Newer indices for predicting macular hole closure in idiopathic macular holes: A retrospective, comparative study.

Purpose: To study the utility and predictive ability of newer macular hole (MH) indices for closure following surgery.
Methods: In this retrospective study, pre- and post-operative optical coherence tomography images of 49 eyes with idiopathic full-thickness MH were reviewed and analysed. Various quantitative parameters of MH like maximum outer diameter (OD), minimum diameter between edges, height, nasal and temporal arm lengths, macular hole angle were noted. Indices including hole form factor, Macular Hole Index, (MHI), Diameter Hole Index (DHI) and Tractional Hole Index (THI) were calculated. Newer area indices like macular hole area index (MAI), cystoid space area index (MCSAI) and tissue area index (MTAI) were calculated using Image J (Ver. 1.51). Receiver operating characteristic (ROC) curves and cut-off values were derived for indices predicting type 1 or type 2 closure. Stepwise regression analysis and binary logistic regression analysis were carried out to predict the chances of hole closure.
Results: ROC curve analysis showed indices like MHI, THI and MCSAI were capable of successfully predicting type 1 closure while OD, DHI and MAI predicted type 2 closure. On stepwise regression analysis, MAI was identified as the most important index in predicting the type of hole closure. Using the binary logistic regression analysis, the predictive ability of the model to identify success or failure following MH surgery was 89.7% and 80% respectively.
Conclusion: MAI measurement could be used as a single important index in predicting hole closure in idiopathic MH. Further research is required to study this area index in detail.

Idiopathic full-thickness macular hole (MH) is a defect in the neurosensory retina at the macula caused by the antero-posterior and tangential tractional forces by the vitreous and internal limiting membrane (ILM), respectively. The original Gass’ classification of MH used the slit-lamp biomicroscopy for estimating the hole size and other characteristics to differentiate between various stages of MH.[1] Spectral domain optical coherence tomography (OCT) is an important investigative tool in staging, diagnosis and monitoring of MH closure after surgery.[2]

The first study to use OCT to analyse macular holes preoperatively was published by Ip et al. in 2002.[3] Since then, various studies have been published describing the role of macular hole measurements and derived indices like hole forming factor (HFF), macular hole index (MHI), diameter hole index (DHI) and tractional hole index (THI) in pre-operatively predicting the anatomic closure and visual gain following MH repair surgery.[456] Wakely et al. compared the different methods of MH measurement in the same cohort of patients.[7] They found no real advantage by calculating derived MH indices from the basic ophthalmic measurements in predicting the anatomic closure and functional success following macular hole surgery. They found the preoperative base diameter as the single most important determinant in predicting the anatomic and functional success. Chhabalani et al. evaluated the MH angle in pre-operatively predicting the successful MH closure. However, no significant correlation was noted between MH angle and closure.[8]

The disadvantage with these indices is that they take into account primarily one of the 2 tractional forces for MH formation. For instance, the DHI which is defined as the ratio of the minimum diameter of the MH to the base diameter indicates the strength of the tangential traction at the fovea and postulates that the strength of traction is maximum when the inner diameter of the MH is equal to the base diameter. The THI, on the other hand, which is defined as the ratio of the maximal height of the MH to the minimum diameter represents primarily the antero-posterior vitreomacular traction and/or retinal hydration responsible for the macular hole formation and less of tangential traction. THI was found to correlate significantly with postoperative visual acuity at 3 months. However, there are a few other indices like the HFF and MHI which take into account both the traction forces in MH formation.

Furthermore, the MH edges and the retinal surface are curved surfaces and hence using linear measurements may not show us the correct picture. With advanced computing, it is now possible to automatically measure the area of structures that are geometrically irregular.

The macular hole size and its elevated oedematous edges on clinical examination mainly occur depending on the extent of antero-posterior and tangential tractional forces acting on the hole. Thus, this OCT-based imaging study on MH is based on the hypothesis that in a predefined region of the macular hole, the ratio of areas occupied by the macular hole, intra-retinal fluid and retinal tissue could be important determining factors in assessing the extent of tangential and antero-posterior tractional forces and thereby predicting the chances of success or failure following MH surgery. No previous studies have evaluated these area indices as predictor of MH closure.

Thus, in the current study, we aim to study the relationship and predictive value of area indices with MH closure and compare it with previously established indices namely OD, HFF, MHI, DHI, THI and macular hole angle.

