Topical Bevacizumab in the Treatment of Corneal Neovascularization: Title and subTitle BreakResults of a Prospective, Open-Label, Noncomparative Study FREE

The cornea has the unique feature (except for cartilage) of being normally avascular, but under pathologic conditions, vessels invade the cornea from the limbal vascular plexus. A wide variety of insults, including infection, inflammation, ischemia, degeneration, trauma, and loss of the limbal stem cell barrier, can cause corneal neovascularization (NV).¹ Although corneal NV can occasionally serve a beneficial role in the clearing of infections, wound healing, and arresting stromal melts,² its disadvantages are numerous. Corneal NV often leads to tissue scarring, edema, lipid deposition, and persistent inflammation that may significantly alter visual acuity.³ Based on data derived from the Massachusetts Eye and Ear Infirmary in 1996, it is estimated that for any given year, 1.4 million patients in the United States develop corneal NV; 12% of these cases are associated with a decrease in visual acuity.⁴ Twenty percent of corneal specimens obtained during corneal transplantation show histopathologic evidence of NV.⁵ Corneal NV accompanies the most common causes of corneal infectious blindness in both the developed (herpetic keratitis)⁶ and developing (trachoma and onchocerciasis) world,⁷ which causes millions to lose their sight. Corneal NV is also notable in extended wear of hydrogel contact lenses.^{8 - 9} The prevalence of NV ranges from 125 000 to 470 000 people in the United States who wear soft contact lenses for refractive correction.⁴ All these data indicate that corneal NV is a significant contributor to eye disease.

Corneal NV may not only reduce visual acuity but it also results in the loss of the immune privilege of the cornea, thereby worsening the prognosis of subsequent penetrating keratoplasty.¹⁰ Preexisting corneal stromal blood vessels have been identified as strong risk factors for immune rejection after corneal transplantation.^{11 - 12} For instance, whereas the success rate of corneal transplantation in low-risk avascular beds surpasses 90%, the survival rates are drastically lower in high-risk neovascularized beds in which corneal grafts suffer from rejection rates far worse than first kidney or heart allografts.^{11 - 12}

Current treatments for corneal NV, including medications (such as steroids or nonsteroidal anti-inflammatory agents), laser photocoagulation, fine-needle diathermy, photodynamic therapy, or restoration of the ocular surface with the use of conjunctival, limbal, or amniotic membrane transplantation, have demonstrated variable and largely limited clinical success.¹ The highly variable efficacy and myriad adverse effects (cataract, glaucoma, and increased risk of infection) of topical and systemic corticosteroids are well known to clinicians who use these agents regularly in trying to arrest these disease processes. Other treatment modalities are often ineffective, or vessel recanalization occurs,requiring multiple treatment sessions, which can lead to serious adverse effects. Furthermore, none of these treatments specifically target the molecular mediators of angiogenesis.¹³

Vascular endothelial growth factor (VEGF) is thought to be a key mediator in the process of NV.¹³ The prominent role of VEGF in the pathophysiology of corneal NV has been demonstrated in experimental models of corneal NV.¹⁴ It has been shown that VEGF is upregulated in inflamed and vascularized corneas in humans and animal models.¹⁵ It has also been shown that inhibition of angiogenesis by neutralization of VEGF can promote corneal graft survival in animal models.¹⁶ Vascular endothelial growth factor inhibitors, such as pegaptanib sodium, ranibizumab, and bevacizumab, are currently used for the treatment of neovascular age-related macular degeneration.¹⁷ The first 2 agents have been approved by the Food and Drug Administration for use in neovascular age-related macular degeneration; the third drug, which is a full-length humanized antibody against VEGF, has been approved for use in oncology but is also widely used off label to treat choroidal NV,¹⁸ central retinal vein occlusion,¹⁹ proliferative diabetic retinopathy,²⁰ and iris NV²¹ with encouraging results. Bevacizumab's use has now been widely adopted and is arguably part of the standard of care for the treatment of neovascular age-related macular degeneration in many patients.^{17 - 18}

