Bacterial otitis media in sub-Saharan Africa: a systematic review and meta-analysis

Background Otitis media is inflammation of the middle ear, comprising a spectrum of diseases. It is the commonest episode of infection in children, which often occurs after an acute upper respiratory tract infection. Otitis media is ranked as the second most important cause of hearing loss and the fifth global burden of disease with a higher incidence in developing worlds like Sub-Saharan Africa and South Asia. Therefore, this systematic review is aimed to quantitatively estimate the current status of bacterial otitis media, bacterial etiology and their susceptibility profile in sub-Saharan Africa. Methods A literature search was conducted from major databases and indexing services including EMBASE (Ovid interface), PubMed/MEDLINE, Google Scholar, ScienceDirect, Cochrane Library, WHO African Index-Medicus and others. All studies (published and unpublished) addressing the prevalence of otitis media and clinical isolates conducted in sub-Saharan Africa were included. Format prepared in Microsoft Excel was used to extract the data and data was exported to Stata version 15 software for the analyses. Der-Simonian-Laird random-effects model at a 95% confidence level was used for pooled estimation of outcomes. The degree of heterogeneity was presented with I2 statistics. Publication bias was presented with funnel plots of standard error supplemented by Begg’s and Egger’s tests. The study protocol is registered on PROSPERO with reference number ID: CRD42018102485 and the published methodology is available from http://www.crd.york.ac.uk/CRD42018102485. Results A total of 33 studies with 6034 patients were included in this study. All studies have collected ear swab/discharge samples for bacterial isolation. The pooled isolation rate of bacterial agents from the CSOM subgroup was 98%, patients with otitis media subgroup 87% and pediatric otitis media 86%. A univariate meta-regression analysis indicated the type of otitis media was a possible source of heterogeneity (p-value = 0.001). The commonest isolates were P. aeruginosa (23–25%), S. aureus (18–27%), Proteus species (11–19%) and Klebsiella species. High level of resistance was observed against Ampicillin, Amoxicillin-clavulanate, Cotrimoxazole, Amoxicillin, and Cefuroxime. Conclusion The analysis revealed that bacterial pathogens like P. aeruginosa and S. aureus are majorly responsible for otitis media in sub-Saharan Africa. The isolates have a high level of resistance to commonly used drugs for the management of otitis media.


Background
Otitis media is an inflammation of the tympanic membrane and middle ear with a spectrum including acute otitis media, otitis media with effusion and chronic suppurative otitis media [1,2]. Otitis media often occur secondary to acute upper respiratory tract infections, can also be caused by allergy and changes of the middle ear or Eustachian tube anatomically or functionally [3]. Worldwide around 1.23 billion people are affected by otitis media, thus it is ranked as the fifth global burden of disease and the second cause of hearing loss [2,4]. Children are the most affected groups and it is one of the commonest disease responsible for receiving antibiotics among children [4][5][6]. Otitis media is a major chronic disease in low and middle-income countries than in high-income countries. The highest incidence rate reported from Sub-Saharan Africa and South Asia [4][5][6][7][8].
Otitis media, especially with a chronic and recurrent form, is associated with complications such as hearing loss, decreased learning capability, and low educational achievement [9]. Around 20,000 people, majorly children under 5 years of age, die every year with the associated complications [7].
Majorly pathogenic bacteria such as non-typable Haemophilus influenzae, Streptococcus pneumoniae, Staphylococcus aureus, Streptococcus pyogenes, Pseudomonas aeruginosa, Proteus mirabilis, Escherichia coli, and Moraxella catarrhalis are the etiological agents of otitis media [10][11][12][13][14][15], although viruses and fungi also associated [14][15][16]. The occurrence of otitis media is directly related to the colonization rate of the nasopharynx by the bacteria [12]. Viral upper respiratory tract infections (URTI), disrupts the mucociliary system, impairs the host's primary mechanical defense for bacterial invasion and predispose children to acute otitis media (AOM) [12][13][14][15]. This study is aimed to quantitatively estimate the isolation rate of bacteria from patients with otitis media (OM), major isolates and resistance of isolates to commonly used antibacterial agents.

Study protocol
Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) flow diagram [17] was followed for identification of records, screening of titles and abstracts, and evaluation of full texts for inclusion. We have followed the PRISMA checklist [18] throughout the review and the published methodology can be accessed from http://www.crd.york.ac.uk/PROSPERO/display_record.php? ID = CRD42018102485. The study protocol is also registered on PROSPERO with reference number ID: CRD42018102485. , "middle ear inflammation", "middle ear effusion", "secretory otitis media", "purulent otitis media" and each sub-Saharan country entry term with Boolean operators (AND, OR), and truncation were used. All articles available online until February 25 -March 5, 2019, were considered. Google Scholar and World Cat were used to access gray literature from organizations and online university repositories. There was no language preference set in search strategy for all databases except for Google Scholar search, where the language preference was set to English and French. This deviation from the protocol (articles only English language) was made after observing some studies in the region that were published in the French language.

