Outcomes of treatments for keratomalacia in dogs and cats: a systematic review of the published literature including nonâ•’randomised controlled and nonâ•’controlled studies

1 British Small Animal Veterinary Association This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. Outcomes of treatments for keratomalacia in dogs and cats: a systematic review of the published literature including non-randomised controlled and non-controlled studies

Systematic review of randomised trials or n of one trial Randomised trial or observational study with dramatic effect Non-randomised controlled cohort/ follow-up study Case series, case-control studies or historically controlled studies ‡ Mechanism-based reasoning † Level may be graded down on the basis of study quality, imprecision, indirectness (study PICO does not match questions PICO), inconsistencies between studies or because the absolute effect size is very small. Level may be upgraded if there is a large or very large effect size. ‡ As always, a systematic review is generally better than an individual study.
inhibitors and are required for normal homeostasis of the ocular surface. Imbalance of these enzymes with their inhibitors leads to over-zealous collagen destruction and corneal degradation. Pre-existing and ocular surface co-morbidities may increase the risk of keratomalacia, including keratoconjunctivitis sicca, neurotrophic keratitis, exposure keratitis due to lagophthalmos and bacterial or fungal keratitis.
Anti-collagenase treatment is instigated to try to mitigate collagen loss and retain as much corneal tissue as possible. Anti-collagenase treatments used in veterinary ophthalmology include: topical: serum, plasma, fresh frozen plasma (FFP), freeze-thaw cycled plasma (FTCP), platelet-rich plasma (PRP), ethylenediaminetetraacetic acid (EDTA), acetylcysteine and tetracyclines (topical and/or oral). Additionally, most cases require antimicrobial treatment appropriate to any identified infections (by cytology, culture and sensitivity or PCR testing) as well as analgesic treatments including oral non-steroidal anti-inflammatory drugs and/or opioid-based analgesia and/or paracetamol (dogs only) and atropine. Any other ocular co-morbidities may require additional treatments (e.g. eyelid surgery, tear replacement and/ or stimulation etc.) Surgical debridement of necrotic corneal tissue may be appropriate and help to stabilise the malacic area by debulking the enzyme-rich malacic tissue. In some cases, the degree of corneal tissue loss is substantial and tectonic support may be required. Surgical treatments described in dogs and/or cats include: conjunctival pedicle grafting, corneal grafting (fresh or frozen; homologous or heterologous tissue), corneoconjunctival transposition (CCT), collagen biomatrix grafts [e.g. porcine intestinal (BioSIS) or bladder (ACell) submucosa, bovine pericardium (Tutopatch) and equine pericardium, without or without additional third eyelid flap (TELF)], amnion grafting (homologous or heterologous), cyanoacrylate glue application, keratectomy with TELF and collagen cross-linking.
A systematic review of the current literature was undertaken to determine the evidence base for the various treatments of keratomalacia described in dogs and cats. The aim of this review was to assess the evidence base and identify recommended treatment(s) based on globe survival, visual outcome, and time to resolution (while maintaining a globe).

MATERIALS AND METHODS
The following question was designed to establish the evidence base for the treatment of keratomalacia in dogs and cats: "Which of the treatment options for keratomalacia in dogs and cats offers the best chance of globe survival, the fastest time to resolution with globe survival, and the best visual outcome." An on-line literature search was undertaken 5 October 2020 for studies and case series/reports related to treatment (both medical and surgical) of keratomalacia in dogs and cats. The search criteria were restricted to the English-language publications over the last 30 years (1990 to 2020).
The search utilised the PubMed (http://www.pubmed.gov/) and Institute for Scientific Information (ISI) Web of Science (http://wok.mimas.ac.uk/) databases. Databases were searched using the following terms: (keratomalacia OR corneal melt* OR corneal malacia) AND (dog OR canine OR canid OR cat OR feline OR felid) AND (treatment OR outcome OR morbidity OR complications). A further search for (cornea* graft*) AND (dog OR canine OR canid OR cat OR feline OR felid) was undertaken on PubMed and cross-referenced against original search to exclude duplicates and assess if grafts undertaken for keratomalacia (grafting for other diseases excluded).
