Tomographic Features of Macula After Successful Macular Hole Surgery

Abstract

Purpose.: We determined the retinal features associated with a dissociated optic nerve fiber layer (DONFL) appearance after successful macular hole surgery with internal limiting membrane (ILM) peeling.

Methods.: A tota of 47 eyes of 47 patients who underwent pars plana vitrectomy with ILM peeling was examined by spectral-domain optical coherence tomography (SD-OCT).

Results.: A DONFL appearance was found in 31 (66.0%) eyes on color fundus photography. The postoperative best-corrected visual acuity and its improvements were not significantly different between eyes with and without this appearance. The SD-OCT images showed small inner retinal defects corresponding to each arcuate stria in all eyes with the DONFL appearance. Of these eyes, 25 (80.6%) included inner retinal defects extending beyond the retinal nerve fiber layer (RNFL) into the ganglion cell layer and inner plexiform layer. Similar inner retinal defects were seen in 14 (87.5%) of the 16 eyes without a DONFL appearance; 10 eyes (62.5%) included the defects that extended deeper than the RNFL. Such deep defects were found more frequently in the temporal than in the other quadrants (P < 0.0001). In 10 eyes that underwent Micro Perimeter-1, a reduction in retinal sensitivity (<10 dB) was found in 20 (10.4%) of the 192 points on arcuate striae, but only in 7 (3.4%) of 208 points away from the striae (P = 0.005).

Conclusions.: Macular hole surgery with ILM peeling may lead to deep inner retinal damage regardless of the development of the DONFL appearance detected on color fundus photography.

Introduction

In 1991, Kelly and Wendel reported that pars plana vitrectomy (PPV) followed by gas tamponade had a success rate of 58% for idiopathic macular hole closure. ¹ Since then, PPV has become a common surgical procedure for the treatment of idiopathic macular holes. Internal limiting membrane (ILM) peeling during the vitrectomy was shown to improve the closure rate and lower the reopening rate significantly. ^2–8 Thus, ILM peeling now is a routine surgical procedure during macular hole surgery. 

After successful macular hole surgery with ILM peeling, some of the retinas were noted to have arcuate striae that were slightly darker than the surrounding retina on biomicroscopy and color or red-free fundus photography. The striae were located in the area of ILM peeling, and they ran in the direction of the optic nerve fibers. ^9–13 This feature was termed a “dissociated optic nerve fiber layer” (abbreviated as “DONFL” in earlier studies) appearance by Tadayoni, et al. ⁹ The incidence of DONFL appearance ranged from 42.6% to 62.2% on fundus photography in eyes that underwent ILM peeling, and usually appears 1 to 3 months after surgery. ^9–12  

The development of the DONFL appearance has been attributed to the ILM peeling. ^11,12,14 Earlier investigators suggested that the DONFL appearance was not due to injury to the retinal nerve fibers, but to cleavages of the retinal nerve fiber bundles due to damage of the Müller cells. This suggestion was made because the time-domain optical coherence tomographic (TD-OCT) images did not show any changes deeper than the retinal nerve fiber layer (RNFL). In addition, no evidence was found that indicated significant reduction of the visual acuity and retinal sensitivity determined by static perimetry or microperimetry. ^9–13 However, it also has been reported that removal of the ILM causes changes in the histology of the retina, ¹⁵ delay in the recovery of the b-wave of the focal macular electroretinograms, ¹⁶ and paracentral scotomata on microperimetry. ¹⁷ Thus, it remains to be determined whether the DONFL appearance and ILM peeling are associated with damage to the retina that can alter the function of the retina. 

It has been reported that the changes in macular retinal thickness after successful macular hole surgery with ILM peeling are different for the four macular quadrants; the temporal quadrant showed the most severe thinning followed by the superior or inferior quadrants, whereas the nasal quadrant had a thicknening, ^18–20 regardless of the presence or absence of the DONFL appearance. ¹⁹ Despite the different responses in the macular quadrants, the regional differences of the development of arcuate striae and severity of the retinal changes associated with the arcuate striae have not been reported fully to our knowledge. 

Spectral-domain OCT (SD-OCT) has better resolution due to its ability to reduce speckle noise, ^21,22 which is the most important artifact that can blur the boundaries of the retinal layers. ^23,24 We have found that examining eyes that had undergone macular hole surgery by SD-OCT frequently showed defects that extended deep into the inner retina, particularly in the temporal macular area. Because most of the earlier studies showed only TD-OCT images of vertical scans through the fovea, ^9–13 it was not certain whether the DONFL appearance was limited to the RNFL. 

Thus, the purpose of our study was to determine whether ILM peeling and the presence of the DONFL appearance had any adverse effects on the postoperative retinal morphology and visual function after successful macular hole surgery, and to determine regional differences of these effects. To accomplish this, we examined the images of the retinas obtained by SD-OCT, and tested the retinal sensitivity associated with and without the DONFL appearance. 