Methods

A retrospective review of electronic medical records and OCT images of consecutive patients diagnosed with idiopathic MH having undergone MH surgery, between January 2017 and December 2017, was performed. Prior approval was obtained from the institutional review board and informed consent was obtained from each subject. Inclusion criteria were subjects who underwent primary MH repair surgery for idiopathic MH with minimum 8-weeks follow-up; availability of good quality OCT scans at baseline, at 8 weeks and at the last follow-up visit. Exclusion criteria were presence of any other ocular pathology contributing to visual loss; macular hole secondary to any other cause such as trauma, high myopia, postsurgical, diabetic macular oedema; history of previous vitreoretinal surgery; history of any treatment related to retinal diseases such as laser photocoagulation. For this study the ethics committee approval was obtained. EC- No - C/2018/06/03 - Dated - 30 June 2018.

All patients underwent a comprehensive eye examination including best-corrected visual acuity (BCVA) testing, dilated fundus examination with slit-lamp biomicroscopy, colour fundus photography and OCT at baseline, at 8-week and as per physician's discretion during the follow-up visits. Data collected included demographics; BCVA at baseline, 8-week and last visit; details of the ocular and systemic exam; imaging with OCT at baseline, 8-weeks and last visit; surgical details and complications, if any. The MHs were grouped on the basis of classification proposed by Duker et al.[2] The BCVA was assessed using Snellen's chart and then converted to ETDRS letters using the formula 85 + 50 × log (Snellen fraction) for analysis. Lens grading was performed as per standard Emery and Little classification.[9]

Spectral domain OCT

Optical coherence tomography measurements were performed with the Spectralis OCT (Heidelberg Engineering, Heidelberg, Germany). Macular volumetric assessments consisting of horizontal axial scans with 512 A-scans per line with scanning area 6 × 6 mm, 25 scan patterns centred at the fovea, were performed. The scan which had the maximum diameter of macular hole (MH) was analysed. Only scans with good signal strength were used for analysis. This scan image was saved in the.jpg/.jpeg format and then exported for further analysis with Image J.

Steps in macular hole surgery

All patients underwent pars plana vitrectomy with ILM peeling with endotamponading agents. All the surgeries were performed by 2 surgeons (NKY, RV) who had at least a minimum of 5 years’ experience in macular hole surgery. All surgeries were performed using the 23G or 25G vitrectomy system and Brilliant Blue Green (BBG) dye was used in all cases to stain the ILM. ILM staining was done after fluid -air exchange to limit the dispersion of the dye and intraocular illumination was switched off to avoid phototoxicity. During the surgery, conventional ILM peeling was performed up to the arcades after staining. Finally, fluid-air exchange was performed, followed by exchange with tamponading agent. 20% sulphur hexafluoride (SF6) gas was used as the endotamponade in all cases. All patients were advised prone position at least for one week. Other surgical details collected included surgical complications, if any, and additional procedure such as cataract surgery.

Follow-up

Patients were seen at 4 and 8 weeks after the surgery, and then at every 3-4 months till the last follow-up as per the physician's discretion. The OCT scans at 8-week's visit were taken for analysis.

Anatomical outcomes

In our study, the patients were divided into two types of anatomical outcomes as defined by Kang et al.[10]: 1) Type 1 in which macular hole was completely closed with a foveal morphology appearance; and 2) Type 2 in which the macular hole was poorly closed or not closed, with the absence of foveal-area neurosensory retina.

Measurements with Image J

The preoperative OCT image was analysed by ImageJ (Open source software - version 1.51) by 2 independent observers to test for inter-observer variation (ASM & SSH). Using the line tool in Image J, the following measurements were made in microns; minimum diameter between the edges, maximum diameter at base, arm lengths, macular hole height and macular hole angle. A scale of 59 pixels per centimetre was set.

From these measurements, the following indices were derived:

Measuring area indices with Image J

Using the polygonal tool, a region of interest (ROI) with the macular hole at centre including the areas of intraretinal fluid was selected and set by the ROI manager in the OCT image. All the images were of the same pixels [685 pixels in width × 246 pixels in height]. Using the measure tool, the total area (TA) of the ROI was calculated. Using the polygonal tool, the space within the macular hole was marked and set by the ROI manager. The area measured was defined as the macular hole area (MA). Then, the image was converted to 8-bit and binarization was done by the auto local threshold of Niblack. The light pixels and dark measured were defined as the stromal or tissue area (MTA) and cystoid space area (MCSA) respectively. Using the measure tool, the MTA was calculated. The MCSA was derived by subtracting the sum of MA and MTA from TA. The binarized image was reconverted to an RGB image, and the cystoid space area was marked using the threshold tool [Fig. 1].

