Efdal Yoeruek and Focke Ziemssen are joint first authors of this paper.
Safety, penetration and efficacy of topically applied bevacizumab: evaluation of eyedrops in corneal neovascularization after chemical burn
Article first published online: 8 NOV 2007
© 2007 The Authors. Journal compilation © 2007 Acta Ophthalmol Scand
Volume 86, Issue 3, pages 322–328, May 2008
How to Cite
Yoeruek, E., Ziemssen, F., Henke-Fahle, S., Tatar, O., Tura, A., Grisanti, S., Bartz-Schmidt, K. U., Szurman, P. and Tübingen Bevacizumab Study Group (2008), Safety, penetration and efficacy of topically applied bevacizumab: evaluation of eyedrops in corneal neovascularization after chemical burn. Acta Ophthalmologica, 86: 322–328. doi: 10.1111/j.1600-0420.2007.01049.x
- Issue published online: 8 NOV 2007
- Article first published online: 8 NOV 2007
- Received on December 28th, 2006. Accepted on August 1st, 2007.
- alkali burn;
- corneal neovascularization;
Purpose: That vascular endothelial growth factor (VEGF) plays a major role in inflammatory angiogenesis has been well established. This pilot study was designed to evaluate experimental treatment with bevacizumab eyedrops in corneal neovascularization induced by alkali burn. The feasibility of topical administration, corneal cell viability and corneal penetration were investigated in an animal model.
Methods: Eighteen chinchilla bastard rabbit corneas injured with 1 m NaOH were divided into three groups: untreated, early and late treatment groups. Eyedrops of bevacizumab solution (25 mg/ml) were administered five times daily. Clinical examination under stereoscopic microscope was performed to evaluate corneal opacity, neovascularization, vessel size and oedema. Histopathology was analysed for vessel density and apoptotic reaction. Additionally, intracameral bevacizumab concentration was measured with enzyme-linked immunosorbent assay (ELISA) after repeated topical applications.
Results: A fast increase in aqueous bevacizumab concentration was achieved when the solution was instilled every minute onto a healthy eye surface. As well as clear anti-angiogenic effects, anti-fibrotic effects were also seen after corneal burn, maintaining corneal transparency. Early treatment of actively growing vessels showed a significantly better outcome, although apoptosis of pre-existing vessels could also be induced by the late treatment. No specific toxicity was seen regarding epithelium, keratocytes or endothelium.
Conclusions: The data from this pilot study suggest that bevacizumab eyedrops can sufficiently penetrate the corneal stroma and anterior chamber. When administered soon after alkali burn, bevacizumab seems to significantly reduce corneal damage. Combinations of established treatment regimens with topical bevacizumab might be considered in severe injuries with otherwise devastating prognoses.
Corneal neovascularization and subsequent opacification remain the most frequent causes of blindness after severe alkali burn trauma (Chang et al. 2001). Despite the routine use of topical steroids (Kuckelkorn et al. 2002; Den et al. 2004), the inflammatory response can lead to oedema, lipid deposition and corneal scarring that may not only significantly alter visual acuity (VA), but also worsen the prognosis of subsequent penetrating keratoplasty (PK).
Previous work has demonstrated that neovascularization is initiated soon after trauma (Gan & Fagerholm 2001). Proliferation of vascular endothelium starts as early as 24 hours after an induced inflammation (Kvanta et al. 2000). Neovascularization involves the sprouting of essential new vessels from capillaries and venules from the pericorneal plexus (Yaylali et al. 1998). During the subsequent remodelling of extracellular matrix components, hyaluronan is deposited in the limbal area. Polymorphonuclear granulocytes invade and initiate further inflammation.
Not only leucocytes, but also corneal epithelial, stromal and endothelial cells have been shown to be important sources of vascular endothelial growth factor (VEGF) (Gan et al. 2004). Corneal avascularity is dependent on VEGF receptor 3 expression of corneal epithelium (Cursiefen et al. 2006); VEGF is known to be required in inflammatory neovascularization of the cornea (Amano et al. 1998; Kvanta 2006).
Therefore, many attempts have previously been made to inhibit corneal neovascularization by blocking VEGF (Lai et al. 2002; Cursiefen et al. 2004a, 2004b). In addition, different anti-VEGF antibodies have been used before PK to reduce the risk of graft rejection following corneal neovascularization (Yatoh et al. 1998).