Methods

Subjects

The medical records of patients who had undergone macular hole surgery at the Kyoto University Hospital between January 2007 and October 2010 were reviewed. The inclusion criteria were: presence of an idiopathic macular hole, successful anatomical closure of the macular hole after 23-gauge transconjunctival three port PPV with ILM peeling, and a follow-up period of more than 6 months from the last macular hole surgery. Eyes were excluded if they had surgical complications, such as peripheral retinal tear and retinal detachment, during and after the primary surgery. Other exclusion criteria included evidence of other vitreoretinal and macular diseases, glaucoma, uveitis, high myopia (>−6 diopters [D]), history of ocular trauma, diabetes mellitus, or a history of vitreoretinal surgery. 

Preoperatively, all patients had a comprehensive ophthalmologic examination including measurements of the refractive error (ARK-700A autorefractor; Nidek, Gamagori, Japan), the uncorrected and best-corrected VA (BCVA) using a Landolt chart at 5 meters, the axial length using A-scan ultrasonography (UD-6000; Tomey Corporation, Nagoya, Japan), and the intraocular pressure with a Goldmann applanation tonometer. In addition, the anterior and posterior segments of the eyes were examined by slit-lamp biomicroscopy. 

All investigations of this study adhered to the tenets of the Declaration of Helsinki. Retrospective review of the patient data in this study was approved by the Institutional Review Board and Ethics Committee of the Kyoto University Graduate School of Medicine. All patients gave informed consent for the procedures performed. 

Surgical Procedures

All eyes underwent 23-gauge transconjunctival PPV. Triamcinolone acetonide (TA) was used intraoperatively in all of the eyes to make the vitreous and posterior hyaloid membrane more visible. The preservative in the diluent was removed from the TA suspension (Kenacort-A, 40 mg/mL; Bristol Pharmaceuticals, Tokyo, Japan) with a filter and rinsed with balanced salt solution (BSS). The TA particles (40 mg) then were resuspended in 2 mL of BSS. After core vitrectomy, 0.1 mL of the diluted TA suspension was injected onto the posterior retina to enhance visualization of the posterior hyaloid membrane and residual vitreous. A posterior vitreous detachment was created by aspiration with a vitreous cutter in eyes that did not have a posterior vitreous detachment. The detached vitreous and peripheral vitreous were excised and removed. ILM peeling was performed after staining the ILM with indocyanine green (ICG) or TA. To prepare the ICG solution, 25 mg ICG powder (Ophthagreen; Santen, Osaka, Japan) was dissolved in 1 mL distilled water, and 9.0 mL BSS was added to obtain a 0.25% concentration. To make the ILM more visible, a small amount of the ICG solution or the resuspended TA solution was sprayed gently onto the macular surface in a BSS-filled eye, and immediately aspirated with active suction by the vitreous cutter. Particles of TA were observed as white specks on the posterior retinal surface. 

The ILM peeling was initiated by grasping the ILM with an ILM forceps at the superior or inferior macular region, but not the temporal region, and then extended in a circumferential manner without touching the retinal surface over 2 to 3 disc diameters around the fovea. Fluid-air exchange was performed, followed by gas tamponade with 40 mL of 25% sulphur hexafluoride or 0.5 to 1.0 mm of 100% sulphur hexafluoride. The patients were instructed to keep their head prone for at least 1 week after the surgery. Simultaneous phacoemulsification with intraocular lens implantation was performed on all phakic eyes. 

Spectral-Domain Optical Coherence Tomography

SD-OCT examinations were performed at the initial examination and during the postoperative period by experienced ophthalmologists using mainly the Spectralis HRA+OCT (Heidelberg Engineering, Heidelberg, Germany). Postoperative SD-OCT examinations were performed at 1, 3, 6, and 12 months after surgery to evaluate the macular hole closure and postoperative retinal structural changes, such as improvement of foveal morphology and development of DONFL. The SD-OCT images taken at 6 months postoperatively were used to evaluate the postoperative retinal structural changes. 

Up to 100 single B-scan images from the same location can be averaged by the real time (automatic real-time [ART] module) software embedded in this instrument. This reduced the speckle noise, which markedly improved the visibility of the boundaries of the retinal layers and the morphologies of the various pathologic lesions. ²⁵ Our routine SD-OCT examination included 24 radial macular scans of 6 or 9 mm lengths, and 6 or 9 mm horizontal and vertical serial scans centered on the fovea for the entire macular area. In each scan, 50 B-scans were averaged to reduce the speckle-noise. 