Identification of records and search strategy
Screening and eligibility of studies ENDNOTE version 8.0.1 (Thomson Reuters, Stamford, CT, USA) was used to manage the search results and to identify duplicate records. Records were also checked manually to avoid duplicates due to variation in reference styles across sources followed by a screening of titles and abstracts with predefined inclusion criteria. Two authors (TT and HM) independently evaluated the full texts of selected articles for eligibility and the discrepancies between the two authors' evaluations were solved by the other authors to come up with a consensus.

Inclusion and exclusion criteria
Patients with clinically confirmed otitis media were the study participants. Otitis media was defined as middle ear inflammation with discharge and without a clear indication of the tympanic membrane, and participants with middle ear inflammation with discharge for more than 3 months with a perforated tympanic membrane were considered as chronic suppurative otitis media. Records, published from 2009 to 2018, on the isolation of bacterial pathogens from ear swab or discharge, and susceptibility testing were included regardless of the patient's clinical characteristics.
Review and original articles conducted outside of sub-Saharan Africa were excluded during initial screening. Articles with data collection period before 2009, having unrelated, missing or insufficient outcomes, and irretrievable full texts (after requesting the corresponding authors) were excluded.

Data extraction
Two authors (TT and HM) independently extracted important data related to study characteristics (country, first author, year of publication, study design, patient characteristics, number of culture-positive results (bacterial), nature of bacterial isolates, and number of isolates) and outcome of interest (number of culturepositive results, number of isolates with prevalence data for each bacterium, and susceptibility test results) using data abstraction format prepared in Microsoft Excel (Data Extraction and Summary sheet).

Critical appraisal of studies
Critical appraisal to assess the internal (systematic error) and external (generalizability) validity of studies and to reduce the risk of biases was conducted according to the Joanna Briggs Institute's critical appraisal tool adapted for prevalence studies [19] and graded out of 9 points. The mean score of the two authors was taken for final decision and studies with a score greater than or equal to five were included.

Outcome measurements
The primary outcome measure is the prevalence of bacterial otitis media, major isolates and the antimicrobial resistance in sub-Saharan Africa. The measurement regarding antimicrobial resistance was conducted for selected drugs (Ampicillin, Sulfamethoxazole-Trimethoprim (cotrimoxazole), Chloramphenicol, Amoxicillin-clavulanic acid, Gentamycin, Ceftriaxone, Cefuroxime, Erythromycin, Tetracycline, and Ciprofloxacin) obtained from patients with presumed or confirmed otitis media.

Data processing and statistical analysis
Stata version 15 was used for analyses of outcome measures after importing data extracted in Microsoft excel format. We have applied the DerSimonian and Laird's random-effects model for the analyses at a 95% confidence level considering the variation in true effect sizes across the population (clinical heterogeneity). The heterogeneity of studies was determined using I 2 statistics. A univariate meta-regression model was performed, with a cutoff value p-value< 0.05, on study characteristics to assess the possible source of heterogeneity. Begg's and Egger's tests were used to evaluate the presence of publication bias, and presented with funnel plots of the standard error of proportions [20,21].

Search results
We conducted a systematic review and Meta-analysis per under the PRISMA statement [17]. In our literature search, a total of 1314 records were identified from several sources. From these, 621 duplicate articles were removed with the help of ENDNOTE and manual tracing. The remaining 693 records were screened using their titles and abstracts, and 607 of them were excluded. Eligibility evaluation of full texts was done for 86 records. From these, 32 articles were also excluded as the outcome of interest was missing, insufficient and/or ambiguous, and 2 articles were not accessible in full text [22,23]. Fifty-two studies were evaluated by a full-text review and 16 studies had data collection period before 2009 and three studies [24][25][26] have no record of the data collection period. Finally, 33 articles passed the eligibility criteria and quality assessment and hence included in the study (Fig. 1). Thirty-two studies were utilized for the analysis of prevalence, one study has reported only the isolates without the prevalence of otitis media, and only 22 studies were utilized for the assessment of susceptibility testing since the other 11 studies have an incomplete report regarding study outcome and in some studies, the test was done with a variable number of isolates for different drugs.

