Association Between Keratoconus and Inflammatory Bowel Disease: A Systematic Review and Meta-Analysis

Supplemental Digital Content is Available in the Text.

Abstract

Purpose:

To investigate an association between inflammatory bowel disease and keratoconus.

Clinical Relevance:

Keratoconus can be difficult to diagnose early, and because of its progressive nature, early treatment is beneficial to preserving vision. Understanding potential associations of systemic disease with keratoconus may aid the development of screening protocols and guide early management decisions.

Methods:

Our search of 5 key databases (PubMed, Ovid Medline, Scopus, Web of Science, and Embase) was conducted on October 16, 2024. The selected papers were categorized as patient studies or genetic association studies for analysis. The primary outcome was odds ratio. Meta-analysis and test of assumptions was conducted on patient studies. Bias assessments were conducted using Q-GENIE and ROBINS-E.

Results:

Our search yielded 299 publication titles, of which 9 were included in our systematic review and 3 were incorporated into the meta-analysis. The aggregate odds ratio of patient-focused studies was 1.39 (95% confidence interval: 0.82–2.37). In the 4 studies investigating Crohn disease, there was no significant causal association (all P-values >0.05). All 5 genetic association studies demonstrated a genetic association between either inflammatory bowel disease (as a collective) and/or ulcerative colitis and keratoconus.

Conclusions:

The literature demonstrated a genetic association between ulcerative colitis and keratoconus. This association has not been consistently replicated in patient-focused studies.

Keratoconus is a progressive ectasia characterized by bilateral thinning and cone-like steepening of the cornea, primarily affecting individuals in adolescence and early adulthood.¹ The end result of these corneal changes is irregular astigmatism and reduced vision. Patients with keratoconus have a worse quality of life than those with retinal conditions such as age-related macular degeneration, diabetic macular oedema, and retinal vein occlusion despite having relatively better visual acuity than in those conditions.^2–4 The prevalence of keratoconus varies by sex and geography at ∼1.38 per 1000 individuals.⁵

Keratoconus can be difficult to diagnose early, and because of its progressive nature, early treatment is beneficial to preserving vision.^6–8 Keratoconus has numerous reported associations such as eye rubbing, atopic disease, asthma, Down syndrome, and a genetic predisposition.⁹ Finding other associations with keratoconus may aid the development of screening protocols and guide early management decisions.

Inflammatory bowel disease (IBD), consisting of Crohn disease (CD) and ulcerative colitis (UC), is characterized by chronic inflammation of the entire digestive tract or limited to the colon, respectively. IBD has a significant prevalence worldwide, especially in Western nations.¹⁰ The age of presentation for IBD varies by age and sex, with men having a greater risk during childhood.¹¹ There exists only a limited amount of literature and no systematic review addressing an association between these diseases and keratoconus.

This systematic review and meta-analysis aimed to determine the association between keratoconus and IBD. The data from the review may be of benefit to clinicians to assist in the early diagnosis of keratoconus and researchers to develop screening tools.

METHODS

A systematic search of the literature pertaining to any study connecting IBD and either UC or CD with keratoconus was performed. The study was registered with PROSPERO (registration ID: CRD42024611694). The review method followed Preferred Reporting Items for Systematic Reviews and Meta-Analyses Statement (PRISMA) guidelines.¹²

Eligibility Criteria for Considering Studies for Review

Inclusion Criteria

Prospective or retrospective studies including patients with an IBD including UC and CD and investigating an association with keratoconus or being a keratoconus suspect were included. Studies with any patient number were included.

Exclusion Criteria

Study cohorts with other ocular morbidities were excluded. No ex vivo or animal studies were included. Studies in nonhumans and in languages other than English were excluded.

Search Strategy

The search terms were developed to encapsulate specific terms and broader terms for the studied conditions (such as corneal ectasia). The main search terms were derived from the studied conditions being IBD, UC, CD, and keratoconus. We performed a systematic search on 5 different databases: PubMed, Ovid Medline, Scopus, Web of Science, and Embase on October 16, 2024 (Fig. 1). Unpublished studies were searched for via the DANS Data Station on life sciences, which houses the now closed OPENGREY database.¹³ No unpublished studies were found to warrant inclusion. Reference lists of studies initially selected for inclusion were manually reviewed. All searches were restricted to English only. There was no time limit placed on publication date.

