Systematic ILM Peeling. 20 Years of Experience

Didier & Yvette Ducournau

lundi 14 avril 2008 par Didier ducournau

During the eighties, there was a general consensus that considered the iodiopathic Epiretinal Membranes (ERM) as a proliferation coming from outside the neuroepithelium and covering the macula. Today, even if most pathologists agree with the idea that these membranes are composed essentially of retinal glia, the old physiopathological concepts still persist among clinicians and surgeons…… What is an idiopathic epiretinal membrane ? Why does it appear ? Why is there vision loss ? How does internal limiting membrane (ILM) removal work ? These questions still have no concrete answers today. My own current answers are the result of a long-standing “relationship” with ILM during the last twenty years, initiated with ILM peeling as early as 1986, then grounded on a conjunction of biomicroscopic observations, analysis of pathological studies, functional results of surgery, diversification of scientific knowledge from ophthalmology to neuropathology, and the will of independent thinking. In this article I will be describe my current concepts, and the sequential building steps that have led to my current concepts.
Didier Ducournau, the writer and the hands

Yvette Ducournau. the brain and the muse

1986 - Starting systematic ILM removal

JPG - 60.4 ko
Fig.1 : Because the intraocular fiber illuminates the macula with an angle of 25 degrees, most of the reflection does not leave the eyeball. With a slit lamp placed at 5° from the axis the great majority of the reflected light will reach the surgeon’s eyes.

Everything began in 1985. During a vitrectomy, I broke an intraocular pipe, the last one in the box. In order to finish the operation, I thought to switch on the operating slit lamp mounted on the microscope, which I had been using for the retinal detachment external procedure. I then discovered that, among all the advantages provided by this unconventional illuminating system (see the article), two were very useful for this kind of surgery :

  • First, the free use of the left hand provided additional forceps stabilization, which facilitated greater manipulation precision video1.
  • Second, the reflection of the membranes was much higher than previously seen with the intraocular pipe (Fig1). I found this advantage to be very useful, as I was associating the ERM with the reflection we were observing during examination video2 & video3.

Using this new illumination system, I easily discovered that after removing a first retracted structure, the reflection sometimes persisted. This forced me to take off a more transparent and unretracted second structure in order to eliminate all remaining reflections video4. I systematically adopted this procedure for all the ERM from 1986 and for macular hole (MH) procedures from 1990.

1987 - What is the idiopathic ERM reflection composed of ?

In 1987, I had a second stroke of good luck : I married an ophthalmic pathologist named Yvette. I naturally asked her to study the two different structures that I was removing. Specimens from 56 eyes were examined. She made an immuno histo chemical study with anti-collagen I,II,III and IV antibodies, anti-vimentine antibody, anti-protein S100 antibody, anti-cytokeratine antibody and anti-GFAP antibody.

GFAP (Glial Fibrillary Acidic Protein) is an intermediate protein fibril found in astrocytes of the central nervous system. It plays a dominant role in astrocytic interactions [1]. Its expression is essential to the preservation of tissue architecture, to myelinisation and to the integrity of the blood brain barrier (BBB) [2].

She discovered that in all the specimens coming from idiopathic diseases, only two kinds of structures were present :

  • Collagen (most of time the collagen II, but sometimes collagen III or IV), more or less present depending on the presence of the posterior hyaloid (PH) (Fig2)
  • ILM mixed with GFAP antibody-positive (GFAP-pos) tissue, present in all cases (Fig3).

Yvette explained to me that GFAP-pos reaction indicated the presence of a gliosis at the ILM level.

The term “Gliosis” is used by pathologists when the nervous tissue shows an increased level of GFAP.

This can be observed whether glial cells (in the retina, the astrocytes and the Müller cells) are filled with glyofibrils of GFAP or the number of glial cells is increased. She also conducted a Scanning Electron Microscope (SEM) study, which showed astrocytic proliferation at the ILM level (Fig4).

JPG - 62.9 ko
Fig 2:Adherent posterior hyaloid stained by collagen II antibody.

JPG - 60 ko
Fig 3 : Removed ILM (in grey) mixed with glial cells and glyofibrils of GFAP stained in brown by GFAP antibody.

JPG - 61 ko
Fig 4 : SEM : the ILM, with a typical porous aspect, is reshaped by astrocytic expansions.

