Herbal medicine use and predictors among pregnant women attending antenatal care in Ethiopia: A systematic review and meta-analysis

Background: The use of herbal medicine among pregnant women is increasing in many low- and high-income countries due to their cost-effectiveness in treatment and ease of access. Research findings across Ethiopia on the prevalence and predictors of herbal medicine use among pregnant women attending antenatal care are highly variable and inconsistent. Therefore, this systematic review and meta-analysis aims to estimate the overall prevalence of the use of herbal medicine and its predictors among pregnant women attending antenatal care in Ethiopia. Method: We searched articles in Medline (PubMed), EMBASE, HINARI, Google Scholar, Science Direct, Cochrane Library, and other sources. The study included a total of eight studies that reported the prevalence and predictors of herbal medicine use among pregnant women from different regions of Ethiopia. Cochrane Q test statistics and I2 tests were used to assess heterogeneity. A random effect meta-analysis model was used to estimate the pooled prevalence. In addition, the association between risk factors and herbal medicine use in pregnant women attending antenatal care were examined. Results: A total of eight studies were included in this review. The pooled prevalence of herbal medicine use among pregnant women attending antenatal care in Ethiopia was 47.77% (95% CI: 28.00-67.55). Subgroup analysis by geographic regions has showed that the highest prevalence (57.49%;95% CI: 53.14, 61.85) was observed in Oromia Region and the lowest prevalence was observed in Addis Ababa (31.39%; 95% CI: 2.83, 79.96). The herbal medicines commonly consumed by women during pregnancy were ginger: 41.11% (95% CI: 25.90, 56.32), damakasse: 34.63% (95% CI: 17.68, 51.58), garlic: 32.98% (95% CI: 22.21, 43.76), tenaadam: 19.59% (95% CI: 7.54, 31.63) and eucalyptus: 4.71% (95% CI: 1.1, 8.26). Mothers’ previous history of self-medication (95% CI: 1.91, 51.35), illness during pregnancy (95% CI: 1.56, 23.91), employment status (95% CI: 3.89, 10.89), educational status (95% CI: 1.52, 2.68), and place of residence (95% CI: 1.86, 3.23) were predictors of herbal medicine use by women during pregnancy. Conclusion: In this study, about half of women attending antenatal care use herbal medicine and it is relatively high. The most commonly consumed herbal medicine during pregnancy was ginger followed by damakasse, garlic, tenaadam and eucalyptus. During pregnancy, it is not known that these most commonly consumed plant species have harmful fetal effects. However, many of the medicinal plant species are poorly studied, and it is not possible to rule out teratogenic effects. Teamwork between healthcare professionals and traditional practitioners to educate on the use of medicinal plants will encourage healthier pregnancies and better health for mothers and infants. Three authors (FA, GS & YM) have identified both published articles and unpublished researches reporting the prevalence and predictors of herbal medicine use among pregnant mothers attending ANC care in Ethiopia. Studies were identified through a literature search of Medline (Pub Med), EMBASE, HINARI, Google Scholar, Science Direct, Cochrane Library, and other sources. The reference list for each included article was also manually searched for search optimization. The search was conducted from January 9, 2019 to October 15, 2019, and it was limited to English language. Unpublished studies have also been searched through Google and Google Scholar. The search terms were predetermined for an extensive search that included all fields in records, as well as Medical Subject Headings (MeSH terms) to expand the search in an advanced PubMed search. We merged keywords with the “OR” in the Boolean operator within each axis and then related the search strategies for the two axes to the “AND” operator. The key terms used for the search were “Prevalence” OR “Epidemiology” AND “Traditional Medicine” AND/OR “Herbal Medicine” OR “Pregnant Mothers” AND/OR “Pregnant Women” OR “antenatal follow up” AND “Ethiopia”. The specific searching detail in pubmed with MeSH terms was (“herbal medicine” [MeSH Terms] OR (“herbal” [All Fields] AND “medicine” [All Fields]) OR “herbal medicine” [All Fields]) AND (“gravidity” [MeSH Terms] OR “gravidity” [All Fields] OR “pregnant” [All Fields]) AND (“mothers” [MeSH Terms] OR “mothers” [All Fields]) AND (“Ethiopia” [MeSH Terms] OR “Ethiopia” [All Fields]) were used. All literatures accessible until October 2019 were included in the systematic review and meta-analysis. The systematic review and meta-analysis was carried out in