Systematic review of the diagnosis and management of necrotising otitis externa: Highlighting the need for high-quality research

Objectives: To present a systematic review and critical analysis of clinical studies for necrotising otitis externa (NOE), with the aim of informing best practice for diagnosis and management. Design: Medline, Embase, Cochrane Library and Web of Science were searched from database inception until 30 April 2021 for all clinical articles on NOE. The review was registered on PROSPERO (ID: CRD42020128957) and conducted in accordance with PRISMA guidelines. Results: Seventy articles, including 2274 patients were included in the final synthesis. Seventy-three percent were retrospective case series; the remainder were of low


| INTRODUCTION
Necrotising otitis externa (NOE) is a serious, progressive and potentially fatal infection of the external auditory canal (EAC), classically associated with elderly diabetic patients and Pseudomonas aeruginosa infection. NOE begins as simple otitis externa (OE), which progresses via the fissures of Santorini of the osteocartilaginous junction of the EAC to cause perichondritis, osteomyelitis and in some cases, skull base osteomyelitis (SBO). 1 Case reports of NOE date from 1838, 2 with recognition of NOE as a unique disease entity in 1968. 3 Initially named 'malignant otitis externa' due to a characteristic triad of aggressive spread, treatment resistance and high mortality, 3 nomenclature has shifted to NOE to reflect its infectious rather than malignant aetiology. 4 NOE is a rare disease globally. Incidence rates, while underreported, range from 0.221 to 1.19 cases per 100 000, [5][6][7] but has risen more recently 5 presumed secondary to increasing age, diabetes prevalence and clinician awareness. Antibiotic resistance, particularly quinolone-resistant Pseudomonas species is also of concern, [8][9][10] prompting a renewed interest in rationalising antimicrobial therapy.
NOE is a complex infection to diagnose and manage, compounded by the lack of agreed diagnostic criteria. Two systematic reviews have been published to date. 8,11 Mahdyoun et al. included literature published up to 2011 and highlighted the absence of strong evidence on diagnosis, treatment and follow-up. 11 A more recent review summarised changes over time, but excluded articles pre-2009, such that only 10 articles and 284 patients were included. 8 In this systematic review, we provide a comprehensive, rigorous and critical analysis of all NOE clinical articles published to date. We place particular focus on the critical evaluation of the available data to guide clinical decision-making and identify key areas for future research.

| Literature search and article selection
The systematic review was registered on PROSPERO on 23 January 2020 (PROSPERO ID: CRD42020128957). 12 A comprehensive search of Medline, Embase, Cochrane Library and Web of Science was performed on 8 November 2019, then repeated on 30 April 2021, using the search strategy in Appendix S1.
Titles and abstracts of identified articles were independently screened by two reviewers using the inclusion and exclusion criteria in Table 1A, with a third reviewer making the final decision if there was disagreement. For SBO, articles were only included if cases were explicitly secondary to EAC pathology. The selected full-text articles were reviewed by a single reviewer to confirm inclusion.

Key points
• No single symptom, sign, or investigation in isolation can categorically diagnose NOE.
• Ten percent of NOE cases have no known immunosuppressive risk factor.
• P. aeruginosa is the commonest isolated pathogen but NOE can be caused by a wide range of organisms.
• There is no clear evidence to support dual antipseudomonal therapy or inform duration of treatment.
Final analysis was restricted to articles of higher quality, with an explicit case definition and ≥6 cases (Table 1B). Articles were further subdivided into 'general', describing unselected NOE cases, or 'specific', describing particular subpopulations. 'Specific' articles were only considered in subsections related to clinical investigations or management.

| Data extraction and analysis
Key variables, shown in Appendix S2, were extracted by a single reviewer into REDCap ® , a secure electronic data capture tool hosted by the University of Oxford. 13,14 Articles were classified by study type using definitions in Appendix S3 and by level of evidence using the Oxford Level of Evidence. 15 Data analysis was conducted in Microsoft Excel (version 16.46), Prism (GraphPad Software, Version 9.0.2), Stata (version 17.0) and RStudio (version 1.3.959). The Joanna Briggs Institute (JBI) checklists were used to assess methodological quality and risk of bias; JBI was chosen due to the availability of tools for all study types included in our review (Appendix S4). 16 Results were synthesised and reported according to PRISMA guidelines (Appendix S5). 17 No ethical approval was required for this study.