Distribution of DONFL Appearance

To determine the most common site for the DONFL appearances, we divided the macular area on color fundus photography by two 45-degree oblique lines (Fig. 1, blue lines) that ran through the center of the fovea. The lines divided the retina into the superior, inferior, temporal, and nasal quadrants. The number of eyes that showed a DONFL appearance for each quadrant of the macula in the color fundus photographs taken at 3 to 6 months after surgery was counted. If a single arcuate stria was located on the border of 2 quadrants, it was counted only in the quadrant where a greater portion of the striae was located. The number of eyes that had defects extending over the RNFL also was counted on the 24 radial macular SD-OCT scans centered at the fovea. Each quadrant included 12 half-scans with an interval of 7.5° between neighboring scans. The scan lines were recorded, and were able to be seen on the fundus image of the confocal scanning laser ophthalmoscopy in the Spectralis system. These scan lines were superimposed on the color fundus photography by referring to the positions of retinal vessels, the fovea, and the optic disc, and then the inner retina defects that extended deeper than the RNFL corresponding to each arcuate stria were identified. All of the arcuate striae were included in the radial scans. 

Figure 1

 

Comparisons of the cross-sectional images of the retina obtained 6 months after macular hole surgery with ILM peeling by SD-OCT and TD-OCT. A 54-year-old woman presented with a stage 4 macular hole in her left eye, and the visual acuity was 20/50 preoperatively and 6 months after surgery (Case 2). ILM peeling was performed after staining with indocyanine green. (A) Color fundus photograph, (B, C) Horizontal B-scan images through the central fovea obtained with SD-OCT (B) and TD-OCT (C). (D, E) Vertical B-scan images through the central fovea obtained with SD-OCT (D) and TD-OCT (E). Horizontal and vertical green lines in (A) indicate the scan lines for the horizontal (B) and vertical (D) SD-OCT images. Horizontal and vertical orange lines in (A) indicate the scan lines for horizontal (C) and vertical (F) TD-OCT images. Blue crossed lines indicate the borders of the 4 quadrants used to count the number of arcuate striae. Scan lengths for SD-OCT and TD-OCT were 9 and 6 mm, respectively. In the SD-OCT images, blue arrowheads point to the inner retinal defects limited to the RNFL, and red arrowheads to the inner retinal defects that extend beyond the RNFL into the GCL and IPL. In the TD-OCT images, the blue and red arrowheads indicate the points corresponding to the points that are pointed to by the arrowheads in the SD-OCT images. The red arrowhead in the temporal macula (B) points to an area with a complete loss of the RNFL and GCL. The depth of this damage is not evident in the TD-OCT image (C).

Macular Area Microperimeter-1 Testing

The retinal sensitivity was determined in the 10 eyes with a DONFL appearance that had the results of Microperimeter-1 (NIDEK, Vigonza, Italy) using the same testing protocol before, and 3 to 6 months after surgery. The MP1 software automatically tracks fundus movements by evaluating every acquired frame for shifts in the x- and y-directions of the fundus with respect to a reference frame obtained by an infrared camera at the beginning of the examination. A 4-2–staircase strategy with a Goldmann III size stimulus was used for the Microperimeter-1 examination of 57 stimulus locations covering the central 10° as described. ²⁵ Each stimulus was matched to a location on a grid that covered 0.71° in the foveal and parafoveal regions, and 1.4° in the perifoveal region. The white background illumination was set at 1.27 cd/m². The difference between the stimulus and background luminances was 127 cd/m² at 0 dB, and the maximum stimulus attenuation was 20 dB. The duration of the stimulus was 200 ms. 

The most central 5 points and circumferential 12 points were excluded in the calculation of the mean retinal sensitivity, and to generate histograms for the number of eyes with each sensitivity value. A test point was taken to be on the arcuate striae when the center of the test point lay within the arcuate striae. 

We determined the locations for Microperimeter-1 test points on the SD-OCT images by comparing the vascular shadows in the SD-OCT image with vessels in the color photographic images from the Microperimeter-1. 

Statistical Analyses

Values are shown as the means ± SDs. χ² tests were used to determine whether the incidences of DONFL appearance, and inner retinal defects that extended beyond the RNFL were significantly different among the 4 quadrants, and whether the number of points with retinal sensitivity of <10 dB were statistically different on and away from the arcuate striae. Fisher's exact test was used to compare incidences of these features between 2 quadrants. The Mann-Whitney U test was used to compare numeric values between eyes with and without these features, and to compare mean retinal sensitivity determined by Microperimeter-1 between points on and outside the arcuate striae. A value of P < 0.05 was considered to be significant difference. Statistical analyses were performed using SPSS software version 17.0 (SPSS, Inc., Chicago, IL). 

Results

Subjects

We studied 47 eyes of 47 patients (women 34, men 13) with a macular hole. The clinical characteristics are shown in Table 1. Their ages ranged from 51 to 83 years, with a mean of 67.2 ± 7.9 years. The preoperative BCVA ranged from 20/32 to 20/600 (median 20/63) in the Snellen equivalent. The BCVA at 6 months after surgery ranged from 20/12.5 to 20/200 (median, 20/32). To make the ILM more visible, ICG was used in 30 (63.8%) eyes and TA in 17 (36.2%) eyes. 

Table 1. 