FIGURE 1.

PRISMA flow diagram for literature search and inclusion.

Study Selection

We included all studies that investigated a potential association between any IBD and keratoconus. All search results were exported on Endnote (version 20). After the deletion of duplicates, 31 titles were screened for potential relevance. These 31 studies were then subject to abstract and full-text review to determine suitability for inclusion, with 9 studies included (Table 1). This process was repeated blind by a second reviewer (R.K.) who presented 7 of the same studies. The 2 differing studies by Hu et al and Zhang et al were included in the study upon consultation with the third reviewer (H.K.).^14,15

TABLE 1.

Characteristics of Studies (n = 9) Included in Our Systematic Review on Keratoconus and Inflammatory Bowel Disease

Study (Author, yrs)	Country/Region	Sex of Cases (% Male)	Mean Age of Cases	Included Patients			Study Type
				Cases	Control	Total
Claessens et al,19 2021	Netherlands	62.8	29.7	2051 KC	12,306, No KC	14, 357	Case–control
Nemet et al,20 2010	Israel	56.8	37.3	426 KC	1704, No KC	2130	Case–control
Lee, Jung & Cho, 202022	South Korea	49.4	31.1	575 KC	2875, No KC	3450	Case–control
Trechot, 201521	France	39.8	38.7	201 IBD	100, No IBD*	301	Cross-sectional
Chen & Chen,26 2024	Europe	N/A	N/A	2116 KC	24,626, No KC	26,742	Mendelian randomization
Jin et al,25 2024	Europe	N/A	N/A	N/A	N/A	N/A	Mendelian randomization
Cheng & Yang, 202424	Europe	N/A	N/A	311 KC	209,287, No KC	209,598	Mendelian randomization
Zhang et al,15 2022	People's Republic of China	93.0	23.3	15 KC	15, No KC	30	Case–control
Hu et al,14 2022	N/A	N/A	N/A	N/A	N/A	N/A	Genetic mining

^*Patients were recruited from ophthalmology department, presenting for various indications. Patients were excluded if they had any family history of keratoconus.

KC, keratoconus; N/A, not available/applicable.

Data Collection, Synthesis, and Analysis

The key data of interest were extracted into an Excel spreadsheet designed for the study. Study ID (author, year of publication) and patient demographics were included for all studies. In patient studies, the primary extracted data were the odds ratio of developing keratoconus if afflicted with IBD. In studies analyzing genetic association, the primary reported and, therefore, extracted metrics were the shared genes between 2 diseases or shared single nucleotide polymorphisms (SNP) and the associated odds ratio. Depending on the study, the uniformity of reporting varied—with some studies reporting on IBD as a whole, whereas others broke down data between UC and CD.

Risk of Bias

The risk of bias approach used to evaluate each study was determined based on the National Health and Medical Research Council (NHMRC) guidelines, using suggested good practice tools where relevant.¹⁶ The risk of bias assessment for all patient-associated studies was conducted using the Risk of Bias in Nonrandomized Studies of Exposure (ROBINS-E) tool, designed for nonrandomized studies of exposures (see Supplementary Table 1, Supplemental Digital Content 1, http://links.lww.com/ICO/C10).¹⁷ NHMRC guidelines did not provide a best practice tool for the risk of bias assessment of genetic association studies. The tool used for the genetic association studies was the Quality of genetic association studies tool (Q-Genie) (see Supplementary Table 2, Supplemental Digital Content 2, http://links.lww.com/ICO/C11).¹⁸

Statistical Analysis

Odds ratios of patient-focused studies were transformed and weighted to calculate summary effect values. Meta-analysis and forest plots were made with R project (v4.4.2) using R-packages Metafor and forestplot. Meta-analysis for patient studies was done using a random effects model, combining both forms of IBD into a crude estimate for IBD as a whole.