Recently, off-label use of topical as well as subconjunctival bevacizumab has also been considered to be a new treatment modality for corneal NV.^{22 - 27} While there is substantial evidence for the intravitreous administration of bevacizumab in the treatment of choroidal NV, data regarding the safety and efficacy of topical bevacizumab in the treatment of corneal NV are, as yet, preliminary. Topical bevacizumab was demonstrated to inhibit corneal NV after chemical injury in an experimental rat model.²⁴ In humans, a small number of studies have shown that topical bevacizumab can reduce corneal NV in a few patients with significant corneal NV.^{25 - 26} However, many aspects of topically administered bevacizumab in the treatment of corneal NV—including long-term safety and efficacy against actively growing as well as established corneal NV, optimal dosing for modulating the neovascular process, and long-term stability of treatment results—have not been well known. The purpose of this article is to report the long-term (6-month) results of the safety, efficacy, and stability of treatment of clinically stable corneal NV in 10 patients using topical bevacizumab in a prospective, open-label clinical study.

This was an open-label, single-site, uncontrolled, single group–assignment, safety/efficacy study of topically administered bevacizumab in subjects with corneal NV. This study was approved by the institutional review board of the Massachusetts Eye and Ear Infirmary. Potential patients signed an informed consent form at the time of the screening visit.

Adult patients of either sex were eligible for participation if they had clinically stable corneal NV. Patients with superficial or deep corneal NV that extended farther than 2 mm from the limbus were considered. However, patients were considered not to have clinically stable corneal NV if they met any of the following conditions: (1) current or recent (≤3 months) episode of corneal and ocular surface infection (bacterial, viral, fungal, or acanthamoebal); (2) ocular surgery in the study eye, including cataract surgery, full-thickness or lamellar keratoplasty, ocular surface reconstruction, limbal stem cell transplantation, or amniotic membrane transplantation within 6 months prior to study entry; (3) current or recent (≤3 months) use of contact lens; and (4) current or recent (≤3 months) persistent corneal epithelial defects (of at least 14 days in duration measuring more than 1 mm²). Other exclusion criteria are listed in Table 1. Qualified patients who met these criteria were invited to participate in the study and, if they consented, were enrolled consecutively to begin the course of topical bevacizumab. The study eye was identified at the screening visit. If both eyes were eligible for the study, the eye most affected by corneal NV was selected for entry.

A topical bevacizumab solution was formulated and aseptically prepared from commercially available intravenous bevacizumab (Avastin; Genentech Inc, San Francisco, California) and transferred into a sterile, light-protected dropper container by the Massachusetts Eye and Ear Infirmary pharmacy staff. Based on the earlier case report on the efficacy and safety of topical bevacizumab,²⁶ a similar formulation of 1.0% of bevacizumab (10 mg/mL) with 0.01% of benzalkonium chloride (0.1 mg/mL) with a pH of 6.2 was used. The patients were instructed to refrigerate the study drugs at 2°C to 8°C (36-46°F). To reduce the chance of systemic absorption of bevacizumab, both puncta of the study eye were temporarily plugged for the duration of treatment (3 weeks). The dosing of topical bevacizumab, 1%, used was 2 and 4 times a day. To avoid the potential systemic and local adverse effects of long-term VEGF blockade, the treatment course was limited to 3 weeks. During the study, all concomitant medication treatment regimens were kept as constant as permitted by accepted medical practice.

Follow-up visits were scheduled at week 1, 3, 6, 12, and 24. At all visits, in addition to a comprehensive eye examination, a detailed review of medical and ophthalmic histories and current medications was recorded. Central corneal thickness was also measured by ultrasonic pachymeter at all visits. Digital corneal photographs at the slitlamp microscope were taken at the screening visit as well as visits in weeks 3, 6, 12, and 24. Systolic and diastolic blood pressure measurements were obtained at all visits. This study did not include blood sampling or any pharmacokinetic measures.

Safety was monitored via occurrence of adverse events. All adverse events (ocular and systemic) were monitored and recorded throughout the course of the study, including their seriousness and severity, action taken, and relationship to study treatment. Ocular adverse events were identified by eye examination, visual acuity testing, intraocular pressure, biomicroscopy, and corneal fluorescein staining through week 24. Systemic adverse events were identified by physical examination, subject reporting, and changes in a vital sign (blood pressure) through week 24.