The humanized IgG antibody bevacizumab is the first Food and Drug Administration (FDA) approved anti-VEGF drug. The drug is commercially available and has been shown to abolish ocular neovascularization (Friedlander & Welch 2006). The purpose of this pilot study was to determine whether topical treatment with the commercial antibody solution can achieve effective anti-angiogenic concentrations within the corneal tissue. Further aims included establishing the solution’s tolerability and efficacy to prevent inflammation-related corneal neovascularization.
Materials and Methods
Eighteen female chinchilla bastard rabbits (ChBB:CH), aged 3–6 months and weighing 1.5–2.5 kg, were obtained from Charles River Laboratories (Sulzfeld, Germany) and acclimatized for 1 week before the experiments started. Only the right eye underwent alkali burn so that no animal was blinded. Animals were handled according to the Association for Research and Vision in Ophthalmology Resolution on the Use of Animals in Ophthalmic Research and in observation of German federal law.
One eye in each rabbit was exposed to defined alkali injury under general anaesthesia with intramuscular injections of 35 mg/kg ketamine (Ketanest; Parke Davis GmbH, Berlin, Germany) and 5 mg/kg xylazine (Rompun; Bayer Schering Pharma AG, Leverkusen, Germany) and local anaesthesia with oxybuprocaine drops (Novesine 0.4%; Novartis Pharma AG, Nürnberg, Germany) on the corneal surface. A round, 4-mm diameter filter paper was soaked in 1 m sodium hydroxide (NaOH) and placed onto the limbal cornea for 20 seconds. The surface was then carefully rinsed with physiological saline solution.
Drug preparation and treatment protocol
The concentration of commercially available bevacizumab (25 mg/ml Avastin®; Hoffmann LaRoche, Basel, Switzerland) was used in sterile ophtioles (RaPha-Tec Typ PP Multi-professionell; Baesweiler, Germany) for administration of 10 μl drops. Approximately 1.678 nmol (molecular weight 149 kDa) were applied with one instillation to secure a relevant amount in excess of VEGF molecules. The drug was administered five times during the daytime, as this dosage had proved itself in first off-label treatments. Control animals received eyedrops of 0.9% NaCl. In the early treatment (ET) group, topical administration of bevacizumab eyedrops started at the time of alkali burn, whereas application in the late treatment (LT) group began on day 7 after injury.
Clinical examination was performed to evaluate both the general appearance of the treated eyes and to assess local toxicity and ocular intolerance. Corneas were examined under the stereoscopic microscope on days 1, 3, 5, 7, 9, 12 and 16 after the alkali burn in a blinded fashion. A scoring system was used to evaluate corneal opacity, neovascularization, vessel size and oedema. Ultrasound pachymetry (I3 System; M & C Medizintechnik & Chirugiebedarf GmbH, Mainhausen, Germany) was available to distinguish corneal oedema from cloudiness, but was not routinely performed. Corneal opacity was scored on a scale of 0−4, where 0 = completely clear; 1 = slightly hazy, iris and pupils easily visible; 2 = slightly opaque, iris and pupils still detectable; 3 = opaque, pupils hardly detectable, and 4 = completely opaque with no view of the pupils. Neovascularization was scored on a scale of 0−4, where 0 = no vessels at the corneal limbus; 1 = vessels within 1 mm of the corneal limbus; 2 = vessels within 2 mm of the corneal limbus; 3 = vessels 4 mm over the corneal limbus to the corneal centre, and 4 = vessels within 2 mm of the corneal centre. Vessel size was scored on a scale of 0−3, where 0 = no vessels; 1 = vessels detectable under the microscope; 2 = vessels easily seen under the microscope, and 3 = vessels easily seen without the microscope. Oedema was scored on a scale of 0−2, where 0 = absence of oedema; 1 = oedema present but mild, and 2 = very significant oedema. The vascularized area was calculated from measurements performed both in vivo (using a calliper under the microscope) and on standardized photographs. Clinical photographs were taken with a digital camera (DCR-TRV50E; Sony, Tokyo, Japan) and a zoom photographic slit-lamp (Zeiss 40 SL-P; Carl-Zeiss Meditec Systems GmbH, Oberkochen, Germany).
