AIMS/BACKGROUND An ideal keratoprosthesis (KPro) would closely resemble a donor corneal button in terms of its surgical handling, optics, and capacity to heal with host tissue in order to avoid many of the complications associated with the KPros which are currently in clinical use. This study was carried out to assess the long term clinical outcomes on implantation of the core and skirt poly(2-hydroxyethyl methacrylate) KPro in animals.
METHODS 20 KPros were made and implanted as full thickness corneal replacements into rabbits and followed for up to 21 months to date.
RESULTS 80% of the prostheses have been retained, with a low incidence of complications such as cataract, glaucoma, and retroprosthetic membrane formation which are frequently associated with KPro surgery.
CONCLUSIONS KPros of this type may offer promise in the treatment of patients for whom penetrating keratoplasty with donor material carries a poor prognosis. Refinement of the KPro and further animal trials, including implantation into abnormal corneas, are however mandatory before human implantation could be planned.
Statistics from Altmetric.com
There is a need for an alternative to donor corneal tissue for transplantation into those patients in whom penetrating keratoplasty (PK) is almost certain to fail. There is a long history of attempts to provide such an alternative through a variety of keratoprosthesis (KPro) designs and materials.1-10 However, although a small number of KPro designs have been shown to offer some hope of visual improvement when carried out by experienced surgeons,11-13 the complication rate remains high.
We are developing a KPro which resembles a donor corneal button in order to minimise the risk of complications. The Chirila KPro has a core and skirt design. Two flexible polymers of 2-hydroxyethyl methacrylate (PHEMA) are joined by an interpenetrating polymer network, a strong but flexible junction, resulting in a transparent optical core and an opaque sponge skirt. The two regions have different water contents and thus different physical characteristics, but are chemically identical. The core allows transmission of light and provides refractive power while the sponge skirt allows fibrovascular ingrowth, helping to secure the device in the long term. This is the first reported KPro which can be sutured into the host cornea as a full thickness implant in a manner analogous to PK.
All 20 KPros in this study were made using divinyl glycol (DVG) as crosslinking agent for the polymer. Ten (50%) additionally incorporated methyl methacrylate (MMA) as comonomer (5% wt HEMA) in the sponge skirt region in order to improve cell adhesion during biocolonisation of the sponge. This study was carried out to determine the handling characteristics of the KPros during surgery and to monitor the clinical outcomes and complications over a long period of follow up. In particular, we wished to assess whether the incorporation of MMA was beneficial overall, as it was expected to improve cell ingrowth but to reduce mechanical strength compared with sponges made without MMA. This is important because mechanical strength (resistance to tearing by sutures) had previously been identified as a crucial determinant of postoperative outcome, and mechanical tests of sponge specimens alone do not adequately predict peroperative performance.
Materials and methods
The method of manufacture has previously been described in detail.14-17 The Chirila KPro is made entirely of the flexible hydrogel PHEMA and comprises an optically clear core surrounded by a peripheral opaque sponge rim which allows tissue ingrowth (Fig 1). The rim is made first, by polymerisation of HEMA in the presence of 80% water, a crosslinking agent (DVG), and an initiator within a mould. PHEMA with a lower water content is then polymerised within the centre of the sponge skirt resulting in a transparent optic which lacks pores. During this process an interpenetrating polymer network (IPN) is formed between the core and the skirt, a flexible but extremely strong junction.14 On removal from the mould the crude button is frozen and cryolathed to the required dimensions (9 mm diameter) and curvature, and autoclaved before use. The 10 KPros made without MMA as comonomer will be referred to as “DVG KPros” and were implanted in rabbits D1 to D10. Those KPros in which MMA had been added will be referred to as “DVG/MMA KPros” and were implanted in rabbits M1 to M10.
Experimental surgery was carried out under general anaesthesia (halothane/oxygen by mask) in 20 half lop rabbits, weighing 2.5 kg each, and conformed with the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes, 1990. Pilocarpine 2% was given preoperatively to the prepared right eye. The cornea was de-epithelialised and a scleral ring was sutured to the limbus. The eye was then proptosed by means of gentle pressure while an 8 mm trephine was used to mark the cornea centrally to three quarters of its depth. The eye was then returned to the orbit and the corneal button removed with scissors. On opening, the anterior chamber was flushed with 5000 U heparin sodium.
