Proliferative vitreoretinopathy: risk factors and pathobiology

Abstract

Proliferative vitreoretinopathy (PVR) is still a major cause of failure of retinal detachment surgery. Despite a dramatic increase in our pathobiologic knowledge of PVR during the last 10 years, little of this information has been used to modify the surgical management of the disease, and, thus, the anatomic and functional results are still unsatisfactory. Collaborative research involving clinicians and basic researchers must be encouraged. PVR must be considered a multifactorial disease caused by interaction of several cells and intra- and extraocular factors. Therefore, therapeutic options based on the inhibition of one factor or phenomenon may be regarded with scepticism. To prevent PVR, it is necessary to determine the factors involved in its development, and because of its relatively small prevalence, large, prospective, multicenter studies seem necessary. In addition, clinical research must not be underestimated. PVR affects both sides of the retina and the retina itself, a point to which little attention has been paid and that is critical for surgical results. Therefore, a new classification that provides information about clinical relevance, such as the evolutionary stages of the disease (biologic activity) and the degree of surgical difficulty (location of the fibrotic process), seems necessary.

Introduction

Proliferative vitreoretinopathy (PVR), a scarring process that develops with some retinal detachments (RDs), is the most common cause of surgical failure in the rhegmatogenous RD treatment. PVR can also be defined as the growth and contraction of cellular membranes within the vitreous cavity and on both retinal surfaces (Fig. 1, Fig. 2, Fig. 3, Fig. 4), and in many cases a fibrotic process of the retina itself (Fig. 5, Fig. 6) (Pastor, 1998; Sebag, 2001).

Data analysis in the literature indicates that PVR is not a specific clinical entity, but rather an end point of a number of intraocular diseases with various stimuli. RD and the associated vitreal alterations are important but not the sole factors in PVR development (Weller et al., 1990). PVR may be present before surgery (referred to by some as primary or preoperative PVR), but its incidence increases after surgery (postoperative PVR), and therefore it may result from intraocular surgical effects (Sebag, 2001).

The results of surgery of this complication are extraordinarily variable in series reported in the literature. The variations result from multiple factors, mainly the inherent variability of the disease but also different criteria to establish the degree of severity. Classifications of PVR (Retina Society Terminology Committee, 1983; Silicone Study Group: Lean et al., 1989) (Machemer et al., 1991) are not useful tools by which surgical decisions can be made, and most surgeons prefer to classify PVR as “light,” “moderate,” or “severe” (Pastor, 1998). No clear divisions exist among the three forms, which do not represent a true classification, but rather a presurgical impression of the degree of difficulty to reattach the retina and hence a clue for performing extrascleral surgery or vitrectomy.

The other crucial question is the lack of information that surgeons have about the degree of activity of PVR at the time of surgery. PVR is related to inflammation and surgery may stimulate it and thus increase PVR. PVR is a dynamic process, but we have little information about its chronobiology, even from a pathobiologic point of view. This knowledge seems crucial for improving surgical results, and some authors consider that if there are clinical signs of activity, it may be useful to delay surgical intervention for some weeks (Glaser, 1994).

Poor results following prophylactic measures can also be explained by this ignorance of chronobiology. A treatment based on the time course of the disease could obtain better results (Hui and Hu, 1999).

All these facts and lack of information contribute to the variability of the surgical results. Anatomic success is now reported in 60–80% of cases, depending on the disease severity (Pastor, 1998; Sebag, 2001). The more advanced the disease, the lower the reattachment rate, and results with anterior PVR are uniformly worse than those with posterior PVR. Surgery is the standard treatment. Light and moderate cases of PVR can be treated by conventional surgery for RD (buckling and encircling procedures), but pars plana vitrectomy is indicated in most cases. Treatment of complex cases involves procedures that are technically complicated, such as retinotomies and/or retinectomies, and the use of substances for vitreous replacement, all of which are still controversial (Pastor, 1998).

Despite the improvement of surgical reattachment rates with vitrectomy, anatomic success does not ensure visual improvement. Only 40–80% of patients who undergo anatomically successful surgeries recover at least ambulatory vision. Functional success may be poor because of macular changes, but it is also related to the number of procedures performed (Pastor, 1998).

