Downloaded from http://bjo.bmj.com/ on March 7, 2016 - Published by group.bmj.com Clinical science Clinical evaluation and characterisation of corneal vascularisation Lana A Faraj,1 Dalia G Said,1,2 Mohamed Al-Aqaba,1 Ahmad M Otri,1 Harminder S Dua1

1Division of Clinical ABSTRACT This is believed to be the single most important Neuroscience, Department of Background/aims To clinically characterise corneal mechanism by which the remains normally , University of 4 Nottingham, Nottingham, UK neovascularisation (CVas) with a view to elaborate avascular. Ectopic VEGF receptor 3 is also 2Department of Cornea and clinical presentations and to standardise descriptors for strongly expressed by corneal epithelium and could External Eye Diseases, Research clinical evaluation and research. work in a similar manner.5 Institute of Ophthalmology, Methods of 165 patients with CVas due to a Yet, the physiological and nutritional needs of Cairo, Egypt variety of corneal pathologies were observed clinically the cornea are no different from those of any other Correspondence to with the slit lamp biomicroscope and photography at tissue. In most tissues, these needs are served by Professor Harminder Singh different time points over the course of their disease. the network of blood vessels. In the cornea, the Dua, Department of Parameters assessed included location, depth, length, fluids that bathe it, the aqueous humour on the Ophthalmology, QMC, branching pattern, colour, lipid leakage, nature of blood inside and the tear film on the outside, together Derby Road, Nottingham, flow and presence of haemorrhage. A clinical grading of with the limbal vascular arcade, serve this func- NG7 2UH, UK; 1 [email protected]. CVas was applied. tion. Blood vessels also play a crucial role in host uk Results CVas can arise from the limbus, defence, responding to paracrine and neuronal and . CVas preferentially travels along planes created stimuli generated by injury or insult. When the Received 2 February 2015 by corneal incisions, suture tracks and lamellar cornea is stressed by such events, new blood vessels Revised 1 May 2015 fi Accepted 21 June 2015 keratoplasty, both deep lamellar and endothelial invade the corneal tissue to ful l this role. A multi- Published Online First keratoplasty. CVas also principally follows the depth of tude of conditions induce new blood vessels to 10 July 2015 pathology. CVas can be classified into active young, invade the cornea, resulting in ‘corneal vascularisa- active old, mature, partially regressed and regressed. tion’ (neovascularisation), a term which has clinic- Herpetic infection was the most common cause of lipid ally become a collective noun for corneal vessels keratopathy. induced the least associated with varied aetiologies. It is more a clin- amount of vascularisation. A simple and efficient clinical ical sign than a diagnosis. grading system for ascertaining the severity of CVas was The inflammation and result- developed and validated. ing from corneal vascularisation (CVas) and the Conclusions The various clinical characteristics of CVas increased risk of immune-mediated rejection in described in this study can be used to standardise the scarred, vascularised corneas requiring transplant- nomenclature and description of CVas and clinically ation, has prompted clinicians to devise means to grade its severity. As modern, effective methods of shut vessels. Topical medication, cautery, treating CVas are being introduced, it is important that argon and yellow dye laser6 and fine needle dia- – there is uniformity in the descriptors used to establish thermy (FND)7 9 have all been advocated with baseline values and evaluate outcomes of treatment. The varying degrees of success. The advent of parameters established in this study can help in this anti-VEGF antibodies has resulted in a surge of regard. interest in using these agents to treat CVas. Doses, regimes and modalities of administration are cur- rently as varied as the authors. Avastin INTRODUCTION (Bevacizumab) has been studied in both animal Transparency and lack of blood vessels are unique experiments and human trials with reported good attributes of the cornea. The latter is in many ways efficacy in shutting CVas though there is a concern a prerequisite for the former. The cornea has about its interference with nerve regeneration and – evolved to keep blood vessels away and many delayed wound healing.10 12 factors contribute to this. The compactness of the There are many other experimental agents and corneal stromal architecture is considered to be an the list is constantly growing.2 The problem rests impediment to vessel invasion. Several molecules, not only with the agent or its administration but singly or in combination, provide the milieu which equally so with the clinical variations in CVas being maintains it in an avascular state.12These include treated. We undertook an extensive study of CVas proteases and their inhibitors, growth factors, Fas by direct clinical examination of patients and ligand, angiostatin, thrombospondin and others.12 images. We were able to define certain fundamental Vascular endothelial growth factor (VEGF), a principles that would help improve the understand- potent inducer of vascularisation, is secreted by ing of CVas and guide development of protocols corneal epithelial and stromal cells and by invading for its management. 3 To cite: Faraj LA, Said DG, macrophages. However, a soluble variant of the Al-Aqaba M, et al. Br J VEGF receptor 1 (VEGFR-1 or sFlt) is also MATERIALS AND METHODS Ophthalmol 2016;100: expressed by corneal epithelial cells.4 This molecule One hundred and sixty-five patients with CVas – 315 322. binds to and sequestrates the available VEGF. were prospectively observed, clinically with the slit