Figure 1

Measuring Area indices using Image J. (a) Pre-operative OCT image exported from Spectralis OCT machine and opened with Image J. (b) Using the polygonal tool, a region of interest (ROI) with the macular hole at centre including the areas of intraretinal fluid was selected and set by the ROI manager in the OCT image. Using the measure tool, the total area (TA) of the ROI was calculated. (c) Using the polygonal tool, the space within the macular hole was marked and set by the ROI manager. The area measured was defined as the macular hole area (MA). (d) The image is converted to 8-bit and binarization was done by the auto local threshold of Niblack. (e) Overlay image showing the light and dark pixels defined as the stromal or tissue area (MTA) and cystoid space area (MCSA) respectively. Using the measure tool, the MTA was calculated. The MCSA was derived by subtracting the sum of MA and MTA from TA

From these measurements, the following indices were derived:

Statistical analysis

Data collected were analysed with GraphPad Prism (8.1.1) and Microsoft Excel 2016 for Windows. Basic demographic data, pre- and post-operative visual acuities and macular hole indices between the two types of macular hole closures were analysed using the Mann-Whitney U test. For categorical variables like gender, PVD status and MH size, difference between the 2 groups were analysed using the Chi-square Test. A receiver operating characteristic (ROC) curve analysis was carried out to evaluate the predictive ability of the different macular hole indices for type 1 and 2 closure. The areas under the ROC curve (AUC) were calculated to judge the efficiency of each index on comparison with the different types of closures. The cut-off values from the ROC curve were obtained for highest possible sensitivity and specificity for the different indices. Correlations between the type of closure as the dependent variable and the computed indices as independent variables were calculated using the Pearsons’ correlation matrix. Stepwise regression analysis was conducted using the Excel to identify the index having the best correlation with the type of closure. Binary logistic regression analysis using excel was done to calculate the role of indices in predicting closure.

Results

In this retrospective study, we analysed the demographic and OCT parameters of 49 eyes in 47 patients with idiopathic macular hole who underwent macular hole repair surgery. Mean age among study patients was 62.5 ± 7.3 years. There were 17 men and 30 women. Preoperative visual acuities in ETDRS letters ranged from 20 to 61 (mean 39, SD 15). One eye had small-sized macular hole, 7 eyes had medium-sized macular holes and 41 eyes had large-sized macular holes. Cataract surgery was combined along with the MH surgery in 37 of the 49 eyes (76%).

Surgical Outcomes: Anatomical and visual outcomes were assessed at the end of the primary macular hole surgery. Type 1 macular hole closure was achieved in 29 eyes (59%). The remainder eyes showed type 2 closure (non-closure 5 eyes and closure with central defect 15 eyes). Failed cases required a second procedure with repeat ILM peeling with either gas or silicone oil tamponade with prone positioning. Mean postoperative visual acuity was 46 ± 14 ETDRS letters. The follow-up period varied between 2 and 12 months (mean 5.36 months, SD 2.27 months).

OCT measurements: The basic ophthalmic measurements on Image J: the mean base diameter was 1882.7 ± 726.9 μ; the mean inner diameter was 938.3 ± 398.9 μ; the mean macular hole height was 1566.6 ± 375.4 μ; the mean nasal arm length was 955.5 ± 360.5 μ; the mean temporal arm length was 917.4 ± 330.2 μ and the mean macular hole angle was 58.6 ± 8.2°. The calculated area indices using Image J: the mean macular hole area was 1.936 mm2 (SD 0.851 mm2); the mean macular hole cystoid space area was 1.857 mm2 (SD 0.971 mm2); and the mean macular hole tissue area was 2.344 mm2 (SD 0.935 mm2).

The derived macular hole indices were: 1) mean HFF was 1.042 (SD 0.299); 2) mean MHI was 0.936 (SD 0.373); 3) mean THI was 2.039 (SD 1.091); 4) mean DHI was 0.505 (SD 0.133); 5) mean MA was 0.326 (SD 0.098); 6) mean MTA was 0.386 (SD 0.053) and 7) mean MCSA was 0.287 (SD 0.068) [Table 1]. Intraclass correlation for interobserver repeatability and reproducibility ranged from 0.89 to 0.93 for all the measurements.

