Distribution, prevalence, and causative agents of fungal keratitis: A systematic review and meta-analysis (1990 to 2020)

Fungal keratitis is a sight-threatening infection with global distribution. In this systematic review and meta-analysis we collected all the articles with data on the prevalence of fungal keratitis among various patient groups from January 1, 1990 to May 27, 2020. The 169 eligible articles were divided into 6 groups. The pooled prevalence was variable with values ranging from 0.05% among post-keratoplasty patients to 43.01% among patients with a clinical suspicion of fungal keratitis. Except for post-keratoplasty cases (yeast: 51.80%), in all patient groups moulds were more common than yeasts. Although more than 50 distinct species of fungi have been found to cause fungal keratitis, Fusarium species followed by Aspergillus species were the most common causes of the disease. In general, 9.29% (95% CI 6.52, 12.38) of fungal keratitis cases were mixed with bacterial agents.

A healthy cornea is rarely infected with fungal agents [13]. Traditionally, FK is considered to 76 be a disease prevalent in rural settings occurring in traumatized eyes by vegetative materials 77 or soil-contaminated objects in middle-aged agriculturists and labourers. These are evidenced 78 to be the most susceptible eyes for fungal infection of the cornea in low-income countries [1,79 14]. Conversely, contact lens usage (CLU) is the primary culprit, predisposing hosts to FK in 80 developed countries [1,7,14]. 81 The most prevailing species implicated in FK, include those belonging to Aspergillus, 82 Fusarium, Candida, Curvularia, and Penicillium genera in descending frequency [15][16][17][18][19]. 83 Most of these species are environmental residents which invade traumatized or 84 immunologically weakened eyes [6]. However, thermally dimorphic fungi, although rarely, 85 have also been reported as causative agents of FK [8]. Yeast ocular infections occur more  100 The protocol of this study was registered in the International Prospective Register of 101 Systematic Reviews (PROSPERO) with the ID number CRD42020188770 and the study was 102 done according to the preferred reporting items for systematic reviews and meta-analyses 103 (supplementary file 1) [21]. Relevant literatures were searched in Web of Science (ISI), 104 PubMed, Scopus, and Google scholar using the main keywords "fungal keratitis", 105 "keratomycosis", and "mycotic keratitis" and a set of other keywords, solely, and in 106 combination (supplementary file 2). To ensure that the search captures all the relevant 107 articles and because of usage of general phrases such as "infectious keratitis" or "microbial 108 keratitis" that include a set of microorganisms i.e. bacteria, viruses, fungi, and amoebae; a 109 wide search strategy was used. The search was limited to "article" as document type 110 (whenever available), "English" as language, and "January 1,1990 to the date of search (May 111 27, 2020)" as the publication date.  113 The resulting articles in database searching were imported into an EndNote X9 software 114 library for de-duplicating and title and abstract screening. After excluding irrelevant citations, 115 full texts of citations were downloaded and checked for eligibility. All studies reporting data 116 of prevalence and causative agents of fungal keratitis were eligible. Studies reporting animal 117 or ex vivo models of keratitis, case reports and case series (without a denominator of the 118 population that the cases had been diagnosed from), case-controls, cohorts, clinical trials, in 119 vitro studies on virulence factors, antifungal susceptibility pattern of keratitis-isolated fungi 120 but without data of prevalence, review articles, letters, and studies on therapy or keratitis 121 caused by non-fungal microorganisms were excluded. Studies on specific populations e.g. specific age groups, those with specific surgical interventions, etc. were also excluded except 123 for situations that several number of articles on the same patient group were available. In this 124 case, the articles were included in the study but were analyzed separately. The quality of the 125 relevant full texts was assessed using a modified version of the Newcastle-Ottawa Scale. All 126 steps of screening and quality assessment were done by two independent researchers and in 127 the case of inconsistency, a third researcher made the final decision.   136 Data were analyzed using Stata software version 14. To determine the heterogeneity, I 2 and 137 Cochran Q test were used. In accordance with the Higgins classification approach, I 2 values 138 above 0.7 were considered as high heterogeneity. In the presence of heterogeneity, a random 139 effect model was used in calculations. The pooled prevalence with a 95% confidence interval 140 (CI) was calculated using the "metaprop" command, and to estimate the pooled prevalence, 141 we used the random-effect model. The exact method was used for calculating pooled 142 estimates, variances, and their confidence intervals. We used Freeman-Tukey double arcsine 143 transformation for variance stabilization.

144
The pooled prevalence of fungal keratitis and the pooled prevalence of yeast and mold 145 keratitis and mixed fungal-bacterial infections were estimated. To determine the pooled 146 prevalence of fungal keratitis in different countries, subgroup analysis was performed. The 147 "metabias" command was used to check the publication bias, and if there was any publication 148 bias, the prevalence rate was adjusted with the "metatrim" command using trim-and-fill 149 method. The meta-regression analysis was used to examine the effect of the year of 150 publication and sample size as factors affecting heterogeneity among studies. In all analyses, 151 a significance level of 0.05 was considered. group III, the denominator was very much smaller than other groups because it did not 177 include patients who were clinically suspected and their etiology was rather non-infectious.

194
Based on the results of meta-regression analysis, no significant change was noted in 195 prevalence over the 30 years of study (p-value= 0.081) ( Figure S3). There was also no 196 association between the prevalence and the sample size studied in each report (p-value:  From a total of 15295 fungal isolates, 13048 were identified. These isolates belonged to 63 202 distinct genera, and Fusarium species (n=5294, 40.57%) followed by Aspergillus species 203 (n=4047, 31.02%) and Candida species (n=582, 4.46%) and these were the most common 204 causes of the disease. As shown in Table 2  to molds ( Figure S6).

220
From a total of 2016 fungal isolates from these patients, 1636 isolates were identified, mainly 221 at the genus level. As shown in Table 3, the identified isolates belonged to 27 distinct genera, 222 and Fusarium species (n=771, 47.13%) were the most common causes of disease followed by   235 In general, 13 articles were categorized in this group. These articles included studies due to molds ( Figure S8).
261 262  268 Studies categorized in this group were those that reported the prevalence of fungal keratitis   277 Generally, 23 articles met the inclusion criteria that were focused on patients who had 278 undergone keratoplasty. We divided these articles into 2 groups, 7 articles that provided data 279 on patients who had undergone keratoplasty due to a variety of indications except infective 280 keratitis, and 16 articles that provided data on patients who had undergone keratoplasty due 281 to infective keratitis. Such was done because the denominator of the latter group was smaller 282 than that of the former. The pooled prevalence of fungal keratitis (reinfection or recurrence) 283 was 0.05% (95% CI 0.00, 0.14) and 8.57% (95% CI 3.89, 14.62) in these groups, respectively ( Figure S13). In general, the prevalence of yeast was higher among these patients with a 285 pooled value of 51.80% (95% CI 14.41, 88.30) (Figure S14). 286 Data of the causative fungi were available only in 13 articles. From the total of 379 isolates, 287 294 were identified. As shown in were recorded in countries such as Paraguay, Ethiopia, Sri Lanka, and Bangladesh (Table 1) 315 which had the lowest eligible studies included (Table S1); these prevalence rates might be 316 less reliable than those documented in countries with high eligible studies such as India, 317 China, USA, Nepal, Taiwan, and Brazil (Table 1 and Table S1).