Co-occurrence of tuberculosis and diabetes mellitus, and associated risk factors, in Ethiopia: a systematic review and meta-analysis

Highlights • DM patients have a high TB risk, and TB patients can develop DM during their treatment.• Around 4.14% of DM patients in Ethiopia have TB.• Around 12.7% of TB patients in Ethiopia develop DM.• Type 1 DM patients have a higher TB risk.• Old age and a family history of DM are risk factors for DM in TB patients.


Introduction
Even though one quarter of the global population is estimated to have been infected with Mtb, only 5-10% have a lifetime risk of falling ill with tuberculosis (TB). The risk of TB is higher among immunecompromised individuals and those with certain underlying diseases ( Walker et al., 2013 ). Those with compromised immune systems, such as people living with HIV, malnutrition, or diabetes (DM), or smokers, have a higher risk of falling ill with TB ( WHO, 2020 ). Many new cases of TB are attributable to five risk factors: undernutrition, HIV infection, alcohol use, smoking, and DM ( WHO, 2020 ). In 2019, an estimated 2.2, 0.76, 0.72, 0.70, and 0.35 million TB cases were attributable to undernutrition, HIV infection, alcohol use disorders, smoking, and DM, respectively ( WHO, 2020 ).
Aside from the traditional risk factors, DM is increasingly being recognized as an independent risk factor for TB, and the two often coexist ( Yorke et al., 2017 ). The interaction between DM and TB is a major Abbreviations: BMI: body mass index, DM: diabetes mellitus, EPTB: extrapulmonary tuberculosis, HIV: human immunodeficiency virus, IFG: impaired fasting glucose, MDR-TB: multidrug-resistant tuberculosis, PTB: pulmonary tuberculosis; PRISMA: Preferred Reporting Items for Systematic Reviews and Meta-analyses, OR: odds ratio, TB: tuberculosis, WHO: World Health Organization. Figure 1. Flowchart describing the selection of studies for the systematic review and meta-analysis of TB and DM co-occurrences in Ethiopia ratio of 17% (9-25%) ( WHO, 2019 ). Moreover, the burden of noncommunicable diseases, including DM, is increasing in Ethiopia, with 3.2% in adults ( IDF, 2019 ), ranging from 2.0% to 6.5% ( Bishua et al., 2019 ).
Diabetic patients are susceptible to TB. Individual studies conducted in Ethiopia have also confirmed this. A TB prevalence among DM patients of more than 5.0% has been reported in Ethiopia ( Gedfew et al., 2020 ;Abera & Ameya, 2018 ). Studies have also reported that DM is a common phenomenon among TB patients in Ethiopia, with a prevalence of up to 15.8% ( Damtew et al., 2014 ). However, the findings of individual studies are inconclusive, and there is a scarcity of updated data on the status of TB and DM co-occurrences in the country. Thus, our study aimed to assess the burden of TB and DM co-occurrences, and associated risk factors, in Ethiopia.

Search strategy
Systematic article searching was conducted using electronic databases (PubMed, CINAHL, DOAJ, African Index Medicus) and other gray literature sources (Google, Google Scholar, WorldCat). This was performed independently by two authors (AA and GD) under the consultation of a senior librarian at the Ethiopian Public Health Institute. Inconsistencies were resolved by a third author (ZWB). The keywords used for searching included tuberculosis, diabetes, co-occurrences, risk factors, associated factors, determinants, predictors, and Ethiopia. These keywords were used in conjunction with the Boolean operators AND and OR. The search string for the PubMed database was (((( "Tuberculosis " [Mesh]) OR (TB)) AND ( "Diabetes Mellitus " [Mesh] OR "Diabetes Mel-litus, Type 1 " [Mesh])) OR (DM)) AND ( "Ethiopia " [Mesh]) (Additional file 1).

Study selection procedure
This study was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) reporting checklist ( Hutton et al., 2015 ;Knobloch et al., 2011 ) (Additional file 2). A stepwise approach was followed to select the eligible articles included in the final analysis. The article selection procedure was conducted independently by two authors (AA and GD) using predefined inclusion criteria, and the inconsistencies were resolved by a third author (ZWB). Initially, all the identified articles ( n = 392) were exported to the EndNote X8 citation manager, and 100 duplicates were removed. Next, 292 articles were screened by title and abstract. Around 18 articles that passed the first stage were assessed through a full-text review. During this review, the study subjects, study design, study quality, and outcome were considered. Finally, 14 articles became eligible for data extraction. During article eligibility assessment, the PICOS (Population, Intervention, Comparison, Outcome, Study design, Study setting) criteria were assessed ( Figure 1 ).

