Hepatitis E virus infection among pregnant women in Africa: systematic review and meta-analysis

We had checked the presence of systematic review and protocol on this topic by searching different data bases. The data bases checked included Cochrane data bases of systematic review, the national health center review and dissemination data base, Joanna Briggs Institute data base a systematic review and implementation of reports (JBI-DSRIR), Health technology assessment (HTA) and the Campbell collaboration library and evidence for policy and practice information (EPPI-center). After checking all the above-mentioned data bases, a comprehensive literature search was conducted starting from September 20–2017-August 30,2018 on PubMed, Science Direct, African Online journal and Google Scholar. The search was carried out by two researchers (MD, FM) independently by using the following key words; name ``Hepatitis E virus seroprevalence``, ``Hepatitis E virus epidemiology`` OR ``Non A non B`` AND ``pregnant women`` AND (Country name _{_}1 OR Country name_2 OR … ..), where these ellipsis represent names of each African countries. Content experts were consulted for additional materials. The references cited by each eligible study were examined to identify additional articles.

Cross sectional and cohort studies published in 1993 to August 30, 2018 in Africa were included in the study. There is no language restriction. Articles that assessed the study seroprevalence of HEV infection in pregnant women, studies both HEV and HIV infections on pregnant women, studies on vertical transmission of HEV from mother to infant were included in the study. There was no age restriction. The studies were included only pregnant women population living in Africa.

This review considered studies that include the seroprevalence of HEV infection among pregnant women in Africa as an outcome. The seroprevalence is calculated by dividing the number of HEV infected positive pregnant women over the total number of pregnant women. Seroprevalence is defined as the presence of IgG antibody in serum/plasma of pregnant women by ELISA method.

The quality of the studies was assessed using Joanna Briggs Institute (JBI) quality appraisal criteria adapted for studies reporting prevalence data [18]. The following items were used to evaluate prevalence studies: (1) appropriate sample frame; (2) appropriate sampling technique; (3) adequacy of sample size; (4) description of study subjects and setting; (5) sufficient coverage of data analysis; (6) validity of method for identification of condition; (7) standard, reliable measurement for all participants; (8) appropriateness of statistical analysis; and (9) adequacy and management of response rate.

The relevant data from each selected study have been extracted independently by two authors (MD, FM) and summarized into an excel spread sheet. Discrepancies were resolved through consensus and discussion with a third author (MT). For each selected study, the following parameters were extracted: First author and reference, year of publication, study country/area, year/s/ of study period, study design, IgG prevalence, total number of pregnant women, method employed for HEV detection (ELISA kits).

Data were analyzed using Stata version 11 software package (Stata Corporation, College Station, TX). A random effects model was used to determine pooled prevalence and the 95% confidence interval (CI), by employing the approach of DerSimonian and Laird [19]. In addition, Freeman Turkey arcsine methodology also used to address stabilizing variances [20]. The heterogeneity of study results was assessed by the use of I² test. Significant heterogeneity was considered for P < 0.10 and I² > 50% [21, 22]. Possible source of variation was explored using sensitivity analysis and sub-group analysis by stratifying studies through predetermined variables; study region, type of method used and year of studies published. Publication bias was measured by Begg’s funnel plot and Egger’s regression [23]. A p-value < 0.05 on the Egger test was considered indicative of statistically significant publication bias. The forest plot with 95% CI pooled the overall seroprevalence of HEV infection was summarized by using figure. This systematic review and meta-analysis was reported based on PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) Statement [24].

A total of 452 research articles were retrieved by electronic search, of these, 350 non-duplicate papers were assessed and 306 records excluded based on titles and study area. The remaining examined by abstract screening of which, 22 articles were excluded because studies had no full text and 2 studies were excluded because they were focused on outbreak studies. After exclusion of duplicates and irrelevant studies based on titles and abstracts, 20 articles were retrieved full text detail analysis. Two additional articles were retrieved from reference list of published articles, finally a total of 22 studies were included in this systematic and meta-analysis (Fig. 1).

The 22 studies included in this systematic review (Table 1) and over all sample size of 8008 pregnant women in Africa [25‐46]. The present paper included studies from 12 (21.4%) of the 54 African countries. The regional distribution of countries Eastern Africa included Eritrea, Ethiopia, Tanzania and Sudan. West Africa included Benin, Burkina Faso, Cameroon, Ghana and Nigeria. North Africa included Egypt and Tunisia. Central Africa included only Gabon. The studies were published between 1993 to 2018 and all of the samples collected from 1988 to 2016. The sample size of the selected studies ranged from 90 to 2428. There were large differences in calculated seroprevalence between countries. In the pregnant women, the highest seroprevalence reported from Egypt 84.4% and the lowest seroprevalence from Gabon 6.6%. In addition to the difference in seroprevalences between countries, differences in seroprevalence have also been reported with in countries. Majority of the studies were cross sectional and only two studies were cohort. All studies used ELISA for the diagnosis of HEV. The studies used different types of ELISA assay methods, two studies used Wanti, two studies Dia. Pro, three studies Euroimmun, two studies International immune diagnostics, two studies TMB and other elven studies used different ELISA assay methods (Table 1). Mean age of the studies were specified in 18 studies ranged from 13 years Adje et al. [38] to 50 years Nigussie et al. [34].

