Value of Internal Limiting Membrane Peeling in Surgery for Idiopathic Macular Hole and the Correlation between Function and Retinal Morphology

Transcription

1 Acta Ophthalmologica Thesis Value of Internal Limiting Membrane Peeling in Surgery for Idiopathic Macular Hole and the Correlation between Function and Retinal Morphology Ulrik Correll Christensen, M.D. Department of Ophthalmology, Glostrup Hospital, University of Copenhagen, Denmark Copenhagen 2009

2 PhD Thesis Value of internal limiting membrane peeling in surgery for idiopathic macular hole and the correlation between function and retinal morphology Ulrik Correll Christensen, MD Department of Ophthalmology, Glostrup Hospital, Glostrup, Denmark ABSTRACT. Idiopathic macular hole is characterized by a full thickness anatomic defect in the foveal retina leading to loss of central vision, metamorphopsia and a central scotoma. Classic macular hole surgery consists of vitrectomy, posterior vitreous cortex separation and intraocular gas tamponade, but during the past decade focus has especially been on internal limiting membrane (ILM) peeling as adjuvant therapy for increasing closure rates. With increasing use of ILM peeling and indocyanine green (ICG) staining, which is used for specific visualization of the ILM, concerns about the safety of the procedure have arisen. At present, it is not known whether ICG-assisted ILM peeling potentially reduces the functional outcome after macular hole surgery. The purpose of the present PhD thesis was to examine whether ICG-assisted ILM peeling offers surgical and functional benefit in macular hole surgery. We conducted a randomized clinical trial including 78 pseudophakic patients with idiopathic macular hole stages 2 and 3. Patients were randomly assigned to macular hole surgery consisting of (i) vitrectomy alone without instrumental retinal surface contact (non-peeling), (ii) vitrectomy plus 0.05% isotonic ICGassisted ILM peeling or (iii) vitrectomy plus 0.15% trypan blue (TB)-assisted ILM peeling. Morphologic and functional outcomes were assessed 3, 6 and 12 months after surgery. The results show that surgery with ILM peeling, for both stages 2 and 3 macular holes, is associated with a significantly higher closure rate than surgery without ILM peeling (95% versus 45%). The overall functional results confirm that surgery for macular hole generally leads to favourable visual results, with two-thirds of eyes regaining reading vision ( 20 40). Macular hole surgery can be considered a safe procedure with a low incidence of sightthreatening adverse events; the retinal detachment rate was 2.2%. Visual outcomes in eyes with primary hole closure were not significantly different between the intervention groups; however, for the stage 2 subgroup with primary macular hole closure, there was a trend towards a better mean visual acuity in the non-peeling group (78.2 letters) compared to the ICG-peeling group (70.9 letters), p = Performing repeated macular hole surgery was associated with a significant reduction in functional outcome indicating that primary focus should be on closing the macular hole in one procedure. 1

3 Morphological studies of closed macular holes with contrast-enhanced optical coherence tomography (OCT) found thinning and discontinuity of the central photoreceptor layer matrix that were highly specific for predicting the likelihood of an eye having regained reading vision 12 months after macular hole surgery. Additionally, healing after macular hole surgery appeared to begin with the contraction of the inner aspect of the retina, forming a roof over a subfoveal fluid-filled cavity, and to end with a gradual restoration of the anatomy in the outer layers of the retina at the junction of the photoreceptor inner and outer segments. We found the more intact this structure was on contrast-enhanced OCT 3 months after macular hole surgery, the better the visual acuity after 12 months, whereas late rather than early resolution of subfoveal fluid had no impact on final visual outcome. The use ILM peeling and intraoperative dyes did not have any functionally important effects on postoperative macular structure. Based on the above findings, we conclude that ILM peeling should be performed in all cases of full thickness macular hole surgery. The use of 0.05% intraoperative isotonic ICG with short exposure time appears to be a safe alternative in stage 3 macular hole surgery, whereas a slight reduction in functional potential not can be excluded when performing 0.05% isotonic ICGassisted ILM peeling in stage 2 macular hole surgery. Key words: indocyanine green internal limiting membrane macular disease Macular hole surgery optical coherence tomography toxicity Trial registration number: NCT Acta Ophthalmol : 1 23 ª 2009 The Author Journal compilation ª 2009 Acta Ophthalmol doi: /j x Introduction Idiopathic macular hole is a retinal disease characterized by a full thickness defect of the foveal retina with associated visual acuity reduction, metamorphopsia and a central scotoma. Macular hole was considered an untreatable condition until 1991 where vitrectomy and intraocular gas tamponade was first introduced as an effective treatment (Kelly & Wendel 1991). The primary aim in macular hole surgery is to close the hole which is an absolute requirement for subsequent visual acuity gain. Since the initial reports on macular hole surgery, several intraoperative modifications have been tried in an attempt to enhance closure rates, and during the past decade, focus has especially been on internal limiting membrane (ILM) peeling as adjuvant therapy for inducing controlled gliosis and hole closure. With increasing use of ILM peeling and indocyanine green (ICG) staining, which is used for specific peroperative visualization of the ILM, concerns about the safety of the procedure have arisen. Especially, ICG has been accused of reducing the functional potential after macular hole closure, primarily as a result of cytotoxic effects on the retinal pigment epithelium and retinal ganglion cells (Jackson 2005). At present, the scientific evidence for the beneficial effects of ICGassisted-ILM peeling on hole closure and final functional outcome is poor, and no well-conducted randomized clinical trials have examined the value and safety of ICG-assisted-ILM peeling in idiopathic macular hole surgery. Hypotheses and aims The primary aim of this thesis was to examine the clinical effect and possible negative influence of ICG-assisted ILM peeling on retinal function and morphology in stages 2 and 3 idiopathic macular hole surgery. Study hypotheses were as follows: (1) Surgery for macular hole with ILM peeling results in a better anatomic outcome than surgery without ILM peeling. (2) Patients with primary hole closure after macular hole surgery without ILM peeling will have the best functional outcome. (3) Structural changes in outer macular layers, as assessed with contrast-enhanced optical coherence tomography, are correlated with final visual outcome in closed macular holes and are associated with the use of ICG-assisted ILM peeling. We conducted a randomized clinical trial to answer the following questions: (1) Does macular hole surgery with ILM peeling induce higher closure rates than macular hole surgery without retinal surface contact? (Paper I) (2) Does macular hole surgery without staining and peeling lead to a more favourable functional outcome than macular hole surgery with ICGassisted ILM peeling in eyes with primary macular hole closure? (Paper I) (3) Is macular hole surgery with ICGassisted ILM peeling associated with an increased risk of per- and postoperative complications? (Paper I) 2

4 Additionally, we wanted to develop methods based on optical coherence tomography for improved visualization and quantification of postoperative photoreceptor layer structure (Paper II) to be able to answer the following questions: (4) Do ILM peeling and ICG-staining have any specific effects on photoreceptor layer integrity in closed macular holes and are postoperative photoreceptor layer changes correlated with final visual outcome? (Paper III) (5) What is the prognostic significance for final visual function of structural changes in outer macular layers 3 months after surgical closure? (Paper IV) Background vitreous cortex and or anterior posterior traction from persistent vitreofoveal attachment acting on the foveal umbo. This may lead to disruption of the cone-shaped zone of Mu ller cells that makes out the primary structural support in the central and inner part of the fovea centralis, Fig. 1 (Gass 1999; la Cour & Friis 2002). Macular hole surgery Surgical treatment for macular hole was not available until 1991 where Kelly and Wendel developed classic macular hole surgery consisting of pars plana vitrectomy, removal of adherent cortical vitreous, peeling of epiretinal membranes, intraocular gas tamponade and several days of face-down positioning. From randomized clinical trials, we now know that surgical treatment is indicated for full thickness macular holes (stages 2, 3 and 4), whereas stage 1 holes should be handled conservatively as many resolve spontaneously (de Bustros 1994; Ezra & Gregor 2004). In the original report of 52 eyes with mainly stage 3 macular holes, Kelly and Wendel reported a primary closure rate of 58%. Six months Macular hole Macular hole is a retinal disease characterized by a full thickness anatomic opening in the fovea centralis leading to loss of central vision, metamorphopsia and a central scotoma in the affected eye. The majority of macular holes are primary idiopathic (85%) with a smaller proportion being as a result of trauma, inflammation or high myopia (Ezra 2001; la Cour & Friis 2002). Idiopathic macular holes are estimated to affect approximately people in Denmark per year (la Cour & Friis 2002; Moss et al. 2006) with a three to one preponderance of females. Patients are usually in their 6th or 7th decade of life. (la Cour & Friis 2002). The pathogenesis of idiopathic macular holes is still incompletely understood, but based on work by Gass (Gass 1988, 1995) in 1988, macular holes are thought to develop through four clinical stages; from an impending or occult lesion (stage 1) to a fully developed macular hole involving all retinal layers (stages 2 4). Data provided from optical coherence tomography (OCT) of impending macular holes have indicated that the first changes in macular hole formation is an intraretinal split in the macula evolving into an intraretinal cyst (Gaudric et al. 1999). These observations have led to the opinion that the major pathogenic mechanisms in macular hole formation are tangential traction from contraction of prefoveal Fig. 1. Schematic- and OCT representation of presumed pathogenic mechanisms in macular hole formation. Anterior to posterior traction from persistent vitreo-foveal attachment and or tangential traction from contraction of prefoveal vitreous cortex leads to an impending (stage 1) macular hole with an intraretinal split or a foveolar detachment; approximately 60% of these eyes undergo spontaneous vitreo-foveal separation and resolution, 40% progress (de Bustros 1994). Stage 2: full thickness macular hole with an aperture diameter less than 400 lm. Stage 3: full thickness macular hole with an aperture diameter larger than 400 lm; 20 40% of eyes with stage 2 or 3 holes progress to stage 4 where complete posterior vitreous detachment (PVD) have occurred (la Cour & Friis 2002). Modified with permission from American Medical Association (Gaudric et al. 1999) and American Journal of Ophthalmology (Gass 1995). 3