Study characteristics
As shown in Table 1, a total of 33 studies with 6034 participants were included for systematic review and metaanalysis. Seven of the included studies were retrospective analyses of secondary data (record review) from regional laboratories and hospitals [28,29,35,37,42,43,48]. Fourteen studies were conducted on chronic suppurative otitis media [27, 31, 33, 36, 40-42, 45, 48, 50, 53, 54, 56, 58] and the rest included studies do not indicate the type of otitis media. Seven studies [30,32,35,39,44,52,57] were conducted in children whereas the rest of the studies included patients of all age groups. The average quality scores of studies ranged from 5 to 8 as per the Joanna Briggs Institute scoring scale for prevalence studies (Table 1). All studies have collected ear swab/discharge samples and applied standard microbiological culturing techniques for isolation of bacterial agents and followed the international standard for performing susceptibility testing and interpretation of results (CLSI standards). Concerning the country of study, 14 studies were from Nigeria, 8 studies from Ethiopia, 2 studies from Zambia, and a single study from Angola, South Africa, Sudan, Burkina Faso, Tanzania, Uganda, Ghana, and Malawi.

Study outcome measures Prevalence of bacterial otitis media
The overall estimate of bacterial isolation was 92% (95% CI: 90.0, 94.0) with a degree of heterogeneity (I 2 ), 93.77%. A change was observed on the degree heterogeneity after excluding the known outliers and, performing subgroup analysis and sensitivity testing based on the type of otitis media and age group. The highest bacterial isolation rate was obtained among patients with chronic suppurative otitis media (CSOM) (98%) with a 66.91% degree of heterogeneity. Among patients with otitis media with inclusion of all age groups the prevalence was 87% (I 2 = 74.13%) and among pediatric patients with otitis media 86% (I 2 = 53.66%) (Fig. 2).
After a univariate meta-regression analysis, the year of study, country of study, study design, and sample size were not found to be statistically significant, however, the type of otitis media was found to be significant (pvalue = 0.001).

Bacterial isolates
From all bacterial isolates, the commonest isolate was Pseudomonas species (majorly P. aeruginosa) (ranging from 23 to 25% across subgroups) followed by S. aureus (18-27%), Proteus species (mainly P. mirabilis), Klebsiella species, Coagulase Negative Staphylococci (CoNS) and Streptococcus species like S. pyogenes, and Enterococcus species. Other gram-negative rods under the Enterobacteriaceae (Citrobacter species, Providencia species, Serratia species, Enterobacter species, and Morganella species) account for 7-11%, but separately they account for less than 5% ( Table 2, and Data Extraction     and Summary sheet: isolates). On the subgroup analysis, S. aureus was found to be the major isolate except for chronic suppurative otitis media and Streptococcus species were observed to have a relatively higher prevalence in pediatric patients with otitis media ( Table 2). Four studies have demonstrated the isolation of Acinetobacter species, and Corynebacterium species were isolated by two studies. Anaerobic bacteria were not commonly isolated from the studies, but a study by Chirwa M et al [31] has reported the isolation of Bacteroides species and Peptostreptococcus species.
Though the main objective of this review is bacterial etiologies we found out that fungal pathogens like Candida species and Aspergillus species account for 6% (3-8%) of the etiology for otitis media; only one study has indicated Aspergillus species.

The resistance of isolates to antibacterial agents
Studies have included various drugs for susceptibility testing. We have selected drugs that have been tested in more than 5 studies and accordingly, the analysis was done for Ampicillin, Cotrimoxazole, Cefuroxime, Amoxicillin-clavulanate, Ciprofloxacin, Gentamycin, Chloramphenicol, Amoxicillin, Ceftriaxone, Tetracycline, and Erythromycin.

Publication bias
There is some evidence of publication bias in studies reporting the prevalence of bacterial isolates as confirmed by funnel plots of the standard error with proportion supplemented by statistical tests (Begg's test, p = 0.0016; Egger's test, p = 0.0211) (Fig. 3). However, there was no significant publication bias in the subgroup analysis.