In summary, most surgeons still believe that prevention or prophylaxis by medical treatment should be considered a challenge for research to improve the final outcome. Numerous drugs have been tested on animal models or cell cultures to inhibit cell proliferation and membrane formation and contraction and most of them have demonstrated their efficacy. However, many of these drugs have potentially severe side effects, and only a few have been used in clinical trials (Pastor, 1998). Recently, clinical researchers have focused their attention on the use of drug combinations trying to act on the different stages of PVR. Some of these works have been made in animals (Pastor et al., 2000), and most recently a clinical trial has been published (Asaria et al (2001a), Asaria et al (2001b)). Results seem positive but further studies with larger samples are necessary to establish the real efficacy and safety of these “cocktails”.

Because of difficulties obtaining therapeutic drugs levels in the eye using conventional routes of administration, the use of the intravitreal pathway to prevent PVR is considered a target to control this vitreoretinal complication. However, no agent reported in the literature is used routinely in clinical practice because of concerns about retinal toxicity and lack of information about intraocular pharmacokinetics. Nevertheless, some efforts are being made in this direction (Jonas et al., 2000; Asaria et al., 2001a). However, if risk factors for the development of PVR could be identified, these potentially toxic intravitreal drugs could be administered to patients at greatest risk (Sebag, 2001).

On the basis of the actual knowledge about the pathobiology of PVR, it is easy to understand that any clinical factor that increases intraocular inflammation or facilitates release of retinal pigment epithelial (RPE) cells into the vitreous cavity may stimulate the development of this complication (Nagasaki et al., 1998).

PVR is reported to develop in from 5% to 10% of all RDs (mean, about 7%) (Pastor, 1998). This incidence increases in several clinical situations, and their identification as risk factors has been an important subject for the last 20 years (Bonnet, 1988; Cowley et al., 1989; Girard et al., 1994; Grizzard et al., 1994; Nagasaki et al., 1998). On the contrary, few reports have been published on this matter during the last 5 years, and few relevant facts have been incorporated into clinical practice (Nagasaki et al., 1998; Schwartz and Kreiger, 1998; Limb and Chignell, 1999; Ahmadieh et al., 2000; Kon et al., 2000; Lleo Perez et al., 2000; Rodriguez de la Rúa et al., 2000).

Since the beginning of the 1990s, it is accepted by surgeons that the incidence of primary or preoperative PVR increases in several, well-identified clinical situations, i.e., the extent of detachment; giant, large, multiple, or undetected retinal breaks; aphakia; vitreous hemorrhage; preoperative choroidal detachment; previous failed attempts at reattachment; and the presence of signs of uveitis (Girard et al., 1994; Nagasaki et al., 1998; Pastor, 1998).

However, the incidence of PVR increases postoperatively and is most strongly associated with additional factors: preoperative presence of PVR grades A and B (Retina Society Terminology Committee classification, 1983); intraocular hemorrhage during or after surgery; choroidal detachment; use of air or sulphur hexafluoride (SF₆); excessive cryotherapy, diathermy or photocoagulation; repeated surgical procedures; loss of vitreous during drainage of subretinal fluid, and the use of vitrectomy (Bonnet, 1988; Cowley et al., 1989; Girard et al., 1994; Grizzard et al., 1994; Nagasaki et al., 1998; Pastor, 1998).

Some of these findings are contradictory and it is difficult to assess the actual role of each factor in the development of PVR. To elucidate this question, Girard et al. (1994), in one of the largest prospective series of RD, confirmed the effect of some factors, but their results added more controversy to this subject. According to those authors, only low-grade preoperative PVR (grade A) is a strong predictor for severe postoperative PVR, whereas PVR grade C1 is not. Furthermore, only minor hemorrhages occurring during or after surgery are associated with a higher prevalence of postoperative PVR. On the other hand, the authors did not consider the number of surgeries, aphakia, or subretinal fluid drainage as predictors of PVR development. According to these authors, the reasons for this controversy are the absence of a clear distinction in previous papers between preoperative and postoperative PVR and the use of inappropriate methods of statistical analysis. They propose the evaluation of clinical variables by univariate analysis and stepwise logistic regression.

However, other authors, using similar statistics, found different predictive variables, such as scleral perforation, unsealed breaks, or poorer preoperative visual acuity (Ahmadieh et al., 2000; Lleo Perez et al., 2000).

The incidence of postoperative PVR after primary vitrectomy for treating RD and the risk factors implicated in its development have also been investigated (Kon et al., 2000; Richardson et al., 2000), but once again different criteria in patient selection caused more confusion.

