Ординатура / Офтальмология / Английские материалы / Current Aspects of Pathogenesis and Treatment in Diabetic Retinopathy_Kroll_2007

which in turn produces a dose-dependent upregulation of nitric oxide generation in human endothelial cells and is involved in signaling the permeability-enhancing effects of VEGF [40, 123]. Whereas nitric oxide levels were higher in PDVR with retinal detachment than in normal controls, Hernandéz et al. [49] found no relationship between VEGF and nitric oxide concentrations in the vitreous fluid of patients with PDVR.

Nitric oxide also regulates the release of other cytokines such as platelet-activating factor, tumor necrosis factor and transforming growth factor 1, but the exact pathways are still uncertain [87]. Oku et al. [86] investigated the involvement of nitric oxide and endothelin-1 by examining the levels of both mediators in the vitreous of diabetic patients with PDVR. After observing a significant difference between patients with PDVR and controls, they concluded that both endothelin-1 and nitric oxide could have an active participation in the pathogenesis of PDVR.

A second endothelium-derived relaxing factor associated with the pathogenesis of diabetic retinopathy is prostacyclin. Prostacyclin is released after inflammation and is closely related to nitric oxide. Whereas early hyperglycemia decreases prostacyclin synthesis, in advanced cases of diabetic retinopathy such as PDVR, normal or elevated levels of prostacyclin have been reported [104].

Endothelium-derived contracting factors may be dysregulated in the progression of diabetic retinopathy and PDVR. The most studied of these mediators is endothelin, a strong vasoconstricting peptide that regulates capillary homeostasis in the retinal vascular cells. Hyperglycemia leads to higher levels of endothelin in the retina and disturbance of vasoactive regulators [18].

The effect of hyperglycemia may also mediate several biochemical pathways in the production of growth factors. High glucose levels produce glycated proteins that are biologically active and that may enhance the expression of growth factors [83]. Studies in the last decade have demonstrated hyperglycemia to stimulate the release of several growth factors including transforming growth factor, fibroblast growth factor, platelet-derived growth factor and VEGF.

A prominent mediator of the effect of hyperglycemia to diabetes damage may be related to the activation of protein kinase C. Protein kinase C is a serine/theorine kinase enzyme with a variety of biological functions, including the modulation of cell structures, receptor responsiveness, gene transcription and cell growth. While protein kinase C is a family of at least 12 isoenzymes, each with different enzymatic properties, only the isoen-

zymes and 2 are present in the retina. Studies in animals revealed that only the 2 isoform becomes activated in the vascular tissue of diabetic models [119]. Protein kinase C exerts its influence in the progression of diabetic retinopathy in 2 ways. First, protein kinase C promotes the activation of several growth factors such as transforming growth factor, VEGF and pigment-derived growth factor [3], which in turn are potent angiogenesis factors. Second, binding of VEGF to the target phosphorylation receptors demands the function of signaling proteins, including protein kinase C itself [39]. Experimental studies have shown that one critical component in the mitogenic and permeability-inducing effects of VEGF is the activation of protein kinase C (PKC- ) [3]. Inhibition of the PKCisoform prevents VEGF-mediated cell growth in vitro and reduces ischemia-induced retinal neovascularization in animal models in vivo [5, 24]. Moreover, oral administration of PKCinhibitor has been reported to slow down the progression of diabetesinduced retinal vascular permeability and to normalize changes in retinal blood flow caused by PDVR [59]. These findings suggested that PKCmay be an important pathogenic factor in the evolution of diabetic retinopathy and PDVR.

Biochemical Effects of Hypoxia

In diabetic retinopathy, whereas acute hypoxia stimulates the release of cytokines, the chronic hypoxia facilitates the expression of the several growth factors involved in the formation of new vessels [15]. Although a mediator called hypoxia inducible factor 1 has been identified as stimulator of neovascularization, the exact molecular mechanism of new vessel formation after hypoxia remains unknown.

