Dr Louis Hoffart, Service d'ophtalmologie, Hôpital d'adultes de la Timone, 264 rue Saint Pierre, 13385 Marseille cedex 5, France. Email: email@example.com
Background: To compare the effects of different doses of bevacizumab with both saline and dexamethasone on inflammatory angiogenesis in the rat cornea induced by small chemical lesions.
Methods: Corneal chemical cauterization was performed on 24 rats. Animals were divided randomly into six groups and received a daily subconjunctival injection for 7 days of: balanced salt solution 0.1 mL or dexamethasone phosphate 4 mg/day or bevacizumab 2.5 mg/day, 3.75 mg/day, 5.0 mg/day or bevacizumab 5.0 mg/day + dexamethasone phosphate 4 mg/day. Clinical examination under slit lamp was performed daily for 7 days to evaluate corneal opacity and vessel size evolution. Computer-assisted quantitative image analysis was used to measure the total corneal area covered by neovascularization.
Results: At final examination, the dexamethasone, bevacizumab 5.0 mg/day and dexamethasone + bevacizumab groups showed a significant lowering in corneal opacity score as compared with control (P = 0.024, P = 0.006 and P = 0.013, respectively). Also, a significant reduction on new vessels size score was observed. Surface of corneal neovascularization was significantly reduced in dexamethasone, bevacizumab 5.0 mg/day and dexamethasone + bevacizumab groups compared with control (P = 0.045, P = 0.047 and P = 0.044, respectively).
Conclusion: Our study demonstrates the ability of a 5.0 mg/day bevacizumab subconjunctival injection, in monotherapy or associated with dexamethasone, to cause a short-term involution of corneal neovascularization after corneal alkali burn. Combination of both of these treatments may have advantages to monotherapy approaches.
Corneal neovascularization remains the most frequent cause of blindness after severe alkali burn trauma.1 Despite the common use of topical steroids,2,3 the inflammatory response can lead to corneal scarring that may significantly alter visual acuity, and also worsen the prognosis of subsequent penetrating keratoplasty. Many attempts have previously been made to inhibit corneal neovascularization by blocking vascular endothelial growth factor (VEGF)4–6. Various chemical compounds and drugs, such as steroids,7 methotrexate,8 heparin9 and thalidomide,10 have been proposed as inhibitors of corneal neovascularization. Steroids are still the first choice in clinical practice, because neovascularization is assumed to be secondary to some degree of inflammation. However, when inflammation is not the cause of angiogenesis, such as in diseases associated with deficiency of limbal stem cells or corneal neovascularization secondary to corneal hypoxia, anti-inflammatory corticosteroids have little or no effect on capillary growth. The side effects of steroids, such as glaucoma and cataract formation, should not be underestimated. For the above reasons, effective alternatives are required.
Vascular endothelial growth factor has been proven to be a major inducer of corneal neovascularization, both in experimental models and in the human cornea. Bevacizumab (Avastin, Roche, Neuilly-sur-Seine, France) is a recombinant, humanized, monoclonal antibody that binds to VEGF and prevents it from linking to its receptor. But so far, no FDA-approved anti-angiogenic agent for use against corneal neovascularization is available. Bevacizumab was the first anti-angiogenic drug which showed topical effects against corneal neovascularization. Previous studies have described the regression of corneal neovascularization by topical bevacizumab administration at various concentrations.11–15 The angiogenic and inflammatory systems contribute to the development of neovascularization in a synergistic manner: Therefore, a simultaneous treatment that counteracts both the angiogenic and inflammatory systems would be interesting and may have advantages to monotherapy approaches. However, there was no dose–effect study on anti-angiogenic activity of bevacizumab or comparison with steroids in this indication.
Our goal was to evaluate if subconjunctival injections of bevacizumab can reduce, in a dose-dependent manner, corneal neovascularization induced by alkalin burn in rats. Moreover, we have compared this reduction with changes observed with combined use of bevacizumab and dexamethasone.
Twenty-four male Brown Norway rats aged 6–8 weeks were used according to a protocol approved by the institutional animal care and use committee.