5 after surgery, the visual acuity had improved by 2 or more lines in 73% of the eyes with closed holes (Kelly & Wendel 1991). At present, surgical management can achieve anatomic closure rates better than 90% when surgery is performed with adjuvant therapy such as ILM peeling (Mester & Kuhn 2000; Kwok et al. 2005; Tognetto et al. 2006). Functional outcomes are more difficult to predict, and despite the high anatomic closure rates, reading vision remains compromised in 30 40% of patients (Christensen et al. 2008). A visual acuity gain of more than 2 lines however can be expected in 60 85% of patients (Mester & Kuhn 2000; Tognetto et al. 2006). The reason for this discrepancy between function and morphology is unclear, but studies of closed macular holes with optical coherence tomography have reported changes in outer retinal layers (subfoveal cysts, photoreceptor defects) as possible explanations of compromised function after anatomically successful macular hole surgery (Kitaya et al. 2004; Ko et al. 2005, 2006; Moshfeghi et al. 2005; Villate et al. 2005; Chang et al. 2008). The outcome of macular hole surgery is thought to depend mainly on pre-operative macular hole size and duration of symptoms (Freeman et al. 1997), but outcomes have also been reported to depend on the used surgical technique (ILM-peeling, ICG-staining) (Smiddy et al. 2001; Sheidow et al. 2003; Al-Abdulla et al. 2004). The most common complication seen after macular hole surgery is cataract formation which will develop in more than 80% of phakic eyes within 2 years postoperatively (Thompson et al. 1995). Intraoperative retinal tears induced during posterior vitreous cortex separation occur in approximately 5% of the eyes, and recent series report a retinal detachment rate of 2 5% (Kwok et al. 2005; Tognetto et al. 2006; Christensen et al. 2008). Postoperative visual field defects, which previously have been attributed to dehydration of the retina during fluid air exchange, now seems to be a complication with low incidence (Gass et al. 2001; Christensen et al. 2008). Small asymptomatic paracentral scotoma and macular retinal pigment epithelium alterations seen postoperatively have been related to intraoperative retinal light toxicity and surgery with ILM peeling (Christensen et al. 2008; Haritoglou et al. 2001a,b). The primary goal of idiopathic macular hole surgery is to induce hole closure which is an absolute requirement for visual acuity improvement. The rationale for surgical management as originally described by Kelly and Wendel is mobilization of hole edges by removing tangential and anterior posterior vitreous traction, activation of marginal glia-cells by vitrectomy and epiretinal membrane peeling, and finally immobilization and apposition of hole edges by intraocular gas tamponade and face-down positioning. Since the initial report by Kelly and Wendel numerous adjuvant surgical techniques focusing on any of these general surgical principles have been tried in an attempt to enhance closure rates, but at present no convincing evidence exists to support the supposition that any of these improvements will result in a better functional outcome. Adjuvant surgical techniques for the treatment of idiopathic macular hole Growth factors Wound healing adjuvants have been placed on the macular hole after fluid gas exchange with the hope that it would promote glial repair and hole closure by stimulating fibroblast proliferation and collagen synthesis. Adjuvants investigated include autologous serum (Banker et al. 1999; Ezra & Gregor 2004); thrombin (Blumenkranz et al. 2001); autologous platelet concentrate (APC) (Paques et al. 1999) and transforming growth factor beta 2 (TGFb 2 ) (Smiddy et al. 1993; Thompson et al. 1998). Randomized trials have evaluated the effect of autologous serum (Ezra & Gregor 2004); APC (Paques et al. 1999) and TGFb 2 (Thompson et al. 1998), but no compelling evidence suggests that the use of any of these adjuvants improves success rates. Laser photocoagulation Laser photocoagulation applied to the retinal pigment epithelium (RPE) at the centre of the macular hole has been performed as adjuvant therapy to promote closure of large (Cho et al. 2006); persistent (Ikuno et al. 1998) or reopened macular holes (Ohana & Blumenkranz 1998). This technique was primarily used before the era of ILM peeling to stimulate the release of cytokines, such as TGFb 2, from the RPE thus promoting macular hole closure (Matsumoto et al. 1994). Retrospective studies report closure rates of up to 90% and no significant detrimental effects on visual outcomes (Ikuno et al. 1998; Ohana & Blumenkranz 1998; Min et al. 1999; Cho et al. 2006); but long-time followup studies have never been performed. One small randomized clinical trial comparing the effect of central laser as adjuvant therapy in macular hole surgery found no beneficial or detrimental effects of laser treatment compared to ILM peeling (Cho et al. 2006). Intraocular tamponade and face-down positioning Intraocular tamponade with gases or silicone oil is performed to facilitate reapposition of hole edges by providing buoyant forces that pushes the retina against the underlying RPE. The forces is greatest at the apex of the arc of contact, and by sustaining an accurate face-down position, the maximum vector forces can be directed against the macular hole. Favourable closure rates (>90%) and functional outcomes (60 85% two lines gain) have been achieved by long-acting gas tamponade (C 3 F 8, C 2 F 6 ) and strict face-down positioning for 1 2 weeks (Mester & Kuhn 2000; Kwok et al. 2005; Tognetto et al. 2006; Christensen et al. 2008). Similar results (>90% closure rate) have been reported with the use of short-acting gases (SF 6 ) (Simcock & Scalia 2001; Kim et al. 2008) or air (Park et al. 1999; Hasler & Prunte 2008) and only 2 4 days of prone positioning. At present, it is not known whether decreasing or eliminating face-down positioning will affect the surgical outcome, but the general trend is towards shorter lasting tamponade and shorter duration of face-down positioning. No randomized trials exist. The buoyant forces of gas bubbles are approximately ten times greater than that of silicone oil, indicating that the primary effect of intraocular silicone oil probably is to help intraretinal dehydration by preventing passage of liquid vitreous through the hole (Tornambe 2003). Anatomic outcomes with silicone oil tamponade are 4

6 usually good (>80%) (Goldbaum et al. 1998; Karia et al. 2001), but visual results appear to be worse with silicone oil (38 72% one line gain) (Goldbaum et al. 1998; Karia et al. 2001; Lai et al. 2003) than after intraocular tamponade with gas. Adverse reactions related to the use of silicone oil tamponade are possible toxic damage to the optic nerve (Budde et al. 2001; Papp et al. 2007) and accelerated development of cataract, excessive intraocular inflammation, glaucoma and keratopathy (Gallemore & Mccuen 2006). Additionally, the use of silicone oil tamponade mandates a second operation for oil removal. For this reason, silicone oil has no place in primary macular hole surgery. It might be considered in selected cases where patients are unable to maintain the postoperative prone position and where primary surgery with complete gas tamponade in a pseudophakic eye has failed (Karia et al. 2001). Internal limiting membrane peeling (ILM peeling) The internal limiting membrane represents the structural boundary between the retina and the vitreous. It is the basal lamina of the Mu ller cells and is formed by projections from the Mu l- ler cell footplates. The ILM is translucent and 1.5 lm thick in the peripheral foveal area (Yamada 1969). Surgical removal of the ILM in macular hole surgery was first performed in 1996 in an attempt to ensure complete removal of all tangential tractional components (glial cells, macrophages, fibrocytes, myofibroblasts) involved in macular hole formation (Yooh et al. 1996). By removing the ILM, it was believed that the scaffold, where upon the cellular proliferation occurred, was removed. Despite lack of scientific evidence for its beneficial effects on anatomical and functional outcomes, the procedure has been adopted by most vitreoretinal surgeons as a supplementary treatment in macular hole surgery (Schaal et al. 2006). Based on case series, primary anatomic closure rates of 60 90% without ILM peeling (Kelly & Wendel 1991; Brooks 2000; Mester & Kuhn 2000; Ezra & Gregor 2004; Tognetto et al. 2006) and % with ILM peeling (Brooks 2000; Mester & Kuhn 2000; Tognetto et al. 2006) have been reported. Only two randomized trials comparing outcomes with or without ILM peeling have been published, one of which is included in this PhD thesis (Kwok et al. 2005; Christensen et al. 2008). Our study reports a primary closure rate of 45% without peeling of any kind and 95% with ILM peeling (Christensen et al. 2008). While ILM peeling may improve anatomic closure rates, its effect on functional outcomes after macular hole surgery is even more controversial with some case studies reporting loss of functional potential after surgery with ILM peeling (Smiddy et al. 2001; Al-Abdulla et al. 2004); others reporting visual outcomes to be better after ILM peeling (Brooks 2000; Mester & Kuhn 2000; Sheidow et al. 2003) and yet others not being able to demonstrate any differences between ILM peeled on non-peeled eyes (Margherio et al. 2000). Possible mechanisms of retinal damage related to ILM peeling Iatrogenic retinal trauma Scientific evidence suggests that ILM peeling can result in mechanical damage to the inner retinal layers. One study reported a selective delay in the recovery of the focal macular electroretinogram b-wave after ILM removal indicating alterations in the physiology of the Mu ller cells that are involved in this response (Terasaki et al. 2001). Other studies have found small nerve fibre layer irregularities in approximately 50% of eyes (Ito et al. 2005; Christensen et al. 2008); small asymptomatic paracentral scotoma (Haritoglou et al. 2001; Beutel et al. 2007) and retinal micro-haemorrhages (Karacorlu et al. 2003) in all patients after ILM peeling. However, none of the studies have been able to correlate these findings with a significant reduction in visual acuity. Dye-associated retinal toxicity Because of its thin and translucent nature, ILM peeling is difficult to perform in a safe and reproducible fashion without the use of a staining agent. Dye-assisted ILM peeling therefore was introduced to reduce operating time and limit the risk of iatrogenic mechanical damage to the retina (Kadonosono et al. 2000) The most commonly used vital stains are indocyanine green (ICG), infracyanine green (IfCG), trypan blue (TB), brilliant blue G (BBG) and triamcinolone acetonoide (TA). Indocyanine green (ICG) is a hydrophilic dye used for angiography because of its properties as a fluorophore. Indocyanine green also acts as a chromophore staining the ILM green because of its affinity to laminin and collagen type IV within the ILM (Weinberger et al. 2002). It is the most effective and specific ILM stain used in macular hole surgery (Stanescu- Segall & Jackson 2008). Since the introduction of ICG to stain the ILM in 2000, multiple case series have discussed potential adverse effects and thus reduction of functional potential related to its use (Weinberger et al. 2001; Al-Abdulla et al. 2004; Ando et al. 2004; Horio & Horiguchi 2004; Ferencz et al. 2006). These reports describe various concentrations, exposure times, osmolarities and volumes of ICG, making it difficult to establish a causal link between the use of ICG and functional damage. The most commonly reported adverse events after ICG-assisted ILM peeling are visual field defects (Poliner & Tornambe 1992; Haritoglou et al. 2001a,b; Kanda et al. 2004; Tsuiki et al. 2007) and retinal pigment epithelium (RPE)- or ganglion cell changes (Sippy et al. 2001; Enaida et al. 2002; Engelbrecht et al. 2002; Gandorfer et al. 2003; Ho et al. 2003; Iriyama et al. 2004; Jackson et al. 2004; Goldstein et al. 2006; Yip et al. 2006). Several mechanisms related to ICGinduced ocular toxicity have been proposed, including (i) a direct dosedependent biochemical injury to RPE cells and ganglion cells (Sippy et al. 2001; Enaida et al. 2002; Ho et al. 2003; Iriyama et al. 2004; Goldstein et al. 2006; Nomoto et al. 2008); (ii) osmolarity effects of the ICG-solution at the vitreo-macular interface, indicating that hypotonic ICG solutions could be harmful to the RPE (Stalmans et al. 2002; Jackson et al. 2004); and (iii) phototoxic properties of ICG inducing photo-oxidative cell damage caused by an overlap in absorption spectra between ICG (peak nm) and different types of endoillumination used in vitreoretinal surgery (Jackson et al. 2004; Yip et al. 2006). 5