There are a few theories, besides the effect of hyperglycemia, in which diabetes mellitus may result in ischemia, such as thickened basement membrane, platelet aggregation and leukocyte activation [15]. First, diabetic retinopathy is reported to co-occur with a retinal basement membrane thickening by an increase in the production of fibronectin and collagen [110]. A second hypothesis concerns the formation of thrombin leading to capillary obliteration and retinal ischemia. Diabetes stimulates protein kinase C, which in turn upregulates the production of endothelial cells, leukocytes and platelets to produce platelet-activating factor [113]. The third mechanism early in diabetic retinopathy by which hyperglycemia causes hypoxia is related to the leukocyte activation and adherence as leukocytes adhere to vascular endothelium. As Joussen et al. [60–62] suggest, the inhibition of

VEGF bioactivity may prove useful in the treatment of early diabetic retinopathy in the future [14].

Regarding the progression of PDVR, recent evidence has indicated that the tissue hypoxia-associated proliferative vitreoretinopathy leads to an upregulation of angiogenic cytokines such as growth factors and vasoactive hormones. Regarding the growth factors, their presence in the vitreous and epiretinal membranes supports their role in the pathogenesis of PDVR. Each angiogenic factor may likely operate coordinated or in cascade with other growth factors [23].

VEGF is probably the most important biochemical agent in the development of diabetic neovascularization. Although several growth factors including insulin-like growth factors 1 and 2 and basic fibroblast growth factor have been implicated in retinal neovascularization [19], experimental and clinical investigations showed that VEGF is the dominant biochemical factor involved in the onset and progression of diabetic retinopathy and PDVR. VEGF possesses the characteristics of a mediator of proliferative retinopathies: it is produced by the retina, induced by hypoxia, is proangiogenic, induces permeability and is diffusible through the eye.

Numerous retinal cells produce VEGF, including retinal pigment epithelial cells, pericytes, endothelial cells, Müller cells and astrocytes [23]. Retinal vascular endothelial cells express VEGF and have numerous high affinity receptors to VEGF. Epiretinal neovascular membranes in PDVR disease demonstrated overall VEGF-A expression. Besides, high oxygen consumption by rod photoreceptors in the dark-adapted state may be a powerful driving force of hypoxia and VEGF stimulation [7]. Hammes et al. [43, 44] postulated that, when disregulated, Müller cells may express VEGF. Intraocular VEGF concentrations are increased during proliferation and diminished after laser treatment for PDVR [4, 58].

Evidence of the role of VEGF in new vessel formation has been based on the experiments in primate models of iris neovascularization and in murine models of retinopathy of prematurity. VEGF exerts a potent stimulus for new blood vessel formation by binding to the high affinity tyrosine kinase receptors VEGFR-1 and VEGFR-2. However, the genetic expression of VEGF precedes the new vessel growth, since animal models have an abundant amount of VEGF before the appearance of diabetic morphological changes. Most importantly, experiments with transgenic mouse models indicated that VEGF expression in the retina is sufficient to determine retinal neovascularization [35, 43, 44, 46, 85]. Additionally, the blockade of VEGF is sufficient to inhibit retinal neovas-

cularization in several experimental models of ischemic retinopathy [1, 5]. Not only may the neuroretinal cells be involved with the new vessel formation in PDVR, but also retinal pigment epithelium cells may induce the progression of PDVR because retinal pigment epithelium cells produce VEFG.

Moreover, several biochemical mediators implicated in the pathogenesis of diabetic retinopathy have been demonstrated to increase the expression of VEGF, including glucose, advanced glycation products, adenosin, cytokines (transforming growth factor- , interleukin-1) and numerous growth factors (fibroblast growth factor and pigment-epithelium-derived growth factor, plateletderived growth factor, insulin-like growth factor-1, transforming growth factor- ) [19, 20, 21, 23, 82, 83, 105, 108, 120]. High numbers of glycation end products, for example, have been seen in the blood of patients with diabetic retinopathy, and this increase is believed to be a stimulator of VEGF release and then a causal factor in the development of diabetic retinopathy [112].