RESULTS

Study Selection

Our search yielded 299 studies, of which 9 were included in the systematic review and 3 were suitable for inclusion in the meta-analysis. The study selection process is shown in a PRISMA flowchart in Figure 1. The characteristics of the included studies are shown in Table 1.

Study Quality

From the 9 studies included, 4 were case–control studies, 3 were mendelian randomizations, one was a cross-sectional study, and one was a gene mining study. All studies were evaluated by 2 reviewers with a summary table of conclusions presented in the supplementary media.

Among the 4 case–control studies, the maximum granularity of data was reported as given that is, odds ratio for each type of IBD separately where possible, but this level was not consistently available.^15,19–21 Definitions for the diseases studied in each paper were not provided by the publishing authors. The data were categorized as patient-focused studies and genetic association studies because of the differing focuses of each type of study for analysis.

Patient-Focused Studies

There were 3 case–control studies looking at patient data across 4 different demographic groups (Table 2). From these 3, 2 studies (Nemet et al²⁰ & Claessens et al¹⁹) analyzed subgroups of IBD, whereas the other studies considered IBD broadly (Lee et al²²). No study reported CD, UC, and IBD overall as 3 distinct groups. In all 3 studies, patients were selected on the basis of keratoconus diagnosis; by contrast, Trechot et al²¹ selected patients with IBD. In these 3 case–control studies, no significant association was found between CD and keratoconus. One study found a significant association between UC and keratoconus [odds ratio (OR) = 12.1, 95% confidence interval (CI): 1.3‒116.4].²⁰ Overall, no study found a statistically significant association with IBD and keratoconus. Only one other study examined isolated UC with keratoconus, this study used a much larger dataset with a similar gender split but it did not corroborate a significant association.¹⁹ No study investigated the order of diagnosis of IBD and keratoconus, the symptom presentation of patients or characteristic features. The included studies focused on the specific prevalence of keratoconus and other comorbidities. There was no heterogeneity found in the study, I² = 0%. The pooled odds ratio for patient data for both forms of IBD was 1.39 (95% CI: 0.82–2.37) (Fig. 2).

TABLE 2.

Results of Patient-Focused Studies (n = 4) Included in Our Systematic Review of Keratoconus and Inflammatory Bowel Disease

Study (Author, yrs)	Disease	Odds Ratio (95% CI)	P
Claessens et al,19 2021	Crohn disease	0.94 (0.33–2.70)	0.942
Nemet et al,20 2010	Crohn disease	1.6 (0.3–8.3)	0.630
Claessens et al,19 2021	Ulcerative colitis	1.46 (0.54–3.86)	0.463
Nemet et al,20 2010	Ulcerative colitis	12.1 (1.3–116.4)	0.03
Lee, Jung & Cho, 202022	Inflammatory bowel disease	1.18 (0.44–3.19)	0.743
Trechot, 201521	Inflammatory bowel disease	N/A*	0.220

^*No test-statistic provided in paper, P-value reflects nonsignificance of association found.

N/A, not available.

FIGURE 2.

Meta-analysis of patient studies combining data for both forms of inflammatory bowel disease to present a unified summary statistic.

Genetic Association Studies

Out of the 5 genetic association studies, 3 of these were mendelian randomizations that obtained Genome Wide Association Studies data from the International Inflammatory Bowel Disease Genetics Consortium (IIBDGC) database involving 86,640 Europeans (Table 3).^23–26 The other 2 genetic association studies were unique in approach and demographic, Hu et al¹⁴ being a genetic mining study and Zhang et al¹⁵ a case–control study; these studies reported P-values suggesting a nonrandom genetic link between the 2 diseases. In the 4 studies that investigated IBD overall, there were shared genes found (P < 0.05).^14,15,24,25 The 2 studies that investigated UC found shared genes (P < 0.05).^25,26 The one genetic study that investigated CD and keratoconus found no significance (P = 0.062).²⁶ For the studies involving mendelian randomizations, the pooled odds ratio for IBD and keratoconus was 1.24 (95% CI: 1.21–1.26). All mendelian randomizations demonstrated no horizontal pleiotropy as indicated by the Egger test. There was no heterogeneity found as I² for mendelian randomizations was 0%.