The JBI criteria’s for assessing the quality of primary studies recommend to include primary studies scored ≥60% of methodological checklists in the meta-analysis (Table 2). We found eight studies scored > 80 (25,27, 28, 34, 35, 42, 44, 46). Six studies scored between 70 and 80% (29,30,31,37,41, 45) and other remaining studies between 60 & 70% (26,32,33,36,38,39,40,43). Almost majority of studies (59%), sample size was adequate, it is greater than 200 (29,31,32,33,39,42,43,45). More importantly, 81% of included studies used ELISA, valid methods for the diagnosis of HEV infection.

There was an old cohort study conducted from 1988 to 1991 and reported 59% in Ethiopia, Eastern Africa [35]. A study in Egypt, North Africa, reported higher seroprevalence 84.3% among pregnant women conducted on 1997–2003 [30]. In 2005 and 2008 a study in Gabon, central Africa reported 14.1 and 6.6% lowest seroprevalence among African pregnant women respectively [36, 37]. Later on, a study conducted in 2010–2011 in Burkina Faso, Western Africa reported 10.6% [27]. Recently, HEV reported 42.4% in Ethiopia [34] and Ghana [39] 12.2% among HIV positive pregnant women (Table 1).

All 22 included studies were pooled for meta-analysis. As presented on the forest plot (Fig. 2), the seroprevalence of HEV among pregnant women ranged from (6.6–84.27%). The overall pooled seroprevalence of HEV among African pregnant women was 29.13% (95% CI 14.63–43.63). Heterogeneity of studies among reported prevalence using level of heterogeneity was assessed using random effect by I statstic (I² = 99.7%, P = 0.001). There was high heterogeneity level. A p value of 0.001 indicates the presence of significant heterogeneity and I² = 99.7% indicates the heterogeneity level was high (Fig. 2). To explore the heterogeneity, we have done sensitivity analysis. We further conducted subgroup meta- analysis to identify the source of high heterogeneity by grouping variables: Study country, study region, assay method, year of publication, sample size.

We had done sensitivity analysis by removing one study with large sample size (37). The overall pooled prevalence was 26.01% (95% CI:19.9–32.1) with I² = 97.5%, P = 0.0001 (Fig. 3). Additional file 1: Figure S1. However; there was substantial heterogeneity.

Stratified analysis of HEV infection on pregnant women based on different regions of Africa. There was difference of seroprevalence in different regions estimated HEV seroprevalence in pregnant women in West Africa as 16.40 (95% CI 11.39–21.41), North Africa 50.01 (95% (4.43–95.58), East Africa 35.0 (95% CI 21.74–48.26), Central Africa 10.45 (95%, CI 3.02–17.88). There was difference across the region (I² = 99.7, P = 0.001) (Fig. 4). In addition to this the North Africa pooled estimate and central Africa are 50.01 and 10.45% respectively, they are out of the overall range (14.63–43.63). Such regional difference has been described between North Africa highest seroprevalence and Central Africa lowest seroprevalence. In addition to differences seroprevalence between countries, there have been also reported differences with in countries. For instance, there have been great differences in seroprevalence in Egypt 45–84.3%, Ethiopia 31.1–58% and Sudan 12.5–61.2% among pregnant women (Fig. 5).

The pooled anti-HEV IgG seroprevalence determined by different commercial ELISA assays showed large variability with reported seroprevalence rates ranging from 10.45 to 32.18 (Fig. 5). The frequently used assays in this systematic review were Euroimmun, Wanti, International immune diagnostics and TMB. For these four assays the pooled seroprevalence rates among pregnant women in Africa were: Wanti 37.15 (26.54–47.76%), International immune diagnostics 34. 80 (22.08–47.52%) Euroimmun 32.33 (5.89–58.77%) and others 31.92(6.98–58.88%) (Fig. 6).

This meta-analysis revealed that seroprevalence of HEV infection among pregnant women differed by publication year (Fig. 7). When we noted the seroprevalence of HEV among pregnant women with time, it decreases from 1992 to 2018. The infection seems to have a decreasing trend over time (Additional file 2: Figure S2). Seroprevalence of HEV infection varies with sample size (Fig. 8).

We had assessed the publication bias of studies by using funnel plot and Egger’s regression, in this systematic review, there was publication bias, evidence of an Egger’s regression p-value< 0.0 was seen when all studies considered (Fig. 9).

Different factors associated with the heterogeneity such as study design, publication date, the sample size of the study, region and assay method were investigated using meta-regression but none of these variables were statistically significant (Table 3).

This study is the first systematic and meta-analysis review among pregnant women in Africa and this meta-analysis has relatively large sample size with good quality studies but also several limitations. The reliability and accuracy of the test depends on the screening method employed. The included studies used different types of ELISA kits for detection of anti- HEV IgG with different specificity and sensitivity. Most studies reported only anti-HEV IgG which shows mostly past infection. In addition, we have included studies more than 24 years of difference. More over almost all studies did not include molecular tests which is important for HEV diagnosis for active infection. Furthermore, data were not available in all 54 African countries, only 12 African countries included in the systematic and meta-analysis.