It was our finding that hyalocytes, blood cells, retinal pigment epithelium (RPE) and inflammatory cells were extremely rare in idiopathic membranes.
Collagen and ILM are acellular structures ; the glial component appeared then to be nearly the only cellular component we found in idiopathic cases whatever the two anatomical findings at the time of surgery :

  • In case of posterior vitreous separation (either spontaneous or surgically induced), we observed video5 :
    • a pre-retinal structure with the typical fibrillar aspect of the PH (in SEM) and a collagenII-pos and GFAP-neg staining as is normal for a non pathological PH (Fig.2).
    • a retinal structure, composed of an acellular ILM lined by glial cells (or gial cell fragments), GFAP-pos and collagenII-neg (or almost negative) (Fig3). Glial cell expansion was obviated in SEM pictures (Fig4).
  • Sometimes (in 30% of cases) the two structures were so adherent to one another that they happened to be removed together ; these specimens therefore showed both GFAP-pos and collagenII-pos stainings ; no other substantial cellular contingent was present video6.

From these pathological findings, we concluded that the biomicroscopical observation of a reflection was therefore more likely to match the presence of glial expansions, whose anarchic proliferation, whether or not mixed with PH collagen fibers, sent back the received light, rather than to additional pathologic tissue covering the macula.

1992 - Publishing the systematic ILM removal

As the GFAPpos reaction at the ILM level was not present in a normal retina and was always present in the 56 specimens, I concluded that we had to remove this tissue .

I started to present the concept of systematic ILM removal during ERM surgery [3], [4], [5]. This happened to be strongly controversial, due to several explanations :

  1. At the time (1989), publications stressed the importance of not removing the ILM in order to preserve vision. [6] Recommendations for intentional ILM removal started in 1994. [7]
  2. Many publications [8], [9], [10], based on pathologic studies, were presenting different conclusions by highlighting the presence of RPE cells, fibrocytes, and myofibroblasts, most often due to the fact that they were mixing idiopathic ERM and post-retinal detachment ERM or misunderstanding Kampik’s magnificent 1981 study [11] (where at the time, the glial component was already described).
  3. The denomination “epiretinal membrane” generated confusion, as it made people think that there was a pathological process coming from outside the interface, then spreading over the retina (as in the case of secondary post retinal detachment ERMs)
    JPG - 79.6 ko
    Table 1 : Improvement in LogMAR according to pre-op V.A.
  1. The strong functional results of PH removal alone, without ILM peeling, generated confusion as well : this can be explained by the fact that, in at least 1/3 of PH removal cases, fragments of the ILM are also removed (the two structures are often highly adherent to one another).

I conducted a statistical analysis (on 1413 idiopathic ERM and 596 MH with a mean follow-up of 45 months) comparing visual improvement (in Logmar lines) after ILM removal, and reports of PH removal alone (Tab1).

I found that visual improvement was two times higher with ILM removal, providing that the analysis included groups of the same pre-operative visual acuity.

Personally I continued to promote systematic ILM removal in idiopathic ERM and MH surgery.

In the last ten years, the observations made by my wife were supported by all the pathological studies [12], [13], [14], [15], [16], [17] (these studies found GFAP-pos astrocytes and remnants of Müller cell foot-plates in most of the specimens of idiopathic ERM and MH ; no other cellular elements were found in the early stages of the diseases.

Later on Yvette happened to have the unique opportunity to study a child’s eye, removed for anterior segment congenital abnormalities and presenting an inflammatory ERM. In HES standard coloration, several cellular elements covered the retinal surface, giving the illusion of an “extra-retinal origin membrane”(Fig 5). However, GFAP staining demonstrated that the proliferation was composed of glial retinal elements coming from the inner part of the retina (Fig6).

JPG - 100.6 ko
Fig 5 : Paramacular retina. The vitreous (arrow) is detached ; pre-retinal proliferation (asterisk) and increased cellularity in the inner retina are observed (HES x 100)

JPG - 123 ko
Fig 6 : The preretinal proliferation and the increased cellarity of the inner retina are GFAPpos glial elements (GFAP X 100)

Our conclusions were that :

  • idiopathic ERM is not a proliferative process which comes from the outside of the vitreoretinal interface and spreads over the retinal surface
    JPG - 71.7 ko
    Fig 7 : Plaque of PH, adherent to the posterior pole
  • the biomicroscopic reflection we observe is produced by a gliosis that develops at the ILM level and mixes with more or less adherent PH collagen fibers. It is dependent on the underlying anatomical condition, regardless of the gliosis level :
    • when collagen fibers are abundant, in absence of PVD or in case of PVD with a residual plaque of vitreous adherent to the macula (Fig7), the reflection is higher , as the number of fibers involved in the anarchic disposition is higher.
    • when there are no residual vitreous fibers, the reflection resembles a regular cellophane membrane.

1992 - 2000 . Why is there gliosis ? Why does vison drop ?

The problem was then to understand why this gliosis was present in the inner part of the retina.