 

Distribution of Total DONFL Appearances (n = 31) and Those in Which Retinal Defects Were Seen to Extend Beyond RNFL After Macular Hole Surgery With ILM Peeling

	Total,* n = 31	With Retinal Defects Beyond RNFL†
Temporal	18 (58.1%)‡	18 (100.0%)§
Superior	30 (96.8%)	16 (53.3%)
Inferior	25 (80.6%)	6 (24.0%)
Nasal	24 (77.4%)	3 (12.5%)

 The number of eyes with DONFL appearance in each quadrant is shown with percentage in parentheses.

*    P = 0.003 among 4 quadrants (χ² test).

†    P < 0.0001 among 4 quadrants (χ² test).

‡    P < 0.0001 compared to the superior quadrant (Fisher's exact test).

§    P < 0.0001 compared to the superior, inferior, and nasal quadrants (Fisher's exact test).

Incidence of DONFL Appearance

A DONFL appearance was found on color fundus photography in 31 of the 47 eyes (66.0%) after macular hole surgery. The differences in the postoperative BCVA (P = 0.765), and improvement in the BCVA after the macular hole surgery (P = 0.329) between eyes with and without a DONFL appearance were not significant. The DONFL appearance was found in 20 (66.7%) of 30 eyes in which IGC was used, and in 11 (64.7%) of 17 eyes in which TA was used. No significant differences were found in the age (P = 0.528), sex (P = 0.318), use of ICG or TA (P = 1.0), preoperative BCVA (P = 0.819), or postoperative BCVA at 6 months (P = 0.744) between eyes with and without a DONFL appearance. 

Tomographic Features of DONFL Appearance

Depressions of the inner retina were seen in the SD-OCT images over the arcuate striae (Figs. 1–3). The depth of the retinal depressions varied; some were limited to the RNFL layer (Figs. 1, 2, 4), and others extended beyond the RNFL into the ganglion cell layer (GCL) and inner plexiform layer (IPL), especially in the temporal macular region (Figs. 1–4). The depth of the inner retinal defects varied with location even within the same arcuate striae (Fig. 2). Examination of the eyes with TD-OCT showed that some of the defects were seen as disruptions of the RNFL and some were not seen (Fig. 1). Even if the defects were detected in the TD-OCT images, it was difficult to determine which layers were involved in the defects. 

Figure 2

 

Varying depths of the inner retinal defects seen in an arcuate striae 6 months after macular hole surgery with ILM peeling. A 66-year-old woman presented with a stage 3 macular hole in her right eye; the visual acuity was 20/63 preoperatively and 20/20 at 6 months after surgery (Case 11). ILM peeling was performed after staining with indocyanine green. (A) Color fundus photograph. Blue, red, and green lines indicate the scan lines for the SD-OCT images in (B), (D), and (F), respectively. (C, E, G) Magnified (×4) views of the area outlined by the red dashed lines in (B), (D), and (F), respectively. Red arrows in the SD-OCT images point to the inner retinal defect associated with the arcuate striate pointed to by the red arrowhead in (A). Blue arrowheads point to the other inner retinal defects visible in the SD-OCT images. Note that the depth of the inner retinal defect differs with the location within the arcuate striate. INL, inner nuclear layer.

Figure 3

 

Two cases showing the most severe inner retinal defects within the arcuate striae with DONFL appearances after macular hole surgery with ILM peeling. (A–C) A 64-year-old woman presented with a stage 2 macular hole in her right eye; the visual acuity was 20/63 preoperatively and 20/16 at 6 months after surgery (Case 14). ILM peeling was performed after staining with triamcinolone acetonide. (D–F) A 74-year-old woman presented with a stage 3 macular hole in her right eye. The visual acuity was 20/125 preoperatively and 20/63 at 6 months after surgery (Case 18). ILM peeling was performed after staining with indocyanine green. (A, D) Color fundus photographs. Green lines indicate the scan lines for the SD-OCT images in (B) and (E), respectively. The red arrowheads point to the arcuate striae of a DONFL appearance. (C, F) Magnified (×4) views of the area outlined by red dashed lines in (B) and (E). In the SD-OCT images, the red arrowheads point to the inner retinal defects that correspond to the arcuate striae pointed to by the red arrowheads in the color fundus photographs in (A) and (D). The defects extend beyond the RNFL into the GCL and IPL.

Figure 4

 

Inner retinal defects seen in 2 eyes without DONFL appearances after macular hole surgery with ILM peeling. (A–C) A 70-year-old woman presented with a stage 2 macular hole in her left eye; the visual acuity was 20/50 preoperatively and 20/20 at 6 months after surgery (Case 35). ILM peeling was performed after staining with triamcinolone acetonide. (D–F) A 67-year-old woman presented with a stage 3 macular hole in her right eye; the visual acuity was 20/63 preoperatively and 20/63 at 6 months after surgery (Case 37). ILM peeling was performed after staining with indocyanine green. (A, D) Color fundus photographs. Green lines indicate the scan lines for the SD-OCT images in (B) and (E), respectively. There are no apparent arcuate striae of the DONFL appearance. (C, F) Magnified (×4) views of the area outlined by the red dashed lines in (B) and (D). In the SD-OCT images, blue arrowheads points at the inner retinal defects limited to the RNFL, and red arrowheads the inner retinal defects that apparently are hollowed beyond the RNFL into the GCL and IPL.

