Correlation of Increased Vascular Endothelial Growth Factor With Neovascularization and Permeability in Ischemic Central Vein Occlusion

Objective  To establish a role for vascular endothelial growth factor (VEGF) during the onset and clinical course of neovascularization of the iris (NVI) in ischemic central retinal vein occlusion.

Methods  Sixteen patients with ischemic central retinal vein occlusion were followed up for 12 months by clinical examination, retinal and iris angiography, and serial anterior chamber sampling of aqueous humor. Aqueous VEGF level was determined by enzyme-linked immunoassay, and permeability changes were estimated by capillary zone electrophoretic assessment of aqueous albumin.

Results  A correlation was found between aqueous VEGF concentrations and the onset, persistence, and regression of NVI; extent of retinal capillary nonperfusion; and vascular permeability. The NVI occurred when aqueous VEGF concentrations were 849 to 1569 pg/mL and regressed fully when they fell below 550 pg/mL. Aqueous concentrations of serum albumin, a marker of increased permeability, correlated with increased VEGF. Placental growth factor was found at low levels only when VEGF levels exceeded 330 pg/mL. The NVI remained VEGF-dependent during the course of the disease, regressing only if VEGF concentrations were reduced after laser ablation of hypoxic retina.

Conclusions  The close temporal correlation between aqueous VEGF levels and the course of neovascularization and permeability in human ischemic central retinal vein occlusion indicates that increased aqueous VEGF level may predict the need for treatment, and that anti-VEGF therapy at an early stage of ischemic central retinal vein occlusion may be therapeutically beneficial.

VASCULAR endothelial growth factor (VEGF) is strongly suspected to be a key mediator of angiogenesis and increased vascular permeability in human diseases, including ocular neovascular disorders. Despite this, there are few opportunities in human disease to observe VEGF concentrations over time and to compare these data against the clinical onset, persistence, or regression of neovascularization, and changes in vasopermeability. Therefore, definitive evidence of the role of VEGF is lacking for many human neovascular diseases.

Ischemic central retinal vein occlusion (iCRVO) is an isolated ocular disorder in which rapid venous obstruction of unknown etiology is associated with decreased retinal perfusion, capillary closure, and retinal hypoxia. When severe, this leads to profound ocular vascular leakage, neovascularization, intractable elevation of intraocular pressure, and a blind painful eye.^1-4 It is the single most common cause of surgical removal of the eye in North America. Human operative or pathological specimens show that VEGF concentrations in aqueous fluid are elevated in the presence of neovascularization of the iris(NVI) and are undetectable in the normal eye^5-9;an association of VEGF with NVI has been reported in animal models of iCRVO,^10-13 but a role for VEGF during the onset of NVI and the clinical course of human iCRVO has not been established.

Ischemic central retinal vein occlusion provides a unique opportunity to investigate human angiogenesis. It is characterized by quantifiable regions of retinal ischemia determined by the area of capillary nonperfusion on fluorescein angiography, the development of new vessels on the iris, a stable volume of ocular fluid into which soluble growth factors are secreted, and the presence of optically clear structures, the cornea and lens, through which changes in the iris and retinal vasculature can be visualized. In addition, because no current therapy can restore vision^14-17 and prophylactic treatment of NVI is rarely practiced, iCRVO provides an opportunity to investigate clinical-molecular correlates in the natural course of neovascular disease.

To establish a role for VEGF in the neovascularization associated with human central retinal vein occlusion, we measured aqueous VEGF, basic fibroblast growth factor, and placental growth factor (PlGF) during the untreated and treated course of iCRVO.

Consecutive patients presenting to the Medical Retina Clinic, Moorfields Eye Hospital, London, England, were diagnosed as having central retinal vein occlusion by the presence of intraretinal hemorrhage and dilated, tortuous vessels in all 4 retinal quadrants. Patients with significant ischemia, defined as greater than 10 disc areas of retinal capillary nonperfusion determined by intravenous fluorescein angiography,^15,17-20 were offered enrollment. Patients with extensive intraretinal hemorrhage that precluded precise assessment of the extent of retinal nonperfusion were also offered entry, as 80% of this indeterminant group are known to proceed to ischemia.¹⁵ Patients with concurrent systemic or ocular disease, such as diabetes mellitus, ocular inflammation, or macular degeneration, were excluded.