In neuropathology, astrocytic gliosis is found in the case of ischemia. In conditions of chronic ischemia in mice, transitional limited hypertrophy of fibrous astrocytes localized on the brain surface without cell death has been observed [18]. On the other hand, in the ischemic area, a decrease in GFAP staining at the astrocyte protoplasmic level signifies increased permeability of the blood brain barrier (BBB) to macromolecules and then quick cell death. In experimental in-vitro studies, reduction of GFAP in perivascular astrocytic endfeet also precedes an increase in BBB permeability [19] .

Regarding our idiopathic ERM, the same ischemic hypothesis can be applied.
This hypothesis is actually reinforced by the following arguments :

  • It is our biomicroscopic observation that most of the idiopathic ERM present vascular anomalies (pathological arterio venous crossings, vascular constriction…), compatible with chronic ischemia
  • Most of the idiopathic ERM present edema, which is the consequence of vascular leakage
  • Multifocal ERG recordings in ERM-affected eyes show preoperative delay of P1 implicit time, usually found in retinal ischemic diseases : this also suggests implication of ischemic mechanism [20].

However, other potential trigger mechanisms can be imagined for gliosis, such as inflammation or vitreous traction (which is likely to also induce inflammation, as any trauma ).

It is also possible that vitreous traction, usually considered to be at the origin of the disease, can also be the consequence of the disease ; since the glial cytoplasmic digitation can penetrate the PH and create a strong adherence, one could hypothesize that the vitreous traction could be on the contrary the result of the initial gliosis, when the PVD begins (Fig8).

JPG - 50.4 ko
Fig8 :

How can we explain the drop of vision ? Certainly the PH contraction can explain the metamorphosia, but how can the glial cells (which are repairing cells) affect vision ? The photons can easily pass through them (even if they are filled with GFAP fibrils), reach the retinal pigment epithelium cells, and stimulate the axoplasmic flow. Does the traction interrupt the axoplasmic flow ? Is it not more logical to consider that the ischemic edema, in disconnecting the synapses, is the origin of vision loss ?

Whatever the primary trigger mechanism, the problem was then to understand how removing the gliosis at the inner retina level, or, in other words, removing a repairing process, could generate a high improvement of visual acuity, and produce a large decrease in retinal thickness .

2000 : A sheatotomy gone bad

While I pondered this question, a third lucky event occurred in February 2000. I had planned to perform a sheatotomy on a patient with a temporal superior venous thrombosis (Fig9). During the operation, I did not realize that the vein was over the artery, and I cut the vein. In order to decrease the potential scotoma (and to make my lawyer’s job easier), I had the idea to remove the ILM to facilitate the cleansing of the intra retinal hemorrhage. Surprisingly, one month later, the BCVA was 20/25 (Fig10) and three months later 20/20 (Fig11) (pre op Va 20/35).

JPG - 58.5 ko
Fig 9 : Veinous thrombosis pre-operative aspect

JPG - 77.5 ko
Fig 10 : One month post-op ; the cut vein rises in the vitreous

JPG - 66.7 ko
Fig11 : three months post-op : no hemorrhage, mild residual edema

Starting from this experience, I decided to proceed similarly, performing ILM peeling for all the edemas following venous thrombosis, edematous diabetic retinopathy, and complicated anterior segment operations ; I published my first series in October 2001 [21] . The visual improvement obtained after 6 months was about the same as the one observed after idiopathic ERM or MH, providing the results were compared with the same pre-op acuity.

2001 Yvette’s edema work

But then a question appeared : since most of these cases of edema did not present membrane reflections as in idiopathic ERM cases, why did removing an ILM without reflection (this is to say without gliosis) induce the same visual improvement as removing an ILM with gliosis ? All evidence showed that removing the additional gliosis did not adding anything to the ILM removal. This evidence added to the previous question of how removing a repairing process could have a beneficial effect. But then, if the beneficial effect was not due to the reflection removal, how can it be induced by ILM removal alone ?

As my wife had discovered and archived two specimens of edema in enucleated eyes, she thought to conduct a study on retinal glial reaction patterns in macular edema, and to compare them to different similar situations where CME was not present. She then analyzed seven eyes enucleated for melanoma or endophthalmitis (see case presentation below).



Seven enucleated human eyes were selected for histopathologic and immunohistochemical evaluation. Five of them had been removed for choroïdal melanoma including one with CME. The remaining two had been enucleated for edophthalmitis predominantly involving their anterior segments ; one exhibited massive CME.