Other growth factors including hepatocyte growth factor, pigment-derived growth factor and insulin-like growth factor-1 probably act in conjunction with VEGF in the development of PDVR. Hepatocyte growth factor is an endothelium-specific growth factor that seems to have an important function in the pathogenesis of PDVR. Hepatocyte growth factor is a mesenchyme-derived pleiotropic factor that regulates cell motility and growth. Canton et al. [16] postulated that preretinal membranes of PDVR synthesize hepatocyte growth factor in diabetic patients. However, regarding the progression of advanced diabetic retinal disease, the production of tumor necrosis factormay, in addition to the effect of VEGF, lead to a breakdown of the blood-retina barrier. Phagocyte invasion and production of mediators of inflammation are important factors leading to worsening of fibrovascular proliferation in severe PDVR. Tumor necrosis factorbasically promotes angiogenesis and adhesion of leukocytes to endothelial cells. Tumor necrosis factoralso stimulates the production of monocyte chemotactic protein by retinal pigment epithelial cells [8]. It is reasonable that hepatocyte growth factor plays a role in retinal neovascularization in cooperation with other factors. Tumor necrosis factoralso stimulates the VEGF production directly.

Given its role in inhibiting angiogenesis and inducing cell differentiation, much investigation arose regarding pigment-epithelium-derived growth factor. Pigment-epi- thelium-derived growth factor is a potent endogenous angiogenic inhibitor with a neurotrophic influence on

the retina and is essential for maintaining angiogenic homeostasis in the retina. Also, this factor favors an inhibitory environment when oxygen concentrations are normal or high [26]. Then, pigment-epithelium-derived growth factor plays a considerable role in protecting the retina from pathological angiogenesis [120]. The production of pigment-epithelium-derived growth factor is downregulated by hypoxia, which is the central pathogenic stimulus of VEGF-A. Pigment-epithelium-derived growth factor possibly participates in the regulation of blood vessel in the eye by creating a permissive environment for angiogenesis when oxygen levels are low.

Insulin-like growth factor-1 was the first growth factor implicated in the occurrence of PDVR, based on the observation of reduced incidence of PDVR on growth hormone or insulin-like growth factor-1 deficient dwarfs [81]. Moreover, others reported an acute increase in serum insulin-like growth factor-1 soon before the onset of PDVR [81]. Also, insulin-like growth factor is present at PDVR vitreous in a higher amount in comparison to patients without neovascularization. However, whereas in- sulin-like growth factor-1 has been recently established to regulate not only neovascularization but also VEGF action in the mouse model of ischemia-induced PDVR, Simó et al. [102] described no correlation between insu- lin-like growth factor-1 and VEGF in the vitreous of patients with PDVR, although both were simultaneously increased.

Regarding the vasoactive hormones, there is some evidence that the renin-angiotensin system may be associated with angiogenesis in the eye in PDVR. The reninangiotensin system is present in the kidney, heart, ovary and adrenal gland [120]. The most important enzyme of this system is angiotensin-converting enzyme, which cleaves angiotensin I to the effector molecule of the re- nin-angiotensin system called angiotensin II. Physiologically, angiotensin II regulates the intraocular blood flow and pressure as renin inhibitors lower the intraocular pressure. Angiotensin II has been postulated to reduce microvascular leakage by unknown reasons. Animal studies revealed angiotensin II to induce contraction of perycites and angiogenesis [80]. The increased levels in the vitreous of the renin-angiotensin system components (mostly renin) in PDVR suggest a pathogenic role [94]. Also, glomerular hypertension and angiotensin II increase the expression of growth factors [64]. There is some evidence of a direct relation between the renin-an- giotensin system and VEGF, as both VEGF and VEGF receptor mRNA are localized in the ganglion cell layer, Müller cells, the outer nuclear layer and the retinal pig-

ment epithelium, all sites of renin and angiotensin synthesis. Besides, both prorenin and VEGF were found to be simultaneously elevated in the vitreous fluid of patients with PDVR [2].