7 Cell culture studies have tried to identify a safe preparation of ICG for the use in macular surgery, and an isotonic preparation of 0.05% ICG has been found to non-toxic to RPE cells in vitro, even after prolonged exposure (Jackson et al. 2004; Kiilgaard et al. 2006). The clinical safety of ICG staining in macular hole surgery has only been examined in one randomized clinical trial (Beutel et al. 2007) besides the one presented in this thesis (Christensen et al. 2008). The study by Beutel et al. compared ILM peeling with 0.05% ICG staining versus 0.15% trypan blue staining and found a larger proportion of central scotoma in the ICG group, but no significant difference in visual outcome (Beutel et al. 2007). Infracyanine green (IfCG) contains the same fluorophore as ICG, but it differs from ICG in not containing iodine and by being prepared in 5% glucose making the standard preparation isotonic. Many of the issues related to toxicity runs in parallel between ICG and IfCG, but the primary argument for using IfCG instead of ICG has been related to avoiding hypotonic ICG solutions which could aggravate toxicity (Stalmans et al. 2002; Jackson et al. 2004). The possible safety benefit from the isotonic IfCG solution seems to be diminished now that many surgeons prepare ICG in an isotonic solution. Trypan blue (TB) is a vital stain used primarily in surgery for removal of epiretinal membranes, but it is also an alternative to indocyanine green in ILM peeling. TB stains cells with damaged cell membranes in a dosedependent matter (Veckeneer et al. 2001), and studies propose that the faint bluish staining seen after TB exposure is probably because of affinity to cellular proliferations on the ILM, rather than to the acellular ILM (Meyer et al. 2004). Some studies have reported on possible TB toxicity in RPE cell cultures in higher concentrations, (Kwok et al. 2004; Rezai et al. 2004), but clinical and experimental studies suggest that TB staining is unlikely to have toxic side-effects when used in low and clinically relevant concentrations (<0.15%) (Haritoglou et al. 2004; Narayanan et al. 2005; Beutel et al. 2007). Brilliant blue G (BBG) is a blue biostain that provides selective staining of the ILM. The high staining ability for ILM makes BBG an alternative option for indocyanine green and infracyanine green. The dye is relatively new (approved in Europe in 2006), and so far only limited toxicity data from clinical and experimental studies exist (Enaida et al. 2006a,b; E- naida & Ishibashi 2008; Remy et al. 2008). So far no retinal toxicity or adverse effects related to the dye in clinical concentrations have been reported, but long-term safety reports of this novel dye waits. Triamcinolone acetonide (TA) is a synthetic insoluble corticosteroid which has been used as an alternative to visualize the posterior vitreous cortex, epiretinal membranes and the ILM (Peyman et al. 2000; Horio et al. 2005; Shah et al. 2005). After intravitreal application, TA crystals deposit onto epiretinal membranes and the ILM and facilitate their identification. The safety profile of TA points in different directions. TA was found to be toxic to cultured RPE cells (Narayanan et al. 2006). RPE toxicity was also found in a rabbit model (Yu et al. 2006), but another study in rabbits did not detect any retinal toxicity at even high concentrations (up to 30 mg) (Ruiz-Moreno et al. 2007). A possible toxic effect has been related to the preservative (benzyl alcohol) present in the TA solvent (Morrison et al. 2006). No randomized trials exist, but some case studies have reported similar outcomes after TA-assisted ILM peeling compared to ICG-assisted ILM peeling (Karacorlu et al. 2005; Nomoto et al. 2008). Phototoxic retinal damage Light damage to the retina can occur through three possible mechanisms: photo-thermal effects, photomechanical effects and photochemical effects (Glickman 2002). Only photochemical damage may be produced by exposure to non-laser sources like the light source from endoillumination during vitrectomy or an operating microscope; however, photo-thermal damage to the retina may occur if the light probe comes in direct contact with the retinal surface (van den Biesen et al. 2000; Glickman 2002). Photochemical or photo-oxidative damage results when incident light interacts with a chromophore in the ocular tissue. Endogenous chromophores excitable by visible wavelengths ( nm) of light are the photoreceptor visual pigments, and melanin and lipofuscin of the RPE (Glickman 2002). Indocyanine green can act as an exogenous chromophore following intravitreal use. Upon excitation of the chromophore, reactive oxygen species are generated, which results in lipid peroxidation and destruction of cell membranes (Glickman 2002). The risk of inducing photo-oxidative damage depends on the effective energy delivered onto the retina. Thus, increased output power and shorter distance between the tip of the fibreoptic and the retina (leading to focal light) increases the risk of retinal phototoxicity (van den Biesen et al. 2000; Yanagi et al. 2006). Optical coherence tomography of macular holes Optical coherence tomography (OCT) is a non-invasive diagnostic imaging modality that enables in-vivo cross-sectional or three-dimensional visualization of the retinal microstructure. The technique is based on interferometry and measures the echo time delay of back-reflected light from the tissue. The resulting OCT image, which is analogous to ultrasound B-scans, relies on intrinsic differences in tissue optical properties to produce image contrast (Drexler & Fujimoto 2008). The light source in commercial OCT instruments is a superluminescent diode (SLD) which is a combination of a laser- and a light diode. First generation OCT systems, timedomain OCT (Zeiss Stratus OCT3), detect the echo delays of light sequentially as a function of time and can produce an image with an axial or depth resolution of approximately 10 lm (Drexler & Fujimoto 2008). Newest generation OCT systems, spectral-domain OCT (Zeiss Cirrus HD-OCT, Topcon OCT, Heidelberg Spectralis OCT), became commercially available in Spectral-domain OCT measures the entire optical echo signal simultaneously allowing an increase in scan rate of a factor 100 enabling A-scans per second. This makes three-dimensional OCT imaging possible with an axial resolution of 5 8 lm (Drexler & Fujimoto 2008). An even higher axial 6