After fixation in 10% formaldehyde, the eyes were sectioned in the horizontal meridian through the macular region and 5 microns thickness sections were stained with HES. Immunohistochemical studies were performed on adjacent sections by the streptavidin biotin method, using a 1/200 polyclonal anti-GFAP antibody (DAKO Rabbit Anti-Cow Glial Fibrillary acidic protein) in order to visualize and study astrocytic and Müller cell glia.
JPG - 55.2 ko
Fig.12 : equatorial melanoma (arrow) with localized RD (asterisk) (HESx10)

Case #1(A426) : 70 year old woman with a small equatorial malignant melanoma associated with a limited RD which, in no case, affected the macula (Fig.12).
  • Macula : No structural alterations were seen in HES stained sections. In GFAP stained sections there is discrete staining of Müller cells endfeet and of a few astrocytes . Astrocytes, in the form of scarce small cells appear quiescent, and fusiform with scanty cytoplasm. They surround small blood vessels and are in contact with the ILM between the Müller cells endfeet.

  • Non detached extra macular retina : No architectural modifications are seen with HES, but GFAP stained sections exhibit Müller cells gliosis, localized to the nerve fiber layer and ILM (Fig.13).

  • Detached extra macular retina : In HES stained sections discrete pigment granules adhering to the external portions of photoreceptors testify to a recent RD. GFAP staining reveals impressive transretinal Müller cells gliosis, extending from the ILM to the ELM, with staining of some perivascular astrocytes (Fig14).

Case #2 (C298) : 32 year old woman with a retromacular malignant melanoma and localized serous detachment of the submacular retina (Fig.15).
  • Detached macula : HES stained sections show alterations of photoreceptors outer segments with RPE clumping. The retina is not oedematous. GFAP stained sections exhibit transretinal Müller cell gliosis without notable astrocytic gliosis (Fig.16).

  • Non detached extra macular retina : HES stained sections show no architectural modifications. In GFAP stained sections the gliosis is limited to Müller cell endfeet.

Case #3 (C399) : 59 year old woman with a huge malignant melanoma occupying all the posterior pole including the peripapillary area. There is retinal vessel distortion and equatorial serous retinal detachment. The para papillary retina has been invaded by tumor and has undergone cystoid degeneration (Fig.17 ).

HES stained sections of the detached macula demonstrate loss of foveolar photoreceptor cells. There is no associated edema (Fig 18).

GFAP stained sections reveal prominent transretinal Müller cell gliosis extended from the ILM to the ELM associated with perivascular astrocytic gliosis predominantly in the internal layers (Figs 19 and 20). Higher magnifications show massive Müller cell gliotic fibrils filling all interneuronal spaces.
Case #4 (B713) : 86 year old man with large nasal equatorial malignant melanoma and angle closure glaucoma . Anterior portion of eye showing angle closure associated with a partially necrotic choroidal melanoma. Extra macular retinal vessels exhibit a congestive pattern, testifying of hemodynamic disturbance. The macula is detached ; discrete alterations are seen in the detached photoreceptors ; we can notice preservation of all retinal layers (Fig.21) ; GFAP stains of the detached paramacular retina show the presence of both astrocytic and massive Müller cell gliosis as well as hyperplasia and hypertrophy of perivascular astrocytes adjacent to the ILM (Fig. 22).
Case #5 (A493) : 53 year old woman with peripapillary and submacular malignant melanoma with focal peripapillary retinal elevation and without peripheral RD (Fig23). The optic papilla and central retinal vasculature have been distorted by the melanoma ; the adherent overlying portion of the macular retina has undergone cystoïd degeneration (Figs 23&24).

GFAP stained sections of the adjacent portion of the macula exhibit discrete alterations in the photoreceptor layer, without RD. There is perivascular astrocytic proliferation in the ganglion cell and inner plexiforme layers, without any change in GFAP stains of the Müller cells. The outer nuclear layer exhibits oedematous changes (Fig.25).
Case #6 (A988) : 58 year old man with a corneal abscess and endophthalmitis following keratoplasty performed 2 years after corneal and conjunctival burn by hydrochloric acid. HES stained sections exhibit acute inflammatory cells and fibrin in the pre-foveal vitreous extending along the ILM. The subjacent oedematous retina and its parafoveal vessels are inflamed ; the fovea is focally elevated by transudate (Fig.26). Discreet GFAP staining of perivascular astrocytes is noted in the ILM and parafoveal ganglion cell layer. There is a near absence of Müller cell staining (Fig 27).
Case #7 (D1109) : 85 year old man with endophthalmitis following keratoplasty. The detached macula is distorted by vitreous traction and inflammatory cell infiltrate. A discrete edema is localized to the fovea in the external layers (Fig.28). GFAP stained sections of the parafoveal area show inner retinal gliosis and perivascular astrocytic proliferation that extends along the internal retinal layers (Fig. 29). Transretinal Müller cell gliosis is readily visualised in the parafoveal zone.
JPG - 71 ko
Table 2 : gliosis type according to the pathology


This limited number of cases is not sufficient to allow significant statistical analysis. But most pathologists can understand the value of this work as they are familiar with the current difficulties (at least in France) in obtaining human eyeballs with pathological conditions that do not require enucleation. In addition, most of the enucleated eyeballs present complete structural disorganisation that prevents obtaining precise transmacular sections. The opportunity to have two cases of CME was fortunate.