Angiostatin, a fragment of plasminogen, has been identified and characterized as a potent inhibitor of neovascularization, as it has been measured in the vitreous of patients without an underlying proliferative retinal disease and in patients with PDVR with or without previous laser photocoagulation. Spranger et al. [107] demonstrated the association between release of the angiogenesis inhibitor angiostatin and diminished production of the angiogenic growth factor VEGF in eyes with previous retinal scatter photocoagulation.

Augustin et al. [9] investigated the vitreous and epiretinal membranes of patients with PDVR to search for oxidative metabolites, i.e. lipid peroxides, and VEGF and to correlate them with retinal coagulation status. They concluded that several oxidative metabolites are able to modulate growth activity and exert this effect via induction of VEGF.

Changes in Retinal Cells in PDVR

There are several hypotheses to explain how hyperglycemia causes tissue and cell damage in diabetic retinopathy. In general, the glucose molecules interact with proteins and cells of the pericytes and endothelial cells. High glucose levels could even damage endothelial cells by hampering the capacity of the cells to eliminate free radicals, and this effect is most likely caused by increased glucose uptake and depletion of glutathione reductase cofactor NADPH reserves. A further mechanism of glucose cell damage, involving the glycation of collagen, changes the basement membranes and affects retinal vascular cell interaction. Others suggest that hyperglycemia activates protein C kinase, leading to an excess of second messengers, which in turn cause plasma membrane damage including loss of tight junctions and abnormal growth factor receptor expression [98].

Diabetic retinopathy is characterized clinically by microaneurysms, cotton-wool spots, lipid exudates, macular edema, capillary occlusion and finally neovascularization with consecutive hemorrhages. While the retinal vascular endothelial tight junction is formed by 2 proteins called occludin and claudin, studies in diabetic animals revealed that diabetes reduces the quantity of occludin at those tight junctions, leading to disorganization in the arterioles and capillaries. This effect was observed to be caused also by VEGF in the retinal cells. However, it is not yet known if the primary effect of VEGF in dia-

betic retinopathy is to increase the vascular permeability or to protect neurons from degeneration [40].

Another important component of diabetic retinopathy is microvascular occlusion. Both intravascular alterations (leucostasis and microthrombosis) and extravascular processes (invasion of Müller cells into the vascular lumen) are related to the microvascular occlusions in diabetic retinopathy. While the presence of vascular alterations induces us to believe that diabetic retinopathy is a pure microvascular disease, several changes such as the neurodegeneration of retinal cells occur in addition to those microvascular changes [78]. The neuroretinal cells are protected from the circulation of inflammatory cells and their cytotoxic products by the blood-retinal barrier, more specifically the tight junctions in the endothelial cells.

Gardner et al. [40] divided the neuroretinal cells into 4 classes and described the changes induced by diabetic retinopathy. The first class of cells are the pericytes and endothelial cells of the capillaries. Pericytes are involved in the stability and control of endothelial proliferation. Retinal capillary coverage with pericytes may be crucial for the survival of endothelial cells in stress-induced PDVR [45]. While pericytes are modified smooth muscle cells which regulate the retinal vascular flow by dilating and contracting, endothelial cells constitute the bloodretinal barrier.

As impairment in the retinal microcirculation results in retinal ischemia, its histological hallmark is the appearance of acellular capillaries. Early in diabetic retinopathy, histological examination of the retina of diabetic patients demonstrated the presence of ‘ghost’ vessels consisting solely of basement membrane [63]. The damage of retinal capillary cells including pericytes and endothelial cells is responsible for microaneurysms and vascular obstruction. These changes, including pericyte loss, evolve over many years before the onset of PDVR [15]. Whereas diabetic retinopathy has been considered a systemic disease, the absence of pericyte changes in the optic nerve of diabetic patients suggests a local ocular disease [65].

There is also evidence that hyperglycemia may directly worsen ischemia by promoting endothelial cell proliferation and diminishing the inhibitory effects of pericytes on endothelial cells [98]. Hyperglycemia could also alter the status of circulating platelets and leukocytes. These cells may adhere to the capillaries in end-stage PDVR and thereby worsen ischemia.