Faraj LA, et al. Br J Ophthalmol 2016;100:315–322. doi:10.1136/bjophthalmol-2015-306686 315 Downloaded from http://bjo.bmj.com/ on March 7, 2016 - Published by group.bmj.com Clinical science

lamp biomicroscope at different time points over the course of An independent subjective assessment of the specific morpho- the disease process. All patients had slit lamp photomicrographs logical characteristics was carried out by three observers. To test taken. Clinical images and notes of 104 of these were prospect- the reliability of the observations a random selection of 75 ively and retrospectively reviewed at different time points. images from the records of 20 patients was examined and Parameters assessed during clinical evaluation of the eyes and scored. Inter-rater agreement was tested using κ measurement, review of the photographs (a combination of diffuse, slit and which is a chance-adjusted measure of agreement.13 Three raters retro illuminated images) were as follows: (LAF, DGS and AMO) were enrolled in the study to judge for Source: conjunctival, limbal or iris. Patients with the presence or absence of vessel characteristics in our cases. and diffuse keratoconjunctival proliferation were excluded. Randolph’s free-marginal multirater κ was used for statistical Location and depth: quadrants affected (1–4) and whether analysis of the intra-rater and inter-rater agreement. SPSS statis- superficial, anterior stromal, posterior stromal or a combin- tical package (V.17) was used for further analysis.14 Values of κ ation of these. can range from −1.0 to 1.0, with −1.0 representing perfect dis- Length: peripheral or extending centrally (see details below). agreement below chance, 0.0 indicative of agreement which is Branching pattern: straight loops, tortuous, arborising. due to chance and 1.0 perfect agreement above chance. A κ of Colour: bright or dull red, grey or white. 0.70 or above indicates adequate or substantial inter-rater agree- Leakage: presence of lipid and corneal oedema. ment. The study was accorded approval from the audit commit- Nature of the blood flow: cattle trucking or fragmented, con- tee of Queens Medical Centre, Nottingham (approval number tinuous, tumbling of red blood cells. Presence of corneal 2882), and by the ethics committee (reference number 10/ haemorrhages. H0408/12). Features of associated pathology were also noted. All patients were evaluated by direct observation on the slit lamp and a RESULTS study of clinical images. Clinical records were also reviewed to Causes of CVas ascertain diagnosis. The corneal images were taken using anter- The cause for CVas included infection, inflammation, degenera- ior segment slit lamp camera (Topcon SL-7F slit lamp, Topcon tions, trauma including chemical burns, corneal grafts and 4059 Flash and Nikon 5207550 Digital Camera) at the magnifi- contact wear. The distribution of different aetiological cat- cation required to delineate the fine details of the vessels. egories among this study cohort of CVas is provided in table 1. Severity of vascularisation secondary to infective keratitis was Viral keratitis was the most common cause of CVas in our ascertained according to the following scoring system which study, contributing to 24.9% of the cases. Corneal transplant- included number of quadrants involved (1–4), depth of involve- ation accounted for almost a similar number (23.7%) with pene- ment (superficial=anterior one-third of corneal thickness or trating keratoplasty accounting for 15.8%, anterior lamellar deep=posterior two-thirds) and location of vessels (periphera- keratoplasty for 7.3% and endothelial keratoplasty for 0.6%. Of l=outer one-third, mid-peripheral=middle third and central=- the cases, 15.8% were related to bacterial keratitis. central third). At the other end of the range were conditions such as acanth- A score of 1 was given for each quadrant affected; a score of keratitis (9.1%) and limbal stem cell deficiency (4.2%). 1 for superficial vessels and 2 for deep vessels; and a score of 1 for peripheral vessels, 2 for mid-peripheral vessels and 3 for Morphology of CVas central vessels. When one quadrant had more than one vessel, Corneal vascularisation usually commenced from the limbus the score attributed to the worst vessel(s) was counted for that (see also section titled ‘Sources’ below). The vessels presented as quadrant. Thus, the maximum score for any quadrant would be straight or coiled loops, with one limb of the loop being nar- 6 and for the whole cornea 24. Each case of CVas related to rower and lighter in colour and the other being wider and corneal infection secondary to viral, bacterial and acanthamoeba darker red in colour. The former represented the afferent arm infections was assessed according to the aforementioned criteria (carrying blood towards the cornea) and the latter, the efferent and a cumulative score ascertained. arm (carrying blood away from the cornea) (figure 1A). Under high magnification (10× objective and ×25 to ×40 eye piece) at the (photo) slit lamp blood flow in the afferent vessels was seen as a rapid pulsatile movement of the blood column but was Table 1 Distribution of different pathologies causing corneal usually difficult to visualise. In the efferent vessels, a relatively neovascularisation slow tumbling or rolling of red blood cells could be seen. Diagnosis Frequency (%) Afferents were usually located posterior to efferents and were straighter compared to the relatively sinuous course taken by 15 (9.1) efferents. The afferents and efferents crossed each other at one Bacterial keratitis 26 (15.8) or more points. There was no specific predilection for the affer- Viral keratitis 41 (24.9) ent to cross the efferent anteriorly or posteriorly. 1 (0.6) During active vascularisation, fine extensions, seen as red lines, Stem cell deficiency 7 (4.2) appeared from the terminal end of the loops (figure 1B). Loops PKP related 26 (15.8) divided and bifurcated and made connections with branches DALK related 12 (7.3) from adjacent loops as vascularisation progressed (figure 1C). It DSEK related 1 (0.6) was also noted that adjacent loops, after running a very short related 10 (6) straight course, curved towards each other (figure 1A, C). These Ocular surface inflammatory conditions 19 (11.5) loops eventually joined up to form an ‘arch’ of vessels. Miscellaneous 7 (4.2) Occasionally, intracorneal haemorrhages were seen, associated Total 165 (100) with the advancing edge of the vessels (figure 1C). The connect- DALK, deep anterior lamellar keratoplasty; DSEK, Descemet’s stripping endothelial ing vessels between adjacent loops were not designated as either keratoplasty; PKP, penetrating keratoplasty. afferent or efferent.