Discussion
This analysis included 32 original studies addressing bacterial isolation rate from otitis media within the specified time-frame. Since all the included studies have a crosssectional study design the pooled analysis could be used to assess the relative importance of various pathogens in Africa as well as worldwide. All the included studies in this review have collected an ear swab/discharge sample for microbiological analysis. The bacterial isolation rate using culture-based techniques ranged from 86% (pediatric patients) to 98% (patients with CSOM). Data coverage was featuring predominantly within some countries like Nigeria while most other countries don't have any study to represent the population.
Bacteria can reach the middle ear in otitis media from nasopharynx through the Eustachian tube or from the external ear canal through the non-intact eardrum.  Aerobic bacteria like S. aureus, P. aeruginosa and Proteus species (majorly P. mirabilis) were the main bacterial pathogens for otitis media in this analysis. The finding of those pathogens as a major isolate provides a more definitive confirmation of what has been reported by other published single-country studies as the result consistent [37,41,45]. Other bacteria are also indicated from studies like Klebsiella species (6-11%), CoNS (6-12%), and E. coli (5-9%). The prevalence of other pathogens like H. influenzae, S. pneumoniae and, M. catarrhalis is very minimal, even though those isolates were indicated from another review [60,61], and commonly reported from AOM and OME episodes, especially during childhood [4,12,14]. Since most of the studies conducted in sub-Saharan Africa are on CSOM and the collected samples are ear swab/discharge these pathogens are less likely to be reported from the studies. This minimal finding could also be due to regional differences or the absence of studies conducted using molecular techniques that had higher sensitivity compared to culture techniques.
On a recent international consensus on the management of otitis media with effusion panelists generally agreed not to consider antibiotics as a non-surgical management option for OME [62]. Several other reviews have indicated that the management of OM is far better with antibiotics and oto-topical agents than without any treatment, especially if it is CSOM [63,64]. Considering this review setting and population, treatment with antibiotics is a primary option as most of the studies are conducted on patients with CSOM. On the contrary, isolates from otitis media were having a high level of resistance to commonly used antibacterial agents in this analysis. The highest resistance is against ampicillin (ranging from 69 to 94%) followed by Amoxicillin (62-85%), cotrimoxazole (41-90%), Augmentin (47-61%) and Cefuroxime (17-64%). Generally, the resistance to erythromycin, ceftriaxone, Tetracycline, and ceftazidime was less than 50% for most of the studies. Those drugs are commonly used drugs for the management of otitis media, especially in low-income country settings without appropriate microbiological diagnostic service. This review also found that relatively higher susceptibility of isolates to ciprofloxacin, gentamycin, and chloramphenicol, less than 35% resistance. The studies included in this review revealed that non-susceptibility to fluoroquinolones ranged between 16 and 34%.
All studies in this review span on the analysis of ear swab/ discharge, which is a late-onset scenario. Early diagnosis and management of otitis media have a paramount effect on the health of the patients, especially children, as it has complications ranging from tympanic membrane perforation to neurological impairments. Early-onset infections will require the collection of middle ear fluid by tympanocentesis, which will be hard to practice in sub-Saharan Africa, without advanced medical facilities. A clear contribution of pathogens, like fastidious and anaerobic bacteria, can be understood by using highly sensitive laboratory techniques.
It is unlikely that this review has missed a relevant study as we have conducted a highly sensitive systematic search, including grey literature. The dearth of appropriate population-based elegant studies, risk of bias in prevalence studies, mainly due to under-reporting; and heterogeneity of data were the main limitations of this review. Etiological data were more consistent with some discrepancies in isolation rates, which can be explained by the quality of the microbiological procedures used, such as sample collection and transportation and bacterial isolation in some studies. One of the recognized limitations in such pooled studies [65] is differences in middle ear fluid collection guidelines in different localities even with efforts to standardize the design across studies. There are likely differences in care-seeking practices for OM among countries, affecting the severity of cases, and one country may collect middle ear fluid only in cases considered to be recurrent OM. The true contribution of these bacteria to OM is likely higher than the reported as bacterial etiology was not determined by molecular techniques in the included studies [66]. Different clinical severity or age tropisms may have been observed from culture-negative samples with molecular testing.
Following accepted methods, all-inclusive searching with date range, put together with a quality appraisal of the included studies are the main strength of this review. Stakeholders who focus on such matters will get a great benefit from compiling all available evidence to make informed decisions on prevention measures.

Conclusions
The analysis revealed that P. aeruginosa, S. aureus and P. mirabilis are the commonest bacterial pathogens responsible for otitis media in sub-Saharan Africa. A high level of resistance was observed to commonly used antibacterial agents such as Ampicillin, cotrimoxazole, Amoxicillin, and Amoxicillin-clavulanate however, isolates were less resistant to Ciprofloxacin, Gentamycin, and Chloramphenicol.
Thus, drugs like Ampicillin, amoxicillin, Amoxicillinclavulanate, and cotrimoxazole should not be used as a first-line treatment in sub-Saharan countries since countries in this geographical location are commonly dependent on clinical data for treatment with the absence of microbiology laboratory. Without proper treatment, otitis media could lead to intracranial and intratemporal complications with higher and complex management.