The protocol for this review has not been published on a repository or in another journal, although follows the same principles out-lined in Tivers et al. 2017. Studies were assessed by one author (CH) and excluded if they were studies relating to non-keratomalacia disease, related to species other than dogs or cats, had less than three keratomalacia cases included, were conference abstract only publications, were review articles with no new data, were experimental treatments or in vitro studies or were duplicated.
Studies were reviewed and assigned a level of evidence base as described in the Oxford Centre for Evidence-Based Medicine (OCEBM Levels of Evidence Working Group 2017), as summarised in Table 1. Each study was assessed for type of study described (e.g. retrospective, prospective, controlled, random/ non-random, cohort study, case series/study), the number of animals included, criteria for assessing outcome (e.g. vision, corneal clarity, anatomic repair) and duration of follow up and time to resolution (see Tables 2 and 3).
Statistical analysis of the data from included studies in this review was not submitted for statistical synthesis (meta-analysis) as study design differences were considered to have made direct comparison of data misleading.

RESULTS
A total of 76 studies were identified in the initial search of ISI Web of Science databases. Studies were excluded if they related to species other than dogs or cats (four), less than three kera-  (41), were conference abstract only publications (two), were review articles with no new data (four), experimental treatments or in vitro studies (four) or were duplicated (one). The initial PubMed database search revealed a total of 35 studies and 24 were excluded as studies relating to non-keratomalacia disease (14), less than three keratomalacia cases (three), experimental treatments or in vitro studies (three) or were review articles with no new data (four). The final PubMed search (cornea* graft*) AND (dog OR canine OR canid OR cat OR feline OR felid) yielded 67 studies of which 59 were excluded as studies relating to non-keratomalacia disease (32), or related to species other than dogs or cats (nine) less than three keratomalacia cases (three), experimental treatments or in vitro studies (nine), were review articles with no new data (three) or were duplicated (one).
Eighteen (18) studies were identified as providing information to answer the proposed question. One study was classified as providing level 3 evidence, 10 as level 4 evidence (three 4(a) and seven 4(b)) Tivers et al. 2012) and seven as level 5 evidence (summarised in Table 2). The findings of these 18 studies with respect to number of animals included duration of follow up, time to epithelial healing (fluorescein negative staining), anatomical outcome, vision and corneal clarity outcomes for keratomalacia cases are summarised in Table 3.

Direct comparison of different treatments
Only one study compared the outcome of two different treatments for keratomalacia ). This was a prospective, non-randomised, controlled cohort study of 49 eyes (35 dogs and 11 cats) over a 3-year period (2009 to 2012) and was classified as level 3 evidence.
Nineteen eyes (19 animals, 12 dogs and seven cats) were treated with corneal collagen cross-linking in addition to standard medical treatment (topical antibiotics, topical and systemic col-lagenase inhibitors, ±topical atropine 1%, systemic meloxicam and buprenorphine) (CXL group). Thirty eyes (27 animals, 23 dogs and four cats) were treated with standard medical treatment alone (control group). Allocation was dependent on clinician and owner discretion. Cases with corneal perforation or descemetocoele were excluded.
Corneal cross-linking (CXL) was performed under general anaesthesia with the eye positioned in a horizontal plane. Application of 0.1% iso-osmolar riboflavin drops (in 20% dextran solution) was performed every 3 minutes for 30 minutes followed by irradiation for 30 minutes with 365-nm ultraviolet A light (irradiance 3 mW/cm 2 , UV-X Peschke Meditrade, Cham, Switzerland) with continued riboflavin drop application every 3 minutes during this period. Irradiation of the limbus was avoided.