316 Faraj LA, et al. Br J Ophthalmol 2016;100:315–322. doi:10.1136/bjophthalmol-2015-306686 Downloaded from http://bjo.bmj.com/ on March 7, 2016 - Published by group.bmj.com Clinical science

Figure 1 Clinical characteristics of corneal neovascularisation (CVas): (A) Shows that corneal vessels grow as straight or coiled loops made of efferent (carrying blood away from the cornea) and afferent (bringing blood to the cornea) arms. The afferent arm is narrower and lighter in colour (white arrows) and the efferent arm is wider and darker red in colour (black arrows). (B) Shows fine extensions, seen as red lines, appearing from the terminal end of the loops during active progression of CVas (white arrow). (C) Intracorneal haemorrhages associated with the advancing edge of the vessels (white arrow). (D) Single established feeder vessel in a patient with . (E and F). Show efferent–afferent (E-A) crossing changes in CVas in two eyes. The calibre of the efferent vessel narrows at E-A crossings and also alters the blood column proximal to the crossing (black arrows).

In 36 patients where CVas was examined from the very begin- In five patients with deep anterior lamellar keratoplasty, the ning it was consistently observed that as CVas progressed, new single radial vessels spread out like a fan in the deep lamellar loops and connections appeared while others underwent attrition interface upon reaching the G–H junction. In one case, a single and disappeared. The network of corneal vessels was best estab- superficial radial vessel dipped sharply into the beginning of the lished at the site of pathology (eg, or abscess) where suture track in the host periphery, travelled along the track to it tended to persist as a vascular complex, fed by one or more reach the G–H junction and spread like a fan in the interface pairs of ‘feeder’ efferent–afferent (E-A) vessels (figure 1D–F). (figure 2E, F). Often, an established vascular complex in the cornea demon- One patient with Descemet’s stripping endothelial kerato- strated multiple fine feeder vessels in addition to the main E-A plasty developed postoperative infectious keratitis in the nasal vessel(s). The calibre of the efferent vessel was noted to alter at cornea and demonstrated vascularisation in the plane of the E-A crossings (figure 1E, F). The E-A complex terminated at the donor–recipient interface in the opposite temporal cornea site of lesion and did not progress beyond the inciting area of the (figure 2G, H). cornea. In patients where the level of pathology—superficial, anterior stromal, deep stromal—could be clearly defined a statistically significant association with the level of vessels was noted, Vessels follow planes p value <0.001. In no case did we observe vertical twigs or A consistent observation related to corneal vascularisation was branches of vessels oriented in the anteroposterior meridian. that the vessels showed a preference to migrate along tissue planes such as those created by suture tracks, corneal lacerations, graft–host junction or in the interface created by lamellar kera- Types/stages of CVas toplasty (figure 2). This tendency to follow planes was noted in Our study enabled us to stage the vessels into five categories 31.5% of our cases. Suture tracks, with or without sutures, was based on their morphological characteristics. The intraobserver the single most common observed plane (27.2%) along which and interobserver variability for the categorisation was 90.1% vessels migrated. Loose sutures, suture knots and broken sutures and 79.3% respectively among the three observers: were associated with vessels in 3.9% of the patients examined 1. Active young vessels (figure 3A, B): These were freshly (figure 2A, B). formed vessels that were full of blood, appeared bright red Patients with penetrating keratoplasty and anterior lamellar ker- in colour, had minimal surrounding fibrous tissue sheathing atoplasty demonstrated radial vessel in-growth (mostly along the and were actively progressing in the cornea with a well- suture track) up to the G–H junction after which the vessel made a defined arborising network of fine (capillary) vessels. sharp almost right-angled bend and extended circumferentially The corneal stroma surrounding the vessels showed signs of along the G–H junction for variable lengths (figure 2C, D). leakage and oedema. Offshoots from some of these entered the graft along the circum- 2. Active old vessels (figure 3C): This represented the stage ference. This was observed in 26% of the cases. when the vessels had reached and surrounded or covered the