On postoperative day 16, the rabbits were killed and the eyes processed for histology. Paraffin sections were stained with haematoxylin and eosin (HE) and Masson’s trichrome. Primary mouse monoclonal antibody AE5 (ICN ImmunoBiologicals, Costa Mesa, CA, USA) was used for immunohistochemical staining of the 64-kDa basic cytokeratin-3 polypeptide according to Di Girolamo et al. (1999). The expression of cytokeratin-3 is regarded as a marker for cornea-type differentiation in the epithelial sheet. Briefly, deparaffinized sections were subjected to antigen retrieval by enzyme digestion (1 mg/ml proteinase; Sigma Aldrich, Taufkirchen, Germany) for 10 mins at room temperature (RT). For immunostaining, primary antibody (1 : 400 dilution) was applied for 30 mins. As a secondary antibody, a 1 : 500 dilution of biotinylated goat anti-mouse immunoglobulin (Dako Corp., Hamburg, Germany) was applied for 30 mins. Controls were performed by omitting the primary antibody. The specimens were then incubated with Streptavidin conjugated to alkaline phosphatase for 30 mins, visualized by chromogen red solution and counterstained with haematoxylin (ChemMate™ Detection Kit; Dako Corp.).
Histological specimens were examined and documented with a Zeiss microscope (Axioskop; Carl Zeiss Meditec Systems GmbH) connected to a digital camera (HC-300Z; Fujix, Kyoto, Japan) and the appropriate hard- and software (Image Access, Version 3.2; Imaging Bildverarbeitung AG, Glattbrugg, Switzerland).
For the detection of apoptosis, following pretreatment with proteinase K (Dako Demark A/S, Glostrup, Denmark) for 10 mins, the TUNEL (TdT-mediated dUTP nick-end labelling) method with alkaline–phosphatase was performed according to the manufacturer’s protocol (in situ cell death detection kit AP; Roche Diagnostics GmbH, Penzberg, Germany). The positive control section was incubated with 1 mg/ml DNAse I in 50 mm TRIS/HCl, pH 7.5, 1 mm MgCl, 1 mg/ml bovine serum albumin for 30 mins at 37 °C to induce DNA strand breaks. The negative control was incubated only with TUNEL solution without TdT. After incubation of the specimens with the included converter alkaline–phosphatase for 30 mins, colour development was performed with red chromogen (ChemMate Detection Kit, Alkaline Phosphatase/RED, K5005; Dako) according to the manufacturer’s protocol. Haematoxylin III according to Gill (Merck, Darmstadt, Germany) was used as a counterstain. The staining was graded on a scale of 0−3, where 0 = no staining, 1 = faint staining, 2 = clear staining, and 3 = intense staining, by an examiner who was blinded to the treatment groups. The average scores of the six animals in each group were calculated.
Anterior chamber sample
Penetration experiments were conducted under the same general anaesthesia as described before. A single bevacizumab eyedrop was applied every minute in 30 untreated eyes of additional healthy rabbits for 15 mins. At time-points of 2, 4, 6, 10 and 12 mins, a sample of the aqueous humour was taken in five animals via paracentesis using a 25-gauge needle (bevel up). The specimen was sampled after extensive rinsing for 15 seconds with physiological saline solution in order to avoid contamination by increased bevacizumab concentration in tear fluid.
Bevacizumab enzyme-linked immunosorbent assay (ELISA)
Aqueous humour samples were diluted with phosphate-buffered saline (PBS) (1 : 101 to 1 : 105) and 100 μl of each dilution was poured in duplicate into wells in a Maxisorp ELISA plate (Nunc, Roskilde, Denmark). A serial dilution of Avastin (6.25–24.4 pg/ml) was used to generate a standard curve. Plates were coated overnight at RT, washed three times with PBS-T (PBS with 0.05% Tween-20) and blocked with 300 μl/well of 1% bovine serum albumin (BSA) in PBS for 1 hour at RT. After three washes with 400 μl PBS-T each, 100 μl of biotin-labelled rabbit-anti-human IgG (diluted 1 : 1000 with 1% BSA in PBS; DakoCytomation GmbH, Hamburg, Germany) was added to each well and plates were incubated for 2 hours at RT. They were then washed three times with PBS-T as above and 100 μl streptavidin-conjugated horseradish peroxidase (1 : 200 in 1% BSA; R&D Systems, Minneapolis, MN, USA) per well was incubated for 20 mins at RT. Washing was performed as described and 100 μl substrate reagent solution (R&D Systems) was transferred into each well. Reaction was terminated after 20 mins by adding 50 μl of 2 n HCl to each well. Absorbance was measured spectrophotometrically at a dual wavelength of 450–570 nm (EAR400ATX; Labindustries, Berkeley, CA, USA).