TABLE 3.

Results of Genetic Association Studies (n = 5) Included in Our Systematic Review of Keratoconus and Inflammatory Bowel Disease

Study (Author, yrs)	Disease	Reported Result (95% CI)	P
Chen & Chen, 202427	Crohn disease	26 shared SNP	0.062
		OR: 1.096 (0.995–1.206)
Chen & Chen, 202426	Ulcerative colitis	21 shared SNP	0.002
		OR: 1.147 (1.054–1.248)
Jin et al,25 2024	Ulcerative colitis	35 shared SNP	0.038
		OR: 1.25 (1.01–1.54)
Jin et al,25 2024	Inflammatory bowel disease	111 shared SNP	0.017
		OR: 1.21 (1.03–1.41)
Cheng & Yang, 202424	Inflammatory bowel disease	111 shared SNP	0.017
		OR: 1.206 (1.034–1.407)
Zhang et al,15 2022	Inflammatory bowel disease	4 upregulated miRNA	0.026
		Enrichment score: 1.59 (N/A)
Hu et al,14 2022	Inflammatory bowel disease	9/65 genes shared (N/A)	<0.001

miRNA, micro ribonucleic acid; N/A, not available.

DISCUSSION

Our study found no consistent evidence of an association between keratoconus and IBD in patient studies. However, it also found evidence of a genetic association between IBD and keratoconus. This association appears most prominent and consistent regarding ulcerative colitis.

In patient-association studies, Nemet's original study showed a 12.1 odds ratio, which was not further substantiated in a further observational study, with Claessens in 2021 showing a lack of statistically significant association in a Dutch population.^19,20 Nemet's original study may be affected by the statistically significant (P < 0.0001) difference in the number of Ashkenazi Jews in the case group if this ethnic group was a confounder.²⁰ However, Trechot's paper highlighted the high prevalence of suspect keratoconus in patients with IBD, which may be indicative of a deeper link between the 2 diseases.²¹ Importantly, we did not find a consensus in observational studies regarding an association between UC and keratoconus. This may be a result of the impact of patient demographics on the association between the 2 diseases.

In mendelian randomization studies, statistically significant odds ratios were present for the association between keratoconus and UC in 2 papers.^25,26 There was also a statistically significant association between IBD and keratoconus in 4 genetic studies.^14,24,25 Upregulated miRNA are further suggestive of an association between IBD and keratoconus.¹⁵ However, the only study that investigated the genetic association between CD and keratoconus found that there was no association between the 2 diseases.²⁶

Given that all mendelian randomizations used the same dataset for IBD, it is possible that the positive association of UC masked the negative association of CD and reduced the odds ratios. An important aspect is the genetic makeup of this dataset, reflecting only Europeans. The possibility of this is supported by the lack of any observational studies linking CD with keratoconus, with no study in our systematic review finding a statistically significant association. The gap between the strong genetic association data and the lack of observational data may be because of confounding variables that masked the association. However, a key advantage of mendelian randomization is being able to bypass some of the effects of confounding and examine causal relationships more clearly.

The affected mechanism driving the causal relationship is not entirely clear. Although the cornea is relatively avascular, it is not spared from systemic immune effects. Research has examined the presence of a unique microbiota within keratoconic eyes²⁷ and innate and adaptive immune responses in keratoconus pathogenesis.²⁸ The tear film in keratoconic patients has been found to have elevated levels of various immune markers, which may play a role in disease progression.^29,30 The genetic studies analyzed in this review did not provide an in-depth exploration of the specific impact of the shared SNPs—this would require further investigation. Overall, the linkage between immune-mediated disorders and keratoconus is not new, with numerous known and reported associations.^19,31