Faraj LA, et al. Br J Ophthalmol 2016;100:315–322. doi:10.1136/bjophthalmol-2015-306686 317 Downloaded from http://bjo.bmj.com/ on March 7, 2016 - Published by group.bmj.com Clinical science

Figure 2 Illustration of corneal neovascularisation (CVas) following tissue planes. (A) CVas (active young) invading the cornea along a suture track. It makes a 90° bend to follow the plane of the graft–host junction (GHJ) (white arrow). (B) CVas incited by overwear of contact lens after all sutures were removed in a patient who underwent deep anterior lamellar keratoplasty (DALK). CVas follows the plane created by the passage of sutures. (C and D) Radial vessel in-growth up to the GHJ after which the vessel makes a sharp almost right-angled bend and extends circumferentially along the GHJ. (E). A CVas complex arising from a limbal vessel sharply dips into a deep suture track (arrow) and continues to the deep lamellar plane offered by the DALK procedure, before fanning out. (F) A CVas complex extends radially along a suture track. On reaching the GHJ it dips down and spreads out like a fan in the deep lamellar interface in an eye that had undergone DALK (G and H). CVas appearing in the plane between graft and recipient in a patient with Descemet’s stripping endothelial keratoplasty. The patient had infectious keratitis nasally (G) and the vessels appeared on the temporal side (H).

offending lesion in the cornea. Their progression ceased but was not oedematous. All ghost vessels were located in the consolidation continued. In no case did we see the vessels stroma. overshooting the lesion. They appeared less bright than they originally did and maintained a brisk circulation. Of the 165 cases studied sequential observation was made in 3. Partially regressed (figure 3D): These were seen when the 91 cases over a mean of 21.64 (2 days–83 months) months corneal pathology had abated in response to therapy or the follow-up. In these cases a mean of 3.3 time point assessments arrival of corneal vessels or following fine needle diathermy of per patient (range 2–19) was achieved. The duration over which corneal vessels. The circulation in the vascular complex was one stage persisted before changing to the next was variable. An relatively slow, the vessels were less engorged and some parts evolving vascular complex did not necessarily pass through all of the complex had become less visible or underwent attrition. stages. Details of the numbers of vessels in each category are Both active old vessels and partly regressed vessels showed an given in table 2. arborising pattern from the limbus towards the centre, with the As the distinction between active young and active old vessels vessels becoming narrower towards the centre. may not always be very clear as there is often a transition as the 4. Mature (figure 3E): These were relatively large vessels, with former regresses to the latter, and it may be practical to group minimal arborisation and regressed or absent capillary net- active vessels together. Similarly, partially regressed vessels and works, seen to persist in scar tissue or in the corneal stroma mature vessels too may be grouped together. after the corneal pathology had healed. These vessels con- tained blood and maintained a circulation. Lipid keratopathy 5. Regressed (ghost vessels) (figure 3F): These presented as fine This feature of CVas was noted with mature vessels, partially white lines mirroring the morphology of the original vessels regressed (progressing lipid deposition) and regressed vessels and were seen as ‘ghost vessels’. These did not have an (residual or persistent lipid in the cornea) (figure 4A–C). The active circulation and the cornea where they were located distribution of lipid keratopathy (LK) across the CVas