Table 1

Basic demographic data, pre- and post-operative visual acuities and macular hole indices between the two types of macular hole closures were analysed using the Mann-Whitney U test and Chi-square Test

Label	Type 1 closure	Type 2 closure	P
Age	65	68	0.081
Sex (M: F)	13:16	6:14	0.377
Pre-op BCVA	43	33	0.214
Fellow eye BCVA	72	74	0.217
PVD (Complete: Incomplete)	25:4	16:4	0.700
Hole Size (Small:Medium:Large)	1:3:25	0:4:16	0.468
HFF	1.14	0.90	0.007
MHI	1.08	0.73	0.001
THI	2.51	1.36	<0.001
DHI	0.46	0.57	0.007
Angle	58.7	58.3	0.839
MTAI	0.41	0.35	0.001
MCSAI	0.31	0.25	0.001
MAI	0.28	0.40	<0.001
Post Op BCVA	54	34	<0.001
Gain in Letters	11	0	0.007

BCVA=Best corrected visual acuity, PVD=Posterior vitreous detachment, HFF=Hole forming factor, MHI=Macular hole index, THI=Tractional hole index, DHI=Diameter hole index, MTAI=Macular hole tissue area index, MCSAI=Macular hole cystoid space area index, MAI=Macular hole area index

Analysis of results

Using the Mann-Whitney U test and Chi-square test, no statistically significant difference was noted between the type of anatomic hole closure and different demographic variables and pre-operative factors like the PVD status or macular hole size. All MH indices except macular hole angle showed significant difference between type 1 and type 2 MH closures.

Macular hole indices and anatomic success

From the ROC curve analysis, it was derived that macular hole indices like MHI, THI and MCSAI were capable of successfully predicting type 1 closure while OD, DHI and MAI predicted type 2 closure [Fig. 2 and Table 2]. For type 1 closure, higher values of AUC (>0.5) was noted for MHI (0.791), MCSAI (0.783) and THI (0.840). The AUC values for MAI (0.878), DHI (0.728) and OD (0.678) were higher for Type 2 closure. From the ROC curve analysis, our aim was to derive cut-off values for each index, so as to predict closure with maximum possible sensitivity and specificity. For instance, a value of MAI <0.323 could predict successful closure with 85% sensitivity and 83% specificity Fig. 3.

Figure 2

Using the receiver operating characteristic curve analysis, the derived area under curve values for various macular hole indices

Table 2

Receiver operating curve analysis corresponding-anatomic success using the different macular hole indices

Index	AUROC	Cut-off value	Highest possible sensitivity (%)	Highest possible specificity (%)	95% CI	P
OD	0.678	>1843	65	75.9	0.517-0.840	0.035
HFF	0.728	<0.97	75	69	0.582-0.873	0.007
MHI	0.791	<0.81	75	75.9	0.662-0.921	0.001
THI	0.840	<1.53	75	79.3	0.725-0.954	<0.000
DHI	0.728	>0.505	65	62.1	0.580-0.876	0.007
Angle	0.517	<57.6	60	58.6	0.349-0.686	0.839
MTAI	0.772	<0.393	70	65.5	0.633-0.911	0.001
MCSAI	0.783	<0.303	80	75.9	0.648-0.917	0.001
MAI	0.878	>0.323	85	82.8	0.781-0.976	<0.000

OD=Outer diameter; HFF=Hole forming factor; MHI=Macular hole index; THI=Tractional hole index; DHI=Diameter hole index; MTI=Macular hole tissue index; MCSAI=Macular hole cystoid space area index, MAI=Macular hole area index; CI=Confidence interval; AUROC=Area under ROC curve

Figure 3

Clinical application of newer area indices for predicting macular hole closure. (a and b) Pre- and post-operative OCT image of a full-thickness macular hole with a THI = 4.154 (red arrow) and MAI = 0.197 (blue arrow) showing favourable anatomical outcome.(c and d) Pre- and post-operative OCT image of a full-thickness macular hole with a THI = 0.797 (red arrow) and MAI = 0.492 (blue arrow) showing unfavourable anatomical outcome

Stepwise regression analysis of the multiple variables was carried out to identify the index showing the best correlation with the type of closure. MAI was identified as the index to show the maximum correlation with the type of hole closure (r2 = 0.378, AIC - -88.784). Binary logistic regression analysis to predict the hole closure was carried out using the 3 most important indices namely THI, MTAI and MAI. Of the three indices, MAI was seemed to be the most important index in predicting hole closure. The log likelihood chi square was 26.935 and the predictive capability of the model was 89.7% and 80% in predicting success and failure of hole closure, respectively.

