The protocol for this systematic review has been registered in the International Prospective Registry of Systematic Review (PROSPERO) with a specific registration number CRD42023415610.

This meta-analysis followed the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guideline (50). Studies were searched through PubMed/Medline, Cochrane Library, Science Direct, CINAHL, African Journals Online, HINARI, Global Health, and Google Scholar. In addition, digital libraries were searched to identify grey literature.

Endnote software was used to collect, organize, and remove the duplications of search outcomes. The search was made using the search terms: “prevalence,” “proportion,” “magnitude,” “incidence,” “tobacco smoking,” “cigarette smoking,” “smoking,” “tobacco,” “cigarette” “substance use” “factors,” “determinants,” “predictors,” “factors associated,” “associated factors,” “risk factors,” “adult,” and “Ethiopia.” All key terms were searched by a combination of Boolean operators “AND” or “OR” as appropriate, and the search was carried out from April 1 to May 15, 2023, by three authors independently (CD, AA, and BD).

Population: this systematic review and meta-analysis includes studies conducted among tobacco-smoking adults in Ethiopia. Exposure: an individual adult who experienced tobacco smoking whereas an individual adult who did not tobacco smoking was considered as a comparison. Outcome: studies assessed tobacco smoking as a primary outcome. All studies conducted with observational studies design (cohort, cross-sectional, and case–control) and institutional and community-based studies were also included in systematic review and meta-analysis. Besides, studies published from 2000 to 15 May 2023 in the English language were also included in the meta-analysis. However, qualitative studies, unretrievable studies, editorial letters, studies with poor methodological quality, and studies that did not report the outcome of interest were excluded from the meta-analysis.

The primary outcome of the study was the pooled prevalence of tobacco smoking among adults in Ethiopia, calculated by dividing the number of smokers by the total sample size and multiplying by 100. Additionally, the study aimed to identify the factors linked to tobacco smoking in the form of a log odds ratio.

After searching for relevant studies from the aforementioned electronic databases, they were exported to Endnote X20 (51) and any duplicates were removed. Three authors (CD, BD, and MG) independently extracted all the required data from the included studies by using a standardized data extraction template. The data extraction template consisted of various study details, such as the author’s name, region, publication year, study area/region, study population, study setting, study design, sample size, response rate, and tobacco smoking prevalence (current and lifetime prevalence). In the case of any disagreements during data extraction, other authors (AA, MD, and SAD) resolved them.

After duplicate files were removed, four reviewers (CD, MG, BD, and SAD) screened the relevant studies for inclusion. The Joana Brigg Institute (JBI) critical appraisal checklist for prevalence studies was employed to evaluate the quality of each article (52) and those studies scored more than 50% were included for analysis (53, 54). Each study’s quality was assessed independently out of 100% by four authors (CD, BD, MG, and SAD). If any discrepancies arose during the quality assessment, the mean score was calculated from the results of all reviewers to resolve the differences.

The extracted data was exported into STATA/ SE version-14 statistical software for analysis (55). The level of heterogeneity among the included studies was statistically evaluated using the Higgs I² test, with values greater than 75% considered high heterogeneity (56). DerSimonian and Liard’s method of random-effects model at a p-value less than 0.05 with a 95% CI was used to estimate the pooled prevalence of tobacco smoking using a forest plot. A p-value of less than 0.05 was considered indicative of the presence of heterogeneity (57).

Subgroup analysis was also conducted by characteristics of the study, such as region (Oromia, Amhara, Tigray, SNNPR (Southern Nation, Nationality, and People Region), all regions, or others), study setting (institutional or community), sample size (large—≥ 600 or small—<600), study population (students or community), and year of publication (before 2015 or 2015 and after). The different factors associated with tobacco smoking were presented using odds ratios (ORs) with 95% CI. Moreover, univariate meta-regression was carried out to identify the source of variations among studies that exhibited high heterogeneity.

Additionally, publication bias was assessed using a funnel plot and Egger’s test method with a significance level of p < 0.05 (58). Likewise, Duval and Tweedie’s ‘trim and fill’ analysis was also conducted to adjust the effect of publication bias among the studies included. A sensitivity analysis was also performed to assess the influence of a single study on the pooled prevalence estimates.

After conducting the electronic database search, a total of 1,063 studies were identified. Out of these, 802 articles were excluded based on their titles and abstracts. Besides, 20 articles were also excluded due to poor quality and did not report the outcome of the interest. Finally, 32 full-text articles were found eligible for this systematic review (Figure 1).

A total of 32 articles were reviewed to estimate the pooled prevalence of tobacco smoking among adults in Ethiopia. The highest prevalence of tobacco smoking was reported in a study conducted in the Oromia region (45.3%) (28) while the lowest (2.4%) was reported in a study conducted among all regions of Ethiopia (33). The total number of study participants included in the systematic review and meta-analysis was 69,897. In this meta-analysis, eight studies were conducted in the SNNPR (20, 26, 31, 34, 38–40, 46), seven in the Oromia region (19, 22, 24, 25, 28, 42, 49), three in Amhara region (18, 35, 48), three in Tigray region (27, 29, 32), three in all regions of Ethiopia (33, 37, 41), three in Addis Ababa (23, 45, 47), two in Somalia region (21, 44), and one each in Harari region (43), Benishangul Gumiz (30) and Dire-Dawa city administration (36). Although attempts were made to search cohort, case–control, and cross-sectional studies, we found only cross-sectional studies (Table 1).

