Prevalence of human respiratory syncytial virus infection in people with acute respiratory tract infections in Africa: A systematic review and meta‐analysis

Aim The epidemiology of human respiratory syncytial virus (HRSV) infection has not yet been systematically investigated in Africa. This systematic review and meta‐analysis are to estimate the prevalence of HRSV infections in people with acute respiratory tract infections (ARTI) in Africa. Method We searched PubMed, EMBASE, Africa Journal Online, and Global Index Medicus to identify observational studies published from January 1, 2000, to August 1, 2017. We used a random‐effects model to estimate the prevalence across studies. Heterogeneity (I 2) was assessed via the chi‐square test on Cochran's Q statistic. Review registration: PROSPERO CRD42017076352. Results A total of 67 studies (154 000 participants) were included. Sixty (90%), seven (10%), and no studies had low, moderate, and high risk of bias, respectively. The prevalence of HRSV infection varied widely (range 0.4%‐60.4%). The pooled prevalence was 14.6% (95% CI 13.0‐16.4, I 2 = 98.8%). The prevalence was higher in children (18.5%; 95% CI 15.8‐21.5) compared to adults (4.0%; 95% CI 2.2‐6.1) and in people with severe respiratory tract infections (17.9%; 95% CI 15.8‐20.1) compared to those with benign forms (9.4%; 95% CI 7.4‐11.5); P‐values <0.0001. The HRSV prevalence was not associated with sex, subregion in Africa, setting, altitude, latitude, longitude, and seasonality. Conclusion This study suggests a high prevalence of HRSV in people with ARTI in Africa, particularly among children and people with severe clinical form. All innovative strategies to curb the burden should first focus on children which present the highest HRSV‐related burden.

each year in children aged less than 5 years. A total of 3.2 (95% CI, 2.7-3.8) million of these infections were severe cases requiring hospitalization. Approximately 40-74.5 thousand deaths were estimated in children <5 with ARTI in this meta-analysis. 2 Another metaanalysis had previously showed that the incidence of HRSV infection in developing countries was more than twice greater than that of developed countries and that 91% of hospitalizations and nearly all deaths (99%) were registered in developing countries where access to health care is limited. 3 Preterm infants, those under 2 years, those over 65 years, and immunocompromised patients, are at higher risk of hospitalization for HRSV infections. 4 Patients with a history of atopy, congenital heart disease, congestive heart failure, chronic obstructive pulmonary disease, neuronal and muscular disorders, and cancer are also at high risk for the development of severe HRSV infections. 5 To date, there are more than 60 HRSV vaccine development programs at various stages. Some of them could be used in the next decade.
Africa, a continent in which most countries are developing with limited resources for health, the ecology in this continent can give it a particular epidemiology regarding HRSV infection. To the best of our knowledge, to date, no systematic review and meta-analysis have been conducted in Africa on the epidemiology and drivers of HRSV infection. Therefore, we conducted a systematic review and meta-analysis to estimate the prevalence and drivers of HRSV infection among people with ARTI in Africa. We have done this work to provide accurate data to guide health policymakers and to identify information gaps to guide future research.