Cases were re-examined at days 7, 14 and 28, and thereafter at various time points in long-term follow up. The primary end point was stabilisation of the keratomalacia, and the requirement for surgical/rescue stabilisation (or enucleation) was considered treatment failure. Rescue treatment [CXL, or conjunctival pedicle graft (CPG) or TELF] was recommended if greater than or equal to 20% additional stroma was lost during follow up.
Rescue treatment was undertaken in nine of 30 control group eyes and five of 19 CXL group eyes, which was not statistically different between groups (for total cases or for dog/cat cases separated). Rescue CXL was undertaken in seven of nine eyes in the control group, and CPG in one of nine and declined by owner in one of nine. Rescue CPG was undertaken in four of five eyes in CXL group and TELF in one of five. The ulcer size and depth was greater in the canine CXL group compared to the canine control group, but not in the feline groups. Although rescue treatment was not more significant in either group, the canine ulcers in the CXL group were deeper and larger than those in the control group at initial presentation. Overall stabilisation rate after CXL was 74% (14/19), and 100% (6/6) for rescue CXL.
Ulcer deepening during the follow-up period was seen in both CXL and control groups, but this was greater in the control group (mean 35% stromal loss >50% stromal loss) than the CXL group (50% > 55%). The time to epithelial healing (negative fluorescein staining) (P=0.02) and the time to stabilisation of the corneal stroma was longer in the canine CXL group compared to the canine control group, but there was no statistical significance between the feline groups. The depth of stromal thinning at the site of previous ulceration was greater in the canine control group (20%) compared to the canine CXL group (2.5%), but this effect was not seen in the feline treatment groups.
Prospective studies reporting the outcome for one treatment Three prospective studies were designed to assess the treatment of CXL (Speiss et al. 2014) and accelerated CXL (Famose 2014 andFamose 2015) for the treatment of keratomalacia in dogs and cats (Speiss et al. 2014), dogs (Famose 2014) and cats (Famose 2015). These were non-controlled, non-randomised unmasked studies and classified as level 4 (4a, Tivers et al. 2012) evidence.

Type of study Papers included
Level 3 Non-randomised controlled cohort/follow-up study Pot et al. (2014) Level 4 (a) Non-randomised prospective cohort study Speiss et al. (2014) Famose (2014) Famose (    Speiss et al. reported a prospective pilot study using the same CXL procedure described in Pot et al. on three dogs and three cats with keratomalacia. This small series had a longer follow-up period than Pot et al. with a median follow up of 9.5 months, and complications in three of six cases of bullous keratopathy, sequestrum formation and corneal pigmentation. Interestingly, examining the Pot et al. study for these complications demonstrated no statistically significant difference in rate of the control group versus the CXL group, suggesting the underlying keratomalacia might be responsible for complications rather than the CXL procedure. Accelerated CXL was used as a treatment for keratomalacia in eight dogs (Famose 2014) and 10 cats (Famose 2015) utilising 0.1% isotonic riboflavin (in 20% dextrose solution, Vibex) applied every 2 minutes for 30 minutes irradiation for 3 minutes with 370-nm ultraviolet A light (irradiance 30 mW/cm 2 , KXL Avedro, Waltham, MA, USA). Both studies showed faster median epithelialisation at 15 days in dogs and 8 days in cats, compared to 33 days in dogs and 20 days in cats in the Pot et al. study. Variable degrees of corneal fibrosis were noted in both cats and dogs, and corneal pigmentation was noted in two of eight dogs at 30 days post-treatment, although all cases were reportedly visual.
Retrospective case studies reporting the outcome for one treatment Seven retrospective case studies reported the outcome for single treatments including lamellar keratectomy, porcine acellular biomaterials (SIS or ACell), or bovine pericardium (Tutopatch) with TELF for varying lengths of time (2 to 5 weeks), or cryopreserved amniotic membrane (human or bovine) or medical treatment alone. These studies were classified as level 4 (4b, Tivers et al. 2012) evidence.