8 resolution (2 3 lm) has been achieved using broader bandwidth femtosecond lasers instead of traditional SLD light sources, ultra-high resolution OCT. Ultra-high resolution OCT systems however are not commercially available. Transverse image resolution in all mentioned systems has been limited by ocular aberrations to approximately 15 lm (Drexler & Fujimoto 2008). As described the OCT technology has developed rapidly, and OCT is currently the gold standard imaging technique for diagnosing and followup of retinal diseases including macular hole. OCT has contributed to our understanding of the pathogenesis of macular hole by establishing the role of anterior posterior and tangential forces (see Fig. 1). Preoperative OCT features have been used to predict the outcome of surgery, finding that holes smaller than 400 lm have the best anatomic and visual prognosis (Ip et al. 2002; Kusuhara et al. 2004; Christensen et al. 2008). OCT is also useful in the postoperative evaluation of macular hole to verify complete hole closure and examine alterations in foveal structure responsible for compromised function. Current management of macular hole in Denmark To determine how macular hole surgery is currently being performed in Denmark, we developed a questionnaire concerning the different aspects of full thickness macular hole surgery (stages 2 4). In January 2007, the questionnaire was sent to surgeons at eye departments in Denmark thought to perform macular hole surgery. Questionnaires were sent to 19 vitreoretinal surgeons at ten eye departments. Results are based on the 13 of 19 (68.4%) returned questionnaires (Table 1). The results show good agreement on the general surgical principles. All surgeons (12 of 12) perform a vitrectomy with posterior vitreous cortex separation, peeling of visible epiretinal membranes, fluid air exchange and intravitreal tamponade. Internal limiting membrane (ILM) peeling is being performed by 9 of 12 surgeons (75%) and always with the aid of a vital dye to enhance visualization. Indocyanine green (ICG) is the preferred dye used Table 1. Strategies for full-thickness macular hole surgery in Denmark in Variable Eye departments performing MHS (Zealand 3, Funen 1, Jutland 4) 8 Surgeons performing MHS 12 Type of anaesthesia Local 7 (58%) General 5 (42%) General surgical method Pars plana vitrectomy + PVD 12 (100%) + ERM peeling (no ILM peeling) 3 (25%) + ILM peeling + ERM peeling 9 (75%) Dye for ILM staining* ICG 8 (89%) TB 2 (22%) IfCG 0 (0%) Triamcinolone 2 (22%) None 0 (0%) ICG concentration 0.05% 4 (50%) 0.1% 2 (25%) 0.2% 1 (12.5%) 0.25% 1 (12.5%) ICG exposure time 30 seconds 5 (62.5%) >30 seconds 3 (37.5%) Fluid-air exchange Yes 12 (100%) No 0 (0%) Growth-factor adjuvants Yes 0 (0%) No 12 (100%) Endotamponade Gas 12 (100%) Silicone oil 2 (17%) Type of gas à C 3 F 8 10 (83%) C 2 F 6 1 (8%) SF 6 2 (17%) Postoperative face-down None 1 (8%) 5 days 6 (50%) 7 days 3 (25%) 10 days 2 (17%) Cataract extraction performed Prior to MHS 3 (25%) Combined with MHS 3 (25%) When cataract develops 6 (50%) Abbreviations: MHS = macular hole surgery; PVD = posterior vitreous cortex detachment; ILM = internal limiting membrane; ERM = epiretinal membrane; ICG = indocyanine green; TB = trypan blue; IfCG = infracyanine green. *9 surgeons perform ILM peeling; three surgeons regularly use two different types of dyes. 2 surgeons regularly use two different types of tamponade. à 2 surgeons regularly use two different types of gas. regularly by 89% of ILM-peeling surgeons in low concentrations (between 0.05% and 0.25%). The majority of surgeons (62.5%) use ICG with a very short exposure time (<30 seconds) on the retinal surface. We did not collect information about the ICG solution osmolarity or how it was applied. Endo-tamponade with gas is performed by all surgeons (100%) and long-acting C 3 F 8 is the most commonly No. used (83%) gas. Silicone oil is used by 2 of 12 (17%) in some cases. Seventyfive per cent recommend face-down positioning for 5 7 days postoperatively. There is no general consensus on how to handle development of postoperative cataract; 25% perform cataract surgery prior to macular hole surgery, 25% perform combined phaco vitrectomy and 50% wait until cataract develops. 7

9 Materials and Methods To assess the value and clinical safety of 0.05% isotonic ICG-assisted ILM peeling in stages 2 and 3 idiopathic macular hole surgery, we designed a randomized clinical trial. Copenhagen Macular Hole (COMAH) study The COMAH study is a randomized clinical trial with 12 months of followup comparing morphological and functional outcomes after primary idiopathic macular hole surgery with or without ILM peeling. Participants Enrolment was conducted between March 15, 2005 and February 1, During this period, 173 patients were referred to the Department of Ophthalmology at Glostrup Hospital, University of Copenhagen for macular hole surgery. Patients were examined to confirm the diagnosis of full thickness stage 2 or 3 idiopathic macular hole according to the classification by Gass (Gass 1988, 1995) and were considered for inclusion according to the criteria given in Table 2. Eighty-nine eligible patients were scheduled for cataract surgery with subsequent macular hole surgery after 4 6 weeks. At the day of baseline examinations, the day before macular hole surgery, 78 patients still fulfilled the inclusion criteria and were subsequently included in the study for randomized surgery. Approval was obtained from the local Committee on Biomedical Table 2. COMAH study inclusion and exclusion criteria. Inclusion criteria Idiopathic macular hole stage 2 or 3 Symptom duration 12 months Visual acuity 34 ETDRS letters Intraocular pressure 23 mmhg Informed consent Exclusion criteria Ophthalmoscopically visible epiretinal fibrosis at the baseline examination Previous ocular surgery (except cataract surgery) Any eye disease affecting visual function Systemic disease affection retinal function, incl. diabetic retinopathy Abbreviation: ETDRS: Early Treatment of Diabetic Retinopathy Study Chart, letters read at 4 m. Research Ethics, and informed consent according to the tenets of the Declaration of Helsinki was obtained from all subjects before entering the study. The study adheres to the Consolidated Standards of Reporting Trials (CONSORT) statement and was registered in the clinicaltrials.gov database (NCT : Trial of Prognostic Factors and Surgical Methods for the Treatment of Idiopathic Macular Holes). Interventions The trial was initially designed as a two-arm interventional study comparing outcomes after vitrectomy alone (non-peeling) versus vitrectomy plus ICG-assisted ILM peeling for stages 2 and 3 idiopathic macular hole without any visible epiretinal fibrosis. However, a planned interim analysis made after enrolling 40 patients showed significant inferiority of anatomical and functional outcome in unpeeled stage 3 patients, and we concluded that continuing the study with patients randomized 1:1 to peeling versus nonpeeling was unethical. Following this, non-peeling of stage 3 eyes was terminated, stage 3 holes instead being randomized to either 0.15% trypan blue (TB)-assisted ILM peeling or 0.05% ICG-assisted ILM peeling. In this case, trypan blue should serve as the presumed non-toxic control dye. Thus, the surgical intervention for macular hole consisted of: (1) vitrectomy alone without retinal surface contact (non-peeling) (2) vitrectomy plus 0.05% isotonic ICG-assisted ILM peeling (3) vitrectomy plus 0.15% TB-assisted ILM peeling A standard 3-port pars plana vitrectomy included verified separation of the posterior vitreous cortex with direct aspiration over the posterior pole. A careful examination of the retinal periphery was then performed after indentation, and any iatrogenic retinal breaks found were treated by laser- or cryopexy. No touch of the retinal surface was performed in the non-peeling group. In the ICGassisted ILM peeling group, 25 mg ml ICG (ICG pulsion, PULSION Medical Systems, Mu nich, Germany) was dissolved in 5 ml isotonic glucose and diluted with 45 ml balanced saline solution (BSS) to obtain a final isotonic solution of 0.05% (0.5 mg ml) ICG. In the TB-peeling group, 0.1 ml of 0.15% TB (membrane blue, DORC international, Zuitland, the Netherlands) was used. The ICG solution was applied in a BSS filled eye, and TB was applied in an air filled eye. Both agents were washed out after an exposure time of 15 seconds. In cases with insufficient staining after TB exposure, a second application of 15 seconds was allowed. Peeling of the ILM was performed over a circular area centred of the macular hole and of a diameter of approximately two optic disc diameters. Finally, a complete fluid air exchange and intravitreal tamponade with 10 15% perflouropropane (C 3 F 8 ) was performed. Then, patients were instructed to maintain a face-down position for at least 10 hr per day for 5 days. Patients were provided a registration sheet for help and documentation of face-down time. Postoperatively a combination of dexamethasone- and chloramphenicol eye drops was administered 4 times daily for 3 weeks. In eyes without macular hole closure after the primary surgery, repeated macular hole surgery was offered as soon a possible with additional 0.05% ICG-assisted ILM peeling and gas tamponade. Phacoemulsification cataract or clear lens extraction was performed in all phakic study eyes 4 6 weeks prior to macular hole surgery to avoid postoperative cataract and postoperative cataract surgery on a vitrectomized eye. After cataract surgery anti-inflammatory treatment with dexamethasone eye drops were administered 6 times daily for 2 weeks and then tapered slowly with one drop per week until macular hole surgery. Main outcome measures Closure of the macular hole 3 and 12 months after surgery (primary endpoint) and best-corrected ETDRS visual acuity (BCVA) after 12 months (secondary endpoint). Other key outcome measures were visual field defects, retinal nerve fibre layer defects, retinal pigment epitheliopathy, per- and postoperative recorded adverse events and postoperative alterations in macular structure as assessed with optical coherence tomography. 8