| Search strategy and selection criteria
We performed a comprehensive and exhaustive search of MEDLINE through PubMed, Excerpta Medica Database (EMBASE), Africa Journals Online, and Global Index Medicus to identify all relevant articles published on HRSV in Africa from January 1, 2000, until August 31, 2017, regardless of language of publication. Both text words and medical subject heading terms were used. The following terms and their variants were used for HRSV: "HRSV," "human respiratory syncytial virus," and "respiratory syncytial virus." Individual country names for the 54 African countries and African subregion names such as "Northern Africa" or "Southern Africa" were also used as additional key search terms for more abstracts on the subject. African country names were introduced both in English and in languages relevant to each country, for example, "Ivory Coast" and "Côte d'Ivoire". Where country names have changed over time, old and new names were included, such as "Zaire" and "Democratic Republic of Congo". Titles and abstracts of all eligible papers were reviewed, and full texts of articles were accessed. Search strategy conducted in PubMed is shown in Table S1. This search strategy was adapted to fit with other databases. A manual search which consists in scanning the reference lists of eligible papers and other relevant review articles was conducted. The search in electronic databases was conducted on September 6, 2017.
We considered observational studies (cross-sectional, casecontrol, and cohort). In the case of duplicate reports, the most comprehensive/complete and up-to-date version was considered.
We considered studies including patients with clinical diagnosis of acute respiratory tract infection as defined in each study. Studies among populations with underlying medical conditions, studies conducted during an outbreak period, case series or studies in which HRSV was imported cases were excluded. The search for HRSV had to be conducted systematically or by sampling of the population in the presence of defined inclusion criteria (respiratory signs) and HRSV detection by polymerase chain reaction (PCR) technique on respiratory samples. Studies lacking or with not extractable primary data and/or explicit method description were excluded. In the case of missing data, we contacted authors of the paper. We planned to use Google Translate in the case of other languages than French, English, or Spanish.
Two investigators independently screened records based on titles and abstracts for eligibility. Full texts of articles deemed potentially eligible were retrieved. Further, these investigators independently assessed the full text of each study for eligibility and consensually retained studies to be included. Disagreements when existing were solved through a discussion.

| Data extraction and management
Data were extracted using a preconceived and tested data abstraction form. In the cases of missing data, authors were directly contacted to provide missing information. Two investigators independently extracted data including name of the first author, publication year, study design, setting, sampling method, respiratory samples collection period, timing of data analysis, number of viruses screened, site of recruitment location (country, city, latitude, longitude, and altitude), clinical presentation, number of patients screened, number of patients infected with HRSV, diagnostic techniques used, and proportion of male participants. We assigned a United Nations Statistics Division (UNSD) African region (central, eastern, northern, southern, and western) to each study regarding the country of recruitment. 6 We considered two groups of clinical presentation; severe respiratory tract infection (SRTI) including severe acute respiratory infection, acute lower respiratory infections, bronchitis, bronchiolitis, pneumonia, severe or very severe pneumonia, and benign respiratory tract infection (BRTI) including upper respiratory tract infection, and influenza-like illness. Using Google Global Positioning System, we assigned altitude, latitude, and longitude according to the cities and country of recruitment. 7 In the case of multicities, we considered the median. Disagreements between investigators were reconciled through discussion and consensus or an arbitration of a third investigator.
Two investigators evaluated risk of bias in included studies using an eight-item rating scale. 8 These items included (a) participation response rate more than 75% agree to participate or analysis to show whether respondents and nonrespondents were similar for the sociodemographic characteristics; (b) ARTI clearly defined; (c) method of inclusion identical for all subjects; (d) description of diagnostic technique; (e) same type of sample collected for all patients (nasopharyngeal aspirate, nasal, or throat swab); (f) standardized method for sample collection (quantity of aspirate or of liquid used for the nasal wash with any virological medium transport for swabs); (g) analysis performed according to relevant subgroups (by age classes, by center, or by symptomatology, for example); and (h) and presentation of data sources (counts are presented, not only percentages). Each item was assigned a score of 1 (Yes) or 0 (No), and each score was summed across items to generate an overall study quality score. The total score was ranged from 0 to 8 with the overall score categorized as follows: 6-8: "low risk of bias," 3-5: "moderate risk," and 0-2: "high risk." Disagreements were solved through discussion and consensus.

| Data synthesis and analysis
Data analyses used the "meta" and "metafor" packages of the statistical software r (version 3.3.3, The R Foundation for statistical computing, Vienna, Austria). Unadjusted prevalence and standard errors of HRSV infection were recalculated based on the information of crude numerators and denominators provided by individual studies. To keep the effect of studies with extremely small or extremely large prevalence estimates on the overall estimate to a minimum, the variance of the study-specific prevalence was stabilized with the Freeman-Tukey double-arcsine transformation before pooling the data within a random-effects meta-analysis model. 9 Symmetry of funnel plot and Egger's test served to assess the presence of publication bias. 10 A P-value <0.10 on Egger test was considered indicative of statistically significant publication bias.