The inner retinal defects that were limited to the RNFL were detected in all of the 31 eyes with the DONFL appearance, and 24 (77.4%) of these 31 eyes included defects that extended beyond the RNFL. 

Similar inner retinal defects were seen in 14 (87.5%) of 16 eyes in which the DONFL appearance was not evident on the color fundus photographs (Fig. 4). Of these 16 eyes, 11 (68.9%) eyes had defects deeper than the RNFL. 

Inner retinal defects deeper than the RNFL were found in 23 (76.7%) of 30 eyes in which IGC was used, and in 12 (70.6%) of 17 eyes in which TA was used. No significant differences were found in age (P = 0.266), sex (P = 1.0), use of ICG or TA (P = 0.733), preoperative BCVA (P = 0.333), or postoperative BCVA at 6 months (P = 0.668) between eyes with and without the deep inner retinal defects. 

Distribution of the DONFL Appearance in the Macula

In the 31 eyes with a DONFL appearance, the distribution of the DONFL in the color fundus photographs was significantly different for the four macular quadrants; arcuate striae were found most frequently in the superior quadrants, and least frequently in the temporal quadrant. This difference between the superior and temporal quadrants was statistically different (P < 0.0001, Table 1). In contrast, the incidence of defects extending beyond the RNFL in the SD-OCT images was significantly higher in the temporal quadrant than in any other quadrant (P < 0.0001, Table 1). 

Comparison of Retinal Sensitivity Map and Corresponding SD-OCT Images

Some of the test points were located within the arcuate striae, and the others outside the arcuate striae (Fig. 5). Some test points located within the arcuate striae were on the sites of the inner retinal defects deeper than the RNFL, whereas others were on the inner retinal defects limited to the RNFL (Fig. 5). The test points were so sparse that they were rarely on the tiny areas with the most severe inner retinal defects (Fig. 5). Some test points located within the arcuate striae and deep inner retinal defects tended to show lower retinal sensitivity than the surrounding test points. However, other test points that were located on the arcuate striae, but not on the deep inner retinal defects, often did not have a reduction in retinal sensitivity (Fig. 5). 

Figure 5

 

Retinal sensitivity map obtained by Microperimeter-1 superimposed on color fundus photograph and corresponding SD-OCT images after macular hole surgery with internal limiting membrane peeling. The same eye shown in Figures 3D to 3F. (A) Retinal sensitivity of 57 measurement points indicated by square points. Yellow lines indicate the borders to divide the macular area into 4 quadrants. (B) Same result with arrowheads. Red and blue arrowheads point at the measured points on and away from the arcuate striae, respectively, on the color fundus photograph. The green lines indicated by an asterisk, dagger, and double dagger show the scan lines for SD-OCT images in (C), (D), and (E), respectively. The measured points that were not used for analysis are masked by orange squares. (C–E) Horizontal SD-OCT images through the green lines in (B). Green and yellow squares indicate the location of the measured points in the Microperimeter-1 examination. Red and blue arrowheads point at the measured points on and away from the arcuate striae, respectively, on the color fundus photograph. The numerical value on each colored square indicates the retinal sensitivity at each point indicated by the square. Note that the testing points are sparse for the narrow arcuate striae and the small inner retinal defects on SD-OCT images; the test points rarely hit the tiny areas with the most severe inner retinal defects (red arrow).

Retinal Sensitivity at Test Points on and Away From Arcuate Striae

A histogram of the measured points and their retinal sensitivities is shown in Figure 6. Before the surgery, 398 (99.5%) of the 400 points had a retinal sensitivity ≥10 dB except 2 for points at 7 dB. After the surgery, only 27 (6.8%) of the 400 points had a retinal sensitivity <10 dB. A retinal sensitivity of <10 dB was found in 20 (10.4%) of the 192 points on the arcuate striae, but in only 7 (3.4%) of the 208 points away from the arcuate striae; this difference was statistically different (P = 0.005). 

Figure 6

 

Histograms showing the number of measured points with different retinal sensitivities measured with the Microperimeter-1 in the central 10° before and after macular hole surgery with ILM peeling in the eyes that developed a DONFL appearance. (A) Before macular hole surgery. (B) After macular hole surgery. Red and blue bars indicate the numbers of measured points on and away from the arcuate striates, respectively, with different retinal sensitivities.

The mean postoperative retinal sensitivity on the arcuate striae was significantly lower (1.6 dB [9.3%] reduction) than that away from the arcuate striae (P = 0.034; Table 2).When the same comparison was made for each quadrant, there were significant differences in the retinal sensitivities in the temporal quadrant (P = 0.0001), but not in any other quadrant. 

Table 2. 