Sixteen patients were entered into the study with informed consent and, in accordance with the guidelines of the Declaration of Helsinki and the Association for Research in Vision and Ophthalmology, followed up for a mean of 62 weeks(range, 28-90 weeks). Patients with no NVI at presentation were assigned to the "untreated history" arm; those with NVI entered the "treated history" arm and received panretinal laser photocoagulation (PRP). After a minimum of 6 months, patients were assigned to 1 of 4 outcome groups (Figure 1). Patients in the untreated history arm either did not(outcome group 1) or did (group 2) develop NVI; patients with NVI demonstrated either vessel regression (group 3) or persistence (group 4) after PRP. Anterior chamber paracentesis for analysis of aqueous humor in the untreated history group was performed monthly for up to 6 months or when NVI developed. Patients who initially had or who developed NVI were sampled up to 6 times during a minimum period of 6 months. The interval between paracenteses in the treated history group was determined by pace of the clinical response.

Iris neovascularization was determined by means of slitlamp biomicroscopy by 2 examiners (S.R.B. and P.G.H.) and graded²¹ by location (pupillary margin, midstroma, or iris base) and extent (number of iris clock hours involved). Neovascularization in the angle was assessed by means of 4-mirror gonioscopy before dilation at all clinical visits.²² Color photography of the iris was performed at every visit, and iris angiography was performed with development of clinically evident NVI, after its regression, or in the presence of normal vasculature at the end of 6 months.

Patients received PRP after NVI developed (either 2 clock hours of the iris, or any in the angle) according to published guidelines.¹⁵ Direct argon green laser (500-µm spot size, 0.1 second, 300-950 mW) was used initially, with subsequent treatments performed by either direct or indirect methods.

After topical anesthesia and cleansing of the ocular surface (20% povidone-iodine solution for 1 minute), 80 to 150 µL of aqueous fluid was withdrawn through clear cornea. Aqueous samples were centrifuged at 12 000g for 3 minutes, and the supernatants were frozen at −70°C.

Blood was collected in sodium citrate tubes to prevent platelet disruption and centrifuged at 5000g for 8 minutes, and the plasma was stored at −70°C. Venous blood samples were obtained at presentation and/or at the time of NVI.

Levels of aqueous VEGF, basic fibroblast growth factor, and PlGF were determined at each time point in triplicate by enzyme-linked immunosorbent assay with the use of commercial kits (R&D Systems, Minneapolis, Minn). Intra-assay and interassay variability was less than 5% to 10%. No detectable cross-reactivity was observed between aqueous VEGF and PlGF up to a concentration of 2000 pg/mL.

Samples (5 nL) of aqueous humor were analyzed by capillary zone electrophoresis using a 214-nm detector (P/ACE system 5510; Beckman-Coulter, Fullerton, Calif). They were run in an uncoated fused silica capillary (27 cm × 20 µm) with 100mM sodium borate buffer, pH 10.2, at 10 kV and 8.5 µA. The location of serum albumin in the profile was determined by running pure albumin (Sigma-Aldrich Corp, St Louis, Mo) and aqueous humor spiked with albumin. The amount of albumin in the sample was estimated by determining the area of the peak with a migration time of 4.5 minutes.

Data were treated nonparametrically. The Wilcoxon rank sum test was performed to assess differences in mean growth factor concentrations between different outcome groups; the signed rank test was used to compare the simultaneous aqueous and plasma aqueous VEGF concentrations for each patient. The association between disc areas of nonperfusion and aqueous VEGF concentration was assessed with the Spearman rank correlation coefficient.

Sixteen patients were recruited; the median age was 69 years (range, 31-82 years), the male-female ratio was 11:5, and 6 right eyes were involved. Patients were followed up clinically for 28 to 90 weeks, and patients in the untreated history arm who did not develop NVI were followed up for at least 1 year, with final aqueous sampling performed after 6 months (range, 30-36 weeks). Patients with persistent NVI were followed up to 65 weeks after treatment.