Discussion

In our study, we measured the newer area parameters like MA, MCSA and MTA on OCT using Image J, calculated the area indices and compared them with other known and existing macular hole indices in predicting the anatomical outcome following macular hole surgery. Preoperative mean values of all the indices except the macular hole angle differed significantly between the two groups. The ROC curve analysis showed that the higher AUROC values of MHI, THI and MCSAI were seen with type 1 closure while higher AUROC values of OD, DHI and MAI were seen with type 2 closure. On stepwise regression analysis, MAI was seen as the most important index showing statistically significant correlation with the type of hole closure. Amongst the other conventional indices, we saw that a few were better at predicting Type 1 closure whereas a few were better at predicting Type 2 closure. This result can be attributed to the specific vectors being studied by the indices. These results are similar to what we found in our current study. The THI and MHI are largely affected by the height of the hole which is indicative of the amount of antero-posterior traction and/or retinal cystic spaces (hydration theory). A larger component of traction means that when eliminated with surgery increases the chances of Type 1 closure manifold and the complete resolution of cystic spaces is associated with better improvement in visual acuity. Ruiz-Moreno et al. also found the basal diameter, minimal hole diameter, THI and MHI corelate significantly with BCVA at 3 months post-surgery.[5] It is interesting to see that the MHI which uses a single measure of height divided by the maximum diameter was a better predictor than the HFF which takes into account the nasal and temporal arm of the hole edges divided again by the maximum diameter. It could have been hypothesised that a single measure of height would give a less complete picture than the nasal and temporal arms added together. Haritoglou et al. studied the correlation between postoperative BCVA and base and minimum diameter of the MH, MH height and MH form factor.[11] They found a significant negative correlation of BCVA with base diameter and MH height and a poor correlation with HFF. They stated that the long-term predictive value of the HFF could be questioned. This indicates that the antero-posterior traction forces are represented better by the macular hole height rather than nasal and temporal arm lengths. We found a significant positive correlation between postoperative anatomic closure and HFF. The diameter hole index evaluates the distance between the edges of the hole. This is largely indicative of the tangential traction and which we observed was better predictive of type 2 closure. The macular hole angle being a unidimensional measure gives very little information about the tractional forces, and hence was the only index to not show any statistical significance.

Amongst the new indices that we studied, we saw that the higher MCSA index was able to predict Type 1 closure. This meant that more the cystic spaces, higher was the chance of better anatomical outcomes. Since this predicts similar to the MHI and the THI we may hypothesise that the presence of retinal cysts is indicative of greater antero-posterior tractional forces and taller macular holes.

The MA index evaluates the total area of the macular hole. The higher value of MAI >0.323 was capable of predicting type 2 closure with 85% sensitivity and 83% specificity. This index predicts very similar to the DHI which is similar in the area of interest – the lumen. The maximum outer diameter (OD) is one of the measurements which could be calculated with greatest ease. Wakely et al. compared the different methods of MH measurement in the same cohort of patients and concluded that the preoperative outer diameter as the single most important determinant in predicting the anatomic and functional success.[7] One could argue that the OD which can be measured by the greatest of ease to be a superior index in predicting macular hole closure in comparison to the other indices like the MAI and DHI. However, in our study, with respect to the area under ROC curve for type 2 closure, the MAI (0.878) occupied the highest area under ROC curve with a narrower 95% confidence interval limits. The MAI accounts for both the tractional forces in a more complete manner as compared to the other indices. Its AUC is superior to DHI (0.728) and OD (0.628). Hence, we would say that the MAI which considers the area of the lumen in relation to the total area to be the best predictor. From the results of this study, we may hypothesise that the area indices account for the various contours of the retina while the other conventional indices assume a more polygonal pattern.

Our study had the advantage that all the MHs were treated by a uniform surgical technique and all measurements were done simultaneously using the same open access imaging software. The major limitations of the present study were its retrospective design and small sample size. We did not correlate the area indices with visual acuity outcomes. In our study, we had holes of larger sizes. This could be due to late presentation of subjects and longer duration of the MH. Also, in all the eyes, the surgery was performed using the conventional ILM peeling technique. Inverted flap peeling of the ILM was not attempted despite the presence of large macular holes. This could have affected the final anatomical closure outcome.

Conclusion

In conclusion, our study measured new parameters on OCT and derived 3 new area indices which correlated well with hole closure. The MAI especially seems to be single most important indicator in predicting anatomic success following macular hole surgery. Further studies are needed in this direction in future. Also, a new software can be formulated and incorporated with the latest OCT machines, so that the macular hole area/volume can be measured easily.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.