The KPro was also flushed with heparin, placed into the recipient full thickness bed and sutured into position with interrupted 10/0 nylon. The KPro was held with lens introducing forceps to prevent scratching or tearing. The sutures were turned so as to place the knots internally. No Viscoelastic was used.
A 360 degree conjunctival peritomy was then performed using scissors, to incorporate Tenon’s capsule, and the conjunctiva-Tenon flap was drawn over the cornea-KPro surface and sutured to the limbus inferiorly using 6/0 Vicryl (Fig 2). A subconjunctival injection of 1 mg dexamethasone and 10 mg gentamicin, and topical chloramphenicol ointment, were given.
The animals were all examined on a daily basis for evidence of complications such as conjunctival retraction, aqueous leakage, infection, or iritis. Chloramphenicol ointment was given once daily for the first month. All animals were returned to theatre 2 months after implantation and the conjunctival flap was trephined open over the optical core under general anaesthesia. Once the flap had been opened, they received chloramphenicol ointment once daily for antibiotic cover and as a lubricant. On three occasions, at 1, 2, and 6 months postoperatively, intraocular pressure (IOP) was determined using a Tono-Pen XL electronic tonometer (Mentor O & O, Inc, MA, USA) during examination under anaesthesia.
Sixteen (80%) of the 20 KPros (10 DVG, 10 DVG/MMA) have been retained successfully to date. Surgery was performed sequentially on rabbits D1 to D10 and M1 to M10 such that rabbits D1 to D10 (KPros without MMA) have now been followed for up to 21 months, and rabbits M1 to M10 (KPros containing MMA as comonomer) for up to 19 months.
Peroperative and early postoperative complications are given in Table1. The KPros made without MMA as comonomer, in rabbits D1 to D10, were felt to be easier to handle and less prone to tearing during suture placement. However, the surgical learning curve counterbalanced this and the peroperative complication rates were not significantly different in the two groups. Other than stretching around the suture tracks in the sponge skirt, only one peroperative complication was seen—a bulging lens in rabbit M1 required lensectomy.
To date four rabbits (20%) have had complications requiring them to be killed. In rabbit M8 suture tears led to wound dehiscence, with the gap being plugged by epithelialised iris. This animal was killed at 13 days. Wound dehiscence also led to rabbit D5 being killed at 13 days. Two further rabbits were lost later during follow up—D4 at 8 weeks, owing to accidental damage to the KPro during flap opening, and M1 at 20 weeks. In the latter, possibly as a result of trauma, the conjunctival flap parted from the inferior limbus and a localised KPro-cornea dehiscence was noted in relation to broken sutures.
Late postoperative complications have been few. There have been no cases of infection, iritis, cataract, retroprosthetic membrane formation, or extrusion of the KPro. A red fundal reflex has been noted in all rabbits (Fig 3) and no rabbit has developed a relative afferent pupillary defect. IOP measurements taken at 1, 2, and 6 months are given in Table 2.
The conjunctiva-Tenon layer was found to be firmly adherent to the skirt, where tissue penetration occurs, but not stuck to the surface of the optical core from which it was easily removed 2 months after KPro implantation. The conjunctival opening was found to contract in a purse string manner over time, requiring retrimming every few weeks. Mild giant papillary conjunctivitis (mucoid production in response to the exposed optic, but without marked papillary changes in the upper tarsal plate) has been noted in approximately half of the animals at some point during follow up and responds to increased lubrication.
White deposits were noted to occur within the optic of the KPro in five animals, reducing the red reflex. These changes were noted from about 4 months after implantation and were progressive. There was no evidence of intraocular inflammation in any of these five animals. The deposits were subsequently shown to represent calcium deposition.