318 Faraj LA, et al. Br J Ophthalmol 2016;100:315–322. doi:10.1136/bjophthalmol-2015-306686 Downloaded from http://bjo.bmj.com/ on March 7, 2016 - Published by group.bmj.com Clinical science

Figure 3 Classification of corneal vessels. (A) Active young: bright red, brush-like vessels invading in to the cornea with . (B) Active young: bright red vessels emerging from previously mature corneal vessels and invading a recurrent herpetic keratitis lesion. (C) Active old: dull red vessels with an active circulation of blood and surrounding oedema of the cornea. (D) Partially regressed: The vessels are dull red with a slow circulation and some parts of the complex are less visible or have undergone attrition. (E) Mature, relatively large vessels with blood circulation but no established fine network (capillary) of vessels (insert). (F) Regressed: Vessels appearing as fine white lines (ghost vessels) displaying the outline of the original vessels.

aetiological groups is summarised in figure 4D. Of the 29 cases that invaded the cornea and spread in a centrifugal manner with LK, 25 cases were related to viral keratitis. There was a sig- (figure 5A, B). nificant association between viral keratitis and LK with a p value fi of <0.001 ( gure 4D). For the clinical diagnosis of LK interob- Severity of vascularisation server and intraobserver variability was 100%. CVas related to keratitis was of the greatest severity and acanthamoeba was the least (p<0.005). In acanth- Sources of vessels amoeba keratitis more quadrants were affected but the vessels Corneal vascularisation occurred from the conjunctiva, limbus tended to remain in the periphery (figure 6). Thus, overall and iris. In all cases bar two, the vessels arose from the limbus. severity of CVas in acanthamoeba keratitis was less. The level at which the vessels invaded the cornea corresponded to the level of location of the pathology as stated above and in DISCUSSION localised lesions, from the sector of limbus closest to the path- In clinical parlance the terms corneal angiogenesis, corneal vas- ology. In chronic cases this pattern was not observed and more cularisation and corneal neovascularisation are used inter- generalised invasion of the cornea, from several sectors of the changeably. The acronym CNV is now established as limbus and at different levels, was seen. In two cases where the representing ‘choroidal new vessels’ and in our opinion, should iris was adherent to the posterior surface of the cornea follow- not be used for ‘corneal new vessels’ to avoid confusion. Hence ing ulcerative perforation, the iris was the source of new vessels it is proposed that the term CVas be used to denote ‘corneal vas- cularisation’. The has a normal vascular system and hence use of the term ‘new vessels’ could be considered appro- priate, whereas the cornea does not have a normal vasculature, Table 2 The number and percentages of types of CVas reported and hence the term ‘new’ is not essential. in this study In this study, we covered a range of aetiologies for corneal fl Type of vessel Frequency Percentage vascularisation that re ect the case mix and referral pattern to our unit. The differences noted in vascularisation pattern were Active young 148 27.1 related to the severity of CVas and presence of LK, which are Active old 227 41.6 discussed below. Also, the level of vascularisation was consist- Mature 128 23.4 ently related to the level of pathology rather than to the aeti- Partially regressed 35 6.4 ology. Superficial corneal pathology resulted in superficial Regressed (ghost vessels) 8 1.5 vascularisation and deep pathology resulted in deep vessels. Total 546 100 Often when the disease process extended through the thickness of the cornea, superficial and deep vessels were seen in the same

Faraj LA, et al. Br J Ophthalmol 2016;100:315–322. doi:10.1136/bjophthalmol-2015-306686 319 Downloaded from http://bjo.bmj.com/ on March 7, 2016 - Published by group.bmj.com Clinical science