All procedures were performed with animals under general anaesthesia induced by intraperitoneal injection of ketamine hydrochloride (35 mg/kg, Imalgène, Rhone Merieux, Lyon, France) and medetomidine (0.5 mg/kg, Domitor, Pfizer, Orsay, France). When all procedures had been completed, animals were injected intraperitoneally with atipazole hydrochloride (1 mg/kg, Antisedan, Pfizer, Orsay, France) to antagonize the effects of medetomidine. A topical anaesthesia by oxybuprocaïne chloryhydrate 0.4% drops (Novésine, Merck Sharp & Dohme-Chibret, Paris, France) was applied to the corneal surface. A modified silver nitrate cauterization technique initially described by Mahoney and Waterbury13 was used to induce corneal neovascularization. Randomly, one cornea of each animal was cauterized by pressing an applicator stick (with a diameter of 2 mm) coated with 75% silver nitrate and 25% potassium Nitrate (Höllensteinstift 10, Pharma Winter, Leipzig, Germany) to the central cornea during 6 s under slit-lamp examination. Excess silver nitrate was removed by rinsing the eye with 5 mL of saline. To increase the reproducibility of the lesions, a single investigator (LH) cauterized all animals. The exclusion criteria were: poor alkali burn extent and corneal perforation.
Following chemical corneal burn, animals were randomized into six groups. On each animal, the cauterized eye was given a subconjunctival injection of either 0.1 mL of balanced salt solution (BSS, Alcon, Reuil-Malmaison, France) for group 1 (n = 4); of 0.1 mL of dexamethasone phosphate (Merck Sharp & Dohme-Chibret, 4 mg/mL) for group 2 (n = 4); of 0.10 mL (2.5 mg), 0.15 mL (3.75 mg) or 0.20 mL (5.0 mg) of bevacizumab (Avastin, 25 mg/mL) for groups 3 (n = 4), group 4 (n = 4) and group 5 (n = 4), respectively; and 0.20 mL (5.0 mg) of bevacizumab associated with 0.1 mL of dexamethasone phosphate in group 6 (n = 4). Treatments started immediately after cauterization and were administered daily for 7 days. The development of corneal neovascularization was observed on a daily basis with slit-lamp examination (Carl Zeiss, Jena, Germany) in a blinded fashion under general anaesthesia as described above. A scoring system was used to evaluate corneal opacity and vessels size. 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. New vessels size were also scored on a scale of 0–4, where 0 = no vessel; 1 = limbal vessels < 1 mm; 2 = vessels < 2 mm away from the limbus; 3 = vessels between 2 and 4 mm from the limbus; and 4 = vessels in the central 2 mm of the cornea.
In addition to the clinical assessment, we used a quantitative method for measuring changes in new vessels spread. Standardized corneal photographs were taken with ×12 magnification using a digital camera (Coolpix 5400, Nikon, Tokyo, Japan) attached to the slit-lamp microscope (Fig. 1a). Image analysis of corneal scar was first performed using an image processing and analysis software (Image J 1.31, Wayne Rasband, Research services brand, National Institute of Mental Health, Bethesda, MD, USA). The surface of each burn was measured in square millimetres. Furthermore, a semi-automatic program was also developed in MATLAB (v7.0, MathWorks, Natick, MA, USA) to quantify the neovascularized corneal area. Using colour distributions of neovascularized regions, we first estimated the direction in the red, green and blue (RGB) space that represent best new vessels colour. RGB images were projected along this new single colorspace dimension using the following transformation: Nv = (R + B/2 − G) * R (Nv, neovascularization colour direction; R, red channel; B, blue channel; G, green channel; Fig. 1a,b). Neovascularization is thus directly quantified as an increase in this single dimension. Quantitative analysis was performed inside a triangular region of interest (ROI) of standardized size (Fig. 1b,c). Triangular ROIs were used because it is most accurate geometrical approximation of a spherical sector (from the 2-D projection images of the cornea that were taken, the best corneal sample is actually a spherical sector from limbus to apex). In order to classify a pixel as being neovascularized or not, we extracted the baseline statistical description of the new colorspace dimension (mean [mcs] and standard deviation [sdcs]) within a control ROI placed in areas without evident neovascularization (Fig. 1b). These values were used to transform, for each pixel, the new colorspace dimension into statistical z-score values. To do so, we normalized by the standard deviation (sdcs) the difference between the neovascularization value (Nv) and the mean (mcs): z-score = Nv − mcs/sdcs (Fig. 1b,c). In this new statistical dimension, we then quantified the percentage of neovascularization within a test ROI of similar size, aligned along the limbus with the innermost vessel of the limbal arcade used as the border (see Fig. 1c). We defined a pixel as being neovascularized if its z-score value exceeded a threshold of 2.96 (Fig. 1d,e) compared with the control ROI (corresponding to a P = 0.05 assuming Gaussian noise distribution). The ROI was manually fit on pictures from time series of each animal and laid over the cornea at the same place on each picture. The reproducibility of ROIs disposition on the eye was based on anatomical features such as vasculature or eyelid borders. Inside the ROI, the number of pixels exceeding threshold value (corresponding to neovascularization) was extracted to compute a vessel ratio (VR) that is the percentage of the total area covered by vessels (Fig. 1f). The quantification of the neovascularization was performed in a blinded fashion.