Of 9 patients in the untreated history arm, 3 did not develop NVI (group 1) and 6 did (group 2). These 6 group 2 patients, and 7 with NVI at initial presentation, entered the treated history arm and received laser therapy. By symptomatic history, the median time to NVI was 13 weeks (range, 6-38 weeks). Of these 13 patients, 10 showed complete vessel regression after PRP (group 3) and 3 had persistent vessels (group 4) (Figure 1).

Consideration of all patients showed that the extent of retinal capillary nonperfusion on baseline fluorescein angiography correlated with initial VEGF levels determined within 1 week of angiography (Figure 2) (Spearman rank correlation coefficient = 0.72, P<.01).

The final aqueous VEGF concentrations of untreated patients who did not develop NVI after more than 6 months of observation (group 1; median, 88 pg/mL; range, 87-109 pg/mL) were markedly less than the VEGF concentrations determined either at the onset of NVI in patients who developed NVI under clinical supervision (group 2; median, 1201 pg/mL; range, 849-1569 pg/mL) or at presentation of NVI in patients who arrived at the clinic with preexistent NVI (group 3; median, 1464 pg/mL; range, 849-4297 pg/mL) (Table 1).

Wilcoxon rank sum analysis identified a significant difference (P<.02) between the final aqueous VEGF concentrations of group 1 patients, as compared with the VEGF levels in patients progressing to or presenting with NVI (median, 1359 pg/mL; range, 849-4297 pg/mL). Among group 2 patients, VEGF had reached concentrations greater than 849 pg/mL at the time of detection of NVI, and PlGF was also detectable in association with VEGF levels greater than 330 pg/mL (median, 88 pg/mL; range, 57-240 pg/mL).

Treated patients showing vessel regression after PRP (group 3) had significantly lower VEGF concentrations (median, 189 pg/mL; range, 135-550 pg/mL) than treated patients (group 4) with persistent vessels after 6 months of observation (median, 1295 pg/mL; range, 1025-1569 pg/mL; P<.03) (Figure 3 and Table 1). Aqueous fluid from control patients undergoing routine cataract surgery (n = 16) had low aqueous VEGF levels, ranging from less than 5 pg/mL to 96 pg/mL. Aqueous concentrations of VEGF showed no correlation with plasma concentrations (mean, 51 pg/mL; range, 28-120 pg/mL). Placental growth factor was detected only when VEGF level was greater than 330 pg/mL in 5 of 51 samples taken during active neovascularization. Samples taken with vessel regression were negative in 46 samples tested. Basic fibroblast growth factor was not detected in any of 49 samples tested.

Group 1 patients (untreated, no NVI) demonstrated 10 to 20 disc areas of retinal capillary nonperfusion and did not develop NVI at any time through follow-up (Figure 4A-D). In 2 of the 3 patients with elevated VEGF levels (270 and 749 pg/mL) at the onset of iCRVO, VEGF spontaneously decreased to a final lower level (88 and 87 pg/mL, respectively).

Group 2 patients (those who developed NVI) had more than 30 disc areas of retinal nonperfusion or had indeterminate amounts (2 cases) because of blockage by intraretinal hemorrhage on baseline fluorescein angiography. In 2 cases, the VEGF level increased to greater than 1100 pg/mL approximately 4 weeks before clinically evident NVI occurred and was similar when NVI was evident (Figure 4H).

Group 3 patients (treated with vessel regression) had greater than 30 disc areas of retinal nonperfusion except for 2 cases (10 and 20 disc areas of retinal nonperfusion, respectively). Within this group, NVI regressed completely in all 10 patients after PRP (Figure 4I-K). Although vessel regression began shortly after first application of PRP in some patients, vessel regression was considered complete only when no perfused vessels could be clinically observed, and this took 5 to 50 weeks, with a median time of 23 weeks. After retinal laser therapy, VEGF levels decreased at variable rates. Six days after laser treatment, the earliest sample taken, the aqueous VEGF concentration from 1 patient had dropped to less than half its original value. By contrast, VEGF levels fell more slowly in patients with extensive retinal edema or hemorrhage who required multiple sessions of laser. Overall, at the time of full vessel regression, VEGF levels were less than 550 pg/mL (median, 204 pg/mL; range, 135-550 pg/mL). Basic fibroblast growth factor and PlGF were not detected after vessel regression.