The dataset was analysed using the statistical software package JMP 184.108.40.206 (SAS Institute Inc., Cary, NC, USA). Statistical analysis was performed to determine the differences between the control and the two experimental groups. Because of the small sample size, non-parametric tests were applied and 95% confidence intervals (95% CI) were given in descriptive distribution analysis of parameters. The differences between treatment groups were tested using non-parametric Kruskal–Willis test. Multiple pairwise comparisons (Dunn’s procedure) included Bonferroni adjustment for multiple testing. In all experiments, p < 0.01 was considered to indicate a statistically significant difference.
Penetration showed a time-dependent increase in aqueous concentration of bevacizumab after repeated applications (Fig. 1). The intraocular drug level was lower to a factor of 1000 in comparison with the eyedrop concentration. After two eyedrops, significant aqueous levels of bevacizumab were detected.
The alkali burn caused a reproducible and fast induction of corneal neovascularization. No difference in epithelial healing was seen in the case of epithelial damage. Corneal clarity did not differ significantly between the treatment groups (p = 0.268). However, at the final examination the ET group showed a clear tendency towards lower scores (95% CI 0.8–2.9) compared with control animals (95% CI 1.8–3.5) (Fig. 2A).
Neovascularization could be significantly inhibited by early treatment (p = 0.001, Kruskal–Wallis test). Scores even exhibited significant differences in pairwise analysis after adjusting for Bonferroni correction (p = 0.0067). Whereas the ET group showed a clear decrease in neovascularization scores (95% CI 0.7–1.6), the late treatment group (95% CI 1.9–3.1) did not achieve a significant effect compared with the control group (95% CI 2.9–4.1). Without treatment the mean neovascularized area increased in a time-dependent manner up to 35 mm2 (95% CI 29.0–41.0 mm2). Analysis of the neovascularized area revealed a significant reduction by early treatment with bevacizumab (7.1 mm2, 95% CI 3.4–10.7 mm2). Although the LT group showed a slight reduction in the second postoperative week from 25.5 mm2 (postoperative day 7, 95% CI 22.3–28.7 mm2) to 23.7 mm2 (postoperative day 16, 95% CI 19.2–28.2 mm2), there was no significant difference compared with the control group at any examination (Fig. 2E).
Although treatment modalities had a significant effect on vessel size (p = 0.005, Kruskal–Wallis), no essential swelling of the corneas was seen (p = 0.492). Minute corneal oedema was present in only six of 18 animals and showed no preference for any treatment modality (Fig. 2D).
Histology and immunohistochemical analysis
There was no sign of cytotoxicity in any of the treatment groups. Endothelial cells and epithelial layers exhibited no histological alteration other than some preparation artefacts (Fig. 3). We tested for the presence of human IgG antibody in corneal sections, but could not detect any positive staining. After 16 days of treatment, no (protein) deposits were found in either clinical or histological examinations.
Apoptosis tests disclosed a marked difference between the LT group and control animals (Fig. 4A) with a high number of vascular endothelial cells in the newly formed blood vessels exhibiting positive staining for the apoptotic marker (Fig. 3I). A median score of 2.9 (95% CI 2.52–3.08) in the LT group significantly differed from that in the control group (median 0.03; 95% CI 0.00–0.10; p < 0.01). Only weak signals were found in the ET group specimens (median 0.69; 95% CI 0.16–1.71).
All corneas in the control group showed large numbers of intrastromal blood vessels. Mean density of vascular endothelial cells (59.5, 95% CI 33.6–85.4) was reduced by 61.2% in the early bevacizumab administration compared with the control sections (153.2, 95% CI 116.5–189.8) (Fig. 4B). In late treatment, 103.2 (95% CI 78.6–127.7) was counted as mean vascular endothelial number within the 400 × 400-μm section, which was not significantly different from that of the control animals (p = 0.02). Neovascularization was associated with a marked formation of collagen fibres surrounding the vessels. The increase in stromal connective tissue was less clearly pronounced in the ET group sections, but was also found in LT animals (Fig. 5).
This pilot study demonstrated that locally applied bevacizumab is able to specifically suppress corneal neovascularization after alkali injury in vivo. The experiments also suggest that the treatment might improve the probability of a better longterm outcome by maintaining corneal transparency and avascularity. Our experimental series showed topical bevacizumab treatment to have a beneficial effect on the size and extent of neovascularization and stromal connective tissue deposits, resulting in improved corneal clarity after alkali burn.
Vascular endothelial growth factor is a known key factor in inflammation-related neovascularization (Amano et al. 1998) subsequent to chemical burn (Gan et al. 2004). One major advantage of anti-VEGF treatment is that it prevents the proliferation of not only blood, but also lymph, vessels (Cursiefen et al. 2004a, 2004b) and therefore promises a better prognosis for subsequent PK surgery.