In this meta-analysis, 25 and 26 studies were included to estimate the pooled prevalence of lifetime and current tobacco smoking, respectively. The pooled prevalence of lifetime tobacco smoking among adults was 16.0% (95% CI: 13.6–18.39). A random-effects model revealed that the included articles have high heterogeneity (I² = 99.1%; p < 0.001) (Figure 2). Similarly, the pooled prevalence of current tobacco smoking was 13.7% (95% CI: 11.40–16.10), with significant heterogeneity (I² = 98.9%; p < 0.001) (Figure 3).

Publication bias was subjectively evaluated by using a funnel plot. The finding indicated that there is asymmetrical distribution of the studies, which suggests the presence of publication bias (Figure 4). The Egger test statistics also revealed the presence of statistically significant publication bias (p < 0.001). Duval and Tweedie’s “trim and fill” analysis indicates significant variation in the newly estimated pooled odds ratio (the adjusted point estimate) [OR = 1.68, (95% CI: 1.36–1.99)] as compared to the initial or observed point estimate [OR = 2.42, (95% CI: 2.11–2.73)] (Figure 5).

The findings of sensitivity analysis suggest no evidence of a single study’s effect on the overall pooled prevalence (Figure 6).

To identify the source of heterogeneity among the included research articles, sub-group analysis was carried out based on the region where the studies were conducted, study population, year of publication, study setting, and sample size. Even though heterogeneity was not resolved among the included articles, the pooled prevalence of lifetime tobacco smoking was relatively higher among students than the general population [16.2 (95% CI: 11.94–20.45)]. In addition, a high prevalence of tobacco smoking was observed in the studies done in the Oromia region as compared to other regions [24.7 (95%CI: 17.87–31.71)], with significant heterogeneity (I² = 96.3; p < 0.001). Similarly, the prevalence of tobacco smoking was high among institutional-based studies [15.9 (95%CI: 13.59–18.39)] as compared with community-based studies [13.1 (95%CI: 9.64–16.60)], with statistically significant heterogeneity (I² = 99.4; p < 0.001) (Table 2). In addition, a univariate meta-regression was conducted using the study population, sample size, and study year as factors. However, neither of them was found to be a statistically significant source of heterogeneity (Table 3).

A total of 17 studies were used to identify factors associated with tobacco smoking (18–24, 26, 27, 29, 30, 34–36, 38, 39, 44). The results from the random-effects model showed that the odds of tobacco smoking were three times higher among males compared with females (OR = 3.22, 95% CI: 2.06–5.03). Significant heterogeneity was observed among the included articles (I² = 93.1%, p < 0.001) (Figure 7). The association between alcohol consumption factors and tobacco smoking was examined based on the results of ten studies (18, 20–23, 26, 29, 31, 34, 39). Eight of the included studies had a positive association (18, 20–22, 26, 29, 31, 34) while a negative association was noted in the other two studies (23, 39). The odds of tobacco smoking were 3.78 times higher among alcohol users than non-alcohol users (OR = 3.78, 95%CI: 1.27–11.24) (Figure 8).

A total of 12 studies were included to identify the association between chewing khat and tobacco smoking (18, 20, 21, 23, 26, 27, 29, 31, 34, 36, 38, 44). Ten of the included studies had a positive association (18, 21, 26, 27, 29, 31, 34, 36, 38, 44) while a negative association was noted in the other two studies (20, 23). The result of the random-effects meta-analysis showed that non-significant association between tobacco smoking and chewing khat (OR = 3.96, 95%CI: 0.96–16.29) (Table 4).

Five studies (22, 26, 29, 36, 44) were included to assess if having tobacco-smoking friends is associated with tobacco smoking. All of the included studies had a positive association. The meta-analysis also showed that the odds of tobacco smoking were seven times higher among adults who have tobacco-smoking friends compared to those who did not have smoking friends [OR = 7.21 (95%CI: 5.56–9.35)], with extreme heterogeneity (I² = 40.1%, p = 0.154) (Figure 9).

A total of 4 studies (22, 35, 39, 44) were included to assess if having a tobacco-smoking parent is associated with tobacco smoking. Three of the included studies had a positive association (22, 35, 44) while a negative association in one study (39). Meta-analysis also showed that there was a non-significant association between parental smoking habits and adult tobacco smoking [OR = 2.16 (95%CI: 0.28–16.83)] (Table 4).

This review included most regions of the country and a diverse target population (students and other community members including healthcare workers and patients). The study followed preferred reporting items for systematic review and meta-analysis guideline. However, studies included in this systematic review and meta-analysis were cross-sectional, which limits the causality of predictors on tobacco smoking. Besides, this study included articles published in the English language only, which could induce publication bias.