 

Comparison of Mean Retinal Sensitivities Between Test Points On and Outside the DONFL Appearances After Macular Hole Surgery With ILM Peeling in 15 Eyes of 15 Patients Who Underwent Micro Perimeter-1 Examinations

	Mean Preoperative Retinal Sensitivities, dB	Mean Postoperative Retinal Sensitivities, dB		P Values*
		On DONFL Area	Outside DONFL Area
Whole	16.4 ± 2.54	15.6 ± 4.61	17.2 ± 3.20	0.034
Temporal*	16.6 ± 2.68	14.2 ± 5.62	17.4 ± 3.31	0.0001
Superior	15.7 ± 2.84	15.1 ± 3.78	15.5 ± 3.16	0.619
Inferior	16.7 ± 2.38	16.9 ± 4.10	18.1 ± 3.07	0.212
Nasal	16.7 ± 1.99	16.9 ± 4.24	17.6 ± 2.61	0.936

*  Comparison between test points on and outside DONFL appearance (Mann-Whitney U test).

Discussion

Speckle noise reduced SD-OCT imaging allowed us to determine the depth of the inner retinal depressions called dimples in earlier studies, ^10–13 in eyes that underwent successful macular hole surgery with ILM peeling. The defects were located in the arcuate striae in all eyes with a DONFL appearance. A large proportion (80.6%) of these eyes included the inner retinal depressions that were seen in the SD-OCT images to extend beyond the RNFL into the GCL and IPL. The inner retinal defects, including those deeper than the RNFL, also were found in 62.5% eyes without an evident DONFL appearance. These findings would indicate that comparisons of the visual function between eyes with and without a DONFL appearance will not necessarily show if the DONFL appearance will alter retinal function. Because the mean retinal sensitivity on the arcuate striae was slightly, but significantly lower than that away from the arcuate striae, the inner retinal changes associated with ILM peeling may be more harmful to postoperative retinal sensitivity than believed. ^9–14  

Earlier studies showed that the DONFL was due to the dissociation (cleavage) of the bundles of the optic nerve fibers in the posterior pole. ^9–14 This conclusion was reached because the depressions appeared to be limited to the RNFL in the TD-OCT images, and because the presence of a DONFL appearance was not associated with a significant reduction of the visual acuity or retinal sensitivity. However, the OCT results of earlier studies must be interpreted with caution. First, the resolution of the TD-OCT (Stratus OCT) was not high enough to identify the different retinal layers that were involved in the defects. Abnormalities of the RNFL could be detected in the TD-OCT images, but the images were not clear enough to determine the status of the GCL and IPL. ^9–13 Second, because earlier studies showed only vertical scans through the central fovea, we were not able to see the tomographic images in the other regions. In our study, the deep defects were seen most frequently in the temporal macula, and these deeper defects were limited to a focal part of each arcuate striae (Fig. 2). The slow imaging speed of the TD-OCT would not be effective in detecting such deep focal defects. 

The DONFL appearance is believed to be caused by the ILM peeling because a DONFL appearance did not develop in eyes without ILM peeling during macular hole surgery. In addition, the DONFL appearance was present only in the area where the ILM was peeled. ^11,12 The toxicity of ICG for ILM staining also may cause the development of a DONFL appearance. However, we could not find any significant differences in the development of a DONFL appearance between eyes exposed to ICG and TA. This result is consistent with the earlier studies, ^11,12,14 in which no differences were found in the incidence of the DONFL appearance between eyes with and without the use of ICG, and between ICG-peeled and trypan blue-peeled eyes. In addition, in our study, no significant differences were found in the frequency of the deep inner retinal defects between eyes treated with ICG and TA. Thus, the ILM peeling procedure itself appears to be responsible for the development of the DONFL appearance and the deep inner retinal defects, rather than the dye used for ILM peeling. However, we cannot deny the possibility that the use of the ILM staining, particularly ICG, might have enhanced the retinal changes associated with a DONFL appearance, such as the number of arcuate striae and the depth of the inner retinal defects. Further studies are needed to clarify this possibility. 

It is known that the ILM has an important role in retinal physiology, because it is the basal lamina of the endfeet of the Müller cells. Tadayoni et al. suggested that that the DONFL appearance may be caused by permanent damage to the part of the Müller cells that maintains the optic nerve fiber bundles together. ⁹ Thus, when the Müller cells are damaged, the optic nerve fibers lose their structural support and a dissociation of the optic nerve fiber layer appears. However, the deep inner retinal defects we observed in our study cannot be explained by only the dissociation of retinal nerve fibers. 

Damages to the Müller cells also may cause degenerative changes of the retinal nerve fibers, which can, in turn, cause atrophic changes in the GCL and IPL. This is supported by the observation that the b-waves of the focal macular electroretinogram are reduced after macular hole surgery combined with ILM peeling, but not in eyes without ILM peeling. ¹⁶ In addition, it required at least 1 month for the DONFL appearance to develop, ^12,26 which is in keeping with the time required for biologic changes, such as tissue remodeling, to develop after the macular hole surgery with ILM peeling. 