The four eyes (from D4, D5, M1, and M8) which were removed because of wound dehiscence or iatrogenic damage between 2 and 20 weeks after surgery were placed in 4% glutaraldehyde. Specimens were processed for light microscopy, which confirmed the presence of fibrovascular ingrowth into the skirt region (Fig 4), as seen in our earlier studies.18-20 Cell invasion from Tenon’s capsule into the anterior aspect of the KPro skirt appears to facilitate KPro-cornea healing. There was no evidence of calcium deposition in the optic region in any of these four KPros but specks of calcium were seen histologically within the KPro skirt region of rabbit M1, which was killed at 20 weeks (Fig 5).
It is essential that a KPro be strong enough mechanically for surgical manipulation and suturing and also to provide long term structural integrity for the eye. Sufficient ingrowth of cells and collagen deposition within the skirt must occur for permanent security after the sutures have lost their tensile strength.
Extensive preliminary work established that the PHEMA sponge contains pores of a suitable size for cell ingrowth,19 that tissue ingrowth occurs20-22 and that lamellar implantation of the intact KPro is possible.23 A pilot study was done in which our prototype KPro was placed as a full thickness implant in eight rabbit corneas,18 now followed for up to 2 years. That demonstrated that the PHEMA sponge skirt of the prototype had inadequate mechanical strength, leading to early failure in three of the eight rabbits. In order to increase the mechanical strength of the skirt the method of manufacture was altered. The crosslinking agent used for the KPros in this study, divinyl glycol (DVG), differs from that used previously (ethylene glycol dimethacrylate, EDMA) in that it is much more hydrophilic—that is, more soluble in water. The utility of this crosslinking agent for high water content biomaterials has been established in the poly(1-vinyl-2-pyrrolidinone) gels evaluated as potential vitreous substitutes.24-26 Mechanical tests using a Syntech 200/M materials test system established that PHEMA sponges polymerised using DVG as a crosslinking agent exhibit superior tensile properties to those formed using EDMA.27 This is attributed to increased crosslinking efficiency, as the DVG is more compatible with the aqueous reaction medium.
In spite of the marked improvement in mechanical strength, further improvement would be desirable as an element of sponge stretching frequently occurs during suture placement. However, this does not necessarily induce long term complications, presumably because such holes become sealed after flap coverage and tissue ingrowth into the sponge skirt.
Ten (50%) of the KPros for this study were made with the addition of MMA as comonomer within the sponge skirt.28Incorporation of MMA is desirable in order to improve cell adhesion to the sponge during keratocyte and fibroblast invasion. Unmodified PHEMA is known to be non-adhesive for mammalian cells,29 but the addition of MMA or other monomers by copolymerisation has been shown to increase cell adhesion considerably.30 31 Further, if a KPro for some clinical conditions was designed to be surfaced with corneal epithelium rather than by a conjunctival flap, a HEMA copolymer might prove to be essential in allowing epithelial cell adhesion to the surface. However, since MMA is insoluble in water, the mechanical strength of DVG crosslinked sponges produced with MMA as comonomer is somewhat compromised. An additional purpose of this study was therefore to assess which sponge formulation gave the best overall performance—that is, whether MMA resulted in poorer resistance to surgical handling, increasing the risk of early postoperative failure to an extent which outweighed any benefits of improved biocolonisation. This is important because we had previously found18 that peroperative problems with suturing the KPro were the most significant predictor of early postoperative failure. From this study, it appears that DVG crosslinked sponges are much stronger than our previous EDMA formulation and that strength is not significantly diminished by incorporation of MMA. There is thus no mechanical contraindication to incorporating MMA as comonomer in a PHEMA KPro skirt.