Figure 4 Lipid keratopathy. (A) Shows lipid deposition in the cornea related to two well-defined efferent–afferent feeder vascular complexes. (B) Lipid in cornea related to mature vessels. (C) Persistent thin layer of lipid after regression of vessels (regressed ghost vessels were present). (D) Association of lipid keratopathy with various causes of corneal vascularisation (CVas). CVas related to viral keratitis was significantly associated with lipid keratopathy (p<0.001). DALK, deep anterior lamellar keratoplasty; DSEK, Descemet’s stripping endothelial keratoplasty; OS, ocular surface; PKP, penetrating keratoplasty. cornea. The planar architecture of the cornea determines vascu- As has been previously reported CVas was observed to be a lar migration. The formation of loops with efferent and afferent dynamic process wherein old vessels regress as new loops limbs appears to be a basic process in CVas. Occasionally single appear and advance.18 This is an important consideration in the red lines, where the loop morphology is not discernible, are evaluation of any agent used to treat vessels. A clinical distinc- seen. These, we believe, represent a stage before the formation tion between regression or ‘remodelling’ and regression due to of loops and are perhaps the earliest clinically visible component the test agent may not be easy but has to be factored when of CVas progression. evaluating outcome. Mild or minimal regression may not be These early vascular channels could be construed as the attributable to the agent tested. extension of an afferent of one loop or the commencement of When distinct feeder E-A pair(s) are present, supporting an efferent of the other. The flow of blood is likely to be deter- mature vascular complexes, they are easy targets for surgical mined by the pressure gradient between the adjacent loops. occlusion with techniques such as FND or laser.67The blood Clinically, sprouting is the most likely mechanism of angiogen- column distal to the E-A crossing is relatively stagnant, allowing esis in the cornea. In sprouting, a solid cord of endothelial cells greater absorption of laser energy (authors experience). The emerges from existing vasculature which then canalises to allow strategy for laser ablation of vessels too employs this principle, flow of blood.15 ‘Bridging’ and ‘intussusception’ are other pos- wherein a few pulses are initially delivered to one point causing sible mechanisms.15 Tiny intracorneal haemorrhages that are focal constriction of the vessel, allowing greater effect of subse- occasionally associated with the advancing edge of a vascular quent pulses applied to the ‘stagnant’ column of blood thus complex point to the relative fragility of the newly formed created.6 endothelial loops. Clearance of blood from corneal tissue is Loose sutures, suture knots and broken sutures seem to associated with influx of macrophages,16 which are known to provide a ‘stimulus’ for corneal vascularisation. The mucus that secrete VEGF,17 which in turn would support the vascularisation collects around loose and broken sutures can trap polymorpho- process. Intracorneal haemorrhages may therefore not be merely nuclear cells and microbes inciting localised inflammation/infec- incidental but could contribute to the pathogenesis of CVas. tion, thus attracting vessels.19 20 Removal of the offending

Figure 5 The iris as a source of corneal vessels following ulcerative perforation of the cornea and iris adhesion to the posterior surface of the cornea. (A) A healed lesion with partly regressed corneal vessels that had originated from the iris. Inset: enhanced view of the origin of the vessels. (B) Active young vessels emanating from the iris adherent to the cornea in an active corneal ulcer.

320 Faraj LA, et al. Br J Ophthalmol 2016;100:315–322. doi:10.1136/bjophthalmol-2015-306686 Downloaded from http://bjo.bmj.com/ on March 7, 2016 - Published by group.bmj.com Clinical science

Figure 6 Scoring of corneal vascularisation (CVas). A standard grid of three equidistant circles and a cross is superimposed on the photograph to evaluate extension of vessels into periphery, mid-periphery or centre of the cornea in each quadrant. (A and B). Scores 12 and 13 of CVas respectively in two eyes affected with acanthamoeba keratitis. (C and D) Scores 17 and 17 of CVas respectively in two eyes with herpetic keratitis. CVas was most severe with viral keratitis and least with acanthamoeba keratitis (p<0.005).