This study demonstrates that there are two different types of gliosis :


Vascular horizontal gliosis

In the seven eyes, five presented a retinal vascular process (Tab2) either vascular invasive melanoma (case #3 and #5), glaucoma (case #4) or inflammatory angiopathy (case #6 & #7) . In each of these five cases, perivascular astrocytes localised to the inner retina became hypertrophic or fusiform and filled with gliofilaments, constructing what we can call a horizontal gliosis (Fig30).

This confirmed the hypothesis that there was a similarity between retinal and brain that hypoxia could be at the origin of idiopathic ERM. Astrocytous cells could possibly act at the retinal level - following the example of astrocytes in the brain - in a neuro-protective goal of reducing neuronal apoptosis by regulating cellular transfers [22] and by expressing receptors that modulate neuronal synaptogenesis and plasticity [23]. Astrocytic cells are known to be dedicated to neuronal-vascular relationships and are considered to be stimulated by vascular events. Presuming that retinal perivascular astrocytic gliosis serves to preserve the blood retinal barrier, we can then assume that the presence of this astrocytic gliosis, localized to the inner retina, explains why the internal retinal layers are not initially affected by the edema (Fig 31).

JPG - 209.4 ko
Fig 30 : Horizontal gliosis in CME (GFAP X 400)

JPG - 209.5 ko
Fig 31 : The paravascular gliosis in the inner retina and the CME in the outer retina (GFAP X 200)

Astrocytic gliosis in the inner retina did not appear to block the development of CME in the outer retina, as demonstrated by the two cases #5 & #6.


Retinal detachment vertical gliosis

In the seven eyes, five presented a retinal detachment (case# 1,2, 3, 4 & 7). In each of these five cases, (Tab2) the detached retina exhibited reactive transretinal Müller cell gliosis from the ILM to the ELM , constructing what we can call a vertical gliosis. None of the eyes that exhibited macular Müller cell gliosis developed massive CME. Müller cell gliosis seems therefore to be very effective in protecting the retina against edema, as it worked even in the three cases where an ischemia was also present :

  • In case #3, we had a fantastic example of the vertical gliosis anti-edematous capacity : case #3 and case #5 were very similar ; both presented melanomas invading the disc and inducing retinal vascular alteration, which then induced an astrocytic gliosis. However, CME present in case #5 was missing in case #3, as in this latter case a Müller cell gliosis, induced by macular detachment, protected it from CME.
  • In case #4, ocular hypertony induced ischemia that explained the horizontal gliosis. However, as the macula was detached, the vertical gliosis protected it from CME
  • In case #7, the initial pathological alteration was retinal vasculitis, in which the leaky inflamed capillaries likely contributed to CME formation. The detachment occurred secondarily, as a result of neovascular tractio,n and CME was only observed on one side of the fovea where Müller cell gliosis was interrupted.

The two only cases where a CME was present occurred when no Müller cell gliosis was observed.


The transretinal Müller cell gliosis observed after RD can be compared to the emergence of radial glia observe in neuropathology after trauma. After contusion of the rat spinal cord, GFAP is initially expressed beneath the pial surface, at the white matter level, and then spreads in a centripetal radial fashion to the grey matter of the central canal. [24] The radial glia is believed to play an important role in the repair and regeneration of injured neural tissue. 

JPG - 106.9 ko
Fig 32 : SEM. Muller cells disposition from the ILM to the ELM

JPG - 140.5 ko
Fig 33 : RD. Vertical gliosis filling all the intraretinal spaces from the ILM to the ELM (GFAP X 200)

Müller cells normally occupy the entire retinal thickness and serve as a structural scaffold that stabilizes retinal neuronal cells (Fig32). Müller cell endfeet consolidate the ILM and, at its apical pole, forms the ELM and surrounds photoreceptor nuclei. It therefore limits vertical expansion of the retina. It is likely that the proliferation of GFAP-stained gliofibrils, observed along the entire radial course of these cells each time the neuroretinal epithelium is detached, preserves the blood-retinal barrier, reinforces architectural cohesion, and opposes the installation of the edema by filling all inter-neuronal areas and leaving little space for fluid accumulation (Fig33). 