One interesting observation in our experiments concerned the superior corneal transparency in the ET group, which was remarkable, although not statistically significant. This might suggest an additional inhibitory effect of topical bevacizumab on concomitant corneal scarring, possibly related to inhibition of leucocyte migration and invasion; the beneficial effect of topical bevacizumab in this alkali burn model might relate not only to the known anti-angiogenic effects of the drug but to a possible anti-fibrotic effect. However, it should be noted that lesser scarring is secondary to the inhibition of neovascularization.
An important disadvantage of our study is the relative species specificity of VEGF and its inhibitors. Thus, no definite conclusions can be drawn regarding the exact efficacy of bevacizumab eyedrops in humans. However, theoretical considerations imply that the drug is at least as effective when directed against human VEGF-A as it is in rabbits. The incomplete inhibition of corneal neovascularization in our model might also be caused by the involvement of other proangiogenic factors, such as TGF-β (Cursiefen et al. 2000). Similar results have recently been described in a rat model (Manzano et al. 2007).
Topical administration is an advantageous route for drug delivery to the cornea because it is non-invasive and results in minimal adverse effects compared with systemic administration. At present, the primary limitations for topical application include difficulty of formulation, low water solubility, low stability in solution and susceptibility to loss of bioactivity during longterm storage (Peters et al. 2007). It has been reported that the repeated topical application of most ocular drugs results in intraocular drug levels comparable with those achieved by subconjunctival injection (Shell 1982).
Because the eyedrops series did not replicate the therapeutic regime in humans, the penetration experiments are indicative of only the actual pharmacokinetics. After applying 250 μg bevacizumab per minute (a single eyedrop contains approximately 10 μl of the 25 mg/ml solution) and a total dosage of 2.5 mg after 10 mins, a concentration of 4 μg/ml was measured in the aqueous humour. Given that active transport mechanisms are very unlikely to contribute during this short time, diffusion might be assumed to be the underlying mechanism. Hence, intracorneal concentration is suggested to be much higher in accordance with the presumed gradient, although the exact tissue distribution – especially in the important niche of the perilimbal zone – remains unknown. Avastin™, the available bevacizumab solution, contains TWEEN buffer (1.6 mg polysorbate 20 per 4 ml solvent carrier). In higher concentrations, this agent acts as a detergent and must be considered as enhancing tissue penetration, as previously discussed for Lucentis™ (Gaudreault et al. 2005). However, compared with other topical drugs, the absorption of bevacizumab seems to be quite low. Therefore, the efficacy of local treatment of intraocular (iris) neovascularization is uncertain.
Evaluation times were – in respect to animal rights – too short to evaluate washing out or drug clearance. Thus no exact statements about stability in the corneal tissue can be made. The detection of bevacizumab in the aqueous humour is only an indirect indicator of general penetration into the corneal stroma. We experienced methodological difficulties in proving the abundance of the protein in the cornea (cryosections, wholemounts). The usual fixation/incubation procedures may have caused the wash-out of the unbound antibody.
Previous studies have described the regression of retinal neovascularization by much lower bevacizumab concentrations (Avery et al. 2006). As we do not know the drug’s therapeutic range, we hypothesize that a higher than necessary dose was chosen. No further conclusions can be drawn in the absence of dose−response data. Further dose-escalation studies should evaluate the optimal regimen, as further dilution of the drug would also drastically reduce costs.
Outwith our study, bevacizumab has already been proven to exhibit excellent water solubility. Its stability in an eyedrop formulation should not differ from that in an injected solution. Despite the risk of bacterial contamination, drug-induced infection should occur less frequently than after intraocular administration. Our method of preparing unpreserved aliquots in small ophtioles allows for easy delivery to patients. However, certain demands in terms of storage (refrigerating, no freezing) must be observed. During the observation, the 148 kDa molecule was well tolerated and showed no deposition in the corneal tissue.
Many studies have suggested occluding newly formed corneal vessels with different techniques, including argon-laser, electrocoagulation and photodynamic therapy. However, in the case of a fresh alkali injury, the prevention of early tissue remodelling might be the better alternative, as it can also avoid concomitant scarring and loss of corneal clarity. The earlier treatment starts, the better the results, at least in the short-term. It has been reported that corneal avascularity is maintained by the differentiated regulation of several growth factors and their receptors (Cursiefen et al. 2006). As increased apoptosis of vascular endothelial cells and the regression of at least some newly grown vessels were achieved by the inhibition of VEGF, we hypothesize an early phase during which proliferating cells are still dependent on VEGF. Blocking all VEGF isoforms within this sensitive period could mean removing an essential survival factor from young corneal neovascularizations.