Three studies (Vanore et al. 2007, Goulle 2012, Balland et al. 2016 reported retrospective data for porcine acellular biomaterials [porcine small intestinal submucosa (SIS) and porcine urinary bladder acellular matrix (ACell)]. Vanore et al. described the successful use of SIS in two cats and five dogs with keratomalacia. In all but one dog, a TELF was used for 15 days postoperatively to protect the underlying graft. Vision and globe maintenance was reported in all cases, with all cases fluorescein negative at 15 days postoperatively (when TELF removed), and at 6 months only two of seven had residual corneal scarring (not graded). Both of these cases had suffered corneal perforation and the remaining five of seven had not perforated.
Balland et al. described the use of ACell with a TELF for 18 days in 10 dogs with keratomalacia as part of a retrospective study of 27 dogs and three cats undergoing corneal reconstruction with this biomaterial. In all keratomalacia cases, the postoperative corneal scarring (opacity) was subjectively graded as moderate, although the grading system (mild, moderate, severe) was not described in detail. Re-epithelialisation (fluorescein negative staining) was complete at 18 days in nine of 10 keratomalacia cases and one of 10 at 45 days.
Goulle 2012 reported a much larger retrospective study of 106 cases of corneal reconstruction using SIS biomaterial and a TELF for 3 weeks, of which 42 dogs and seven cats had keratomalacia. Time to fluorescein negative staining (epithelial healing) was not given. Successful anatomical repair was reported in all canine cases, with corneal transparency or a discrete scar at 3 months in 27 (64%) of 42 keratomalacia cases. A mild scar was reported in four (10%) of 42, a marked scar in 11 (26%) of 42, faint pigmentation in eight of 42 (19%) and mild pigmentation in five (12%) of 52. Twenty canine cases were followed for more than 3 months, and of these, five (25%) of 20 developed visual impairment as a result of marked corneal pigmentation. The feline cases were all visual at 3 months postoperatively, with corneal transparency or discrete scar in three (43%) of seven, a mild scar in three (43%) of seven, a marked scar in one (14%) of seven and sequestrum formation in one cat (14%). The corneal opacity grading system (discrete, mild, marked) was not described in detail. Dulaurent et al. 2014 reported on the outcome of a retrospective study of the use of bovine pericardium (Tutopatch) for corneal reconstruction in three dogs with keratomalacia, as well as three feline cases of corneal sequestrum. This was successful in two of three dogs with epithelial healing at 2 weeks, an opalescent scar at 4 weeks and translucent with a focal scar at 2 months in both cases. The remaining dog underwent a rescue surgery of conjunctival bridge graft placement and was blind in this eye. It is difficult to draw any conclusions on the suitability of bovine pericardium grafting for keratomalacia with such a small sample size.