Statistical analysis was performed using SPSS software v13.0 for Windows (SPSS, Inc., Chicago, IL, USA). Statistical significance was defined as P < 0.05. 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–Wallis test. The Mann–Whitney U-test then was used to compare groups pairwise at each time point; and the Wilcoxon signed-rank test then was applied to compare treated and untreated eyes. The sample size calculated for a 0.05 level of significance with 80% power, based on a standard deviation of 10% of surface neovascularization and a minimal difference of 20% between groups was four animals in each group.
Enucleation was performed at the seventh day. The eyes were prepared for histological examination using 10% formaldehyde. Sections were examined by dividing the corneas in two halves through the centre of the lesion and were evaluated with regard to the intensity of new vessels. Light microscopic examination was performed by an examiner who was blinded to the treatment groups.
Chemical injury resulted in induction of corneal neovascularization in all animals tested. The onset and progression of neovascularization followed a similar pattern in all groups: new vessels appeared at the second day, so that mild peripheral neovascularization can be seen on day 3. The vessels appeared short, fine and dense, and emerged from the adjacent limbus. One week after cauterization, there was clear evidence of corneal neovascularization with major vessels excepted in the dexamethasone and dexamethasone + bevacizumab treated groups. We first report the outcomes of clinical analyses using objective rating scales and then describe in a more quantitative way the effects of the different treatments on corneal neovascularization.
Clinical analysis of corneal scarring, opacity and vessels size
The mean surface of scarring in cauterized eyes immediately after chemical burn was 22.3 ± 7.9 mm2. There was no significant difference of corneal cauterized surface between groups (P = 0.068, Kruskal–Wallis).
Neovessels size (Fig. 2) was significantly lowered by dexamethasone (95% CI 0–1.42) and dexamethasone + bevacizumab 5.0 mg/day (95% CI 0–1.10) treatments compared with control (95% CI 2.58–4.42; P = 0.024), bevacizumab 2.5 mg/day (95% CI 2.92–4.28; P = 0.006) and bevacizumab 3.75 mg/day (95% CI 1.90–4.77; P = 0.013) groups from day 5 to day 7.
Regarding the corneal opacity estimation, the dexamethasone and dexamethasone + bevacizumab 5.0 mg/day treated groups showed a significant tendency toward lower scores (95% CI 0–3.25 and 95% CI 0–3.10) at the final examination when compared with both control (95% CI 2.95–4.54; P = 0.025 and P = 0.016, Mann–Whitney U-test) and bevacizumab 2.5 mg/day treated groups (95% CI 3.92–4.00; P = 0.007 and P = 0.010, Mann–Whitney U-test). Significant difference was not achieved at day 7 between dexamethasone and bevacizumab 3.5 mg/day (95% CI 3.24–4.35) and 5.0 mg/day (95% CI 0–4.25) groups.
Quantitative analysis of corneal neovascularization
To obtain a quantitative description of the effects of the treatments, we used our quantification algorithm on corneal images obtained for each eye, each group and each day (Fig. 3). As illustrated in Figure 4, the corneal neovascularization surface as estimated by the VR index increased monotonously over the 7 days in both control and bevacizumab 2.5 mg/day groups, reaching similar maximal final levels at 52.2 ± 27.4% and 56.1 ± 19.2%, respectively, of corneal neovascularization surface within the ROI. By contrast, VR indexes for dexamethasone, bevacizumab 5.0 mg/day and dexamethasone + bevacizumab 5.0 mg/day groups decreased significantly at day 7 (15.2 ± 13.2%, 20.0 ± 17.0% and 14.6 ± 10.2%, respectively) from the control (P = 0.045, P = 0.047 and P = 0.044, respectively, Mann–Whitney) and to the bevacizumab 2.5 mg/day group (P = 0.016, P = 0.045 and P = 0.043, respectively, Kruskal–Wallis). No significant difference was observed between the bevacizumab 3.75 mg/day group from all other groups.