10 Sample size calculation Sample size calculations were performed based on the initial two-arm study design comparing macular hole surgery with or without ILM peeling. A standard power calculation was performed to calculate sample sizes based on a priori assumptions of an anatomic success of 85% after vitrectomy alone and of 100% after vitrectomy plus ICG-assisted ILM peeling. With a power of 90%, it was calculated that a sample size of 40 eyes per surgical arm (n = 80) was required to detect a difference at the 5% significance level. Masking Investigators and patients were masked to intervention, and only the surgeons were aware of the surgical method being used. All examinations through baseline and follow-up visits were performed by to examiners (Ulrik Christensen, Kristian Krøyer) who were masked to the allocation status; however, assessment of 6 and 12 months BCVA was performed by a certified masked person (Birgit Sander) in case that accidental demasking had occurred during the fundoscopicor photographic evaluations 3 months postoperatively. Postoperative evaluations of visual fields, fundus photographs, fundus autofluorescence images and OCT scans were also performed by a masked examiner (Ulrik Christensen) without prior knowledge of the corresponding functional outcome and before breaking the study randomization code. Functional outcome assessment Preoperative examinations were performed the day before macular hole surgery, and follow-up was scheduled at 3, 6, and 12 months after macular hole surgery. All patient contacts included a standard ophthalmologic evaluation with indirect ophthalmoscopy, applanation tonometry, fundus photography and automated perimetry. Best-corrected visual acuity testing (BCVA) was performed using Early Treatment of Diabetic Retinopathy Study (ETDRS) charts with an initial testing distance of 4 m. The ETDRS chart is a logmar chart where ETDRS letters correspond to a log- MAR value of 0.0 and a Snellen ratio of (Early Treatment Diabetic Retinopathy Study Research Group 1987). Visual field testing was performed before pupil dilatation using 30-2 automated perimetry (Humphrey Field Analyzer 750, Humphrey Instruments, San Leandro, CA), and registration of fundus autofluorescence for assessment of retinal pigment epitheliopathy was performed using the Heidelberg Retina Angiograph II (Heidelberg Engineering GmbH, Heidelberg, Germany). Morphological outcome assessment with OCT Preoperatively the macular holes were classified according to Gass (Gass 1988, 1995). Hole size was recorded using one of six radial OCT scans where the hole presented with the largest diameter by measuring its aperture diameter from where the distance across the full-thickness defect was the shortest (Tadayoni et al. 2006) and its base diameter from the aspect of the hole right in front of the retinal pigment epithelium. Sizes were recorded using the build-in caliper of the OCT software. Three, six and twelve months after surgery, macular hole closure was verified by OCT and defined as complete adaptation of the hole margins. OCT examinations were performed using a commercial instrument (Stratus model 3000; Carl Zeiss Meditec Inc., Humphrey Division, Dublin, CA, USA) by recording of 6 radial transfoveal scans. Additionally, 20 nominally identical vertical transfoveal linear B-scans were made. The best 8 out of these 20 tomograms in terms of image quality and consistency of findings were selected and subjected to an automated digital alignment and averaging to obtain a contrast-enhanced OCT image with reduced speckle noise and better definition of the outer layers of the retina (Sander et al. 2005; Jorgensen et al. 2007; Christensen et al. 2008). OCT scanning was performed after pupil dilatation using fixation on an internal target to define the centre of the fovea. Quantitative analyses of postoperative photoreceptor layer structure (paper III and IV) as well as quantitative analyses of macular structure in normal eyes (paper II) were performed on contrast-enhanced OCT images, using a semi-automatic windows-based image processing software that we developed in C++ Builder (Paper II). Additionally, characterization of foveal photoreceptor layer integrity (complete or incomplete photoreceptor inner outer segment (IS OS) junction) was performed by measuring the diameter of loss of homogeneous reflectivity from the IS OS line in the vertical direction on the contrast-enhanced OCT images. This measure was obtained by applying the measure tool from Adobe Photoshop (version 7.0) to the contrast-enhanced OCT scan consisting of 1024 pixels over a depth of 2 mm and 512 pixels over a nominal width of 6 mm (Paper III and IV). Results and Discussion Paper I: Value of internal limiting membrane peeling in surgery for idiopathic macular hole stage 2 and 3: a randomized clinical trial Aim To determine the effect of internal limiting membrane (ILM) peeling on anatomical and functional success rates in stages 2 and 3 idiopathic macular hole surgery. Results (Tables 3 5) Seventy-eight eyes (of 75 patients) with stages 2 and 3 idiopathic macular hole and without any visible epiretinal fibrosis were randomly assigned to (i) vitrectomy alone without retinal surface manipulation, (ii) vitrectomy plus 0.05% isotonic ICG-assisted ILM peeling or (iii) vitrectomy plus 0.15% TB-assisted ILM peeling. Primary anatomic macular hole closure rates were significantly higher with ILM peeling than without ILM peeling for both stage 2 holes (ICGpeeling 100%, non-peeling 55%, p = 0.014) and for stage 3 holes (ICG-peeling 91%, TB-peeling 89%, non-peeling 36%, p < 0.001). Sixteen of 18 patients without primary macular hole closure underwent repeated macular hole surgery with additional ICG-assisted ILM peeling resulting in a final anatomic success rate at 12 months of 24 of 25 (96%) in the non-peeling group, 33 of 34 (97%) in the ICG-peeling group and 18 of 18 (100%) in the TB-peeling group, p > None of the closed holes reopened during the 12 months follow-up period. 9

11 Table 3. Anatomical outcome. Non-peeling group (n = 25) ICG-peeling group (n = 34) TB-peeling group (n = 18) p Value All (n = 77) Primary anatomical success at 3 months (%) 11 of 25 (44) 32 of 34 (94) 16 of 18 (89) <0.001* Final anatomical success at 12 months (%) 24 of 25 (96) 33 of 34 (97) 18 of 18 (100) 1.0* Stage 2 (n = 23) Primary anatomical success at 3 months (%) 6 of 11 (55) 12 of 12 (100) Final anatomical success at 12 months (%) 11 of 11 (100) 12 of 12 (100) 1.0 Stage 3 (n = 54) Primary anatomical success at 3 months (%) 5 of 14 (36) 20 of 22 (91) 16 of 18 (89) <0.01 Final anatomical success at 12 months (%) 13 of 14 (93) 21 of 22 (96) 18 of 18 (100) 0.55 *Statistical analysis (x 2 ) was only performed for non-peeling versus indocyanine green (ICG) peeling. Multiple comparison analyses between the three groups found a significant difference between non-peeling and ICG peeling (p < 0.001), and between non-peeling and trypan blue (TB) peeling (p = 0.003). No difference between ICG peeling and TB peeling (p = 1.0) was found. Table 4. Functional outcome after 12 months in eyes with primary macular hole closure. Visual acuity Non-peeling group ICG peeling group* TB peeling group p Value All (n = 59) ETDRS, mean (SD) 74.9 (8.2) 72.4 (6.5) 72.2 (9.1) 0.32 Gain, mean ETDRS (SD) 20.7 (8.3) 21.5 (7.7) 22.3 (5.4) ETDRS (20 40 Snellen), no (%) 8 of 11 (73) 24 of 31 (77) 12 of 17 (71) lines gain, no (%) 9 of 11 (82) 26 of 31 (84) 16 of 17 (94) 1.0 Stage 2 (n = 18) ETDRS, mean (SD) 78.2 (6.9) 70.9 (7.4) 0.06 Gain, mean ETDRS (SD) 21.8 (9.4) 16.9 (6.3) ETDRS (20 40 Snellen), no (%) 5 of 6 (83) 8 of 12 (67) lines gain, no (%) 5 of 6 (83) 9 of 12 (75) 0.21 Stage 3 (n = 41) ETDRS, mean (SD) 71.0 (8.6) 73.4 (5.9) 72.2 (9.1) 0.76 Gain, mean ETDRS (SD) 19.4 (7.7) 24.4 (7.2) 22.3 (5.4) ETDRS (20 40 Snellen), no (%) 3 of 5 (60) 16 of 19 (84) 12 of 17 (71) lines gain, no (%) 4 of 5 (80) 17 of 19 (90) 16 of 17 (94) 0.64 *One patient (ICG peeling group) is excluded from analyses because of development of a complicated macula-off retinal detachment with formation of proliferative vitreoretinopathy and dense posterior capsular opacification. One patient (TB peeling group) who experienced late primary closure of the macular hole between 3 and 6 months of follow-up is included in the analysis. à Statistical analysis is only performed for non-peeling versus ICG peeling. ETDRS, Early Treatment of Diabetic Retinopathy Study Chart, letters read at 4 m. Overall mean BCVA 12 months after surgery was 70.4 ETDRS letters (SD = 9.8), 51 of 77 eyes (66.2%) reached BCVA 69 ETDRS letters (corresponding to Snellen fraction) and 60 of 77 (77.9%) gained 3 lines or more on the ETDRS chart (corresponding to a halving of the visual angle). In eyes with successful macular hole closure, there was a significant visual acuity gain from baseline until 3, 6 and 12 months of follow-up (p < 0.001) and a trend of continuous visual acuity gain from 3 to 6 months (p = 0.03 but non-significant because of multiple comparisons) and from 6 to 12 months of follow up (p = 0.08). To answer the question whether ICG staining and ILM peeling negatively affected the functional outcome, we compared visual results of primary closed holes (n = 59). Holes in the non-peeling group theoretically would be closed with the minimal amount of retinal trauma, whereas holes in the ICG-peeling group would be closed with a possible risk of mechanical retinal damage by ILM peeling and toxic damage by ICG staining. Holes in the TB-peeling group would be closed with ILM peeling but with a reduced risk of dye-related toxicity, assuming that trypan blue is a safe dye. Overall, eyes with primary macular hole closure reached a mean of 72.8 ETDRS letters (SD = 7.6) 12 months after surgery. This was significantly better than the 15 eyes that needed a second surgery to close the macular hole (mean BCVA: 66.4 letters, SD = 8.6), p = Of primary closed holes, the nonpeeling group reached a mean of 74.9 letters and the ICG-peeling group reached a mean of 72.4 letters, p = Considering the stage 2 subgroup with primary macular hole closure, there was a trend towards a better mean BCVA in the non-peeling group (78.2 letters) compared to the ICG-peeling group (70.9 letters), p = In the stage 3 subgroup, there were no significant differences in visual results between the non-peeling group (71.0 letters), ICG-peeling group (73.4 letters) and TB-peeling group (72.2 letters) at the final follow-up visit 12 months after surgical closure, p = Intraoperative and postoperative adverse events were systematically registered. Iatrogenic retinal tears 10