Costa et al. 2019 published a larger multicentre retrospective study using cryopreserved amniotic membrane for corneal reconstruction in 111 dogs, of which 51 had keratomalacia. Data specific to the keratomalacia cohort was not given except that nine of 51 keratomalacia cases suffered complications (e.g. suture dehiscence, graft failure, graft pigmentation) but the specific complications were not reported. The authors reported epithelial healing in a mean of 26 days (15 to 45 days) over all 111 dogs but was not reported for the keratomalacia cases in isolation. Vision in the keratomalacia cases was described as good in 46 (90%) of 51 and poor or absent in five (10%) of 51, which was not statistically different from the overall rate of 92% vision and 8% poor or absent vision. Corneal opacification was graded as 0 to 2 in 30 (59%) of 51 and 3 to 4 in 21 (41%) of 51 of keratomalacia cases, which was similar to the overall grade 0 to 2 (53%) and 3 to 4 (47%). Grading was subjectively ascribed as 0 -transparent, 1 -faint opacity, 2 -mild opacity, 3 -moderately opaque and 4 -severely opaque. The authors noted that those cases that had longer epithelial healing times (mean 26 days) had less opacity than those with shorter healing times (mean 22 days). The authors also found that larger defects, those with concurrent ocular disease (e.g. anterior uveitis, keratoconjunctivitis sicca, trichiasis, distichiasis etc), those with perforations or descemetocoeles and those utilising human amniotic membrane (rather than bovine) had higher complication rates, but specific data relating to the keratomalacia cohort were not given. The authors also noted that larger defects were significantly associated with poor or absent vision (visual eyes median 5 mm versus poor/ absent vision median 9 mm). Demir et al. (2020) reported a retrospective study utilising lamellar keratectomy with TELF in place for 4 to 5 weeks for the treatment of 20 cats with keratomalacia. All corneas were fluorescein negative at removal of the TELF, and all animals were reported to be visual. Scarring was graded at 1 to 1.5 months as transparent (6/20, 30%), mild (6/20, 30%) or thick and vascularised (8/20, 40%). Follow up was reported to have continued periodically (monthly) thereafter but these data were not given. Guyonnet et al. 2020 reported the outcome of 57 eyes of 53 dogs with keratomalacia treated with medical therapy alone. Medical treatment consisted of topical tobramycin and equine serum each q2 to 4 hours, with topical atropine as deemed appropriate, and systemic meloxicam. Cases were considered as successful or failures at day 15 dependent on fluorescein staining (negative=successful) and whether rescue surgical intervention (conjunctival graft, porcine SIS graft, porcine ACell graft, ovine amniotic membrane graft or enucleation) was required (if required=failure). Thirty one keratomalacic eyes were successfully treated with medical treatment (52%). Median time to fluorescein negative staining was 6 days (range 2 to 15 days). Rescue surgical intervention was undertaken in 26 (48%) of 57 eyes where greater than20% progression of stromal loss was witnessed. Twenty two eyes in this rescue group were visual (three enucleated, one lost to follow up), although the degree of corneal opacity "varied greatly depending on the surgical technique." Of the successfully medically treated group, 30 of 31 eyes were visual at day 15, and 14 of 15 eyes followed to day 60 were visual. At this point, corneal opacity was graded as mild (9/15, 60%), moderate (3/15, 20%) or severe (3/15, 20%).
Seven level 5 studies were identified reporting retrospective case series of single treatments for corneal reconstruction where keratomalacia cases were not easily separated from other cases. Additionally, in four level 5 studies (Chow & Westermeyer 2016, Dorbandt et al. 2015, Hansen & Guandalini 1999, Watte et al. 2004) data on time to epithelial healing or vision were not overtly stated so these outcome parameters were less easily compared with other studies.

DISCUSSION
This systematic review of the literature pertaining to the treatment of keratomalacia in dogs and cats reveals that the evidence base for recommending any one type of treatment is very weak. In recommending the most effective treatment, the decision should be based on the most reliable evidence available. The Oxford Centre for Evidence-Based Medicine system to rank evidence has been revised by the OCEBM Levels of Evidence Working Group in 2017. The revision sought to reflect clinical decision making and was simplified whilst avoiding making definitive recommendations. This allows this system to be used when no systematic reviews are available, and is more appropriate for application to the veterinary literature (Tivers et al. 2012, Tivers et al. 2017. Systematic reviews of randomised controlled trials are considered to provide the most reliable evidence on which to base recommendations for treatment(s). Sadly, randomised controlled trials are infrequent in the veterinary literature, and this review only identified one non-randomised controlled trial classified as level 3 evidence, and no level 1 or 2 studies on the outcome of British Small Animal Veterinary Association treatments for keratomalacia in dogs and/or cats. Most studies were level 4 evidence reporting the outcome of a single treatment, with three prospective studies (4a) and seven retrospective studies (4b). The remainder were level 5 evidence providing minimal evidence to answer the question posed in this review. This is not a criticism of those studies, but merely noting that they did not provide good evidence for answering this particular question.
