**Purpose**:
To test the hypotheses that: (1) structure–function (SF) relationships between visual fields (VF) and Bruch's membrane opening-based minimum rim width (BMO-MRW) measurements are superior to those for peripapillary retinal nerve fiber layer (pRNFL) in perimetric glaucoma; (2) BMO-MRW measurements may extend the utility of structural measurement across the range of glaucoma severity; and (3) to estimate the influence of Bruch's membrane opening (BMO) size on BMO-MRW measurements.

**Methods**:
One hundred eight perimetric glaucoma eyes (68 patients) with good quality spectral-domain optical coherence tomography images of the optic disc and pRNFL, and reliable VF within 6 months were recruited. Relationship of global and sectoral BMO-MRW and pRNFL thickness with corresponding VF parameters and the influence of normalizing BMO-MRW (on BMO circumference, nBMO-MRW) on SF relationships were investigated. Broken-stick models were used to compare the point at which pRNFL and BMO-MRW parameters reached their measurement floor.

**Results**:
The median (interquartile range) of VF mean deviation was −5.9 (−12.6 to −3.6) dB. Spearman correlation coefficients between pRNFL, BMO-MRW, and nBMO-MRW measures and corresponding VF cluster average deviations ranged between 0.55 to 0.80, 0.35 to 0.66, and 0.38 to 0.65, respectively. Bruch's membrane opening–MRW parameters demonstrated weaker SF relationships compared with pRNFL globally and in temporal, temporal-superior, and nasal-inferior sectors (*P* < 0.03). Normalization of BMO-MRW did not significantly influence SF relationships.

**Conclusions**:
Structure–function relationships were somewhat weaker with BMO-MRW parameters compared with pRNFL in eyes with perimetric glaucoma. Bruch's membrane opening–MRW normalization did not significantly change SF relationships in this group of eyes with mild departures from average BMO size.

^{1}The BMO-MRW is defined as the minimum distance between the internal limiting membrane (ILM) on the surface of ganglion cell axons as they enter the optic disc and the termination of Bruch membrane, known as the BMO.

^{2}This parameter is independent of any arbitrary reference planes and is geometrically more precise than ‘cup' or ‘rim' measurement based on an operator-defined ‘contour line' around the visible disc margin.

^{2,3}There is controversy whether BMO-MRW is superior to pRNFL for detection of early glaucoma or with regard to structure–function (SF) relationships.

^{1,2,4}One caveat is that the BMO-MRW measurements can vary as a function of BMO size: eyes with a smaller disc size would be expected to have thicker BMO-MRW measurements compared with eyes with a larger disc size given the same number of ganglion cell axons.

^{5}The magnitude of this confounding issue has not been well explored, although a new parameter, minimum rim area (i.e., the area through which the axons would pass to enter the optic nerve) has been proposed to address this issue.

^{6}

^{7}This residual thickness (at measurement floor) is attributed to nonneural tissues, such as blood vessels and glial cells.

^{7}However, clinically, neuroretinal rim loss is frequently observed to be total with no neural or nonneural tissue remaining where notching or rim loss extends to what clinically is known to be the scleral rim as seen on examination or stereoscopic disc photographs. This raises the question of whether the BMO-MRW measurements could reach a lower floor of measurement later than pRNFL during the course of the disease, and hence, BMO-MRW measurements might be more useful in more advanced stages of the disease. Another important factor to consider in this context is the dynamic range of any given parameter (i.e., not only is it important when the floor is reached, but one has also to consider the rate of change of the parameter down to the final floor).

*P*less than 5% on the pattern deviation plot, both confirmed at least once.

^{8}

^{9,10}The two points immediately above and below the physiologic blind spot were excluded from the analysis and the remaining 52 points were used to calculate the average deviation (AD) in each sector based on the total deviation numerical plot. To calculate the AD in each sector, defect depth in decibel units was first converted to a linear scale (1/Lambert) for each test point and then averaged and converted back to the logarithmic scale in decibels.

^{11–13}

^{14}in a normal database in which the average BMO circumference was calculated to be 4.8 mm. Second, we stratified the sample into small, average, and large discs based on the BMO area. As there are no widely accepted criteria for classifying the BMO area and because the BMO area had a normal distribution (

*P*= 0.733, Wilk Shapiro test), we used the mean ± 1 SD as the boundaries for average disc size in our sample. Eyes with a BMO area more than the mean + 1 SD were considered to have large discs and those with BMO area lower than the mean − 1 SD were considered to have a small disc.

_{BMO}pRNFL is the normalized pRNFL based on BMO.

^{15}we used the following formula to normalize pRNFL measurement based on axial length in phakic subjects:

_{AXL}pRNFL is the normalized pRNFL based on the patient's axial length (AXL) and 24.46 is the average AXL (in mm) assumed by the device database.

^{16}A bootstrap method (with >1000 iteration) was used to compare the strength of correlation coefficients in a pairwise manner. We used the Benjamini–Hochberg method to address multiple comparisons.