12 Table 5. Intraoperative and postoperative recorded adverse events. Adverse event Number of events* (%) Intraoperative Small retinal haemorrhages 6 of 93 (6.5) Retinal tear 5 of 93 (5.4) Postoperative Posterior iris synechiae 24 of 77 (31.2) Posterior capsule opacification 10 of 77 (13.0) Cystoid macular oedema 12 months 3 of 77 (3.9) Vitreous haemorrhage 3 of 93 (3.2) Fibrinoid intraocular reaction 3 of 93 (3.2) Retinal detachment 2 of 93 (2.2) Visual field defects Absolute scotoma 2 of 77 (2.6) Relative scotomas à 0 of 77 (0.0) Retinal pigment epithelium changes Foveal 9 of 77 (11.7) Macular (% fovea) 32 of 77 (41.6) Peripheral 2 of 77 (2.6) Dissociated optic nerve fibre layer appearance 33 of 77 (42.9) *77 primary surgeries + 16 reoperations. Absolute scotoma: complete loss of sensitivity in one stimulus point at 12 months compared with baseline. à Relative scotoma: loss of >10 db in threshold luminance in two neighbouring stimulus points at 12 months compared with baseline. occurred in 5 of 77 patients (6.5%) after 93 vitrectomies (5.4%). Postoperative retinal detachment occurred in 2 of 77 patients (2.6%) after 93 vitrectomies (2.2%). There were no differences in incidence of recorded adverse events between the three treatment groups, except for macular RPE changes and dissociation of the optic nerve fibre layer which were seen only in the two ILM peeling groups; however, these findings did not affect final visual outcome or mean total deviation of sensitivity in the central 10 as assessed by 30-2 automated perimetry. The most prevalent postoperative complication was formation of posterior iris synechiae, which were present in 24 of 77 eyes (31.2%) at the final follow-up visit. Discussion This randomized clinical trial aimed at examining the value of ILM peeling in stages 2 and 3 idiopathic macular hole surgery. The results show that surgery with ILM peeling, for both stages 2 and 3 macular hole, is associated with a significantly higher closure rate than surgery without ILM peeling. Visual outcomes in eyes with primary macular hole closure were not significantly different between the groups, indicating that surgery with short exposure 0.05% (isotonic) ICG-assisted ILM peeling does not significantly affect the functional outcome negatively. The primary closure rate of >90% after ILM peeling (using either ICG- or TB- staining) is similar to results from previous studies (Brooks 2000; Margherio et al. 2000; Mester & Kuhn 2000; Smiddy et al. 2001; Al-Abdulla et al. 2004; Ezra & Gregor 2004; Tognetto et al. 2006); but the primary closure rates after surgery without ILM peeling (45%) is lover in this study than previously reported (Brooks 2000; Al-Abdulla et al. 2004; Ezra & Gregor 2004). This is probably attributed to our radical protocol where removal of epiretinal tissue as well as any instrument manipulation with the retinal surface was forbidden in the non-peeling group. Even though our study only included holes with short symptom duration and no visible epiretinal fibrosis at the baseline examination, the anatomical results may indicate that some invisible or fine cellophane maculopathy exerting tangential traction on the internal limiting membrane may have been present in some of the patients. Thus, our study shows that ILM removal and or epiretinal tissue dissection seems to be a necessary manoeuvre in the majority of macular hole surgeries to obtain acceptable closure rates. Additionally, we cannot exclude the theoretical possibility that missed residual prefoveal vitreous from a split in the posterior vitreous cortex (vitreoscisis) may have exerted tangential traction on the retinal surface. All surgeries were performed by two experienced vitreoretinal surgeons. The overall functional results confirm that macular hole surgery generally leads to favourable visual results, with 66.2% of eyes achieving visual acuity of 69 ETDRS letters ( 20 40) and 77.9% improving more than 3 lines (= halving of the visual angle) in this study. These results are in the better end of previously published data (Brooks 2000; Margherio et al. 2000; Mester & Kuhn 2000; Smiddy et al. 2001; Al-Abdulla et al. 2004; Beutel et al. 2007) and may be attributed partly to the fact that this study is the first where all included eyes were pseudophakic at baseline, thus eliminating cataract as a confounder for functional outcome, partly to the fact that only stage 2 and 3 lesions with a relatively short symptom duration (<12 months) were included. To answer the question whether ICG-assisted ILM peeling negatively affected the functional outcome, we compared visual results of primary closed holes in the three groups. One patient with obvious visual loss caused by a complicated macula-off retinal detachment was not included in these analyses. We did not find any significant differences in mean BCVA, mean visual acuity gain, percentage of eyes reaching or better, or percentage of eyes with halving of the visual angle between the non-peeling group, the ICG-peeling group and the TBpeeling group. Additionally, we found no significant difference in the mean total deviation of sensitivity in the central 10 between non-peeled eyes, ICG-peeled eyes and TB-peeled eyes. Some indications that ICG-assisted ILM peeling may compromise the functional outcome slightly was found in the subgroup of primary closed stage 2 holes, where we found a trend (p = 0.06) towards a better mean BCVA in non-peeled eyes compared to ICG-peeled eyes. For the stage 3 subgroup, the visual outcomes for primary closed holes in the nonpeeling group, ICG-peeling group, and the TB-peeling group were identical (p = 0.96), indicating that a reduction in functional potential is more likely to occur in small macular holes which usually are considered to have the best functional prognosis. 11

13 Given the change in study protocol after the interim analysis which led to termination of the non-peeling stage 3 arm, the statistical power of the performed analyses for final visual outcome in primary closed holes were lower than desired. However, the study had sufficient power (0.77 and 0.83) to detect differences of 10 ET- DRS letters in final mean BCVA in primary closed stage 2 and stage 3 holes. Our observation that reoperated eyes performed significantly worse in all visual outcome parameters compared to eyes with primary macular hole closure indicates that primary focus should be on closing the hole in one procedure, rather than seeking a 50:50 chance of closing the hole without ILM peeling in the search for a (non-significant) slightly better functional outcome. Macular hole surgery using any of the techniques described in this study is a safe procedure with a low incidence of sight-threatening adverse events. Iatrogenic retinal tears were observed in 5.4% of the surgeries, and the retinal detachment rate was 2.2%. Postoperative visual field defects which previously in part have been attributed to the use of ICG-assisted ILM peeling were also rare (2.6%). Some RPE changes, both in the fovea at the location of the previous macular hole and in the macular area at locations corresponding to grasping the ILM were noted. Foveal RPE changes were equally distributed between the three groups and therefore are unlikely to be specific for ICG-assisted ILM peeling. Macular RPE changes were found in 41.6% of all eyes at the final visit, but they were seen only in eyes which had undergone ILM peeling and are likely to correspond to ILM grasping marks. A dehiscence of the retinal nerve fibre layer (DONFL), defined as small superficial irregularities in the retinal nerve fibre layer, was seen on colour fundus photographs in 50.0% of the ILM peeled eyes. DONFL was observed with the same incidence in ICG-peeled and TB-peeled eyes, indicating that the ILM peeling procedure itself rather than ICG staining is responsible. Visual outcomes were identical in ILM peeled eyes with and without DONFL and no difference in central retinal sensitivity were detected in these eyes, indicating that DONFL appearance is not detrimental to functional outcome. All phakic patients underwent cataract surgery 4 6 weeks prior to macular hole surgery. Besides giving us a unique opportunity to follow functional outcome prospectively without the confounding effect of inevitable development of cataract, we prevented patients from having to return for potentially difficult cataract surgery with unstable posterior capsules, loose zonules and a higher risk of posterior capsular rupture as a result of reduced vitreous gel support. The risk of inducing late reopening of a previously closed macular hole was also reduced (Bhatnagar et al. 2007). Unfortunately, we observed a high incidence of posterior iris synechiea (31.2%). This is higher than previously reported in pseudophakic eyes after combined phaco-vitrectomy and face-down positioning (10%) (Tornambe et al. 1997; Lahey et al. 2002); however, in this study we registered all degrees of posterior synechiae (from single strands to sectoral synechiae). Incomplete face-down positioning could be a disposing factor for posterior synechiae development because of forward compression of the intraocular lens by the gas bubble in an upright position. All patients stayed in our department for at least 24 hr after surgery for instruction and observation of keeping a complete face-down position. The high incidence of posterior synechiae is probably because of an unresolved inflammatory response present after the cataract surgery that was performed only 4 weeks prior to macular hole surgery. All patients received anti-inflammatory treatment with dexamethasone eye drops from the day after cataract surgery until 3 weeks after macular hole surgery. The high incidence of posterior iris synechiea indicates the need for increased anti-inflammatory treatment and or periodic pupil dilatation in pseudophakic eyes during face-down positioning. The major strengths of the present study are the randomized design minimizing selection bias and increasing generalizability, the strict clinical criteria for evaluating pre-operative hole size and postoperative hole status, and the fact that all included patients were pseudophakic at baseline eliminating the effect of cataract on functional outcome. The main methodological limitation of our study was the change in study protocol made after the interim analysis which showed unexpectedly poor closure rates after surgery without ILM-peeling. By terminating the non-peeling stage 3 arm and instead introducing the TB-peeling arm, the conditions for the initial sample size calculation were changed. In summary, this article described a randomized clinical trial with masked observers to evaluate the clinical effect, and possible negative influence, of ICG-assisted ILM peeling on retinal function and morphology. ILM peeling was found to induce significantly higher closure rates than non-peeling for both stages 2 and 3 idiopathic macular holes. The use of 0.05% isotonic ICG-assisted ILM peeling with short exposure time is a safe alternative in stage 3 macular hole surgery, whereas a slight reduction in functional potential not can be excluded when performing 0.05% isotonic ICG-assisted ILM peeling in stage 2 macular hole surgery. However, because of a significant reduction in visual outcome after repeated macular hole surgery, we conclude that ILM peeling should be performed in all cases of full thickness macular hole surgery, and the use of 0.05% isotonic ICG-assisted ILM peeling seems to be a safe alternative. Paper II: Normative data of outer photoreceptor layer thickness obtained by software image enhancing based on Stratus optical coherence tomography images Aim To present normative data of photoreceptor layer thickness to illustrate the possibility of obtaining objective thickness measurements of photoreceptor layer structure from contrastenhanced OCT images. Methods Methodological study including 25 normal eyes of 17 persons. Subjects were examined with OCT3 (Stratus model 3000; Carl Zeiss Meditec Inc., Humphrey Division, Dublin, CA), and 20 nominally identical vertical (90 ) transfoveal linear B-scans were made. The best 8 out of these 20 tomograms in terms of image quality 12