Histological examination was carried out only in the four failures in this study, and the degree of cellular infiltration was not found to be higher in the MMA-containing KPro skirts of rabbits M1 and M8 than in the KPro skirts lacking MMA of rabbits D4 and D5. It was not the purpose of this study to kill clinically satisfactory animals at predetermined intervals to assess whether there is a significant difference in cell ingrowth obtained in the two types of KPro skirt (with or without MMA), as this would have prevented our objective of long term follow up to exclude late complications. Therefore to analyse the nature and time course of biocolonisation in uncomplicated cases a separate study involving another 20 rabbits, 10 of whom had KPros made with MMA as comonomer, was undertaken in which animals were sacrificed at predetermined end points for histopathological evaluation.32
The most appropriate formulation in terms of allowing superficial epithelial cell growth over the skirt and optic is undergoing in vitro evaluation, as MMA would be expected to improve epithelial cell adhesion; this might be relevant to any future KPro model designed for implantation without a conjunctival flap in a near normal eye. The current method of implantation precludes epithelialisation of the optic because the flap skirt adhesion prevents ingress of any epithelial cells from the peripheral cornea. Cornea-skirt healing at the KPro edge, which is rapid, prevents epithelial down growth into the AC.
White calcium deposits developed in the optical core of a minority of KPros after implantation. Being such a variable finding, we suspected that it might be due to the method of manufacture. We have found that Soxhlet extraction (to remove toxic monomer) after polymerisation may also remove any remnants of initiator which had come out of solution, so leaving behind tiny spaces. These spaces could become supersaturated with calcium ions thereby forming nidi for calcification within the gel. Histological and in vitro studies in a non-biological environment support this assumption (Fig 6) and the method of manufacture has subsequently been altered to prevent such spaces arising. Dystrophic calcification is thought to be a less plausible explanation for the calcification seen in this instance, although it has been proposed as an explanation for an isolated case of calcification within a PHEMA IOL.33 No similar calcification was seen in an extensive clinical trial of PHEMA IOLs34 but has been noted with other biomaterials such as silicone, especially in relation to infection.35 There was no clinical evidence of infection in any of the eyes with calcification of the KPro optic. In subsequent studies, following improvement in the manufacturing technique, no calcium deposits in the optic have been observed.
We used the two rabbits with the most significant opacification of the optical core (M5 and M10), to assess the ease of KPro exchange, since it would be reassuring to know that KPro implantation could be repeated. The KPro was exchanged for a new one 218 days (M5) and 323 days (M10) after original KPro insertion. This involved dissection of the flap from the inferior limbus and from the sponge skirt and folding it superiorly. The KPro-cornea wound was opened with scissors to remove the KPro, and a new KPro sutured into place with 10/0 nylon before reforming the conjunctival flap. The anterior chamber was flushed with heparin on opening and postoperative antibiotics were given as with primary surgery. Postoperative findings (up to 12 months after replacement surgery) were satisfactory and the optics show no evidence of calcium deposition.
Specks of calcium were identified histologically in the sponge skirt of the KPro removed from rabbit M1. Calcification is commonly seen in implanted PHEMA sponges and other biomaterials.36-38Progressive calcification would be expected to reduce the stability of an implant in the long term. Cell death due to insufficient nutrition39 or to cytotoxicity, with calcification of necrotic cells, has been offered as an explanation for this. However, we have no evidence for such necrosis or of cytotoxicity on in vitro testing. Calcification may simply relate to concentrations of calcium within the pores of the sponge reaching a critical level. Alternatively, it might result from a low grade foreign body response. Occasional macrophages may be found in relation to PHEMA sponges, without evidence of polymorphonuclear infiltration. The size of the particulate sponge material appears to make it amenable to phagocytic attack and cell-polymer interactions may lead to macrophage activation. The finding highlights the importance of further studies on cellular behaviour and interactions within porous biomaterials and into factors which influence calcification. We are using histological techniques including live-dead cell analysis over time, and an accelerated in vitro calcification system for further investigations. Strategies to reduce calcification include altering the physical/chemical nature of the polymer to reduce crystal nucleation sites, incorporation of “competitive” cations which inhibit crystal growth, or modifications of the early inflammatory response. This is a key area of ongoing research.
Currently we are lathing the buttons to a diameter of 9 mm, of which the central optic has a diameter of 7 mm. The final thickness is 0.5 mm, the radius of curvature of the anterior surface 8.00 mm and that of the posterior surface, 8.50 mm.