2002 How does ILM removal work ?

Let’s try to understand and interpret these facts in order to propose a hypothesis explaining how ILM removal works. .

Yvette’s study shows that when hypoxia occurs, GFAP over-expression is observed in the inner retina (horizontal gliosis) which preserves, at that level, the integrity of the blood retinal barrier. However, it does not inhibit CME formation in the outer retina. In the same conditions of hypoxia, the retinal detachment induces a vertical gliosis whose effects overpower the horizontal gliosis, as it acts on the entirety of the retinal structure. ILM removal could act in a similar way as retinal detachment.

Unfortunately, we do not have patients’ eyeballs enucleated after ILM removal which would allow us to confirm Müller cells GFAP over-expression. This is why we are forced to hypothesize that ILM removal could work in inducing this cellular response. Based on this hypothesis, let us try to understand how ILM removal could work in the same manner as the retinal detachment process.

JPG - 138.2 ko
Fig 34 : paramacular region : Arrow shows the outer segment of photoreceptors mixted with Müller cells expansions beneath the ELM (HES X 400)


What occurs during a retinal detachment ?


The detachment of the neuroepithelium from the retinal pigment epithelium exposes the former to ischemia and fluid entry (this is possibility is even greater in the macular area, where the ELM is fenestrated). The Müller cell gliosis is the defensive response to this trauma. The signal that induces this vertical gliosis could be the interruption of the connections between Müller cells and photoreceptors, and retinal pigment epithelium cells (Fig 34) video7.


What are we doing in removing the ILM ?

  1. We first remove all the PH collagen fibers adherent to the ILM. In fact, even after a PVD, a certain amount of PH fibers may remain attached to the ILM (Fig35). Removing the ILM is the best way to ensure the removal of all these fibers. This could explain why this operation always decreases the metamorphopsia.
JPG - 104.3 ko
Fig 35 : SEM : Adherent collagen fibers spread over the ILM surface

JPG - 81.8 ko
Fig 36 : A retina cut. One micron thickness ILM (arrow) ; Müller cell endfeet (asterisk)

  1. We remove the ILM. This is a one-micron acellular structure, which cannot grow back (Fig 36).
  2. We also remove a part of the Mûller cell endfeet and the horizontal gliosis (when it is present). This has been confirmed in 2004 by Wolf [25]. We can then easily imagine that this Müller cell aggression caused during the ILM removal (similar to the trauma of the rat spinal cord) could be the initial signal that induces Müller cell gliosis video8.

We can then hypothesize that the success of ILM removal could be due to the induction of a higher over-expression of GFAP within the entire retina thickness at the Müller cells level. This could be the case despite the excision of an initial repairing process (the astrocytic gliosis) video9.

The GFAP over-expression is probably not the only cellular response involved, as it most likely represents a sole link in the chain reaction that participates in the preservation of homeostasis. An example of this chain reaction can be found in studying EGF-R.

EGF-R : Epidermal Growth Factor Receptor is a protein which EGF binds, activating then tyrosine kinase triggering reactions that cause the cells to grow and multiply. In the central nervous system, it is produced by neurons and glial cells and is involved in
Maintaining tissue homeostasis
Regulating gliosis and injury response
Role in angiogenesis and vasoconstriction
Anti apoptotic factor
Regulating neural/glial precursor cell or stem cell proliferation, migration, differentiation and survival
It is a trans-membrane receptor stimulated by others ligands as TGFalfa, Amphiregulin, Heparin Binding, Epiregulin, Betacellulin, Neuregulin...

In an attached retina, EGF-R is not over-expressed. The same thing is observed when isolated ischemic disease with CME occurs. However, in the case of RD, even if an ischemic component is present, EGF-R is over-expressed (Fig37). In the eyeball with inflammatory case membrane, enucleated for anterior segment complications (see fig), the EGF-R was also over-expressed on the entire thickness of the retina (Fig 38).

JPG - 133.9 ko
Fig 37 : EGF-R antibody X 400. Detached retina. Inner nuclear layerand ganglion cell layer. High membrane staining

JPG - 151.7 ko
Fig 38 : EGF-R antibody X 200. Inflammatory membrane. High staining of the entire retina

EGF-R seems therefore to be involved in the cellular reaction found in ERM. In addition, we know that EGF interacts with other growth factors such as VEGF. Consequently, we do not know the initial cellular process at the origin of this reaction. Whatever this initial process, it seems that Müller cells play a predominant role. This is not surprising given that Müller cells are to known to have similar properties as stem cells. We can then understand how ILM removal could induce a visual improvement due to its neurogenesis properties.

This study and hypothesis allow us to better our understanding of what takes place in all of these macular diseases, and to answer several questions, including :

“Why are the functional results so different from one case to another ?”