The best evidence available for the treatment of canine and/ or feline keratomalacia exists for the use of CXL in the management of keratomalacia but is limited to one level 3 study . This study demonstrated no statistically significant difference in outcome either anatomically or with regards to vision between CXL and control groups. However, seven of nine treatment failures in the control group were successfully rescued with CXL treatment (cross-over for selected patients). The limitations of this study include small treatment groups (based on this preliminary data and assuming the same patient recruitment rate, a power calculation suggested the study would need to run for 10 years to demonstrate a statistically significant difference in outcome between groups) and selection bias (nonrandomisation; clinicians/owners appearing to favour CXL treatment for larger and deeper ulcers in dogs) and unmasked cases (potential bias in follow up assessment). The follow-up period was also relatively short in this study, particularly in the control group patients (control group median follow up 1 month, CXL group median follow-up 3 months). This study had a high risk of bias due to clinician allocation of cases to treatment groups. Additionally deep keratomalacia cases (descemetocoeles and perforations) were excluded due to the nature of the collagen crosslinking treatment considered contraindicated for these cases.
Level 4 evidence in the form of studies reporting on outcome following a single treatment provide significant information but are less able to distinguish a leading treatment in terms of time to epithelial healing, anatomical and vision outcomes. Assessment of vision in dogs and cats remains crude, using the menace response in the studies considered in this review as a positive indication of vision. It is therefore not surprising that differences in vision outcome were elusive given this low bar. Corneal clarity might be indirectly indicative of visual compromise; however, a standardised objective measurement of this was not established in any of the studies, and relied on author grading (e.g. transparent, mild, moderate, marked/severe) that conceivably would vary between studies, making comparison between studies challenging.
Studies in this current review had small to modest numbers of canine keratomalacia cases (range 3 to 53 dogs, median 9 dogs), and small numbers of feline cases (range 3 to 20, median 8 cats), with two studies not separating cases based on species (15 to 25, median 20 animals). These small case numbers make the likelihood of identifying statistically significant differences slim unless dramatic effect(s) of treatment were present.
Length of follow up varied between studies, and given that corneal remodelling may continue for an extended period (months) it is also possible that differences in corneal clarity may have been more obvious with longer follow-up periods. In two prospective studies (Famose 2014, Famose 2015, the follow up was only 30 days. Both studies reported on the outcome of accelerated CXL on kera-tomalacia cases in dogs (Famose 2014) and cats (Famose 2015). In both studies, corneal opacity was described as variable between cases, and data on grades of opacity were not given.
Goulle 2012 demonstrated an apparent increased corneal opacity severity in the feline keratomalacia cases, which is somewhat at odds with the general consensus that the feline cornea scars less than the canine cornea in response to surgery. However, the number of feline keratomalacia cases was small and only three of seven were followed for longer than 3 months. It is possible that a longer follow-up period may have demonstrated further corneal clearing in the feline cases. Demir et al. 2020 reported corneal scarring as transparent in 30%, mild in 30% or thick and vascularised in 40% at 1 to 1.5 months follow up. Follow up was reported to have continued periodically (monthly) thereafter but these data were not given. It seems likely that some of the cats with thick or vascularised scars at 1 to 1.5 months would have had significant clearing at later follow up examinations.
In conclusion, the evidence for recommending any one treatment for keratomalacia in dogs and/or cats over another is very weak. As it stands, a combination of the treatments outlined in this review may be the most appropriate (medical and surgical) depending on the individual case. Whilst no study exists comparing no treatment to any one treatment (for understandable ethical reasons), level 5 evidence (based on physiology and first principles, i.e. mechanism based reasoning) would suggest that medical treatment with anticollagenase treatment is a minimum requirement for these cases to prevent globe loss through perforation (with attendant pain and suffering). Future studies that are randomised and controlled would be warmly welcomed to expand the evidence base in this field.