Considering the anterior and posterior surfaces of the optic separately as refracting surfaces, ignoring lens thickness, and taking the refractive indices of air, the hydrated polymer gel, and aqueous to be 1.00, 1.43, and 1.33 respectively, allows the following approximations to be made: anterior surface power = 1.43 − 1.00/0.008 = 53.75 DS posterior surface power = 1.33 − 1.43/0.0085 = −11.76 DS refractive power of the KPro optic = 41.99 DS.
An optic of about 42 DS would be appropriate for the phakic human patient, and the surgery for implantation does not require routine lens extraction. In an aphakic patient or one requiring concurrent cataract extraction, the KPro can be lathed so as to provide more refractive power. For example, an anterior radius of 6.5 mm and a posterior radius of 9.5 mm would give a lens power of 55.62 DS.
KPros for human clinical use could be selected for the individual’s requirements after A scan ultrasound measurements, with a range of optic powers being available. Further, we have recently demonstrated in the rabbit model (unpublished data) that it is possible to carry out post-implantation adjustment of the optic surface by means of excimer laser ablation. This makes optical fine tuning a possibility and improves the surface quality, removing the fine ridges incurred by the cryolathing process. The ridges would be expected to degrade the image and may predispose to the mild giant papillary conjunctivitis seen in some of the rabbits, and are attributable to our unsophisticated laboratory lathe. For KPros for human implantation, higher standards of manufacture would have to be employed.
There is no clinical or histological evidence of retroprosthetic membrane (RPM) formation. This is partly attributable to the PHEMA forming the posterior surface of the KPro being non-adhesive for cells. However, the other commonly quoted complications of KPro surgery (iritis, cataract, retinal detachment, glaucoma) are also absent in this series. It appears that the risk of all these complications is minimised by the relatively non-invasive surgery involved. Postoperative steroids were not required. The absence of corneal melting around the KPro and of resultant extrusion is satisfactory and is thought to reflect both the lack of mechanical stress at the implantation site and also the protective effects of the conjunctival flap, which excludes collagenolytic enzymes from the healing cornea-KPro wound.
Measurement of intraocular pressure
In general, assessment of intraocular pressure (IOP) in KPro recipients is limited to digital estimation because of rigid KPro components occupying the cornea, often covered by eyelid skin. The use of a flexible KPro which occupies only a part of the large rabbit cornea in this series meant that “para-KPro” measurements could be taken with an electronic tonometer under sedation, to give a more accurate estimation of IOP than by digital assessment alone. The pressure in the right, operated eye, could be compared with that in the left eye, which was assumed to be a normal value for the tonometer under the anaesthetic regimen used. The mean value for the unoperated left eyes measured under these conditions was 14.4, range 7–23. The mean values for intraocular pressures following surgery in the right eyes were 11.2, 15.0, and 13.6 at 1, 2, and 6 months in rabbits D1 to D10, and 15.1, 24.2, and 16.1 at 1, 2, and 6 months in rabbits M1 to M10. A total of four rabbits (M1, M2, M7, and M10) out of the 17 who had their IOP recorded at 2 months had an IOP of 25 mm Hg or more, and three of these rabbits (75%) had had complicated surgery or suture related tears. Conversely, of the 13 rabbits who had “normal” IOPs at 2 months, only five (38%) had had suture related tears in the KPro sponge. The absence of clinical signs of iritis suggests that the higher IOP measurements may have been due to narrow angles pending tissue ingrowth into the sponge and restoration of a normal anterior chamber. It appears that a clinically useful estimation of IOP can therefore be obtained with this KPro.
The large diameter of the rabbit cornea allows the IOP to be taken through the cornea peripheral to a 9 mm implant using a Tono-Pen. Readings taken through the implant itself using a Tono-Pen were found to be highly variable. Thus in the human patient, unless the KPro were reduced in size to allow para-KPro measurements, an electronic tonometer of this kind could not be used. In order to assess the possibility of using a Schiotz applanation tonometer to estimate IOP, measurements were taken by both Tono-Pen and Schiotz tonometer in the right eye of a rabbit 4 months postoperatively. Using the Schiotz tonometer, the 10 g weight gave a scale reading of 4, and the 5.5 g weight a reading of 1, which using the Friedenwald nomogram gives an IOP estimate of 13 mm Hg. The mean Tono-Pen reading taken on the same occasion was 11 mm Hg. This suggests that the Chirila KPro allows clinically useful intraocular pressure estimations to be made by either method, the method chosen depending upon the relative diameters of the KPro and the cornea.