| Study characteristics
A summary of the search is presented in the PRISMA diagram ( Figure 1). 17 Of 1429 articles identified, 422 (30%) articles of 16 528 patients were included in the initial synthesis. The majority were single case reports (49%, 207/422) or formal case series (28%, 120/422; Figure S1A). Since the first article in 1968, 3 a bimodal pattern in the number of articles published per year was seen, with an increase in the last 5 years; however, there remains a paucity of high-quality articles that are not case reports or case series ( Figure 2). Seventeen percent (70/422) of articles of 2274 patients were included in the final synthesis (Table 2), with a median of 23 patients included per article (IQR 14.8-36.0). Seventy-three percent (51/70) were formal, consecutive retrospective case series ( Figure S1B) and increased in number over time ( Figure S2), while quasi-experimental studies tended to be earlier. Most case series (75%, 42/56) were methodologically high quality with low risk of bias; in contrast, only 29% (4/14) of other study types had low risk of bias (Appendix S3).
The majority of articles were from Israel (31%, 22/70), the USA (16%, 11/70) and the UK (7%, 5/70; Figure 3), while none were from South America, Sub-Saharan Africa, South-east Asia, or Oceania.   (Table S2) and varied across articles, but commonly referenced bony erosion.   and no patients were diabetic. 70 Another 'specific' case series of non-diabetic patients identified age, ischaemic heart disease and hypertension as risk factors. 69 In one 'specific' case series of temporal bone osteomyelitis, 27% (15/55) of patients had prior ear surgery and presented more atypically with younger ages, lower diabetes prevalence and lack of otalgia, EAC granulation or isolation of P. aeruginosa. 74

| Clinical presentation
Not all 'general' articles reported signs and symptoms and findings were biased according to the case definition (Table 3). Acknowledging this, the commonest presenting symptom was otalgia (96%, 1249/1307) followed by otorrhoea (78%, 972/1255). Fever was infrequently reported F I G U R E 1 Modified PRISMA 2020 flow diagram of search output. Articles were classified as 'general' if they described unselected NOE cases. Articles were classified as 'specific' if they selected a particular sub-population, such as those who received a particular treatment (e.g., hyperbaric oxygen) or those with skull base osteomyelitis.

| Imaging
Fourteen percent (7/51) of 'general' articles did not describe radiological assessment; for the remainder, imaging modalities and findings were poorly reported. CT was the commonest modality, used in 95% scanning (four articles each), followed by CT and MRI (three articles each); however, the time interval between scans or indication for follow-up imaging was poorly described.

| Microbiology
The commonest method of microbiological sampling across 40 'gen- Note: Articles were classified as 'general' if they described unselected NOE cases. Articles were classified as 'specific' if they selected a particular subpopulation, such as those who received a particular treatment (e.g., hyperbaric oxygen) or had skull base osteomyelitis. Articles were classified for their level of evidence according to the University of Oxford's Centre for Evidence-based medicine (CEBM)'s Levels of Evidence. Abbreviations: AIDS, acquired immune deficiency syndrome; NOE, necrotising otitis externa; RCT, randomised controlled trial; SBO, skull base osteomyelitis.
Topical antimicrobials were used in 51% (26/51) of 'general' articles, including antibiotics and antiseptics such as boric acid and ascetic acid. No articles evaluated the benefit of topical antimicrobials over IV/oral therapy alone; optimal duration; ideal agents; or the effect of topical antimicrobials in driving antimicrobial resistance.
Duration of antimicrobial therapy was poorly documented, and no articles assessed the association between duration and outcome.

| Outcomes
Eighty percent (41/51) of 'general' articles reported mortality data but varied in the time interval for mortality assessment (unknown to 5 years post-diagnosis). All-cause mortality was recorded for 24% (12/51)  Staphylococcus epidermidis were also grown in an additional, unspecified number of cases in these 967 patients. Therefore, the number of times these pathogens were isolated is underestimated in this table (see Table S3 for further detail on species/organisms isolated). Forty-one percent of articles used a mono-APAM regimen, while topical therapy was used in 51% and anti-fungal agents in 31%. There was, however, no clear evidence on optimal antimicrobial choices, the timing of antimicrobial addition or switching, or route. Additionally, the median duration of therapy was 7.2 weeks but ranged 10-fold, reflecting the lack of consensus. Given growing concerns regarding antimicrobial stewardship and antimicrobial resistance, 9,10 along with the increased risk of adverse reactions to antimicrobials in older adults, [88][89][90] identifying an optimal treatment regimen is crucial. Our analysis confirms that NOE is far from benign, with a notable disease-specific 1-year mortality of 2% and high treatment failure and relapse rates (22% and 7%, respectively). Due to the heterogeneity of relapse and treatment failure definitions, as well as the small sample size, it was not possible to discern meaningful temporal trends, and suggests an urgent need for further, high-quality work to improve patient outcomes.

| STRENGTHS AND LIMITATIONS
Our review is the most comprehensive to date, encompassing literature from 1968 to 2021. Moreover, the delineation of 'general' from 'specific' articles reduces selection bias. There are, however, some limitations. Due to the lack of robust, non-retrospective data, it was not possible to systematically determine patient-level detail on pathogens isolated, risk factors, clinical manifestations, or outcomes to allow meta-analysis or assess associations between risks and outcomes.
Some patients may also have been 'double counted' in analyses if data from the same patient population were reported in multiple articles.