The primary efficacy variables were the size and extent of corneal NV. By comparing baseline corneal photographs with follow-up photographs, the efficacy of bevacizumab in treatment of corneal NV was evaluated. Other efficacy variables included measuring the changes in best-corrected visual acuity and central corneal thickness from baseline to the last visit.

Three primary metrics for corneal NV (Figure 1) were considered. The first, referred to as neovascular area(NA), involves measuring the area of the corneal vessels themselves when projected into the plane of a photograph. The second metric, referred to as vessel caliber(VC), involves determining an approximate mean diameter of the corneal vessels. The third metric, referred to as invasion area(IA), measures the fraction of corneal area in which vessels are present. Digital slitlamp corneal pictures were analyzed morphometrically using graphics editing software (Photoshop CS2; Adobe Systems Inc, Berkeley, California) and a mathematical program (written using Matlab; MathWorks Inc, Natick, Massachusetts). After the total corneal area was outlined, by applying the same graphical editing procedure the blood vessels were enhanced and traced by using Photoshop tools and filters. By setting a threshold level, the nonvessel area was erased, and the remaining NA was then pixelated and measured (Figure 1A). Finally, the calculated blood vessel area was normalized to the whole corneal area to obtain the NA score for each corneal picture. We also estimated VC by using a computational technique to measure the largest diameter circle centered at each pixel inside a blood vessel. The mean value across all pixels within blood vessels was taken as an estimate of the mean VC for a given image. Last, the IA was also quantified; the ends of all vascular sprouts were marked, and by connecting all these marks, the contour of the IA was traced and the measured area was again normalized to the whole corneal area.

To assess changes in the 3 metrics, 3 different time points were considered: (1) the initial (screening) visit, (2) the 3-week visit (end of treatment), and (3) the last follow-up visit. Paired ttests were performed with 1-sided alternatives comparing cohort scores for each metric individually. In each case, the 1-sided alternative hypothesis was that there was a reduction in cohort scores for a given metric from the earlier time point to the later time point. P ≤ .05 was considered statistically significant.

Ten eyes of 10 patients (4 men and 6 women) with stable corneal NV were included in this study. The demographic characteristics of the study population, including age, sex, eye, background disease for corneal NV, and the frequency of topical bevacizumab use, are listed in Table 2. The mean age was 46.7 years (standard deviation [SD], 13.7 years); age ranged from 23 to 71 years. The mean follow-up period was 22.8 weeks (SD, 3.8 weeks), ranging from 12 to 24 weeks.

The patients showed a significant reduction in NA from the screening visit to the last visit (P = .001) (Figure 2A). The 95% confidence interval for change in NA was 25.9% to 100% reduction. The mean reduction seen across the cohort in NA was 47.1% (SD, 36.7%), ranging from 11% to 98%. A significant decrease was also found in NA from the screening visit to the 3-week visit (end of treatment) (P = .03). The mean change in NA from the initial visit to the 3-week visit was 27.8% (SD, 41.4%). Lack of significant change in NA from the 3-week visit to the last visit when tested against a 2-sided alternative (P = .27) indicates stability of the treatment result.

The patients showed a significant reduction in VC from the screening visit to the last visit (P < .001) (Figure 2B). The 95% confidence interval for change in VC was 37.8% to 100% reduction. The mean reduction seen in VC was 54.1% (SD, 28.1%), ranging from 22% to 99%. The reduction in VC from the initial visit to the 3-week visit was not found to be significant (P = .12). However, there was a significant decrease (P = .008) seen in VC from the 3-week visit to the last visit, indicating a slightly delayed, but sustained treatment outcome. The mean change in VC from the 3-week visit to the last visit was 35.2% (SD, 37.2%).

The mean change in IA was 4.4% (SD, 52.0%) and 12.2% (SD, 42.0%) from the baseline visit to the 3-week visit and to the last visit, respectively. No statistical significant change was seen in IA, whether from the initial visit to the 3-week visit (P = .30) or from the initial visit to the last visit (P = .19) (Figure 2C).

No significant difference was seen in the therapeutic responses (for any of the 3 metrics used) between patients using bevacizumab 4 times a day (n = 5) and those using the drug twice daily (n = 5) (P = .17 for NA, P = .07 for VC, and P = .09 for IA).