The second category comprises the glial cells, either Müller cells or astrocytes. Whereas Müller cells mainly

span the thickness of the retina from the retinal pigment epithelium to the internal limiting membrane, the astrocytes are limited to wrapping the small retinal blood vessels. Astrocytes undergo severe changes in diabetes, as the production of their intermediate filament and glial fibrillary acidic protein are markedly decreased. As the Müller cells have a direct effect in the formation of diabetic epiretinal membranes in PDVR, Müller cells may be affected much earlier in the course of the disease [10]. In conclusion, the structural glial retinal cells may be severely damaged by diabetic retinopathy when the bloodretinal barrier function is impaired [40].

A third class of cells, the neurons, may be subdivided into photoreceptors, bipolar cells, amacrine cells and ganglion cells. These neurons physiologically convey the electric impulses to the brain. As they are the cells responsible for vision itself, any loss of visual acuity in diabetic retinopathy necessarily implies disturbance of their function. There is evidence that retinal ganglion cells and inner nuclear cells die by apoptosis early in the course of diabetes. Moreover, it has been shown that a continuous atrophy of the thickness of the inner retina and reduction of the number of ganglion cells occurs up to the late stages of diabetic retinopathy including PDVR [40].

The last class of retinal cells, the microglia, are responsible for phagocytosis in the retinal environment. Diabetes has been shown to activate the normally quiescent microglial cells. Because microglial cells may release several mediators of inflammation such as VEGF and tumor necrosis factor, microglial cells seem to play an active role in the progression of diabetic retinopathy to PDVR [91].

Although diabetic retinopathy is mainly characterized as a microvascular disease with concomitant lesions of the neuroretinal cells, there is much evidence that diabetic retinopathy is also an inflammatory disease [60– 62]. As observed in ophthalmoscopic examination, diabetic retinopathy frequently presents with tissue destruction and attempts of tissue repair. Experimental studies revealed the presence of several mediators of inflammation in diabetic retinopathy, including leucostasis, adhesion molecule activation, prostacyclin upregulation, VEGF expression and retinal accumulation of macrophages [12, 40, 47]. Tissue loss is exemplified by neuroretinal cell apoptosis.

Limb et al. [77] examined the vitreous of patients with PDVR in order to search for the intravitreous presence of vascular cell adhesion molecules that mediate steps of inflammation. They found increased numbers of several molecular inflammatory molecules including vascular

cell adhesion molecule 1 and sE-selectin in patients with PDVR compared to controls. Inflammatory mechanisms may contribute to the onset of new vessels and fibrosis as an endothelial, Müller and retinal pigment epithelial cell activator.

The Vitreous in PDVR

The human vitreous is constituted of several intercellular connective tissues including type II collagen, hyaluronic acid and hyalocytes. Vessels are normally excluded from the vitreous, a compartment which has been shown to have antiangiogenic properties [107]. Several clinical and experimental studies have demonstrated that the vitreous plays a primary role in the pathogenesis of PDVR. The studies have shown the presence of new vessels microproliferating inside the vitreous cavity, indicating the vitreous to be active in the angiogenesis process [34]. There is also evidence that epiretinal membranes in the vitreous of patients with PDVR produce a high amount of growth factors as the retinal cells do [43, 44, 46, 79].

There must be a biochemical factor in the vitreous tissue that directly regulates the ischemic retina in diabetic retinopathy. An increased presence of advanced glycation products has already been found in the vitreous of diabetic eyes. In diabetic retinopathy, proteins in the vitreous that inhibit angiogenesis undergo marked nonenzymatic glycosylation of collagen and other proteins and this may represent an initial and trigger finding for proliferative changes in diabetic retinopathy [48, 100].

Sebag [99] described presenile changes in the vitreous in diabetics. Hyperglycemia eventually leads to changes of type II collagen and consequently to liquefaction and syneresis of the vitreous. The instability of the vitreous resulting from this loss of the gel state without dehiscence at the vitreoretinal interface may also induce traction on the retina, which in turn might not only lead to retinal tears but can also contribute to the neovascular process itself. Thus, in addition to providing a scaffold for retinal capillary endothelial and other vasoproliferative cells as postulated by Faulborn and Bowald in 1985 [34], the vitreous may aggravate the process of neovascularization because of changes in the rheologic state.

