(Table 1).
Quality assessment
We assessed the quality of studies using the Joanna Briggs Institute Critical Appraisal (JBI) tools designed for cross-sectional and case-control studies (Moola et al., 2020). The critical appraisal checklist for the cross-sectional and case-control studies contains eight and ten questions respectively. Each question was scored out of 100% and finally, the sum of all questions was turned into 100%. The quality score was graded as low if < 60%, medium if 60–80% and high if > 80% (Porritt et al., 2014; Munn et al., 2019). Two authors (AA, GD) independently assessed the quality of the studies and the third author (ZWB) resolved the inconsistencies through discussion (Additional file 3) .
Outcomes
The primary outcome of this systematic review and meta-analysis study was drug-resistant TB. Whereas, the risk factors associated with DR-TB were the secondary outcomes. The pooled odds ratio along with their 95% CIs were estimated to assess the risk factors associated with DR-TB in Ethiopia.
Data analysis
Data that were summarized in Microsoft Excel 16 spreadsheet were exported to STATA version 15 for statistical analysis. The pooled OR with 95%CI of each risk factor was estimated by assuming the true effect size varies between studies. We presented the pooled results using a forest plot. We used the forest plot and I 2heterogeneity test to assess the heterogeneity among the studies. The I 2 values of 25%, 50%, and 75% were interpreted as the presence of the low, medium, and high heterogeneity, respectively (Sterne and Egger, 2001; Riley et al., 2011). In this study, we used a random-effects model for all risk factors to perform the analysis by considering substantial variability among the studies (Riley et al., 2011). We explored the presence of publication bias through visual inspection of the funnel plot and statistical significance of Egger’s regression test.
RESULTS
Study characteristics
After the systematic article searching in the available databases and other literature sources, we identified 2238 articles. After removing 244 duplicates, 1994 articles were screened by title and abstract. Then, full-text screening was conducted for 43 articles. Finally, after the full-text screening, 27 eligible articles were included in the study (Abay et al., 2020; Abdella et al., 2015; Adane et al., 2015; Alene et al., 2019; Arega et al., 2019; Assefa et al., 2017; Babure et al., 2019; Bedewi et al., 2017; Biru et al., 2020; Deressa et al., 2014; Desissa et al., 2018; Dessalegn et al., 2016; Fikre et al., 2019; Gobena et al., 2018; Hamusse et al., 2016; Hirpaet al., 2013; Jaleta et al., 2017; Mehari et al., 2019; Mekonnen et al., 2015; Mesfin etal., 2018; Mulisa et al., 2015; Mulu et al., 2015; Seyoum et al., 2014; Tadesse et al., 2015; Tesfay et al., 2016; Tsega et al., 2017; Welekidan et al., 2020). We presented the detail using the PRISMA flow diagram (Figure 1). The studies were conducted in different administrative regions across the country with the most frequent studies were from Addis Ababa (7 studies) followed by the Amhara region (6 studies), the Oromia region (5 studies), and the Tigray region (4 studies). However, the studies were reported from the majority of the regional administrative states. The studies were based on data collected from TB patients in health care facilities. The study period for these studies ranges from 2008 (Tadesse et al., 2015) to 2019 (Welekidan et al., 2020). Based on the study designs, 15 studies were cross-sectional studies while the remaining 12 studies used a case-control study design. The majority of the studies included all the age group categories while three studies and two studies included TB patients above 15 and 18 years of age respectively. The studies assessed the risk factors/determinates of either MDR-TB (18 studies), RR-TB (3 studies), or any type of drug resistance (6 studies). The sample size of individual studies ranges from 65 in a study conducted by Babure et al. (2019) to 1876 in a study done by Arega et al. (2019) (Table 1).