In our study, the inner retinal defects were found more frequently than the DONFL appearance. The rate of DONFL appearance on color or red-free photography after ILM peeling was 66% in our study, and ranged from 54% to 62% in the earlier studies. ^10–12 The rate of DONFL appearance was higher in our patients than in those of the earlier studies, but not largely different. In contrast, the rate of DONFL appearance after MH surgery with ILM peeling was 73% and 100% in the 2 earlier studies, ^26,27 in which a DONFL appearance was detected by retinal surface or en face SD-OCT imaging in a 3-dimensional cube scan of Cirrus HD-OCT. Importantly, the HD-OCT images showed a DONFL appearance pattern even in eyes without a DONFL appearance in the color fundus photographs, consistent with our observation. ^26,27 Photography may be less sensitive in detecting certain inner retinal abnormalities after MH surgery with ILM peeling. The DONFL appearance on photography is seen where the optical reflectance from the RNFL is changed abruptly, whereas the retinal surface or en face imaging on SD-OCT is based on the three-dimensional structures of the retinal surface. This difference in imaging principles appears to be the cause of the disagreement between color fundus photographs and SD-OCT images in some cases. ^26,27 If the inner retinal defects do not lead to such sharp changes in the RNFL reflectance, they would not be seen as a DONFL appearance in the fundus photographs. In addition, the DONFL appearance is recognized easily because it shows the characteristic arcuate striae pattern of retinal nerve fibers. If the inner retinal defects do not form the characteristic arcuate striae pattern, it may be more difficult to recognize the defects as a DONFL appearance. 

Earlier studies did not find any evidence that the DONFL appearance was associated with a reduction of the visual acuity or the visual sensitivity. However, it is necessary to interpret these functional results of the earlier studies with caution. First, no significant difference in the postoperative BCVA or the improvements of the BCVA improvements between eyes with and without DONFL appearances was reported as we found. ^9–13 Ito et al. compared the retinal sensitivities using the Humphrey 10-2 program, and reported no significant difference in the mean deviations between eyes with and without DONFL appearances. In our study, however, the inner retinal defects were found in 87.5% eyes without a DONFL appearance, which generally is consistent with the results of an earlier study. ²⁶ In addition, we found that the inner retinal defects extended deeper than the RNFL in 62.5% of eyes without a DONFL appearance. Thus, it is uncertain whether comparisons of the visual functions between eyes with and without DONFL appearances can validate actual damages to the inner retinal layers associated with a DONFL appearance. 

We found a slight, but significant, difference in the sensitivities between the points on and away from the arcuate striae, which is not consistent with the results of earlier studies. ^11,13 Mitamura et al. used microperimetry to determine the retinal sensitivities at 4 points in the nasal quadrant, and reported no significant differences in the threshold values between the points on the arcuate striae and those on the surrounding normal appearing retina. ¹¹ Our results showed that the nasal quadrant had the lowest incidence of inner retinal defects deeper than the RNFL, which may explain why Mitamura et al. did not find any significant difference. Imai et al. compared the retinal sensitivities determined by Microperimeter-1 on and away from the arcuate striae, and did not find any significant difference in any of the 10 eyes. ¹³ As shown in our histogram, a reduction of retinal sensitivity was detected in the points over and also away from the arcuate striae. Ganglion cell bodies are displaced laterally from the cone photoreceptors in the macula by the elongation of the cone axons, which connect to the bipolar cells. ^28–30 Thus, it is possible that a stimulus on the tiny area of the inner retinal defects will not detect a reduced sensitivity because the photoreceptors connecting to the ganglion cells in the area with the inner retinal defects are displaced from this area. Such displacements of the retinal ganglion cells may be responsible, at least in part, for not detecting significant differences between measurement points on and away from the arcuate striae. 

Ito et al. did not detect a scotoma in eyes with a DONFL appearance by scanning laser ophthalmoscopy (SLO) microperimetry. ¹² In contrast, Haritoglou et al. observed paracentral scotomata by SLO microperimetry in 56.2% patients after macular hole surgery with ILM peeling without ICG. ¹⁷ Of these, 27.1% were relative and 72.9% were absolute scotomata. The scotomata were located temporally (74.6%), inferiorly (62.7%), superiorly (50.8%), or nasally (23.7%). The distribution of the scotomata in the 4 quadrants was similar to that of the inner retinal defects deeper than the RNFL in our patients. Although the investigators were not aware of the phenomenon of a DONFL appearance, they mentioned that 57.6% of these scotomata appeared similar to a RNFL defect. It is likely that they unknowingly detected scotomata associated with a DONFL appearance. It is known that the receptive field of a retinal ganglion cell overlaps those of surrounding retinal ganglion cells. Functionally, damage to retinal ganglion cells can be compensated for by the surrounding retinal ganglion cells. ^31,32 This compensation can prevent the development of a scotoma when the area of damage is small, and may be responsible for the inconsistency of our results with that of the earlier studies. 