The vitreous cortex seems to be involved in early stages of the diabetic disease, whereas the vitreous gel influences the late stages. At the posterior pole the vitreous cortex is composed of parallel collagen fibers that are attached to the retinal surface formed by the foot plates of the Müller cells. This attachment consists of extracellular matrix proteins, mainly fibronectin and laminin [71].

Table 1. Development of PDVR

Non-PDR f

PDVR f

1st step of pathogenesis:

Thickening of vitreoretinal interface

Ingrowth of newly formed vessels into the posterior vitreous cortex

f

2nd step of pathogenesis:

Shrinkage of the vitreous gel through crosslinking of collagen fibers (possibly induced by factor 13)

New vessels reach the vitreous by invading the internal limiting membrane, although the exact mechanisms are not totally clear yet. Previous clinical findings indicated that PDVR is rare if the vitreous cortex has detached completely in myopics with PVD and after vitrectomy, since the scaffold for proliferating cells is destroyed [97]. In high myopia lesser diabetic involvements of the fundus oculi have been observed [57], which may serve as a proof for the involvement of the vitreous in the development of diabetic retinal changes and therefore justifies the term PDVR.

Because of a long-standing diabetes-induced breakdown of the blood-retinal barrier, serum proteins, especially fibronectin, accumulate up to 10-fold at the vitreoretinal border region [70]. Fibronectin mediates the migration and adhesion of proliferating endothelial cells supported by growth factors like transforming growth factor- , which bind at specific domains [17]. In the late stages of the disease, the vitreous gel contracts, leading to vitreous hemorrhage due to the rupture of the proliferative vessels, vitreoschisis, as well as tractive or rhegmatogenous retinal detachment. Shrinkage of the vitreous indicates crosslinking of collagen fibers probably induced by transglutaminase (factor 13a) in the presence of fibronectin [6]. Angiogenic cells migrate and then neovascular proliferation arises in this vicious circle.

Morphological Evolution of the PDVR Disease

In the above chapter we learned that PDVR develops as new vessels arise and proliferate at the border region between the retina and the vitreous cortex accompanied by a fibroglia scaffolding (table 1).

Multifactorial reasons, such as hypoxemia and ischemia followed by the accumulation of growth factors lead

to a thickening of the posterior vitreous cortex, which occurs as a first step prior to an ingrowth of proliferating new vessels into the posterior vitreous cortex (fig. 1) [37, 38].

In this process Faulborn and Bowald [34] found small proliferations arising multifocally and growing within the vitreous cortex. They described the fibrous material of the vitreous cortex being densely interconnected with and obviously being incorporated into the newly formed proliferated tissue and thus providing a scaffold for proliferating cells.

In 1990 Yu et al. [122] found in animal experiments that vitreal PO2 increased as a function of the distance from the internal limiting membrane, if inspired oxygen tension was increased. Vice versa, hypoxemia leads to a decrease in vitreal oxygen tension.

In 1991 Vlodavsky et al. [118] described that basic fibroblast growth factor promotes the formation of new blood capillaries. It binds to heparan sulfate, both on the cell surface and in the extracellular matrix. Enzymes such as heparanase lead to a degradation of the basic fibroblast growth factor/heparan sulfate complex and thus regulate the growth of capillary blood vessels in normal and pathological situations. The extracellular matrix also serves as a storage depot for other growth factors and enzymes which generate the proliferation of newly formed vessels growing from the retina into the vitreoretinal interface.

Sebag et al. [96] in 1992 analyzed vitreous samples from patients with PDVR and from patients without diabetes for collagen crosslinks, as well as for the early glycation products glucitolyllysine and glucitolylhydroxylysine. They found that early glycation products were elevated in diabetic vitreous, while the levels of advanced glycation end products were even 20 times higher in diabetic vitreous compared with the vitreous of controls. These diabetes-induced alterations of human vitreous were regarded as particularly important for proliferations of vessels into the vitreoretinal interface. That is why the authors give the vitreous in PDR an important role and agree on the term PDVR instead of PDR [100].