Three treated patients had persistent extensive NVI (group 4; Figure 4M-O) despite extensive laser treatment. The VEGF levels demonstrated an initial decrease after PRP that spontaneously returned to higher levels in excess of 1000 pg/mL (Figure 4P) and persisted in 1 case more than 1 year after the onset of iCRVO. In all group 4 patients, VEGF remained greater than 1000 pg/mL at all times of sampling. These patients had extensive accumulation of subretinal fluid and secondary macular detachments as demonstrated on ocular ultrasound and optical coherence tomography that precluded uptake of laser energy (data not shown). The PlGF was also detected. The clinical changes and the corresponding VEGF concentrations in a representative patient are shown (Figure 4D).

Changes in vascular permeability associated with iCRVO were assessed by measurement of intraocular serum albumin by capillary electrophoresis in a subset of patients (n = 8). With the use of pure albumin, the migration time of this protein was found to be 4.5 minutes (Figure 5), and the amount of albumin in samples was estimated by determining the area of the peak at this position. The amount of albumin in samples of aqueous humor correlated with VEGF concentrations, rising and falling in parallel (Figure 6). In a group 1 patient (Figure 6A), albumin decreased from an initially higher level in parallel with the spontaneous decrease in intraocular VEGF, whereas in a patient in groups 2 and 3, albumin level increased concomitantly with the increase in VEGF before the appearance of NVI and fell after PRP in parallel with a decline in VEGF (Figure 6B). Angiographically demonstrable leakage from the retinal and iris vessels also correlated with elevated concentrations of VEGF (data not shown). Patient 8, the single patient to develop neovascular glaucoma, had the highest levels of albumin, and aqueous VEGF level was 1320 pg/mL at the onset of NVI, although high intraocular pressure precluded further sampling.

In the present study we have shown that aqueous VEGF concentrations correlate temporally with the onset, persistence, and regression of NVI in patients with iCRVO. The median concentration of VEGF in group 2 patients who developed NVI under clinical observation was 1201 pg/mL at the onset of neovascularization, and the apparent threshold for NVI in this group varied from 849 to 1569 pg/mL. These threshold concentrations are similar to that found for NVI in the primate model (>30 pmol/L [approximately 1380 pg/mL])¹² and to the dissociation constant (K_d) for binding of VEGF to VEGF receptor 1, reported as 16 to 114 pmol/L but somewhat lower than its dissociation constant (K_d) of 400 to 1000 pmol/L for VEGF receptor 2.^23-25 While it is likely that measurements in aqueous humor underestimate local, intraocular concentrations, these findings indicate that the onset of ocular neovascularization is associated with VEGF concentrations considerably lower than those previously reported with single time-point analysis of patients with advanced disease.^26-28 We also detected VEGF levels of more than 10 000 pg/mL in the vitreous humor of diabetic patients with proliferative retinopathy and from aqueous and vitreous of eyes containing intraocular tumors (S.R.B. and I.A.C., unpublished results, 1999). The lower VEGF concentrations found in the present study are likely to reflect the close surveillance and evaluation of aqueous rather than vitreous humor.

Measurements in patients who responded to PRP treatment (group 3) suggest that once NVI has become established, VEGF must return to relatively low concentrations(at least 378 pg/mL) to ensure complete vessel regression. This indicates a lower VEGF threshold for vessel disappearance than that associated with the onset of neovascularization. In contrast, patients with persistent NVI after PRP maintained VEGF levels above 1000 pg/mL. Panretinal photocoagulation is believed to reduce aqueous VEGF by ablation of retinal tissue, thereby reducing vitreous and aqueous VEGF levels. Taken together, data from groups 3 and 4 suggest that, during the time course of this study, NVI remains VEGF-dependent, persisting or regressing in parallel with, respectively, chronically elevated or decreased VEGF concentrations.²⁹

Among all patients, there was a positive correlation between disc areas of retinal capillary nonperfusion and VEGF concentration, suggesting that cells in the hypoxic retina are the likeliest source of local VEGF production. Consistent with this possibility, up-regulation of VEGF messenger RNA has been reported in the retinal ganglion cell and nuclear layers in human central retinal vein occlusion.^5,8 A systemic source of VEGF was highly unlikely, as intraocular levels of VEGF showed no relation to concentrations found in blood.