This study shows that the treatment does not act in removing a pathological process, as in the case of cataract surgery. This means that the results depend on the following conditions :

  • The Müller cell’s capacity to fill itself with gliofibrills of GFAP, and then to fight against blood-retinal barrier breakdown, and to repair the interruptions of axoplasmic flow. Moreover, this explains why the visual improvement takes 6 months or more to reach its highest level
  • The aggravation of the original pathological process. This is why long-term results on diabetes edema demonstrate lower improvement than post-vein occlusion edema, as the diabetic microangiopathy is usually increasing.
  • The preoperative visual acuity (the preoperative amount of destroyed cells). Synaptogenesis properties cannot act in large macrocystic edema.
“Why do most of the ERM, treated with ILM removal, show a cellophane membrane reflection one month after surgery ?“

This is not the ILM reflection that we see, since the ILM cannot reproliferate. However, as the initial vascular process still persists, the astrocytic proliferation reappears and reproduces the initial reflection.

“Why does PH removal alone (without ILM peeling) work in ERM and MH ?”

First, we must consider that when we remove the PH in a certain number of cases we also take off plaques of adherent ILM and/or exert traction on the ILM and Müller cell endfeet. Second, this study shows that ILM removal does not act in removing a pathological tissue but in inducing an additional action. Some small MHs do not need ILM removal to be occluded. However, inducing this vertical gliosis will allow to better consolidate the retina (thus decreasing the recurrence risk, as it is now demonstrated) and to have higher post-op functional results as compared with cases after PH removal alone.

“ How can a MH appear after macular surgery for MH or ERM ?”

This question highlights the fact that PH cannot be the only original cause for a MH, as after such a surgery one has at least removed the vitreous fibers. This is an additional argument for an intraretinal process at the origin of the disease ; this process can appear or reappear after a macular surgery. We can also reconcile this with the stage1 macular hole surgery effect. De Bustros’ study showed that PH removal did not change the natural evolution towards the full thickness MH as it should have been if PH traction was the only original process. I demonstrated that ILM removal in stage1 MH stopped this evolution [26]. The results of this study show that by consolidating the retinal architecture, the Müller cell stimulation allows the retina to resist all tractional processes. In addition, the synaptogensis properties explain the visual improvement of almost all the operated eyes (the eyes that would have developed a full-thickness macular hole as well as the ones that would not have)

“ Why is the peeling more difficult in edema than in ERM ?”

We saw that in idiopathic ERM there is a horizontal gliosis. This gliosis is, most of the time, a dissecting gliosis, which facilitates the dissection : the cleavage is already present. When the horizontal gliosis is not present, we have to create the cleavage plane ourselves. “Why are there edemas with reflection and edemas without ?” I think this depends on the speed of vascular disease development. In the case of sudden onset, the retina does not have the time to react with an astrocytic proliferation, and therefore no reflection appears. I think that astrocytic proliferation appears when a slow hypoxia occurs and not in the case of anoxia.


This 20 year adventure, marked by fortuitous circumstances, questions and incertitude, would not have led to a theory if I had been limited by current mainstream ideas. This shows the importance of maintaining an independent state of mind and an openness towards other scientific contributions (in our case ophthalmic histology and neuropathology studies), in order confirm clinical and surgical intuitions. This interdisciplinary interaction allowed us to open the door to a new concept : ILM removal would work beyond the simple adherent collagen fibers ablation (responsible for metamorphopsia), by inducing a cellular response at the Müller cell level. This permits the retina to fight against edema in a more effective manner than astrocytic gliosis alone. This new concept allows us to respond to several questions to which the extra retinal cellular proliferation theory does not.


[1] Tanaka H, Katoh A, Oguro K, Shimazaki K, Gomi H, Itohara S, Masuzawa T, Kawai N. Disturbance of hippocampal long-term potentiation after transient ischemia in GFAP deficient mice. J Neurosci Res. 2002 Jan 1 ;67(1):11-20.

[2] Liedtke W, Edelmann W, Bieri PL, Chiu FC, Cowan NJ, Kucherlapati R, Raine CS. GFAP is necessary for the integrity of CNS white matter architecture and long-term maintenance of myelination. Neuron. 1996 Oct ;17(4):607-15.

[3] Ducournau D, Ducournau Y. Etude anatomo-clinique de 56 membranes prémaculaires. Cours de microchirurgie rétino-vitréenne Professeur M. Bonnet. Lyon, 1992, Oct 2nd – 3rd.