All visual acuity data were converted to logMAR equivalents of Snellen acuity for the purpose of analysis. Mean corrected logMAR visual acuity was 0.65 (SD, 0.49) at the screening visit, 0.68 (SD, 0.52) at the 3-week visit, and 0.63 (SD, 0.60) at the last visit. Changes in visual acuity from baseline to any of the follow-up visits were not found to be significant. Mean values for central corneal thickness were 481.9 μm (SD, 90.6 μm) at baseline, 487.3 μm (SD, 94.3 μm) at the 3-week visit, and 499.7 μm (SD, 93.2 μm) at the last visit, with no statistically significant difference observed among these.

Mean arterial pressure ([(2 × diastolic blood pressure) + systolic blood pressure] ÷ 3) at week 1, week 3, and the last visit were compared with that at the baseline visit. The mean of mean arterial pressure for all patients was 90.8 mm Hg (SD, 8.9 mm Hg) at baseline, 88.6 mm Hg (SD, 6.4 mm Hg) at week 1 (P = .42), 84.4 mm Hg (SD, 6.2 mm Hg) at 3 weeks (P = .09), and 93.5 mm Hg (SD, 11.9 mm Hg) at the last visit (P = .50). No significant changes were found in mean arterial pressure at any follow-up visit. No systemic or ocular adverse events, including thromboembolic events, hemorrhage, allergic reaction, ocular surface toxicity and epitheliopathy (superficial punctate keratopathy, epithelial erosion, or defect), and burning on instillation, were reported. Interestingly, self-reported compliance was extremely favorable; only 3 patients missed 1 or 2 doses of bevacizumab throughout the entire treatment period.

Vascular endothelial growth factor has demonstrated an intimate connection with the pathogenesis of corneal NV.¹⁵ In animal models of corneal NV, increased expression levels of VEGF and VEGF receptors have been confirmed.^{14 ,28 - 29} In humans, pathologic studies have confirmed that VEGF and its receptors are present in higher concentrations in corneal buttons with NV than in normal corneas, irrespective of the cause of NV.^{15 ,30} Furthermore, VEGF blockade, at both the protein and messenger RNA level, has been shown to reduce corneal NV and improve corneal graft survival in experimental animals.¹⁶

Bevacizumab is a full-length, recombinant humanized monoclonal immunoglobulin G1 (IgG1) that binds to and inhibits the activity of VEGF-A, thereby inhibiting angiogenesis.³¹ It was the first anti-VEGF antibody to be approved by the US Food and Drug Administration specifically for the treatment of metastatic colon cancer and recently for non–small-cell lung cancer and metastatic breast cancer. Off-label intravitreal administration of bevacizumab for treatment of choroidal NV¹⁸ has gained wide and rapid acceptance because of its safety, efficacy, and lower cost compared with other anti-VEGF drugs.

In the aggregate, our study shows that topical bevacizumab, 1%, is effective in the treatment of clinically stable corneal NV, as evidenced by a nearly 50% reduction in 2 corneal NV metrics (NA and VC). In the 2-dimensional plane of a corneal photograph, if vessel area (NA) is regarded as a function of mean vessel width (VC) and total vessel length, it would appear that the reduction in VC accounted for most of the improvement that was seen in NA. Furthermore, the absence of meaningful change in IA in our study indirectly supports the conclusion that significant narrowing of blood vessels rather than reduction in blood vessel length is the main outcome of anti-VEGF therapy in corneal NV. Vascular endothelial growth factor acts at several levels on vascular beds: it is a survival factor for endothelial cells, it is a potent vasodilator, and it increases microvascular permeability.³² Vascular endothelial growth factor, once considered to be a vascular permeability factor,³³ was later found to promote endothelial cell growth.³⁴ Vascular endothelial growth factor, therefore, renders the microvasculature hyperpermeable to circulating macromolecules with a potency about 50 000 times greater than histamine.³³ Therefore, the vascular stabilization (reduction in vascular permeability) affected by anti-VEGF therapy may diminish the vascular flow rate, causing reduction in the caliber of blood vessels.