Classically, immune-mediated disorders of the cornea such as marginal keratitis or peripheral ulcerative keratitis affect predominantly the periphery and limbal area—distinctly different to the paracentral nature of keratoconus. Although the included papers do not provide justification for this, it may be because of a difference in the immune response between keratoconus and peripheral corneal immune conditions. Although conditions such as peripheral ulcerative keratitis are vascularly mediated,³² keratoconus may instead be driven by corneal stromal cell–produced matrix metalloproteinase (MMP) that is upregulated centrally and downregulated peripherally.³³ In keratoconus, tear levels of MMP-9, among other immune markers, have been found.^34,35 In UC, serum MMP-9 levels are significantly more elevated, compared with nondiseased and patients with CD.^36,37 Whether the reported genetic link between UC and keratoconus is because of this shared mechanism of raised MMP-9 is uncertain. Further research is needed to establish if a pathway in the disease pathogenesis exists that involves MMP-9 and connects keratoconus and UC, with the exclusion of CD. This hypothesis is shared by the authors of the studied mendelian randomizations and Hu et al.^14,24–26 In addition to MMP-9, Hu et al¹⁴ highlighted that interleukin-17 (IL-17) signaling pathways have been hypothesized as another shared upregulated pathway.

Reconciling the reasons for a mismatch in genetic linkage and obvious clinical association is difficult and speculative. As the SNPs highlighted by each study were not individually assessed and not discussed of their potential role and function, it is difficult to ascertain the exact requirements for expression of both diseases. It is possible that the identified genetic links are made apparent clinically when combined with specific ethnicities or other unknown variables—reflecting the lack of uniformity in observational analysis based on cohort. For reference, many of the established genes associated with keratoconus (of unknown significance) present only in some ethnicities.³⁸ Further epigenetics may also have a role with environmental factors linked to keratoconus pathogenesis and progression.

The utility of finding and establishing genetic associations between diseases is significant. First, it fundamentally improves our knowledge of the disease and its process—which may be an initial point for future research into treatment and management. Second and most directly, if a genetic association is proven in patient studies with significance, it can be used to develop risk stratification for screening. A high enough risk ratio may prompt newly diagnosed patients with IBD to seek ophthalmologic screening for keratoconus if they fit other criteria such as age, family history, and visual function.

Although our systematic review and meta-analysis were comprehensive based on the published literature, numerous limitations must be noted. First, there was a significant paucity of published studies on the topic and the publications found had differing goals, lack of disease definitions, methods, and outcomes assessed. This is reflected in the pragmatic and nonideal nature of the meta-analysis of patient studies that required grouping of both forms of IBD under one risk ratio because of a lack of specific studies separating their results. In addition, we elected to keep studies with poor reporting such as Trechot et al²¹ that lacked a test statistic (the author was contacted with nil response) in favor of a broader consideration of all available evidence. This made corroboration of results and cross-comparisons difficult. This cross-comparison was only possible because of the segmentation of studies into different groups based on method. Naturally, segmenting an already small pool of available studies has strong implications for the strength of the evidence generated and this is a significant limitation. Meta-analysis was not conducted on genetic studies because of all mendelian randomization studies making use of the same IBD genetics database while the remaining 2 genetic studies were unique in method. In addition, Q-Genie, the risk of bias tool used for our genetic association studies, was arguably limited in usefulness because of the variety in genetic association studies that made some assessments irrelevant, and the lack of a “not-applicable” selection and the points-based assessment of risk. There was also a level of technical specificity (eg algorithms used) required for assessment that was generally not available in the studies. Finally, our exclusion of non-English studies may bias results because of social factors such as consanguineous marriage, which are less present in the Anglosphere.

Avenues for further research exist based on potentially promising results from the observational and genetic association studies. Specifically, there is a need to understand further the pathophysiology and patient impacts of an association between keratoconus, suspect keratoconus, and UC. Any such study would need to account for numerous potential confounding variables with a large sample size as the effect size may be small. There is further research needed to understand the clinical presentation of patients presenting with both diseases (keratoconus and UC) in chronicity, predominant symptoms, and the impact of systemic treatment on keratoconus progression.

In summary, there is evidence of a genetic relationship between UC and keratoconus. Clinical studies are now needed to confirm these findings in patient cohorts and understand any underlying pathophysiology.

Supplementary Material

cornea-45-1467-s001.docx ^{(23.9KB, docx)}

cornea-45-1467-s002.docx ^{(19.6KB, docx)}