Heterogeneity was evaluated by the chi-square test on
Cochrane's Q statistic, 11 which was quantified by H and I 2 values.
The I 2 statistic estimates the percentage of total variation across studies due to true between-study differences rather than chance.
In general, I 2 values greater than 60%-70% indicate the presence of substantial heterogeneity. The value of H close to 1 is indicative of some homogeneity between studies. 12 Subgroup analyses were performed for the following subgroup: age groups (0-5/>5 years, children/adults), sex, clinical presentation, setting, hemisphere (north/ south), location according to Greenwich meridian (East/West), by median altitude, and UNSD African Regions. Univariable and multivariable meta-regressions were used to test for an effect of study and participants' characteristics. To be included in multivariable meta-regression analysis, a P value <0.25 in univariable analysis was considered. For categorical variables, the global P value was considered for the inclusion in multivariable models. A P value <0.05 was considered statistically significant. We reported the explained heterogeneity (R²) for the residual heterogeneity of HRSV prevalence.
Following crude overall prevalence, two sensitivity analyses were conducted: one considering only studies with low risk of bias and another one considering only studies conducted in complete season(s).
The Centre for Reviews and Dissemination guidelines was used for the methodology of this review. 13 The Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines served as the template for reporting the present review (Table S2). 14 This review is registered in the PROSPERO International Prospective Register of Systematic Reviews, registration number CRD42017076352.

| Review process
We identified 718 records; after elimination of duplicates, 683 records remained. After screening of titles and abstracts, we found 539 records to be irrelevant and excluded them. We assessed full texts of the remaining 144 papers for eligibility, of which 78 were excluded with reasons ( Figure 1). The inter-rater agreement for study F I G U R E 1 Review process selection was high (κ = 0.81). Finally, 66 full texts including 67 studies were included (one paper included two studies).     Table 2 summarizes results from meta-analyses. The prevalence varied widely from 0.4% to 60.4% across countries. The distribution of the prevalence was not uniform in the continent (Figure 2). The overall prevalence was 14.6% (95% CI 13.0-16.4) in a pooled sample of 154 000 participants with ARTI ( Figure 3). The funnel plot did not suggest any publication bias ( Figure S1), and this result is confirmed by the Egger test (Table 2). The prevalence in studies with low risk of bias was not different to the overall prevalence (Table 2).

| Prevalence of HRSV infection among ARI in Africa
Substantial heterogeneity was present for overall and within all subgroups (Table 2). Publication bias was found for the following subgroup analyses: SRTI, North Hemisphere, and Northern Region of Africa (Table 2).

| Subgroup analyses
The prevalence was higher among children (

| Factors associated with prevalence of HRSV infections
In univariable meta-regression analysis, the prevalence was associated with clinical presentation (R 2 = 8.99%) and age groups (R 2 = 38.84%). In multivariable meta-regression analysis, the prevalence was higher in children compared to adults (adjusted odds ratio:  adults aged more than 65 years. 83 One should note that these three systematic reviews included studies using less sensitive and less specific immunofluorescence and immune-chromatographic We found in this review that the prevalence of HRSV infection was higher in children compared to adults. This is consistent with other reviews. 4,5,81-83 Similar data were found in developed countries. 84  To the best of our knowledge, this article is among the first systematic reviews that use meta-analysis to summarize data on prevalence of HRSV infection in people with ARI in Africa. Strengths of this systematic review and meta-analysis include the use of predefined protocol, a comprehensive search strategy, and involvement of two independent investigators in all stages of the review process. We included only studies that identified HRSV using a reference standard method for diagnosis, the reverse transcriptase PCR.

| D ISCUSS I ON
Although we found publication bias in some subgroup analyses, no publication bias was found in the main analyses suggesting that we were unlikely to have missed studies that could alter the findings.

ACK N OWLED G EM ENTS
None.

CO N FLI C T O F I NTE R E S T S
The authors declare that they have no competing interests.

DATA S H A R I N G A N D DATA ACCE SS I B I LIT Y
All data generated or analyzed during this study are included in this published article and its supplementary information files.