Our study shows a highly variable efficacy across the cohort treated with bevacizumab (Figure 3, Figure 4, and Figure 5), which is evidenced by the high SDs of the computed corneal NV metrics. The level of therapeutic response in case 6 (almost a complete resolution of corneal NV, as shown in Figure 3) and case 8 (a very modest therapeutic response to the anti-VEGF treatment, as shown in Figure 5) exemplifies the therapeutic spectrum of topical bevacizumab in the treatment of corneal NV. Several hypotheses may explain this variability in response to topical bevacizumab, including heterogeneity in corneal NV etiologies, variable levels of VEGF expression in the pathobiology of diverse cases, and variable levels of drug penetration.

Topical application is the preferred method of administration of a drug to the cornea and the ocular surface. However, topical treatment will only be effective if the drug can penetrate the corneal epithelial barrier to reach the target tissues within a therapeutic level. Topical administration of full-length immunoglobulins, such as bevacizumab with a molecular weight of 149 kDa, is typically considered ineffective because such molecules are too large to penetrate the intact corneal epithelium. However, epithelium over the neovascularized area can be defective, particularly in patients with ocular surface disease, which interferes with normal corneal epithelial function and results in incompetent barrier function.³⁵ Our recent work in a mouse model of corneal NV has clearly demonstrated that bevacizumab can penetrate the neovascularized cornea after topical application.³⁶ In view of the size of the bevacizumab molecule, however, the degree of inadequacy of the corneal epithelial barrier likely varies to a large extent from one vascularized cornea to another.

Our study did not show any systemic or local adverse effects. Systemic blood pressure remained stable at the baseline level, and no serious life-threatening adverse effects occurred during the follow-up period. This shows that topical bevacizumab, 1%, in both twice daily and 4 times a day regimens, is safe. This finding, however, may be due in part to the precautions applied to our study, such as the placement of punctal plugs and the exclusion of any patient aged 75 years or older or with a history of hypertension, diabetes mellitus, or a thromboembolic event. Similarly, from an eye standpoint, topical bevacizumab, 1% (with benzalkonium, 0.01%, as a preservative), was tolerated very well in all patients. No local irritation, allergic reaction, or surface epitheliopathy was observed. This is in contrast with a 60% rate of spontaneous loss of epithelial integrity as recently reported by Kim et al.²⁵ In this article, the investigators used topical bevacizumab at a slightly higher concentration (1.25%) twice daily for a much longer period (3 months), and adverse effects generally appeared during the second month of treatment.²⁵ This suggests that the duration of treatment may well determine the safety of topical bevacizumab.

While anti-VEGF therapy shows efficiency in treating corneal NV, VEGF has desirable effects that may be blocked by bevacizumab therapy. These include the capacity to promote the formation of collateral vessels, control vascular tone, and affect corneal nerve regeneration,³⁷ and a substantial role in wound healing.³⁸ In this regard, it is important to note that the long-term neutralization of VEGF may have unintended local or systemic consequences that our study has not yet determined. Prolonged blockade of VEGF may impair wound healing and the regeneration of corneal nerves, which may cause a loss of epithelial integrity in the cornea. Though delivered in a small dose on the surface of the eye, anti-VEGF drugs could also pass into the systemic circulation. Hypertension, proteinuria, and various cardiovascular events are potential consequences of the systemic inhibition of VEGF.

In summary, the significant narrowing of corneal blood vessel diameter and diminishing NA in response to topical bevacizumab therapy provides evidence that anti-VEGF therapy could offer an alternative or adjunctive measure to conventional therapies in the treatment of stable corneal NV. However, further research using larger patient cohorts is warranted to determine the exact dosage and indications for use.

Correspondence:Reza Dana, MD, MSc, MPH, Cornea Service, Massachusetts Eye and Ear Infirmary, 243 Charles St, Boston, MA 02114 (reza_dana@meei.harvard.edu).

Submitted for Publication:October 12, 2008; final revision received December 29, 2008; accepted December 31, 2008.

Financial Disclosure:None reported.

Funding/Support:This study was supported in part by grant K24 EY019098 from the National Institutes of Health, by Prevent Blindness America, and by the Massachusetts Eye and Ear Infirmary Development Office.

Previous Presentation:This study was presented in part at the Association for Research in Vision and Ophthalmology Annual Meeting; April 30, 2008; Fort Lauderdale, Florida.