14 and consistency of findings were selected and subjected to an automated digital alignment and averaging to obtain a contrast-enhanced OCT image with a 3 times increased signalto-noise ratio because of reduced speckle noise and thus better definition of the outer layers of the retina (Sander et al. 2005; Jorgensen et al. 2007). We developed a windows-based image processing software in C++ Builder capable of tracing and quantifying retinal layers on contrastenhanced OCT images. The software provided measurements of retinal thickness, photoreceptor layer thickness (consisting of photoreceptor outer segments plus RPE and corresponding to the hyperreflective layers seen in outer part of OCT scans) and relative reflectivity of the outer nuclear layer centrally in the macula (obtained by registering the OCT pixel light intensity in the central part of retina at the apex of the Mu ller cell cone compared to a reference pixel light intensity from the outer nuclear layer in the peripheral part of the macula. This ratio may be a clinically relevant pseudo-measure of the amount of glia tissue present in the macula after surgical intervention for macular hole or macula-off retinal detachment). Results In healthy subjects, mean photoreceptor layer thickness in the foveal centre was 77.2 lm (SD = 3.95). A thickness profile of photoreceptor layer thickness along the 6 mm long OCT scan in the vertical direction could be generated, and the thickness profile showed a rise in the central 1000 lm of the scan, corresponding to the long cone photoreceptor outer segments in this region. The relative reflectivity of the outer nuclear layer centrally in the macula was 1.06 (SD = 0.11) in healthy subjects. Mean central foveal thickness was lm (SD = 15.2). Validity of the algorithm for measuring retinal thickness and photoreceptor layer thickness was tested by assessing the degree of agreement between the automatic algorithm generated measurements versus thickness measurement obtained manually by the build-in caliper of the Stratus OCT software. We found discrepancies between the manual caliper method and the automatic algorithm of up to 11 lm in retinal thickness measurements and up to 9 lm in photoreceptor layer thickness measurements. Discussion Contrast-enhanced OCT imaging was capable of quantifying thickness of the outer photoreceptor layer in healthy subjects. The relative reflectivity of the outer nuclear layer centrally in the macula was close to 1 in healthy subjects. Validation analyses found discrepancies between manually and automatically generated thickness measurements of approximately 10 lm which are acceptable in daily clinical evaluations. The major advantage of the new algorithm over the Stratus caliper measurements is that is automatic and independent of observer variations. The methodology of this study is limited by the risk of introducing sampling errors into the automated digital alignment procedure when sampling the enhanced OCT images. Aligning scans which are slightly decentred will lead to misinterpretations of subtle intraretinal changes for example underestimation of central photoreceptor layer thickness. Therefore, it is important that each scan used in the digital alignment procedure is selected manually after inspection for qualitatively identical images. Scans which are obviously decentred should be rejected. Because the Stratus OCT used in this study does not contain an eye-tracking system, we have no possibility to test for 100% identical retinal scan location, except than to visually inspect and compare the 20 (presumed) identical scans before running the software alignment procedure. However, the basis for noise reduction is that the recorded speckle patterns will wary only in cases when the image geometry is slightly modified and therefore small movements of the retina relative to the imaging geometry during image recording is necessary for increasing the signal-to-noise ratio of the enhanced image. This means that the loci of the A-scans scatter around the ideal scan path, and some transverse resolution will be lost in the combined image when compared to that of a singe B-scan. However, because of the increase in signal-to-noise ratio by 3 4 times segmentation and characterization of the retinal layers is facilitated and previous studies have shown that the lateral resolution of the average image is close to 25 lm and the axial resolution below 10 lm, indicating a highquality image (Jorgensen et al. 2007). New spectral-domain OCT (SD-OCT) systems are now available with simultaneous capture of the OCT image and the fundus image (Zeiss Cirrus HD-OCT, Topcon OCT, Heidelberg Spectralis OCT). These systems ensure precise registration between the OCT scan and the fundus image and are potentially capable of performing these measurements in a more satisfying way by shortening the time over which images are acquired. With such a system one needs only to perform the axial registration when combining a series of recordings to form an enhanced image. SD-OCT systems however were not commercially available at the time of commencement of our studies. After considering these possible confounders it is our opinion that quantification of contrast-enhanced OCT images is useful in research and clinical decision-making of retinal diseases where the integrity of the photoreceptor layer and thus central vision may be compromised. A decrease in photoreceptor layer thickness as assessed in this study may indicate misaligned, atrophic or absent photoreceptor outer segments in the affected region. An increase in the relative reflectivity of the outer nuclear layer centrally in the macula may indicate an increased amount of glia tissue which also could compromise function. On this basis, the objective thickness measurements on contrastenhanced OCT images as presented in this study, hopefully can aid assessing and understanding better the relationship between functional outcome and photoreceptor layer structure after resolution of retinal diseases involving the macula. Paper III: Macular morphology and visual acuity after macular hole surgery with or without internal limiting membrane peeling Aim To examine postoperative macular morphology and visual outcome after 12 months in relation to internal 13

15 limiting membrane peeling versus no peeling, indocyanine green staining, and reoperation in eyes that achieved macular hole closure after surgery. Methods The study included 74 consecutive eyes in 72 patients from the COMAH study that had achieved macular hole closure 12 months after macular hole surgery. Macular hole closure was defined as 12 months OCT showing complete adaptation of the margins of the hole. Macular structure 12 months after macular hole surgery was characterized by contrast-enhanced OCT and involved registration of central foveal thickness, central photoreceptor layer thickness (the distance from the photoreceptor IS OS junction to the outer border of the RPE), foveal photoreceptor layer discontinuity diameter (defined as incomplete photoreceptor IS OS junction) and relative OCT reflectivity of the outer nuclear layer centrally in the closed macular hole, Fig. 2. Outcomes were correlated with best-corrected visual acuity 12 months after surgery. Results Central foveal thickness was significantly increased in primary closed ILM peeled eyes (ICG- and TB-assisted) compared to non-peeled eyes (p = 0.001); however, postoperative central foveal thickness was not correlated with final BCVA (r = 0.05, p = 0.67). There was no significant difference in 12 months central photoreceptor layer thickness (p = 0.26), photoreceptor layer discontinuity diameter (p = 0.40) or relative reflectivity of the outer nuclear layer (p = 0.26) between any of the four subgroups (non-peeling, ICG-peeling, TB-peeling, reoperation) and when visual outcomes neither were significantly different between groups with primary macular hole closure (see Paper I) subsequent analyses were performed on the entire patient population rather than on intervention-type subgroups. In Paper I we reported that preoperative visual acuity, pre-operative macular hole size and repeated macular hole surgery were factors significantly associated with final BCVA. The present study found statistically significant correlations between preoperative hole aperture diameter and postoperative central photoreceptor layer thickness (r = )0.39, p < 0.001) and postoperative photoreceptor layer discontinuity diameter (r = 0.36, p = 0.002). Additionally, central photoreceptor layer thickness (r = 0.59, p < 0.001) and foveal photoreceptor layer discontinuity diameter (r = )0.57, p < 0.001) 12 months after macular hole surgery were highly significantly correlated with final visual outcome, Fig. 3. To assess the independent predictive value of postoperative photoreceptor layer appearance for final visual acuity after macular hole surgery we performed a logistic regression analysis incorporating central photoreceptor layer thickness, photoreceptor layer discontinuity diameter and preoperative visual acuity, pre-operative macular hole size and repeated macular hole surgery. By successive reducing the regression model we found that postoperative central photoreceptor layer thickness larger than 33 lm (Odds ratio = 12.5) and postoperative photoreceptor layer discontinuity diameter smaller than 177 lm (Odds ratio = 9.86) were the only significant factors of an eye having regained reading vision ( 69 ETDRS letters) 12 months after macular hole surgery. The specific cut-off values for central photoreceptor layer thickness and photoreceptor layer discontinuity diameter were obtained by selecting the value with optimal sensitivity and specificity on receiver operating characteristic curves. Fig. 2. Postoperative vertical transfoveal contrast-enhanced optical coherence tomograms after macular hole surgery. Operated eyes where the macular hole had closed as intended were characterized by anatomical abnormalities of the outer layers of fovea, with the subfoveal photoreceptor layer often showing a central discontinuity of the hyperreflective band representing the junction between photoreceptor inner and outer segments. Postoperative anatomy was described by four quantitative parameters: (1) central foveal thickness, (2) central photoreceptor layer thickness, (3) photoreceptor layer discontinuity and (4) relative reflectivity of the foveolar outer nuclear layer, defined as the optical density of this layer relative to that of a peripheral reference 1500 lm from the centre of the fovea. RNFL = retinal nerve fibre layer; ONL = outer nuclear layer; ELM = external limiting membrane; IS OS = photoreceptor inner- and outer segment junction; OS = photoreceptor outer segments; RPE = retinal pigment epithelium. 14

