The genetics of cholesteatoma. A systematic review using narrative synthesis

A cholesteatoma is a mass of keratinising epithelium in the middle ear. It is a rare disorder that is associated with significant morbidity, and its causative risk factors are poorly understood; on a global scale, up to a million people are affected by this each year. We have conducted a systematic literature review to identify reports about the heritability of cholesteatoma or any constitutional genetic factors that may be associated with its aetiology.


| INTRODUCTION
A cholesteatoma is a self-perpetuating erosive lesion composed of stratified keratinising squamous epithelium in the middle ear. 1 Cholesteatoma has both an acquired and a congenital form. It activates osteoclasts and so will erode through bone, which may include the endocranium, with an attendant risk of life-threatening intracranial infection.
The acquired form of cholesteatoma originates as an inward growth from the lateral epithelium of the tympanic membrane. A typical sequence of events in the onset of the disease includes a history of chronic otitis media (COM) in childhood, subsequent development of retraction of the tympanic membrane and then a cholesteatoma developing within and perforating through this retraction. This seems to particularly occur if the retraction is located in the superior tympanic membrane (pars flaccida). [2][3][4] In children with a history of chronic otitis media with effusion (COME), 15-35% will develop a retraction of the pars flaccida (at up to 25 years of follow-up), but only 0.1-2% will develop a cholesteatoma (at up to 8 years of follow-up). [4][5][6][7] Both presence and duration of COME are predictive of tympanic membrane retraction, 3,4 but tympanic retraction has been documented to occur in the absence of preceding COME. 4 However, histological studies suggest that in such cases there is nevertheless chronic middle ear inflammation, it is just not clinically apparent. 8 Thus, cholesteatoma is often preceded by COM, but only a small proportion of those with COM will develop cholesteatoma. What determines the transition from COM to cholesteatoma is not known, but could be due to environmental factors, heritable factors or random effects. But those who develop cholesteatoma have been reported to have between a 7% and a 20% chance of developing disease in the contralateral ear, 9,10 highlighting the importance of shared genes and shared environments.
Cholesteatoma can also be found behind an intact tympanic membrane. 11 This form is thought to be congenital, and may result from persistence of the foetal epidermoid formation, a small collection of squamous epithelial cells in the middle ear that normally undergoes apoptosis before or shortly after birth. Congenital cholesteatoma can grow laterally and erode through the tympanic membrane, and at that point, it can be difficult to differentiate congenital from acquired disease.
Cholesteatoma is a rare disorder (1:10 000 per year), 1 and therefore, epidemiological studies are difficult to conduct, and causative risk factors are still poorly understood. The citations about cholesteatoma in the definitive catalogue of genes and genetic diseases, Online Mendelian Inheritance in Man, 12 document minimal evidence for the Mendelian inheritance of this disorder. 13 However, reports of familial clustering of disease and of association with genetic syndromes (reviewed here) suggest underlying, but as yet unidentified genetic risk factors. Identifying these could enhance our understanding of disease biology, and open up pathways for diagnostic, screening and therapeutic interventions.
One way to identify candidate genetic factors is through analysis of products of gene expression in pathological specimens.
There are two published large-scale analyses comparing RNA transcript expression in cholesteatoma to that in skin of the external auditory canal. These have shown several hundred genes are differentially regulated in cholesteatoma samples, including genes with products involved in growth, differentiation, signal transduction, cell communication, protein metabolism and cytoskeleton formation. 14,15 However, the results from these studies are inconsistent, and are measuring gene expression once cholesteatoma has formed, and so have failed to significantly further our understanding of constitutional risk.
Here, we describe findings from a systematic review of the genetics of congenital and acquired cholesteatoma. Our aims from this review were to describe how susceptibility is transmitted within families showing disease clustering, to better understand the genetic architecture of disease, and to document any genotypes shown to co-segregate with the cholesteatoma phenotype. We also aimed to classify genetic syndromes associated with increased risk of cholesteatoma, which may implicate candidate genetic loci for further investigation.

| Objectives
To synthesise published evidence that addresses the following questions: 1. Can the development of a cholesteatoma be described as a heri- to July 2015 using the terms "Cholesteatoma" AND "famil* (OR Gene* OR hered* OR inherit* OR syndrom* OR kindred OR pedigree OR oncogene* OR tumour suppressor OR tumor suppressor OR epigenetic* OR mutat* OR somatic OR homeobox)." We supplemented the search with relevant references identified in the citation lists at the article review stage.

| Inclusion and exclusion criteria
Studies were identified from the titles and abstracts by the primary reviewer (BAJ) and secondary reviewer (GW) using the following inclusion criteria: 1. Primary studies of kindreds that provide information about familial clustering.

2.
Primary epidemiological studies that provide evidence of heritability including ethnic differences.

3.
Relevant systematic reviews that provide information about genetics or heredity for cholesteatoma.

4.
Case reports that refer to familial clustering of the cholesteatoma phenotype (>1 family member affected).

5.
Case reports or epidemiological studies that provide evidence of association between cholesteatoma and syndromes Studies were excluded if they were general narrative reviews or opinion pieces, about non-human or experimental disease models, or described pathologies other than cholesteatoma.

| Study selection & data extraction
Full reports of potentially relevant articles were retrieved, and data were extracted by the primary reviewer (BAJ). The study design, patient characteristics and nature of the outcomes were collated and coded red for exclusion, green for inclusion and amber to indicate uncertainty (RAG review). When there were uncertainties about inclusion or data interpretation, the articles were discussed by the reviewers (GW, MB and BAJ) to reach consensus. All studies that met the inclusion criteria were included regardless of quality, which was subsequently appraised (see Risk of bias and quality assessment below).