accordance with the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines [18]. Articles included were those reporting the prevalence of herbal medicine use and predictors among pregnant mothers attending ANC care in Ethiopia. Only articles conducted in Ethiopia. All observational studies (cross-sectional, case controls, and cohort) that contain original data reporting the prevalence and predictors of herbal medicine use among pregnant mothers attending ANC in Ethiopia were considered. Literatures written in English language. Studies that have been considered among pregnant mothers in ANC in Ethiopia. A consideration was given to both published articles and unpublished research. Non-accessible researches which are unpublished, irretrievable from the internet or failed to receive replies to email from corresponding authors were excluded. In addition, research that did not report our outcome of interest was excluded after reviewing complete texts (by three authors (FA, GS and YM). All the data required were extracted using a clear data extraction format prepared by two authors (FA and MD) in Microsoft Excel™. For the prevalence of herbal medicine use, the data extraction format prepared based on first author, the region where the study was carried out, publication year, sample size, and prevalence of herbal medicine use stated for the target group. For predictors, the data extraction format was prepared for each specific predictor (maternal residence, maternal Educational status, maternal occupation, maternal illness and previous self-medication). The researchers selected these variables because they are the most commonly reported associated risk factors in the studies included in this meta-analysis. In this systematic review and meta-analysis, the investigators considered additional variables as risk factors if two or more studies investigated them as risk factors. For every associated risk factor, to compute the odds ratio, the data from the primary studies were extracted in the form of two by two tables by three authors (FA, GS and BS). This systematic review and meta-analysis have two main outcomes. The primary outcome was the prevalence of herbal medicine use among pregnant women attending ANC in Ethiopia. The second outcome of the study was predictor of herbal medicine use among pregnant women attending ANC in Ethiopia. The prevalence was computed by dividing the number of pregnant mothers attending ANC who use herbal medicine by the total number of pregnant mothers attending ANC in the study (sample size) multiplied by 100. The researchers (FA & YM) used the Newcastle-Ottawa Scale adjusted for the quality evaluation of the cross-sectional studies to determine the quality of the studies included in this review [19]. The tool consists of three basic parts; the first section has five stars, and assesses each study’s methodological excellence. The second part of the instrument tests the research comparability and gives two points. The last component measures the consistency of the original articles with respect to their statistical analysis and can be rated out of 3 stars. The qualities of each of the original articles were measured using the tool as a checklist. Articles included in this study have medium to high quality scores (6 out of 10 stars). The necessary data were extracted using a Microsoft Excel™ format and analyzed using the program STATA Version 15.0. The original studies were described using forest plots and tables. The researchers determined the standard error prevalence by the binomial distribution method for each original article. Heterogeneity among the recorded prevalence of studies has been confirmed by the use of test heterogeneity x2, I2 and p-values [20]. The above statistical analyses suggested a considerable heterogeneity among the studies (I2 = 99.2%, p-value < 0.001). As a result, a random effect meta-analysis model was conducted to estimate the combined effect of Der Simonian and Laird. Additionally, univariate meta-regression model was performed by taking year of publication and sample size to identify the likely source of heterogeneity, but none was found to be statistically significant. Possible publication bias was tested objectively using Egger’s correlation and Begg’s regression intercept tests at 5% significant level, respectively [19, 21]. The Egger’s weighted regression and Begg’s rank correlation test methods were also used to assess publication bias (P > 0.05), revealing statistically insignificant publication bias. In addition, to reduce the random differences between the primary study’s point estimates, subgroup analysis was performed based on the area in which the studies were conducted.