[4] Ducournau D, Ducournau Y. Macular epiretinal membranes ablation and analysis. Oral presentation, XIXth Meeting of the Club Jules Gonin. Versailles, France, September 1994, program abstract n° 116

[5] Ducournau Y. Aspects anatomopathologiques des membranes épirétiniennes idiopathiques et secondaires. [Histopathological study of idiopathic and secondary epiretinal membranes] Bulletin des Sociétés d’Ophtalmologie de France, Rapport Annuel La Chirurgie de la Macula. 1996 Nov ; pp87-119

[6] Sivalingam A, Eagle RC Jr, Duker JS, Brown GC, Annesley WH, Federman. J Visual prognosis correlated with the presence of internal-limiting membrane in histopathologic specimens obtained from epiretinal membrane surgery. Ophthalmology. 1990 Nov ;97(11):1549-52.

[7] Morris, R., Kuhn, F., Witherspoon, C.D. Retinal folds and hemorrhagic macular cysts in Terson’s syndrome, Ophthalmology 101:1, 1994

[8] Clarkson JG, Green WR, Massof D. A histopathologic review of 168 cases of preretinal membrane. Am J Ophthalmol. 1977 Jul ;84(1):1-17.

[9] Trese M, Chandler DB, Machemer RG. Macular pucker. Graefes Arch Clin Exp Ophthalmol. 1983 ;221(1):12-5.

[10] Smiddy WE & all. Idiopathic epiretinal membranes. Ultrastructural characteristics and clinicopathologic correlation. Ophthalmology. 1989 Jun ;96(6):811-20.

[11] Kampik A, Kenyon KR, Michels RG, Green WR, de la Cruz ZC. Epiretinal and vitreous membranes. Comparative study of 56 cases. Arch Ophthalmol. 1981 Aug ;99(8):1445-54.

[12] Snead DR & all. Hyperconvolution of the ILM in vitreo-maculopathies. Greafes Arch Exp Ophthalmol. 2004 Oct.

[13] La Heij EC & all. Immunohistochemical analysis of the ILM peeled with ICG. Am J Ophthalmol. 2005 Dec.

[14] Kono T & all. Immunohistochemical study of retinal Muller cell response in experimental ERM formation. Nippon Ganka Gakkai Zasshi. 1998 Jan.

[15] Schumann RG & all. Ultrastructure of the vitreomacular interface in full thickness idiopathic MH. Am J Ophthalmol 2006 Nov.

[16] Hisatomi T & all. Cellular migration associated with MH. Arch Ophthalmol. 2006 Jul.

[17] Tari SR & all. Functional and structural measurements for the assessment of internal limiting membrane peeling in idiopathic macular pucker. Retina. 2007 Jun ;27(5):567-72.11-16.

[18] Lukaszevicz AC, Sampaio N, Guegan C et all. High sensitivity of protoplasmic cortical astroglia to focal ischemia. J Cereb Blood Flow Metab. 2002 Mar ;22(3):289-98.

[19] Reuss B, Dono R, Unsicker K. Functions of fibroblast growth factor (FGF)-2 and FGF-5 in astroglial differentiation and blood-brain barrier permeability : evidence from mouse mutants. J Neurosci. 2003 Jul 23 ;23(16):6404-12.

[20] Becquet F, Zanlonghui X, Libeau L. Pre and postoperative functionnal and anatomical changes in patients with idiopathic ERM. Investigative and Visual Science, Supplement, 2005, Poster Board Number : 5442/B645

[21] Ducournau D. Oral presentation. Retina biarritz 2001 Oct 19th

[22] Nakase T, Fushiki , Naus CC. Astrocytic gap junctions composed of connexin 43 reduce apoptotic neuronal damage in cerebral ischemia. Stroke. 2003 ;34(8):1987-93. Epub 2003 Jul 03.

[23] Krebs C, Fernandes HB, Sheldon C, Raymond LA, Baimbridge KG. Functional NMDA receptor subtype 2B is expressed in astrocytes after ischemia in vivo and anoxia in vitro. J Neurosci. 2003 Apr 15 ;23(8):3364-72.

[24] Shibuya S, Miyamoto O, Itano T, Mori S, Norimatsu H. Temporal progressive antigen expression in radial glia after contusive spinal cord injury in adult rats. Glia. 2003 Apr ;15;42(2):172-83.

[25] Wolf S, Schnurbusch U, Wiedemann P, Grosche J, Reichenbach A, Wolburg. H Peeling of the basal membrane in the human retina:ultrastructural effects. Ophthalmology. 2004 Feb ;111(2):238-43.

[26] Ducournau D. Internal limiting membrane removal in stage1macular hole. Xxth annual Vitreous Society Meeting. 2002 Sep ; San Francisco.


Accueil du site | Contact | Plan du site | Espace privé