In a second step of the pathogenesis of PDVR, shrinking of the vitreous body occurs, especially of the posterior vitreous cortex. This step occurs for reasons not yet very clear, although Akiba et al. [6] hypothesized that the factor 13 (transglutaminasae) of the hematopoietic system may play a central role. The clinical alterations in the PDVR stage of diabetic retinopathy are a consequence of both thickening and shrinking at the vitreoretinal interface. Due to vitreous shrinking, neovascu-

larizations in the vitreous are torn and typical vitreous bleeding or – if there are more extended adhesions between vitreous and retina – tractive retinal detachment results, which may be worsened by rhegmatogenous retinal detachment.

Classification of PDVR

Starting with the introduction of photocoagulation and vitrectomy as therapeutic options of diabetic retinopathy to avoid blindness, several classification systems were presented.

The Airlie House classification, which was introduced purely for assessing results of photocoagulation studies in the 1960s, suggested 2 types of diabetic retinopathy: non-PDR and PDR [25]. The division into 2 patterns has remained until now because it is simple and informative.

Diabetic retinopathy is probably one of the diseases in which medicine-based evidence principles were applied early and very extensively. While major clinical trials started in the 1970s and 1980s to evaluate the benefits of laser treatment and surgery for diabetic retinopathy, they also determined the evolution and consequently the classification systems were developed. The 5 multicenter clinical trials which established the basic concepts for the current classification and treatment of diabetic retinopathy are the Diabetic Retinopathy Study, the Early Treatment Diabetic Retinopathy Study (ETDRS), the Diabetic Retinopathy Vitrectomy Study (DRVS), the DCCT and the UKPDS.

The ETDRS study group classified the proliferative form of diabetic retinopathy in early, high-risk and severe PDR. The ETDRS severity scale was based on the modified Airlie House classification of diabetic retinopathy [29, 32]. The limitation of this scale is that it has proven to be useful in clinical practice only for the outcome of panretinal photocoagulation. Several contemporary surveys have documented that most physicians managing patients with diabetes do not use the full ETDRS severity scale because of its complexity [119]. PDR was classified by the ETDRS severity scale into mild, moderate and high-risk PDR for the initial cases of the disease. If either retinal detachment, traction, rubeosis iridis or fundus obscuration were visualized, it was called advanced PDR (ETDRS, 1991). Further on, the ETDRS confirmed the need of scatter laser coagulation for cases of high-risk PDVR.

Fig. 1. The vitreoretinal interface is believed to play a key role in the development of PDVR (see text).

Although at first pars plana vitrectomy was used for cases of critical forms of PDVR like advanced tractional detachment, in 1983 Shea [101] recommended vitrectomy at earlier stages of the PDVR disease in order to improve surgical outcomes and help in the preservation of useful vision. The DRVS used morphological criteria to define ‘early’ as a stage of the disease prior to extensive contraction causing retinal detachment [30]. The DRVS postulated that early vitrectomy increased the chance of visual

acuity restoration to over 20/40, at least for eyes with very severe new vessels. However, this major clinical trial was conducted prior to the introduction of surgical advances such as endolaser now commonly employed during vitreoretinal surgery in patients with diabetic retinopathy, and caution is therefore needed to interpret its results. For instance, bimanual surgery with delamination or en bloc dissection of epiretinal membranes were not mentioned in that study.

Fig. 3. a–c PDVR, stage B: this stage is characterized by shrinkage of the posterior vitreous cortex. In places where the vitreous adheres to the retina circumscribed retinal detachments are found. b If a tractive detachment is nasal to the optic disc, this is described as stage Bn. c Proliferative and tractive changes in the area of the temporal superior and inferior vascular arcade, which may be followed by a macular detachment, are categorized as stage Bt.

Fig. 4. a PDVR, stage C. Stage C is – similarly to the PVR classification – characterized by a tractive retinal detachment which includes the macula.