suture resulted in stopping the progression of the vessels, treating active vessels, the activity of the inciting lesion should although all such patients were also treated with topical also be addressed. and drops. Although most, if not all, CVas are abnormal some leak and There is a tendency for CVas to preferentially grow through some do not. This is a clear sign that all CVas is not the same and suture tracts (even after suture removal), G–H junctions and differences in function exist that are as yet not understood. An healed lacerations. This suggests that the collagen of wound important and interesting observation was that most LK was sec- healing that is contained in such scars and planes may be more ondary to herpetic keratitis-related CVas. Similarly, the stroma vulnerable to matrix metalloproteinase digestion, facilitating around some CVas showed localised oedema suggesting endothelial cell migration.21 The suture knot appears to be ‘leakage’. All active young vessels ‘leak’ fluid but lipid was more more prone to attract vessels than the rest of the suture. There commonly associated with mature and partially regressed vessels. was a good correlation between the level of lesion and level of Apart from the increased incidence of LK, herpetic keratitis vessels. That vessels progress at the level of corneal pathology was also associated with the greatest severity of CVas. would suggest a very fine and sensitive response of CVas to the Interestingly, acanthamoeba keratitis demonstrated the least chemokines and/or nerve stimuli generated by the corneal severity of CVas. There is no ready explanation for this but the lesion. This is best observed in the early stages of the disease fact that acanthamoeba feed on keratocytes would suggest that process. In persistent pathology vessels invade the entire cornea. keratocyte-derived factors may play an important role in stromal We also noted that the first CVas response was from the limbus CVas, which would be reduced when the keratocytes are closest to the lesion. This again supports the notion that a che- eliminated. mokine gradient is important in driving vessels towards the This study has demonstrated some novel clinical aspects of corneal lesion. CVas and re-enforced others. It is clear that CVas is a heteroge- There are several methods used to quantify corneal vessels. neous group of conditions with some stark variations deter- Early measures were simple and descriptive, for example, count- mined by aetiology and stage of the vessels. The increased ing the number of ‘red lines’ in each quadrant.22 Manual and association of herpetic keratitis with LK and with severity of semiautomated methods using computer software were later CVas is interesting. With emerging and exciting new treatment described but relied on good quality images and had the inher- modalities, clinical studies need to take the observations ent limitation related to overlapping vessels in different planes, reported here into account in the design and assessment of effi- which cannot be differentiated in two-dimensional cacy of intervention. – images.271012 23 Also, for sequential comparisons one would have to rely on similar images for accuracy. Contributors All patients belonged to HSD and were examined by him in his fi clinic. HSD, LAF and DGS collected patient data. HSD and LAF wrote the bulk of the Our proposed classi cation of vessels is clinically useful and manuscript. MA-A, DGS and AMO graded the severity of images. DGS, LAF and should enable clinicians to compare the stage of vascularisation MA-A made the image montages. HSD and LAF analysed the data and approved when determining response to any intervention, medical or sur- the final draft before approval by all authors. gical. Active vessels usually indicate that the stimulus that Competing interests None declared. 24 incited the vascularisation is still operative. Thus, treatment Ethics approval Nottingham Research Ethics committee 10/H0408/12. Approval with anti-VEGFs is likely to be effective on active vessels, the No. 2882. fi mature ones requiring other options such as laser or ne needle Provenance and peer review Not commissioned; externally peer reviewed. diathermy occlusion.7 Moreover, the scoring system is simple, clinically useful and has good intraobserver and interobserver correlation. REFERENCES 1 Azar DT. Corneal angiogenic privilege: angiogenic and antiangiogenic factors in Distinction between stages can be improved by provision of corneal avascularity, vasculogenesis, and wound healing (an American standard photographs together with detailed examination. In Ophthalmological Society thesis). Trans Am Ophthalmol Soc 2006;104:264–302.

Faraj LA, et al. Br J Ophthalmol 2016;100:315–322. doi:10.1136/bjophthalmol-2015-306686 321 Downloaded from http://bjo.bmj.com/ on March 7, 2016 - Published by group.bmj.com Clinical science