Patients who did not develop NVI (group 1) had an initial, transient elevation of VEGF level immediately after onset of iCRVO. The cause of this initial peak is unclear, but it may reflect the immediate cellular response to reduced oxygen delivery due to sudden venous occlusion. The subsequent decline in VEGF could be due to relatively increased perfusion, death of hypoxic VEGF-secreting cells,³⁰ or a degree of unknown"metabolic adjustment" that compensates for reduced oxygen delivery.³¹

Consistent with the ability of VEGF to increase vasopermeability,^32-35 intraocular serum albumin and VEGF levels were concomitantly elevated, both in the absence and in the presence of NVI. Increased permeability was associated with relatively lower concentrations of VEGF than was neovascularization, as indicated by the correlation of an initially elevated VEGF concentration in a group 1 patient with increased intraocular serum albumin level (Figure 6A). Aqueous albumin level (Figure 6B) and angiographic demonstration of vascular leakage paralleled the rise and fall of VEGF. Transient elevations of VEGF level noted in patients who did not develop neovascularization indicate that increased permeability can occur 2 to 4 weeks after the onset of symptoms (the earliest time points measured). Clinically, vascular leakage is well-known to occur within hours of the symptomatic onset of iCRVO. These data suggest that increased permeability is an early event in VEGF-mediated vasculopathy, preceding the cellular events necessary for angiogenesis.

The PlGF was detected in patients with iCRVO if active neovascularization was present and VEGF level was greater than 330 pg/mL, in contrast to observations of experimental iCRVO in the nonhuman primate, where PlGF was undetectable.¹³ However, PlGF has been described late in the course of human diabetic retinopathy.³⁶ Since PlGF/VEGF heterodimers have been shown to be angiogenic,³⁷ it is possible that PlGF may cooperate with VEGF in the stimulation or perpetuation of neovascularization associated with human iCRVO. Although we were unable to detect basic fibroblast growth factor in our study, it may be that other members of the fibroblast growth factor family are present³⁸ or that basic fibroblast growth factor is transiently expressed at times not assessed.

The identification of a close correlation between aqueous VEGF and the clinical course of NVI in iCRVO is strong evidence of the role of VEGF in the pathogenesis of human ocular angiogenic disease. Furthermore, we have determined the intraocular VEGF concentrations at which pathophysiologic neovascularization will occur in iCRVO. These findings contribute to the understanding of the role of VEGF in ischemic retinal disease and could have implications for clinical management. Aqueous VEGF measurement may be appropriate in selected iCRVO cases as a predictor of NVI and the need for retinal laser and possibly antiangiogenic therapy. Our data has implications for patients receiving systemic pro-VEGF therapy. It suggests that ophthalmic surveillance should include early assessment of ocular permeability, within 4 to 6 weeks of treatment, rather than late assessment of neovascularization. Patients with systemic vasculopathy, such as with advanced diabetes mellitus, are those most likely to require VEGF therapy and are also most likely to have concurrent eye disease. We believe these patients will be at significantly higher risk of blinding complications. A sound evidence-based approach to their timely and appropriate assessment will be invaluable for informed decision making, by both patient and physician.

Submitted for publication March 8, 2002; final revision received July 11, 2002; accepted August 6, 2002.

This study was supported by Moorfields Eye Hospital, Special Trustees. Drs Zachary and Martin are supported by the British Heart Foundation, London.

This study was presented at the Association for Research in Vision and Ophthalmology Meeting, Fort Lauderdale, Fla, May 11, 1999.

We thank Sir John Vane, FRS, for critical reading of the manuscript and Catey Bunce for statistical advice.

Corresponding author: Philip G. Hykin, FRCS, Moorfields Eye Hospital, City Road, London EC1V 2PD, England.