Light microscopy evaluation of the histological preparations was consistent with the slit-lamp examinations. Corneas in the control group showed large numbers of intrastromal blood vessels (see Fig. 5a,g). Neovascularization was associated with inflammatory cells surrounding the vessels. Dexamethasone (Fig. 5b,h) and dexamethasone + bevacizumab 5.0 mg/day (Fig. 5f,l) treated groups showed less neovascularization and inflammation than the control eyes. However, there was still evidence of peripheral corneal neovascularization. Density of vascular endothelial cells was reduced in the bevacizumab 3.75 (Fig. 5d,j) and 5.0 mg/day (Fig. 5e,k) groups compared with the bevacizumab 2.5 mg/day (Fig. 5c,h) group and control sections.
Inhibition of vision-threatening corneal neovascularization is a major challenge following corneal chemical insult or inflammation and in many clinical situations corneal anti-angiogenic treatment would be helpful. Previous studies have investigated the on corneal neovascularization of topical bevacizumab administration at various concentrations.11–15 Despite these previous investigations, no further conclusions could be drawn in the absence of dose–response data.
The present study suggests a clear dose–effect relationship with a significant rise in the decrease from control (−32.1 ± 31.3%, P = 0.049) of corneal neovascularization after 5 days of 5.0 mg/day of subconjunctival bevacizumab injections. Both 3.75 and 5.0 mg/day doses resulted in a decrease of neovascularization surface when compared with control and lower doses of bevacizumab. However, our study was not sufficiently powered to achieve significance with a dose of 3.75 mg/day dose. Our data suggest that dose escalation from 2.5 mg to 5.0 mg/day may further increase a partial response obtained at the lower dose level without any increased side effects.
A possible caveat in our result and their comparison with previously published studies is the putative species-specific effects of VEGF and its inhibitors.16 Although bevacizumab was previously described as binding specifically to human VEGF-A, it was shown later that it can bind to murine VEGF-A by molecular biological assays.17 Thus, no definite conclusions can be drawn regarding the exact efficacy of bevacizumab subconjunctival injections in humans. Other limitations of our experiment were, that opposed to severe ocular surface alkali burns, there was no limbal stem cell loss involvement resulting in large subsequent corneal neovascularization and only the acute inflammatory phase following injury was concerned.
Steroids seem by far to be the best therapy to inhibit experimental corneal neovascularization18 and remain the mainstay of clinical therapy in prevention of corneal neovascularization.19 The mechanisms by which steroids inhibit angiogenesis include both direct effects on vascular endothelial cells, and indirect effects through actions on other cells in tissues that may produce angiogenic stimuli.20,21 The significant lower opacity score and smaller inflammatory infiltrates surrounding the new vessels in the dexamethasone group are related to these anti-inflammatory properties.
Eyes included in this study treated with the combination of subconjunctival bevacizumab and dexamethasone had a statistically significant decrease in manifestations of corneal neovascularization after chemical burn. Nevertheless, comparison of this combination therapy versus bevacizumab 5.0 mg/day or dexamethasone monotherapy showed no significant difference in neovascularization surface. These results are encouraging. However, there are limitations. This was a short-term study with a small sample, and cases will need to be followed long-term to determine whether and when repeat injections are necessary.
In this study, eyes with neovascularization secondary to chemical burn treated with subconjunctival bevacizumab 5.0 mg/day alone or combined with dexamethasone showed evidence of short-term resolution of neovascularization. In this indication, the use of a combined treatment that counteracts both the angiogenic and inflammatory systems may have advantages compared with monotherapy approaches. Long-term results are not known, and therefore caution should be exercised until the numerous outstanding questions with regard to safety, dosing, efficacy and duration of effect can be answered by prospective clinical trials. Further studies to evaluate the role of subconjunctival bevacizumab in the treatment of corneal neovascularization are warranted.
The authors gratefully acknowledge the Assistance Publique des Hopitaux de Marseille (APHM) and the Centre National de la Recherche Scientifique (CNRS) for supporting this study. We would like to acknowledge Dr Ivan Balansard, DVM for providing infrastructure and his helpful suggestions all along this study.