Similar gaps between clinical functional testing and local structural damage have been well documented in preperimetric glaucoma. Approximately 50% of retinal ganglion cells have been lost when the visual field defects at the corresponding test point are detectable in the central 10° by standard automated perimetry in eyes with glaucoma. ^33–35 Loss of more than 60% of retinal ganglion cells has been associated with a 3 dB sensitivity loss in the central 10° on standard automated perimetry in glaucoma. ³⁶ In addition to the reasons mentioned above, this gap between retinal ganglion cell loss and functional testing results also has been attributed to the limited number of testing points in standard automated perimetry and microperimetry. In our study, the test points were so sparse that they only occasionally hit the tiny areas of severe inner retinal defects deeper than the RNFL. 

Regardless of whether retinal sensitivity loss associated with the inner retinal defects can be detected with these clinical functional tests, deep inner retinal defects involving the GCL are not favorable for healthy retinal sensitivity in the macula. When treated eyes have other fundus diseases that cause retinal sensitivity loss in the macula, such as glaucoma, retinitis pigmentosa, and rhegmatogenous retinal detachment, the local structural damage may no longer be subclinical, but rather additive to the retinal sensitivity loss possibly causing a worsening quality of vision. These diseases are not rare, and indeed, it also is possible that in the future, the treated eyes may suffer other retinal diseases threatening retinal sensitivity in the macula. When this possibility is considered, the damage to the inner retina caused by ILM peeling should be minimized. In this regard, a lesser extent of ILM peeling may be better, although studies examining the effects of less extensive ILM peeling on macular hole closure rate must be done. Macular holes can be closed only by creating posterior vitreous detachment, particularly in eyes with small macular holes, although the success rate is lower than that for vitrectomy combined with ILM peeling. It may be valuable to consider minimally invasive surgery without ILM peeling or less extensive ILM peeling in eyes with prior or existing ocular diseases that can affect retinal sensitivity in the macula. In addition, enzymatic vitrectomy has been explored to treat macular holes, particularly for small macular holes. Information on adverse effects of current standard procedures of macular hole surgery will be important to guide and motivate the development of new treatments, and to facilitate future decisions to select surgical or medical treatments. Although macular hole surgery appears to be established in terms of macular hole closure, further efforts to sophisticate the case-by-case procedure for macular hole surgery to save all patients from unnecessary retinal sensitivity loss in the macula still are needed. Thus, it is important to be aware of the possible adverse effects of ILM peeling that cannot be known by routine clinical examinations. 

The reason why deep inner retinal defects were found most frequently in the temporal macula was not determined. However, the temporal macula is where the RNFL is thinnest. It is possible that a thinner RNFL is more vulnerable to the effects of ILM peeling. 

A limitation of our study is that we could not differentiate completely between inner retinal damage caused by ILM peeling and the mechanical damage of the surgical procedures. However, it is difficult to believe that mechanical damage during ILM peeling can generate the unique pattern of the DONFL appearance. In addition, the initial grasping of the ILM, which most likely touched the retinal nerve fibers, was done in the superior quadrants and not in the temporal quadrant. A second limitation is that we used color fundus photography. Blue light photography may be better for the detection of the DONFL appearance than color fundus photography. Three previous studies used blue light photography, ^9–11 and 2 used color fundus photography, ^12,19 and the incidences of DONFL appearance in these studies were not largely different. However, we cannot rule out the possibility that some eyes without a DONFL appearance on color fundus photography would have had DONFL appearance on blue filter photography. A third limitation is the small number and possible selection bias of the eyes we used for the MP-1 analysis. Surgeons might have used MP-1 because they found a marked DONFL appearance. This bias would be minimal when comparison was made on and away from the arcuate striae within each eye. 

In conclusion, inner retinal defects that extended deeper than the RNFL were found frequently in the area of the DONFL appearance, especially in the temporal macula, after successful macular hole surgery with ILM peeling. A reduction of retinal sensitivity associated with the DONFL appearance was not common, probably because the deep inner retinal defects were limited to small focal areas. These inner retinal defects also were found in eyes without an evident DONFL appearance, which prevented us from comparing the retinal sensitivity loss between eyes with and without a DONFL appearance. Thus, the DONFL appearance in eyes after successful macular hole surgery with ILM peeling is associated, at least in part, with inner retinal damage, although this had negligible effect on the postoperative retinal sensitivity. It remains to be determined whether the local damage to the inner retinal layers after the macular hole surgery lead to clinically significant adverse effects in eyes with the other ocular diseases that can cause retinal sensitivity loss in the macula. 

Acknowledgments

The authors alone are responsible for the content and writing of the paper. 

Disclosure: K. Nukada, None; M. Hangai, Nidek (F, S), Topcon (F, C), Canon (F), Heidelberg Engineering (R), Santen (R); Sotaro Ooto, None; M. Yoshikawa, None; N. Yoshimura, Nidek (F, S), Topcon (F, R), Canon (F, R)