Risk factors of drug-resistant tuberculosis
In the current study, we extracted data to assess the risk factors of drug-resistant TB in the Ethiopian setting. The risk factors include; socio-demographic characteristics (age, sex, marital status, residence, occupation, and family size), behavioral characteristics (smoking status, alcohol consumption, and khat chewing), and clinical characteristics (HIV serostatus, DM co-occurrence, contact history, imprisonment, previous TB treatment, number of TB episodes, and site of TB infection). We performed a pooled analysis for each variable using a random effect model by considering substantial variability among the individual studies. Based on the pooled analysis of the odds ratio, unemployment (OR; 2.71, 95% CI; 1.64, 3.78, I2; 0.0%) (Figure 2) , having a history of previous TB (OR; 4.83, 95% CI; 3.02, 6.64, I2; 69.6%) (Figure 3) , having contact with a known TB patient (OR; 1.72, 95% CI; 1.05, 2.40, I2; 73.8%) (Figure 4) , having contact with a known MDR-TB patient (OR; 2.54, 95% CI; 1.46, 3.63, I2; 0.0%) (Figure 5) , and having pulmonary TB (OR; 1.80, 95% CI; 1.14, 2.45, I2; 71.4%)(Figure 6) were found to be the risk factors associated with drug-resistant TB in Ethiopia. While, old age individuals (OR; 0.77, 95% CI; 0.60, 0.95, I2; 47.2%) (Figure 7)including above 45 years of age (OR; 0.76, 95% CI; 0.55, 0.97, I2; 47.6%) (Figure 8) , and males (OR; 0.86, 95% CI; 0.76, 0.97, I2; 19.0%) (Figure 9)had lower risk of DR-TB compared to their counterparts (Table 2) .
However, statistically significant association was not found for the following variables; urban residence (OR; 0.86, 95% CI; 0.58, 1.15, I2; 75.2%), being single (OR; 1.12, 95% CI; 0.84, 1.40, I2; 46.0%), being a house wife (OR; 0.86, 95% CI; 0.50, 1.21, I2; 0.0%), being a farmer (OR; 0.81, 95% CI; 0.42, 1.19, I2; 48.0%), being a daily laborer (OR; 0.97, 95% CI; 0.16, 1.78, I2; 33.4%), family size above three members (OR; 0.87, 95% CI; 0.61, 1.14, I2; 0.0%), alcohol consumption (OR; 0.96, 95% CI; 0.53, 1.38, I2; 70.01%), khat chewing (OR; 1.01, 95% CI; 0.63, 1.38, I2; 0.0%), smoking (OR; 0.75, 95% CI; 0.34, 1.16, I2; 58.3%), imprisoned (OR; 1.00, 95% CI; 0.43, 1.56, I2; 0.0%), being HIVpositive (OR; 1.35, 95% CI; 0.95, 1.74, I2; 73.6%), having DM (OR; 0.85, 95% CI; -0.85, 1.93, I2; 0.0%), and having two or more number of TB episodes (OR; 1.03, 95% CI; -0.03, 2.99, I2; 79.2%) (Table 2) (Supplementary figure 1) .
Accordingly, those individuals who were unemployed had 2.71 times the odds to develop DR-TB compared to the employed ones. Likewise, those individuals who had a history of previous TB had 4.83 times the odds to develop DR-TB compared to new TB patients. Similarly, those who had a contact history with a known TB patient had 1.72 times the odds to develop DR-TB compared to their counterparts. Also, those who had a contact history with a known MDR-TB patient had 2.54 times the odds to develop DR-TB compared to individuals who didn’t have a contact history with a known MDR-TB patient. Besides, individuals with pulmonary TB had1.80 times the odds to develop DR-TB compared with patients with extrapulmonary TB. While the risk of DR-TB decreased by 23% among older age groups, i.e. individuals who were above 45 years had a 24% decreased risk of DR-TB compared to individuals below 45 years of age. Likewise, the risk of DR-TB decreased by 14% among males compared to females (Table 2) .
Publication bias assessment
Based on the funnel plot and the Egger’s regression test publication bias was detected for older age (P<0.001), above 45 years age (P<0.001), male sex (P=0.0047), being single (P=0.0036), being farmer (P<0.0299), urban residence (P<0.001), HIV seropositive (P<0.001), smoking (P<0.001), alcohol consumption (P<0.001), previous TB history (P<0.001), having two or more TB episodes (P<0.001), and contact with known TB patient (P=0.001). While publication bias was not detected for age above 40 years (P=0.0608), family size above three members (P=0.7071), being a daily laborer (P=0.7341), being a house wife (P=0.3798), unemployed (P=0.8805), imprisoned (P=0.3692), having contact with known MDR-TB patient (P=0.3425), khat consumption (P=0.2586), having DM (P=0.3082), and having pulmonary TB (P=0.0622)
DISCUSSION
In the current study, we explored in detail the risk factors associated with drug-resistant TB in Ethiopia using 27 eligible articles. After performing a pooled analysis for 18 variables extracted from individual studies, we found that five variables such that unemployment, having a history of previous TB, having contact with a known TB patient, having contact with a known MDR-TB patient, and having pulmonary TB were the risk factors associated with drug-resistant TB in Ethiopia. However, older age and male individuals had a lower risk compared to their counterparts.