| Data synthesis
A narrative synthesis was conducted to explore the review questions about heritability and genetic associations reported for the cholesteatoma phenotype. We tabulated the date of the study, first author, study design, number of subjects, subtype of cholesteatoma, genetic investigations (including family history), associated congenital syndromes, gene nomenclature and direct quotations from discussion or conclusions.

| Risk of bias and quality assessment
We appraised quality of epidemiological studies by reference to the Strengthening Reporting of Observational Studies (STROBE) guidelines 17 and the Strengthening Reporting of Genetic association Studies (STREGA) guidelines. 18 We mapped the evidence for each study to the five levels of evidence described by the Oxford Centre for Evidence Based Medicine. 19 3 | RESULTS

| Study selection & data extraction
Our search identified 449 unique studies, of which 36 met the initial inclusion criteria. Most studies were excluded at the abstract or primary manuscript review stage, but six manuscripts were excluded at the data extraction stage because there were no relevant primary data identified about cholesteatoma or genetic phenomena, 20-23 or because the study described external auditory canal cholesteatoma. 24 The studies identified in the initial search were supplemented by five additional reports identified by hand-searching citation lists. [25][26][27][28][29] Thirty-five studies were finally included in this narrative synthesis (see Figure 1 for a flow chart which summarises these steps).

| Congenital syndromes and cholesteatoma
Twenty-two case reports and epidemiological studies describe the occurrence of cholesteatoma in patients affected by congenital and malformation syndromes, 13,26,30,35-52 several of which have a known underlying genetic aetiology. These are summarised in Table 2.
Some of these reports are of cholesteatoma occurrence in a single case of a particular syndrome, for example Beckwith-Wiedemann syndrome, granulomatosis with polyangiitis, Nager syndrome, primary ciliary dyskinesia, Tolosa-Hunt syndrome, Treacher Collins syndrome and Wolf-Hirschhorn syndrome. Single occurrences of a disease, whether associated with a syndrome or not, are susceptible to publication bias and so do not add to understanding of disease risk in isolation. | 57 with the cholesteatoma phenotype. One is a case report of a 6-yearold boy with a congenital cholesteatoma who was shown to have a deletion in the APC tumour suppressor gene. 53 The other is a candidate gene association study of polymorphisms of the GJB2 and GJB6

| Candidate genes and gene variants
loci that encode connexins 54 in a cohort of 98 children undergoing surgery for cholesteatoma. These studies are also described in Table 3.

| Risk of bias and quality assessment
We identified only a small body of literature that was relevant to our questions about a heritable component for cholesteatoma aetiology. Many of the studies provide some indirect evidence only, given that the authors' objectives were to describe cholesteatoma management or associated environmental factors.
Most of the studies identified in the literature search, and described here, are case reports and so represent the lowest level of evidence. Case reports were automatically categorised as level 5 (see Tables 1, 2 and 3). The remaining observational studies include case series, cross-sectional surveys, case-control studies and cohort studies; each of these manuscripts was reviewed by BAJ and GW to define the level of evidence presented; STROBE and STREGA guidelines were referred to in classifying the quality of the methodology used in the case-control and cohort studies. The level of evidence ranged from 4 (for low-quality case-control studies, surveys and case series) and 2b for a high-quality cohort study 35,49 (see Tables 1, 2 and 3).

| DISCUSSION
This is the first systematic review to explore the constitutional

| Heritability
We have summarised the published evidence about the heritability of acquired and congenital cholesteatomas. We only identified a few case reports and case series that show two or more affected firstdegree relatives; therefore, there is insufficient evidence to describe cholesteatoma as a heritable trait.
However, there are some compelling individual observations to consider, including affected monozygotic 27 and dizygotic twins, 28,33 families with two or more affected generations, 30,32,33 and high rates of bilateral disease in affected families. 28   APC. The APC protein is expressed in many tissue types, influencing cell migration, adhesion and morphogenesis. Loss of APC expression in the colonic epithelium leads to an imbalance of cell growth over cell death, 56 but whether this is relevant to cholesteatoma biology is not known. 53 The second study was a candidate gene association study of 98 children with cholesteatoma for variants in the connexin gap-junction encoding genes, GJB2 and GJB6, 54 some mutations of these loci are known to lead to recessive congenital deafness. Although the authors suggest a high frequency for some GJB2 gene variants associated with cholesteatoma, no conclusions can be safely drawn from this study, because it lacked a control population and had a small sample size, placing it at risk of false discovery.

| Limitations
We excluded non-English manuscripts and studies published before 1980 from our initial search (the earlier and/or non-English articles were subsequently included in the narrative synthesis because they were identified by hand-searching citation lists); it is therefore possible that we have missed relevant publications.
The over-representation of case reports, case series and historical epidemiological studies is unsurprising given that cholesteatoma is a rare disease, but such studies provide low-level evidence in the research hierarchy because they are usually retrospective with incomplete data collection or follow-up, and are subject to author bias, ascertainment bias and publication bias. In addition, such findings may not be generalisable, and should be interpreted with caution, particularly with respect to theories about the underlying aetiology of cholesteatoma.

| CONCLUSION
Cholesteatoma is a complex and heterogeneous clinical phenotype.
In a handful of case reports or case series, congenital and acquired cholesteatomas have been shown to segregate within families in the pattern typical of a monogenic or oligogenic disorder with incomplete penetrance. The liability threshold for the observed cholesteatoma phenotype could therefore depend on a combination of environmental and genetic factors of variable penetrance. Evidence from syndromic cases suggests that genes controlling ear morphology may be risk factors for congenital or acquired cholesteatoma formation.
We should accommodate the hypothesis that a range of aetiological pathways exist for cholesteatoma and that these may result in disease subtypes that differ in both severity and tractability.