Richardson et al. (2000) analysed the so-called “uncomplicated” RD, excluding those cases with preoperative PVR greater than grade B. They reported an incidence of postoperative PVR of only 4%.

Kon et al. (2000) considered primary vitrectomy in selected cases of RD, which included giant retinal tears, posterior retinal breaks, and the presence of preoperative PVR and media opacities. In that series, 39% of patients presented with preoperative PVR, and postoperative PVR developed in 29.4%. In their introductory statements, those authors pointed out the above-mentioned problems with previous studies, namely, that the results are often contradictory and inconclusive, and they added bias associated with any retrospective study (Kon et al., 2000). Through multivariate regression analysis of a prospective study, they tried to identify independent predictive risk factors and included as objective data measurement of total protein concentration of the vitreous. This measure is one of the significant risk factors for development of PVR. In addition, they proposed two mathematical models to predict the probability of developing postoperative PVR. Nevertheless, this interesting paper was biased by the preoperative selection of cases. This series must be considered as a sample of complicated RD, and therefore conclusions cannot be extrapolated to the general population of RD.

Assuming that PVR results from the interaction of multiple factors, we recently suggested classifying patients with RD into different risk groups according to the likelihood of developing PVR (Rodriguez de la Rúa et al., 2000). Based on the results of a retrospective study of 298 RDs, and using multiple logistic regression analysis, we identified five risk factors: presurgical PVR grades A and B, RD affecting four quadrants, endolaser application, intraocular use of gas, and previous intraocular surgeries. According to the obtained odds ratio, a score was assigned to each factor and three levels of risk for developing PVR were established: low, medium, and high. The distribution of patients with RD following these criteria showed a clear positive correlation with the incidence of postoperative PVR.

And very recently a British group has published the results of a prospective study in patients undergoing primary vitrectomy for RDs, and using the above-mentioned mathematical model (Kon et al., 2000), they have showed that it is possible to identify subjects at greater risk of PVR developing after primary vitrectomy (Asaria et al., 2001b).

Despite the obvious limitations of these small prospective studies, we believe that a multifactorial approach to the problem could be useful to identify which patients must receive specific pharmacologic agents to avoid the development of postoperative PVR.

In addition to the identification of clinical risk factors by multivariate analysis, several authors have tried to establish possible relations between PVR and several independent factors. In the 1990s, studies focused on clinical factors, such as aphakia or pseudophakia (Girard and Karpouzas, 1995), giant tears (Yanyali and Bonnet, 1996), choroidal detachment (Dumas and Bonnet, 1996), vitreous hemorrhage (Duquesne et al., 1996), cryopexy (Bonnet et al., 1996), and incomplete posterior vitreous detachment (Capeans et al., 1998). These studies provided valuable clinical information, but they did not consider the multifactorial origin of PVR, and although most of the studies were prospective they could not assure in their samples the similarity of the remaining factors of RD that might influence the development of PVR. Besides, this kind of study did not facilitate modifications in surgeons’ attitudes about disease management that would result in better outcomes.

After this review of the literature, it is difficult to have a clear idea of the factors really associated with this complication, or the factors real contribution to the final outcome. In addition, it is difficult to clarify these questions.

The prevalence of PVR is relatively low, occurring in 10% of RDs, and RD prevalence is estimated to occur in one new case per 10,000 inhabitants per year. Therefore, published samples are relatively small and many factors cannot be clearly associated despite statistical analysis. Prospective or retrospective studies that are sufficiently large occur over a long period of time, with the potential to modify the criteria on the surgical approach to the RD treatment.

Case-control studies may be useful. Surprisingly, we found only one such paper (Yoshino et al., 1989). Nevertheless several problems exist because the authors included only preoperative PVR (named primary), which occurs less frequently than postoperative PVR, and there is a clear disproportion in the number of patients of the control group (n=1353) compared with the case group (n=57). The authors reported that the following factors are related to the development of preoperative PVR: RD of long evolution (more than 1 month), aphakia, vitreous hemorrhage, giant tears, large tears (higher than 3 disc diameters), and horseshoe tears.

PVR has three major locations, pre-, sub-, and intraretinal sites. All forms are considered to be the same disease, but almost all the available information is related to preretinal forms.

The risk factors leading to the development of the different types of membranes found in PVR (preretinal, subretinal, and intraretinal) are also confusing and not well defined. Among the membranes in PVR, subretinal membranes have received relatively little attention. Their clinical significance has been emphasized by some investigators (Lewis et al., 1989), and although it is true that in many cases they do not need a direct surgical procedure, in other cases subretinal strands prevent proper flattening of the retina or cause macular distortion and need to be removed (Machemer, 1980; Lewis et al., 1989).

