (C, R); Novartis, Inc. (C); Thea Phar Inc. (C); Perfuse Therapeutics, Inc. (C); Reichert, Inc. (C, R); Ora Clnical, Inc. (E); E. Tsamis, Topcon, Inc. (R)