^{17,18}

^{19}Structure–function plots were constructed as bivariate plots with the VF sectoral AD on the X-axis and the corresponding BMO-MRW or pRNFL sectoral thickness (in μm) on the Y-axis. A broken stick model was fit to the SF data as follows:

^{20}

*P*> 0.05 for all the differences estimated with bootstrapping). Figure 2 summarizes the results.

*P*> 0.05 for all comparisons estimated with bootstrapping with Benjamini–Hochberg correction; Fig. 4).

*P*< 0.03, with Benjamini–Hochberg correction). The same results were found for corresponding pairs of pRNFL/VF coefficients and nBMO-MRW/VF (

*P*< 0.03 for global and temporal, TS and NI sectors; Fig. 4). Normalization of pRNFL based on axial length or BMO circumference did not improve the SF relationship (

*P*> 0.05 for all sectors and global measurements).

^{2}; therefore, we considered a disc to be small when the BMO area was less than 1.50 mm

^{2}(17 eyes) and large when BMO area was greater than 2.29 mm

^{2}(18 eyes). Seventy-three eyes were considered to have an average BMO area (1.50–2.29 mm

^{2}). In regression analyses with the VF MD or ADs as the dependent variable and the BMO-MRW, the BMO size (categorical variable) and their interaction as predictors, the BMO size category was not a significant predictor of VF MD or AD. The BMO area was inversely correlated with global BMO-MRW (Spearman's

*ρ*= −0.259,

*P*= 0.033). The scatter plot in Figure 5 shows changes in BMO-MRW and pRNFL as a function of BMO circumference. It can be observed that as the BMO area deviation from the average value increases, corrections to BMO-MRW and pRNFL become progressively larger.

*P*< 0.0001) after the change point for global measurements as well as sectoral ones. The change points on the broken stick SF models did not vary as a function of BMO area (data not shown). We compared the ratio of sectoral pRNFL, BMO-MRW, and nBMO-MRW thickness parameters divided by the corresponding average or global thickness. Significant differences were observed in the nasal and NI sectors between pRNFL and BMO-MRW parameters (lower ratio for pRNFL compared with BMO-MRW and nBMO-MRW;

*P*< 0.002 for all comparisons).

^{14}We also explored using the BMO circumference and axial length to normalize average pRNFL thickness. Analyses were repeated after data were stratified according to BMO area. We found that either approach did not significantly change SF relationship for BMO-MRW or pRNFL (

*P*> 0.05 for all comparisons). The change point for global BMO-MRW extended to the left of that for the average pRNFL thickness denoting that global BMO-MRW might be useful for detection of structural change after average pRNFL has reached its measurement floor; however, this needs to be confirmed in larger scale, longitudinal studies as the difference was not found to be statistically significant.

*ρ*= −0.26,

*P*= 0.03); this is in agreement with the findings by Tun et al.

^{21}in healthy subjects, and could be attributed to the fact that with larger BMO area, the retinal ganglion cell axons would be distributed along a larger circumference, and therefore would be relatively thinner. This is the rationale behind normalizing the BMO-MRW. To account for the influence of BMO area on SF relationships, we explored SF relationships after dividing BMO area into three groups or after normalizing BMO-MRW parameters based on BMO circumference. Neither of these methods consistently or significantly improved the correlation between various structural and functional parameters. This is in contrast to the findings of the study by Gmeiner and colleagues

^{5}who reported a better diagnostic performance of BMO-MRW in small discs with BMO area less than 1.84 mm

^{2}. This difference could be due to different characteristics of the two cohorts and disease severity, as all of their cases were preperimetric or early glaucoma cases.

^{5}We used the average BMO circumference from Patel et al.

^{14}to normalize BMO-MRW measurements.

^{5}and Gardiner and colleagues

^{6}reported similar performance whereas others did not.

^{2,4,22}This could be attributed to pRNFL thickness measurements being less influenced by the shape and anatomic variability of the optic disc along with more demanding segmentation of the BMO in addition to the factors mentioned above. However, as reported previously,

^{5}at various stages of the disease a subset of cases may be detected by one parameter and missed by the other and vice versa and therefore, the two parameters should be considered complementary.

^{5}

^{15}demonstrated that optic disc size did not influence pRNFL thickness and the reported effect could be explained by differences in axial length and magnification error. The Spectralis device software addresses the magnification error by incorporating corneal curvature and power into its calculations.

^{23,24}Moreover, the focusing mechanism of the device at least partially compensates for refractive error and reduces the magnification error.

^{25,26}We normalized pRNFL based on both BMO circumference and axial length and, as expected, there was no significant change in the results.

^{9,10}There are several important issues to consider. Different types of glaucoma could vary with regard to severity, location, and extent of damage and this could affect the presentation of VF defects. Garway-Heath and colleagues suggested that based on the current body of knowledge, a similar map could be applied to all types of glaucoma.

^{27}The second issue is that, as mentioned by these investigators, their proposed mapping is valid in average eyes and it would be suboptimal in eyes with markedly shorter or longer axial length or those with an anomalous disc.

^{10}Most of our cases had near average axial lengths. Spectralis SD-OCT measurements are aligned along the FoDi axis, while the VFs are aligned on the geometric axis; however, this was not a major confounding factor in a previous study.