The current study revealed that those individuals who were unemployed had 2.71 times the odds to develop DR-TB compared to the employed ones. It was also supported in a global pooled estimate performed by Pradipa et al. (2018). The poor living condition of unemployed individuals that leads them to pathological changes might be the possible cause (Przybylski et al., 2014). The present study also revealed that previous history of TB treatment is a major risk factor associated with DR-TB in Ethiopia, such that those individuals who had a history of previous TB treatment had 4.83 times the odds to develop DR-TB compared to their counterparts. In line with this study, pooled estimates conducted at different countries across the globe reported a statistically significant association of previous TB treatment with DR-TB (Pradipta et al., 2018; Lukoye et al., 2015; Jimma et al., 2017; Faustini et al., 2006; Zhao et al., 2012). Studies reported that poor treatment adherence during anti-TB treatment results in the subsequent emergence of drug resistance (Zhao et al., 2012).
The other risk factor identified based on the pooled estimates in this study was having contact with a known TB patient whether with an MDR-TB patient specifically or with a TB patient as a general. Our study revealed that those individuals who had a contact history with a known TB patient had 1.72 times the odds to develop DR-TB compared to their counterparts. The risk becomes higher among individuals who had contact with an MDR-TB patient. Such that those who had a contact history with a known MDR-TB patient had 2.54 times the odds to develop DR-TB compared to individuals who did not have a contact history with a known MDR-TB patient. Studies from Burkina Faso and Bangladesh also supported it (Flora et al., 2013; Diande´ et al., 2009). Screening contacts could help to early detect DR-TB cases before disseminated across the population. Besides, the present study revealed that the site of infection was associated with DR-TB. Based on the pooled estimate in this study, individuals with pulmonary TB had 1.80 times the odds to develop DR-TB compared with individuals with extrapulmonary TB. A higher risk of DR-TB among PTB cases was also reported from India in a study based on a 13 years retrospective hospital-based analysis (Raveendran et al., 2015). Another study also supported this (Peto et al., 2009). Difficulties in detecting EPTB cases with lower bacterial loads might be the reason. The mycobacterial strains circulating in the country might be also a reason. For example, a study from China revealed that the DR-TB is higher among extrapulmonary TB cases compared to pulmonary TB cases due to the high prevalence of the Beijing strain (Pang et al., 2019).
The present study also revealed that older individuals had a lower risk to develop DR-TB. The pooled estimate revealed that individuals who were above 45 years had a 24% decreased risk of DR-TB compared to individuals below 45 years of age. Such that those productive age groups are at high risk of DR-TB in Ethiopia. Likewise, a pooled estimate in Jimma et al’s. (2017) study based on a systematic review and meta-analysis conducted in Iran and its neighboring countries revealed that individuals below 45 years of age had 1.57 times the odds to develop MDR-TB compared with those individuals above 45 years old. Higher pooled odds of MDR-TB among individuals below the age of 65 was also reported from a systematic review performed in Europe (Faustini et al., 2006). Besides, a study conducted at Northeastern China reported that those individuals between 28-54 years of age had double odds of MDR-TB when compared with those 65 years or older (Liu et al., 2013). The treatment adherence in the productive age group and the working style of these groups who have a higher contact chance to DR-TB patients might also contribute. The findings of the present study revealed that males had a lower risk of DR-TB compared to females. The risk of DR-TB decreases by 14% in males compared to females. Likewise, individual studies also reported a higher risk of DR-TB among females (Lomtadze et al., 2009; Shivekar et al., 2020). However, the reason behind it should be explored in detail in future works. In the end, this study was based on studies published in the English language that might affect the true estimates. Besides, publication bias was confirmed by the Egger’s regression test for about half of the study variables that might bias the true estimates.
In conclusion, a previous history of TB treatment is a major risk factor for acquiring DR-TB in Ethiopia that might be due to poor adherence during the first-line anti TB treatment. Besides, having contact with a known TB patient, having contact with a known MDR-TB patient, having pulmonary TB, and being unemployed were the risk factors of DR-TB in Ethiopia. Thus, active screening of TB contacts for DR-TB might help to detect DR-TB cases as early as possible and could help to decrease its transmission across the population.