Table 2. Classification systems of diabetic retinopathy

1969 Airlie House classification [25]

1981 Modified Airlie House classification (ETDRS) [29]

2003 International Clinical Diabetic Retinopathy Severity Scale (American Academy of Ophthalmology) [119]

In 2003, Wilkinson et al. [120] reported the results of a workshop for a proposed international diabetic retinopathy severity scale. It was agreed among the participants that, in addition to mild, moderate and severe, there is a level for ‘no retinopathy’ and second, a ‘PDR’ level for the presence of any neovascularization. However, the international clinical diabetic retinopathy disease severity scale proposed no division of PDR into subgroups. The inclusion of no apparent retinopathy and minimal non-PDR were seen as disagreement in the discussion (table 2).

It could be shown that most publications nowadays are based on vitreous hemorrhages with macula on or off, tractive retinal detachments and progressive fibrovascular proliferations as indication criteria for surgery. Therefore, a standard definition and classification of diabetic retinopathy is necessary. It should be clear and critical for the clinical decision process and for communication among ophthalmologists, internists and diabetologists and also for communication with the patient. Furthermore, a classification of PDVR may also be essential to define the indications of surgery and to serve as a predictive factor for surgical outcomes.

Kroll’s Classification

As the vitreous has a defined role in the pathogenesis of PDVR, vitreous removal, either by pars plana vitrectomy [13, 33, 56, 74, 95] or by enzymatic means [52, 53], is an appropriate technique to interrupt this process and to prevent final stages. Various studies have demonstrated that vitrectomy earlier in the course of the disease prevents the onset of severe complications [50, 101]. However, a good postoperative visual outcome is difficult to predict even though various factors like preoperative visual acuity, short duration of visual loss, absence of iris neovascularization, a clear lens and partial panretinal photocoagulation are known [50]. Therefore, Kroll et al. [73] established a classification of PDVR that clearly differentiated between early and late stages of the disease and was based on the described hypothetic pathogenesis of PDVR and on clinical observations with preoperative examination techniques as well as intraoperative observations.

Since the posterior vitreous cortex is mainly responsible for this disease entity, the term PDR has been changed to PDVR in analogy to the term proliferative vitreoretinopathy [73, 100].

The dynamic morphological stages are differentiated as follows:

Stage A (fig 2a, b) is characterized by proliferative changes in the vitreous throughout the retina, especially around the optic disc but also elsewhere; however, the retina is totally attached.

Stage B (fig. 3a–c) consists of circumscribed tractive retinal detachment around the optic disc and at the retinal arcade vessels, as a result of shrinkage of the vitreoretinal interface. However, the macula is attached, so good visual acuity is still present, as long as no hemorrhage or diabetic macular disease occurs. As stage Bn (fig. 3a) one recognizes the tractive detachment nasal to the optic nerve and Bt (fig. 3b) temporal to it.

Stages C1–C4 (fig. 4a–e) are characterized by tractive retinal detachment involving the macula depending on the number of quadrants involved. Due to macular detachment visual function is always reduced [72].

In further studies, our group investigated the importance of Kroll’s classification as a prognostic value regarding the postoperative results of vitreoretinal surgery. We analyzed 563 patients who underwent a pars plana vitrectomy for PDVR and calculated the operative risks in a multivariate logistic regression analysis. The results showed that postoperative increase of visual acuity of 13 lines was significantly less frequent in stages B and C in comparison to stage A. It may be concluded that Kroll’s classification for PDVR has a high prognostic value for postoperative visual outcome and surgical management indications [51, 54].

Final Considerations and Conclusions

While pathogenesis and morphological changes of non-PDR seem to be clear (a classification was set up by the ETDRS), currently applied classification systems of PDR are still insufficient.

Molecular biological examinations show a very complex system of growth factors and multiple other factors involved in the pathogenesis of the disease. As we could show above, in our opinion, the posterior vitreous cortex seems to be the main cause of morphological changes in PDVR. First, in diabetic eyes vessels grow into the thickened posterior vitreous cortex, maintained by molecular processes in the vitreoretinal interface. Second, a shrink-