Search strategy
Databases such as: PubMed/Medline, CINAHL, AJOL and SCOPUS were searched using the search terms: (khat OR Catha edulis OR khat use OR khat consumptions OR khat chewing OR khat abuse OR khat dependence OR khat use disorder OR substance OR substance use OR psychoactive substance OR substance use disorder OR substance abuse) AND (prevalence OR incidence) OR magnitude) AND (factors OR risk factor OR risk OR determinant) AND (maternal OR mother OR primigravida OR wom*n OR pregnant OR prenatal) AND Ethiopia for PubMed/MEDLINE. A Google scholar search was also performed. We formulated a comprehensive search strategy by the combination of Medical Subject Headings (MeSH) thesaurus, text words, and appropriate Boolean Operators tailored to each database. No time and language restrictions were imposed on the search strategy. Reference lists of all articles were searched. Moreover, Google Scholar was searched for gray literature. All searches were performed until August-14-2019. Similarly, CINAHL, AJOL and SCOPUS were searched. using similar search terms tailored to each database. We adhered to the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) (38).
Selection of studies
All studies retrieved through search strategy were imported to EndNote X7 (Thomson Reuters, New York, USA) then, duplicated studies were removed from EndNote Library. The title and abstracts of the remaining articles were assessed independently by two reviewers (BBB and YDG). Disagreements were resolved by taking the mean score of the two reviewers.
Definition of concepts
In this study, khat use during pregnancy is defined as use of khat (any amount) at least once.
Eligibility Criteria
Setting: This review included studies carried out in Ethiopia.
Participants: This review included the results of khat use and its associated factors that assessed during pregnancy.
Outcome measure: This systematic review and meta-analysis have two main outcomes. The first outcome was to determine the pooled prevalence of khat use during pregnancy. The second outcome of the study was to identify factors associated with khat use. Prevalence was calculated by dividing the number of khat chewer during pregnancy to the total number of pregnant women included in the study (total sample size) multiplied by 100. For the associated factors, the reported odds ratio, 95% confidence interval and p-value were used to tabulate and described the results thematically.
Study design: Observational studies (cross-sectional and cohort/longitudinal) were included. Studies that focused on case reports, conference and abstracts were excluded if the required data are not accessed.
Data extraction
Data were extracted from the eligible studies using a pre-conceived and piloted data abstraction form. Data were simultaneously extracted by two independent reviewers (BBB and YDG). The extracted data include: name of the first author last name, year of publication, study area/region, study design, sample size, number of cases/prevalence and associated factors.
Quality assessment
The Newcastle-Ottawa quality assessment tool, adapted for cross-sectional studies (39) was used for quality assessment. This tool had three main parts (selection, comparability and outcome). The first part (selection) has five stars and assesses the methodological quality of the study. The second part of the tool evaluates the comparability of the study. The third part of the tool assesses the quality of the original article’s outcome with respect to the statistical analysis. Individual paper was graded with score ranged from zero to ten stars. The overall quality of each article was determined using the sum of each star of the three parts and defined as high quality for score ≥ 6 out of 10, medium (fulfilling 50% of quality assessment criteria) and poor for < 4.
Data synthesis and statistical analysis
The extracted data were entered into a Microsoft Excel Database and then imported into STATA version 14 (Stata Corp LLC, Texas, USA) software with packages of Meta-analysis for analysis. Meta-analysis was performed with STATA 14 using the “Metaprop” command with random-effects models to calculate the pooled effect size at 95% confidence intervals (40) with the Dersimonian and Laird method based transformed values and their variance (41). The Freeman–Tuckey variant of the arcsine square root transformation of proportions was fitted to avoid variance instability when handling proportions close to one (42). Heterogeneity between studies was measured by the index of heterogeneity (I2 statistics) test (43). I2 statistics values of 25%, 50% and 75% was used as low, medium and high heterogeneity respectively. Sub-group analysis was performed by years of publication, study setting, study design, sample size and quality of included studies. Publication bias was examined by the visual inspection of funnel plot (44) and egger test (45) with p-value of < 0.05 was considered as statistically significant. Sensitivity analyses was conducted to examine influential studies (46). However, we planned for meta-analysis of associated factors, because of the inconsistent and insufficient of data; results were analyzed thematically and presented using texts and table.