The incidence of subretinal membranes is 15.7% of all RDs (Miura and Ideta, 2000). Logistic regression analysis showed a positive association with the number of quadrants detached, longer duration of the RD, the presence of atrophic retinal breaks, and younger patient age. Subretinal forms of PVR occur more frequently after surgery, and subretinal strands and bands are relatively common after penetrating intraocular surgery (Lewis et al., 1989).

Unfortunately, no information about prevalence or risk factors is available on the third form of PVR, intraretinal proliferation.

With the idea of using more objective parameters than clinical evaluation for establishing risk factors, some authors measured the vitreous levels of several substances. Vitreous proteins have been proposed as a marker of the state of inflammation, the breakdown of the blood-retina barrier, and the severity of the wound-healing process (Kon et al., 2000). In fact, higher concentrations of vitreous proteins have been found in patients who developed postoperative PVR compared with those who did not (Kon et al., 2000).

The vitreous levels of intercellular adhesion molecule 1 (ICAM-1) have also been proposed as a risk indicator for PVR (Limb and Chignell, 1999). Those authors recognized that cytokines and several inflammatory factors are found in the vitreous of patients with PVR, but they stated that none has been shown to be an indicator of the risk for development of this complication. Results show that patients with a high risk of developing PVR had significantly higher levels of soluble ICAM-1 than patients at low risk. Nevertheless, the individual values of ICAM-1 showed marked variations, and although the authors consider it a highly specific test they recognize its relatively low sensitivity. However, they encouraged the development of laboratory tests to identify those patients at higher risk of developing PVR, which would assist the vitreoretinal surgeon in the management of RD.

Flow cytometry is useful to measure the proliferation of cells recovered from the vitreous cavity in an experimental model of tractional RD (Yang et al., 1992). The same group of researchers compared findings of flow cytometry in patients with RD complicated with PVR, and the ability of a surgeon to predict the surgical outcome (Cousins and Rubsamen, 1994). Results showed that flow cytometry quantified the concentration of total and proliferating cells, but its accuracy to predict anatomic success did not show differences with the surgeon impression.

For now these and some other measurements of inflammation-related factors are not routinely used for decision making to modify the surgical approach or administer adjunctive therapy other than the standard postoperative antiinflammatory treatment (Pastor, 1998).

Finally, if PVR is accepted as an exaggerated wound-healing process, a genetic approach could be considered. To our knowledge, this question has not been addressed in the literature, and the only paper about the development of PVR in the fellow eye reported a prevalence of 5.4% of bilateral PVR that increased to 10.1% in a subgroup of patients who developed vision-threatening pathology during follow-up (Schwartz and Kreiger, 1998). This subject needs further investigation.

In summary, identifying clinical risk factors for PVR seems crucial for establishing pharmacologic prophylaxis that today cannot be acceptable for all patients with RD. In addition, it could help surgeons select the most appropriate surgical procedure for each case of RD.

As mentioned previously, PVR is considered a scarring process amplified by several factors, some of them likely related to a high level of intraocular inflammation. PVR also can be considered as a wound-healing process that occurs after RD (Pastor, 1998). Therefore, the critical factor seems to be the presence of a retinal break. Its healing process, with some modifications, must be the phenomenon that triggers PVR onset (Goldaracena et al., 1994; Pastor, 1998).

After any tissue injury, a three-phase cascade of mechanisms repairs the wound: inflammation, proliferation, and modulation of the scar (Wilkins and Kulwin, 1979). These phases are recognized in PVR, with some modifications because of the characteristics of the ocular tissues (Pastor, 1998).

Nagasaki and colleagues (Nagasaki et al., 1998) pointed out the importance of the breakdown of the blood-ocular barriers as the first step in the pathogenesis of PVR. It seems obvious that alteration of these barriers modifies the intraocular environment and allows creation of intraocular conditions for the development of PVR. Nevertheless, this barrier breakdown is a constant in all RDs and only a small proportion develops PVR. In addition, intraocular barriers are disrupted in many other conditions (uveitis, aphakia, diabetic retinopathy), and PVR has been never reported as a complication in these diseases. Therefore, the real role of the breakdown of the blood-ocular barriers in this disease must be elucidated.