^{3}Given the large size of the sectors, it is unlikely that the results of this study were significantly influenced. Finally, it is worth noting that the derivation of the Garway-Heath

^{9,10}map is based on pRNFL defects that were detected clinically, and therefore, locations of defect were identified with respect to the clinical disc margin. A more appropriate way to define sectors and spatial relationships would be as a function of the BMO. This could be the topic of an interesting study in future.

^{28}In addition, Bruch's membrane may be absent in the area of peripapillary atrophy or might have decreased reflectivity in glaucoma and automated algorithms might no longer detect the BMO accurately.

^{29–31}All OCT images were reviewed by two of the authors (KNM and FS) to ensure proper segmentation of ILM and BMO, and hence, accurate BMO-MRW measurements. Almobarak and colleagues

^{30}compared the measurement error with automatic versus manual segmentation of ILM and BMO, and reported a median (IQR) difference in BMO-MRW measurement of 12.1 (10.1, 16.8) and 13.4 (10.6, 16.8) μm in healthy and glaucomatous subjects, respectively; neither was statistically significant.

^{13,32}Only the global BMO-MRW was found to have a change point to the left of that for the average pRNFL. We compared the ratio of sectoral RNFL, BMO-MRW, and nBMO-MRW thickness divided by the corresponding average or global thickness measurements to determine whether a larger residual thickness in nasal sectors could potentially explain the different change points for global BMO-MRW (and nBMO-MRW) thickness compared with the average RNFL thickness. We found a significantly higher ratio for MRW parameters in the nasal and NI sectors compared with pRNFL (denoting more prominent pRNFL loss as compared with BMO-MRW,

*P*< 0.002 for all comparisons). These sectors are underrepresented in the Garway-Heath and colleagues' map as a smaller number of locations are measured in the corresponding VF clusters.

^{9,10}This finding needs to be confirmed in future studies before its utility can be fully realized in clinical practice. Normalization of BMO-MRW based on BMO area, did not significantly change the SF relationships. This could be attributed to the limited number of eyes with extreme BMO measurements in our study; it is also possible that larger discs could contain a higher proportion of nonneural tissues (blood vessels, glia, etc.) or that remodeling of neural tissues could slowly occur as glaucoma advances.

**N. Amini**, None;

**R. Daneshvar**, None;

**F. Sharifipour**, None;

**P. Romero**, None;

**S. Henry**, None;

**J. Caprioli**, None;

**K. Nouri-Mahdavi**, Heidelberg Engineering (F, R)

*. 2015; 56: 6886–6891.*

*Invest Ophthalmol Vis Sci**. 2013; 120: 535–543.*

*Ophthalmology**. 2014; 56: 98–105.*

*Invest Ophthalmol Vis Sci**. 2014; 55: 2953–2962.*

*Invest Ophthalmol Vis Sci**. 2016; 57: OCT575–OCT584.*

*Invest Ophthalmol Vis Sci**. 2014; 157: 540–549. e2.*

*Am J Ophthalmol**. 2007; 24: 1426–1430.*

*J Opt Soc Am A Opt Image Sci Vis**. 2002; 134: 177–185.*

*Am J Ophthalmol**. 2002; 43: 2213–2220.*

*Invest Ophthalmol Vis Sci**. 2000; 107: 1809–1815.*

*Ophthalmology**. 2002; 120: 1672–1681.*

*Arch Ophthalmol**. 2000; 100: 115–137.*

*Doc Ophthalmol**. 2007; 26: 688–710.*

*Prog Retin Eye Res**. 2014; 55: 6802–6816.*

*Invest Ophthalmol Vis Sci**. 2012; 53: 4990–4997.*

*Invest Ophthalmol Vis Sci**. 2011; 30: 3181–3191.*

*Stat Med**Companion for the Handbook of Biological Statistics. Version 1.3.0. 2013*. Available at: http://rcompanion.org/rcompanion/f_01.html. Accessed July 12, 2017.

*. 1990; 9: 811–818.*

*Stat Med**. 2012; 53: 2740–2748.*

*Invest Ophthalmol Vis Sci**. 2015; 56: 3337–3344.*

*Invest Ophthalmol Vis Sci**. 2015; 56: 3320–3328.*

*Invest Ophthalmol Vis Sci**. 2016; 57: 1468–1475.*

*Invest Ophthalmol Vis Sci**. 2015; 253: 733–738.*

*Graefes Arch Clin Exp Ophthalmol**. 2014; 55: 3439–3446.*

*Invest Ophthalmol Vis Sci**. 2008; 28: 21–28.*

*Ophthalmic Physiol Opt**. 2001; 108: 1714.*

*Ophthalmology**. 2017; 55: 87–94.*

*Comput Med Imaging Graph**. 2016; 168: 237–249.*

*Am J Ophthalmol**. 2014; 55: 1161–1168.*

*Invest Ophthalmol Vis Sci**. 2016; 25: 873–878.*

*J Glaucoma**. 2016; 57: 4815–4823.*

*Invest Ophthalmol Vis Sci*