The rate of conjunctival flap regrowth has been found to be reduced by increased lubrication. A variant of the KPro now under evaluation has an elongation of the optic designed to discourage conjunctival overgrowth. Trimming of the conjunctival flap opening over the optic is required every few weeks in most rabbits. The flap tissue does not thin sufficiently to allow reasonable vision without its removal from the optic surface.
Giant papillary conjunctivitis in response to the exposed optic is seen in some rabbits, presumably aggravated by the ridges on the surface. However inflammation is mild, perhaps because the rabbit’s excellent tear film provides good lubrication, and its low blink rate ensures that the palpebral conjunctiva is rarely brought over the optic surface. In addition, the regular regrowth of conjunctiva prevents continuous exposure. More sophisticated manufacturing techniques, with the option of photoablation, should reduce this problem further.
Given that the PHEMA used in the optic has a relatively high water content, approximately 38%, there is presumed to be some passage of water and solutes from the aqueous across the optic, and evaporation from the exposed surface. The similarity of intraocular pressure readings in operated and unoperated eyes suggests that this occurs at a similar rate to that of the net movement of fluid across a natural cornea. It may be necessary to lubricate the exposed optic surface in the human patient to prevent superficial drying of the gel. We are evaluating potential surface active medications, such as methoxypolyethylene glycol (MW 750), 40% (v/v) in a chloramphenicol base, which discourages adhesion to the exposed optic surface by tear film proteins and pathogens. We are also evaluating the provision of a tougher protective layer on the external surface, to make it more resistant to minor trauma.
We envisage two very different groups of potential KPro recipients; firstly, those with severe ocular abnormalities involving limbal damage and dryness in whom a KPro offers the only hope of visual restoration. In most of these patients an epithelium could not be maintained and a non-epithelialised optic exposed through mucous membrane or skin would be appropriate. However, if a true “artificial donor button”, able to sustain a stable multilayered epithelium could be devised, ultimately it could have a much wider application, especially in countries where eye bank tissue is in short supply. We are now working on methods to achieve epithelialisation of PHEMA in both core and skirt regions. Corneal epithelium has been grown on the PHEMA copolymer gels and modified sponges in vitro (Fig 7) but long term epithelial maintenance in vivo, being dependent on mesenchymal interaction, may be difficult to achieve over the optic. Modification of the optic design may be required.
The Chirila KPro can be implanted in a manner analogous to PK, in a relatively simple, repeatable procedure. When covered with a conjunctival flap which is opened 2 months postoperatively the complication rate in this animal study compares favourably with other KPro models; cataract, iritis, retroprosthetic membrane formation, glaucoma, retinal detachment, and extrusion were not seen with up to 21 months’ follow up. Intraocular pressure may be estimated by tonometry.
This PHEMA KPro, when DVG is used as crosslinking agent and whether or not MMA is incorporated as comonomer in the sponge, is much less prone to suture related tears than our earlier prototype. Even so, three (15%) KPros failed because of early wound dehiscence because of sponge tears or presumed ocular trauma, and one failed at 8 weeks as a result of iatrogenic damage. Sponge tears may also predispose to temporary flattening of the AC and elevation of IOP in the early postoperative period. Subjectively, individual KPros vary considerably in their surgical handling characteristics and it is clearly important to improve the quality control of KPro manufacture.
The study has identified specific areas for further research. In particular, cellular responses to and interactions with the porous, particulate PHEMA within the sponge skirt are of interest. Animal trials continue, with particular emphasis on determining the long term complication rate and on improving the optical performance and surface quality of the KPro. Assessment of the KPro in animal eyes with pre-existing pathology is also currently under way.
The authors gratefully acknowledge Annette Dawes and staff for anaesthetic assistance and Chris Barry for help with photography.