16 Fig. 3. Scattergrams of correlations between postoperative visual outcome and postoperative macular structure 12 months after anatomically successful macular hole surgery (pooled data). For definition of parameters, Fig. 2. Discussion Despite high anatomic closure rates after macular hole surgery reading vision remains compromised in 30 40% of patients (Christensen et al. 2008). This present analysis of OCT data from the Copenhagen Macular Hole (COMAH) study examined the relation between postoperative macular morphology and visual acuity in a population of closed macular holes 12 months after macular hole surgery with or without ILM removal to assess specific alterations responsible for the compromised function. Using contrast-enhanced OCT imaging we found thinning of the foveal photoreceptors and larger areas of photoreceptor layer discontinuity in the vast majority of patients 12 months after surgical closure. This occurred independently of the use of internal limiting membrane peeling and indocyanine green staining and was strongly correlated with final visual outcome. Previous studies have shown that pre-operative hole size is correlated with final visual outcome (Freeman et al. 1997; Christensen et al. 2008) and the present study confirm that pre-operative hole size rather than the surgical method influence postoperative photoreceptor layer morphology. To examine whether the strong correlation between postoperative photoreceptor structure and function was just an expression of the effect of pre-operative hole size, we performed a logistic regression analysis including previously identified prognostic factors and found that postoperative structural recovery in the form of photoreceptor layer thickness larger than 33 l and photoreceptor layer discontinuity with a diameter of less than 177 lm was associated with an eye having regained reading vision 12 months after macular hole surgery. The ultrastructural correlate for these photoreceptor layer changes as assessed by contrast-enhanced OCT is not clear, but the OCT signal from the photoreceptor inner outer-segment line is thought to arise from the abrupt change in optical index of refraction between photoreceptor inner segments and the rhodopsin rich outer segments (Hoang et al. 2002). The photoreceptor layer inner outersegment discontinuity diameter as measured in the present study therefore may not reflect completely defect photoreceptors but probably is a combination of misaligned and absent photoreceptor outer segments in the affected region. Histological studies however are needed to verify this. The surgical method did affect postoperative foveal thickness which was significantly increased in ILM peeled eyes compared to non-peeled eyes; however, postoperative foveal thickness was not correlated with final visual acuity. The microstructure of thickened retinas in ILM peeled eyes 15

17 were characterized by a diffuse thickening of the inner retina at the level of the outer nuclear layer, leading to flattening of the foveal pit. Correspondingly non-peeled retinas often were characterized by deeper foveal pits indicating that surgery with ILM peeling increases mobility of the hole edges. The relative reflectivity, or optical density, of the outer nuclear layer centrally in the closed macular hole, which was developed as a pseudo measure of the amount of glia tissue present after hole closure, was only weakly correlated with 12 months BCVA but not with the used surgical method. This might indicate that the intraretinal glial response to macular hole surgery with ILM peeling is not significantly different than after macular hole surgery without ILM peeling. In summary, this study found that attenuation and disruption of the foveal photoreceptor matrix were (A) present in the majority of patients with surgically closed macular holes 12 months after surgery and this was not associated with whether ILM peeling and ICG-staining had been performed or not. These postoperative changes in photoreceptor morphology were highly predictive for the likelihood of an eye having regained reading vision 12 months after macular hole surgery. The findings in this report thus adheres to the results of the randomized clinical trial on functional outcome after macular hole surgery reported in Paper I, where we did not find any significant detrimental effects on functional outcome when performing 0.05% isotonic ICGassisted ILM peeling. Structural details of the postoperative optical coherence tomogram is potentially helpful in the assessment of the retinal visual acuity potential after macular hole surgery, which may be helpful when the cause of subnormal vision is unknown or when competing causes of reduced vision are present. Paper IV: Prognostic significance of delayed structural recovery after macular hole surgery Aim To assess the prognostic significance for visual function of persistent subfoveal fluid and persistent photoreceptor layer discontinuity in eyes in which hole closure had been obtained 3 months after macular hole surgery, with special focus on the effect of ILM peeling, ICG-staining and reoperation. Methods Participants were recruited from the COMAH study and included 74 eyes of 72 patients in which a contiguous retinal surface or a full attachment with a flat neuroretinal rim had been reestablished after macular hole surgery. The closure status 3 months after macular hole surgery was assessed by examining OCT images for closure defects. Characterization involved gross foveal anatomic features (normal with foveal depression, flat edges and photoreceptor-rpe attachment with persistent gap, contiguous retinal surface with persistent subfoveal fluid), Fig. 4. In patients with a normal foveal depression, characterization included foveal photoreceptor layer integrity (complete or incomplete photoreceptor inner segment outer segment (IS OS) junction) quantified by measuring the diameter of loss of homogeneous reflectivity from the IS OS junction line in the vertical direction on the contrast-enhanced OCT, Fig. 5. (B) (C) Fig. 4. Optical coherence tomography images showing postoperative foveal configuration 3 months after macular hole surgery. (A) Normal gross anatomic features with an attached fovea. (B) Flat edges with persistent neurosensory defect. (C) Contiguous foveal surface with persistent subfoveal fluid. Results The gross postoperative foveal configuration 3 months after macular hole surgery was normal in 46 eyes (62.2%), flat-edged with a persistent central gap in one eye (1.3%) and contiguous but with persistent subfoveal fluid in 27 eyes (36.5%). Six months after surgery, 59 eyes (79.9%) had a normal foveal configuration, and after 12 months, the foveal configuration was normal in 69 of 74 eyes (93.2%) as a result of the gradual disappearance of the subretinal fluid in all but five eyes. The single eye with flat-edged fovea with a persistent 16

18 Fig. 5. Contrast-enhanced optical coherence tomography images of three surgically closed macular holes with an attached fovea and with variable degrees of photoreceptor layer discontinuity 3 months after surgery. Arrows are placed just below the discontinuous photoreceptor inner segment outer segment junction. Nominal width of scan, 6 mm. Fig. 6. Graph showing the diameter of foveal photoreceptor layer discontinuity in eyes with an attached fovea as a function of time after macular hole surgery. Box intersection lines represent median, box edges represent the 25th and 75th percentiles, and whiskers represent the 5th and 95th percentiles. foveal gap at 3 months demonstrated a fully closed and hence normal foveal configuration at 6 months. The foveal configuration 3 months after macular hole surgery was statistically independent of the pre-operative inner diameter (aperture) of the macular holes (normal configuration, 446 lm; persistent subfoveal fluid, 445 lm, p = 0.98). Subfoveal photoreceptor layer discontinuity diameter 3 months after macular hole surgery was correlated with the pre-operative macular hole aperture diameter (r = 0.59, p < ). The diameter of the discontinuity zone decreased from a mean of 1366 lm after 3 months to a mean 694 lm after 6 months (p < 0.001) and a mean of 286 lm after 12 months (p < 0.002), Fig. 6. With a mean pre-operative value of 909 lm, the base diameter of the macular holes was smaller than the postoperative photoreceptor layer discontinuity diameter after 3 months of 1366 lm (p < 0.001). The four subgroups (non-peeling, ICG-peeling, TB-peeling, reoperation) demonstrated comparable outcomes in terms of macular hole closure type after 3 months (p = 0.53, chi-square), but photoreceptor discontinuity was more extensive in eyes that underwent repeated macular hole surgery (mean 1739 lm, SD 694 lm) than in eyes that did not undergo ILM-peeling (687 lm, SD 568 lm, p < 0.001, post hoc test). As reported in Paper III there was no effect of intervention type on the extent of photoreceptor layer discontinuity after 12 months (p = 0.40). Therefore, subsequent analyses were performed on the entire patient population rather than on intervention-type subgroups to assess the significance of these changes 3 months after surgery for final visual outcome. Although mean BCVA 3 months after macular hole surgery was better in eyes with a normal foveal configuration than in eyes with persistent subfoveal fluid (68.3 letters versus 64.0 letters, p = 0.04), the difference between the two groups was small and not significantly different after 12 months (72.0 letters versus 70.9 letters, p = 0.59), Fig. 7. Photoreceptor layer discontinuity 3 months postoperatively was correlated strongly with final visual outcome (r = )0.50, p < 0.001). The significant overall gain in BCVA from 3 to 12 months (15.6 letters to 20.3 letters, p < 0.001) reported in Paper I and the concomitant overall reduction in photoreceptor layer discontinuity from 3 to 12 months showed no significant correlation (r = )0.04, p = 0.81), Fig. 8. To assess the predictive value of 3 months photoreceptor layer discontinuity for regaining reading vision ( 69 ETDRS letters) after 12 months, a receiver operating Fig. 7. Significance of persistent subfoveal fluid 3 months after macular hole surgery for final visual outcome: 3- and 12-month functional outcomes in eyes with an attached fovea and in eyes with a fovea with persistent subfoveal fluid 3 months after macular hole surgery. Error bars, standard error of the mean. BCVA = best-corrected visual acuity; ETDRS = Early Treatment Diabetic Retinopathy Study. 17