Although quantitative evaluation was restricted by the limited material, our study gave a clear impression that scarring and formation of intrastromal collagen fibres were reduced in animals that were treated early. Although inhibition of VEGF is mandatory at the time of inflammation and wound healing, reperfusion is more important in terms of alkali-induced ischaemia and necrosis. At this point, VEGF might also be beneficial in supporting the restoration of the ocular surface.
In summary, topical anti-VEGF therapy may represent a promising mode of treatment which can be applied in addition to established chemical burn regimens and which is worth further evaluation in clinical studies. Early treatment and dose management seem to be the determining factors.
- 1998): Requirement for vascular endothelial growth factor in wound- and inflammation-related corneal neovascularization. Invest Ophthalmol Vis Sci 39: 18–22. , , , & (
- 2006): Intravitreal bevacizumab (Avastin) in the treatment of proliferative diabetic retinopathy. Ophthalmology 113: 1695.e1–e15. , , et al. (
- 2001): Corneal neovascularization. Curr Opin Ophthalmol 12: 242–249. , , & (
- 2000): Immunohistochemical localization of vascular endothelial growth factor, transforming growth factor α, and transforming growth factor β1 in human corneas with neovascularization. Cornea 19: 526–533. , & (
- 2004a): VEGF-A stimulates lymphangiogenesis and haemangiogenesis in inflammatory neovascularization via macrophage recruitment. J Clin Invest 113: 1040–1050. , , et al. (
- 2004b): Inhibition of haemangiogensesis and lymphangiogenesis after normal-risk corneal transplantation by neutralizing VEGF promotes graft survival. Invest Ophthalmol Vis Sci 45: 2666–2673. , , et al. (
- 2006): Non-vascular VEGF receptor 3 expression by corneal epithelium maintains avascularity and vision. Proc Natl Acad Sci 103: 11405–11410. , , et al. (
- 2004): Efficacy of early systemic betamethasone and cyclosporine A after corneal injury via inflammatory cytokine reduction. Acta Ophthalmol Scand 82: 195–199. , , & (
- 1999): Culture and characterization of epithelial cells from human pterygia. Br J Ophthalmol 83: 1077–1082. , , et al. (
- 2006): Vanishing disc neovascularization following intravitreal bevacizumab (Avastin) injection. Arch Ophthalmol 124: 1365. & (
- 2001): Leucocytes in the early events of corneal neovascularization. Cornea 20: 96–99. & (
- 2004): Vascular endothelial growth factor (VEGF) and its receptor VEGF R-2 in the regulation of corneal neovascularization and wound healing. Acta Ophthalmol Scand 82: 557–563. , & (
- 2005): Preclinical pharmacokinetics of ranibizumab after a single intravitreal administration. Invest Ophthalmol Vis Sci 46: 726–733. , , , & (
- 2002): Emergency treatment of chemical and thermal eye burns. Acta Ophthalmol Scand 80: 4–10. , , & (
- 2006): Ocular angiogenesis: the role of growth factors. Acta Ophthalmol Scand 84: 282–288. (
- 2000): Expression of matrix metalloproteinase-2 (MMP-2) and vascular endothelial growth factor (VEGF) in inflammation-associated corneal neovascularization. Exp Eye Res 70: 419–428. , , et al. (
- 2002): Inhibition of corneal neovascularization by recombinant adenovirus mediated antisense VEGF RNA. Exp Eye Res 75: 625–634. , , , & (
- 2007): Inhibition of experimental corneal neovascularization by bevacizumab (Avastin). Br J Ophthalmol 91: 804–807. , , et al. (
- 2007): Anti-permeability and anti-proliferative effects of standard and frozen bevacizumab on choroidal endothelial cells. Br J Ophthalmol 91: 827–831. , , et al. (
- 1982): Pharmacokinetics of topically applied ophthalmic drugs. Surv Ophthalmol 26: 207–218. (
- 1998): Effect of topically applied neutralizing antibody against vascular endothelial growth factor on corneal allograft rejection of rat. Transplantation 66: 1519–1524. , , et al. (
- 1998): In vivo confocal imaging of corneal neovascularization. Cornea 17: 646–653. , & (