2 Ellenberg D, Azar DT, Hallak JA, et al. Novel aspects of corneal angiogenic and 13 Randolph JJ. Online Kappa Calculator. Secondary Online Kappa Calculator. 2008. lymphangiogenic privilege. Prog Retin Eye Res 2010;29:208–48. http://justus.randolph.name/kappa 3 Maruyama K, Nakazawa T, Cursiefen C, et al. The maintenance of lymphatic vessels 14 Randolph JJ. Free-marginal multirater kappa: an alternative to Fleiss’ fixed-marginal in the cornea is dependent on the presence of macrophages. Invest Ophthalmol Vis multirater kappa. The Joensuu University Learning and Instruction Symposium. Sci 2012;53:3145–53. Joensuu, Finland, 2005. 4 Ambati BK, Nozaki M, Singh N, et al. Corneal avascularity is due to soluble VEGF 15 Carmeliet P. Mechanisms of angiogenesis and arteriogenesis. Nat Med receptor-1. Nature 2006;443:993–7. 2000;6:389–95. 5 Cursiefen C, Chen L, Saint-Geniez M, et al. Nonvascular VEGF receptor 3 16 Wagoner MD, Islam SI, Riley F. Intracorneal hematoma in Mooren ulceration. Am J expression by corneal epithelium maintains avascularity and vision. Proc Natl Acad Ophthalmol 2000;129:251–3. Sci USA 2006;103:11405–10. 17 Li ZR, Li YP, Lin ML, et al. Activated macrophages induce neovascularization through 6 Nirankari VS, Dandona L, Rodrigues MM. Laser photocoagulation of experimental upregulation of MMP-9 and VEGF in rat corneas. Cornea 2012;31:1028–35. corneal stromal vascularization. Efficacy and histopathology. Ophthalmology 18 Duke-Elder S, Leigh AG. Diseases of the outer eye. In: Duke-Elder S, ed. System of 1993;100:111–18. ophthalmology. London: Kimpton, 1965:682–91. 7 Pillai CT, Dua HS, Hossain P. Fine needle diathermy occlusion of corneal vessels. 19 Siganos CS, Solomon A, Frucht-Pery J. Microbial findings in suture erosion after Invest Ophthalmol Vis Sci 2000;41:2148–53. penetrating keratoplasty. Ophthalmology 1997;104:513–16. 8 Faraj LA, Elalfy MS, Said DG, et al. Fine needle diathermy occlusion of corneal 20 Christo CG, van Rooij J, Geerards AJ, et al. Suture-related complications following vessels. Br J Ophthalmol 2014;98:1287–90. keratoplasty: a 5-year retrospective study. Cornea 2001;20:816–19. 9 Trikha S, Parikh S, Osmond C, et al. Long-term outcomes of Fine Needle Diathermy 21 Ye HQ, Azar DT. Expression of gelatinases A and B, and TIMPs 1 and 2 during for established corneal neovascularisation. Br J Ophthalmol 2014;98:454–8. corneal wound healing. Invest Ophthalmol Vis Sci 1998;39:913–21. 10 Dastjerdi M, Al-Arfaj K, Nallasamy N, et al. Topical bevacizumab in the treatment of 22 Cursiefen C, Wenkel H, Martus P, et al. Standardized semiquantitative analysis of corneal neovascularization: results of a prospective, open-label, noncomparative corneal neovascularization using projected corneal photographs--pilot study after study. Arch Ophthalmol 2009;127:381–9. perforating corneal keratoplasty before immune reaction. Klin Monbl Augenheilkd 11 Kim TI, Chung JL, Hong JP, et al. Bevacizumab application delays epithelial healing 2001;218:484–91. in rabbit cornea. Invest Ophthalmol Vis Sci 2009:4653–9. 23 Usui T, Yamagami S, Kishimoto S, et al. Role of macrophage migration inhibitory 12 Koenig Y, Bock F, Horn F, et al. Short- and long-term safety profile and efficacy of factor in corneal neovascularization. Invest Ophthalmol Vis Sci 2007;48:3545–50. topical bevacizumab (Avastin®) eye drops against corneal neovascularization. 24 Chang JH, Garg NK, Lunde E, et al. Corneal neovascularization: an anti-VEGF Graefes Arch Clin Exp Ophthalmol 2009;247:1375–82. therapy review. Surv Ophthalmol 2012;57:415–29.

322 Faraj LA, et al. Br J Ophthalmol 2016;100:315–322. doi:10.1136/bjophthalmol-2015-306686 Downloaded from http://bjo.bmj.com/ on March 7, 2016 - Published by group.bmj.com

Clinical evaluation and characterisation of corneal vascularisation

Lana A Faraj, Dalia G Said, Mohamed Al-Aqaba, Ahmad M Otri and Harminder S Dua

Br J Ophthalmol 2016 100: 315-322 originally published online July 10, 2015 doi: 10.1136/bjophthalmol-2015-306686

Updated information and services can be found at: http://bjo.bmj.com/content/100/3/315

These include: References This article cites 20 articles, 7 of which you can access for free at: http://bjo.bmj.com/content/100/3/315#BIBL Email alerting Receive free email alerts when new articles cite this article. Sign up in the service box at the top right corner of the online article.

Topic Articles on similar topics can be found in the following collections Collections Ophthalmologic surgical procedures (1189) Cornea (511) Ocular surface (605) Eye () (691)

Notes

To request permissions go to: http://group.bmj.com/group/rights-licensing/permissions

To order reprints go to: http://journals.bmj.com/cgi/reprintform

To subscribe to BMJ go to: http://group.bmj.com/subscribe/