Inclusion and exclusion criteria
Articles that reported TB prevalence among DM patients, or DM cooccurrence among TB patients, or articles that reported associated factors for TB and DM co-occurrences in Ethiopia, and that were published in the English language were included. Articles without full text, or that did not separately report the prevalence for each group, or that were not objectively designed to assess TB prevalence among DM patients or DM prevalence among TB patients were excluded.

Data extraction
Data were extracted independently by two authors (AA and GD), and the inconsistencies were resolved by a third author (ZWB). The extracted data included study characteristics, such as author, publication year, regional state, study setting, data collection period, study participants' age range, sample size, number of DM patients with TB, and number of TB patients with DM. Demographic and behavioral data, and clinical factors for TB and DM co-occurrences, were also extracted. The data were summarized using Microsoft Excel 2016 spreadsheets ( Tables 1 and 2 ).

Risk of bias (quality) assessment of studies
The quality of individual studies was assessed using the Joanna Briggs Institute (JBI) critical appraisal checklist for prevalence studies and cohort studies ( Moola et al., 2020 ). The questions in the checklist were equally scored and then totalled to give a final score out of 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 ) (Additional File 3). Two authors (AA and GD) independently assessed the quality of the studies and the inconsistencies were resolved by a third author (ZWB). The symmetry of the funnel plots was assessed visually, and Egger's regression test was performed to assess publication bias.

Outcomes
The presence of TB and DM co-occurrences was the primary outcome. This was assessed by estimating the pooled prevalence of TB among DM patients and the pooled prevalence of DM among TB patients. The associated factors contributing to the presence of TB and DM co-occurrences formed the secondary outcome. The pooled odds ratio for each factor was estimated.

Operational definitions
For this study, diabetes was operationalized as having a fasting blood glucose level of 126 mg/dl or more, while prediabetes/impaired fasting blood glucose was operationalized as having a fasting blood glucose level of 100-125 mg/dl. Older age was taken as the higher age group in the primary studies; in the majority of the studies this was over 45 years.

Data synthesis and statistical analysis
The data summarized in Microsoft Excel 2016 were exported to STATA version 2016 for statistical analysis. The pooled proportion of DM patients infected with TB and the pooled proportion of TB patients who had DM were estimated, along with 95% CIs. To assess the associated factors for TB and DM co-occurrences, the pooled ORs along with 95% CIs were estimated. The pooled estimates were presented as forest plots. The forest plots and I 2 heterogeneity tests were examined to assess heterogeneity among the studies. I 2 values of 25%, 50%, and 75% were interpreted as the presence of the low, medium, and high heterogeneity, respectively ( Sterne & Egger, 2001 ;Riley et al., 2011 ). In addition, the presence of publication bias was assessed through visual inspection Table 1 Characteristics of the individual studies on tuberculosis infection among diabetes patients in Ethiopia included in the current systematic review and meta-analysis    of the funnel plots and Egger's regression test. Asymmetry of the funnel plots and statistical significance of Egger's regression test ( p < 0.05) were considered to represent the presence of publication bias.

Risk factors for tuberculosis and diabetes mellitus co-occurrences
The associated risk factors for developing TB among DM patients and DM co-occurrence among TB patients were estimated using the available studies conducted so far in Ethiopia. The pooled OR was estimated for the 13 variables associated with developing TB among DM patients, while the pooled OR was estimated for the 12 variables associated with DM co-occurrence among TB patients. These variables were chosen based on their appearance in the primary studies. The variables included socio-demographic, behavioral, and clinical characteristics.

Discussion
This systematic review and meta-analysis study assessed the burden of TB and DM co-occurrences, and associated factors, in Ethiopia. Based on data extracted from the seven available studies, the pooled prevalence of TB among DM patients was estimated as 4.14% (95% CI 2.45-5.83%), while the pooled prevalence of DM among TB patients was estimated as 12.77% (95% CI 6.91-18.62%). Our study also revealed that type of DM was associated with developing TB among DM patients, while older TB patients (OR 2.25, 95% CI 1.38-3.13) and TB patients who had a family history of DM (OR 3.65, 95% CI 1.89-5.41) had a higher risk of developing DM compared with their counterparts.
Our study revealed that around 4.14% (4140 per 100 000 population) of DM patients in Ethiopia had TB. This was higher than the estimated national TB prevalence among the general population (140/100 000 population) ( WHO, 2020 ). In a study carried out by Wagnew et al. (2018) , using studies conducted in African and Asian countries, the pooled prevalence of TB among DM patients was esti-    ( Wagnew et al., 2018 ). Workneh et al., in their systematic review, revealed that the median overall global prevalence of TB among DM patients was 4.1% ( Workneh et al., 2017 ). Another global pooled estimate revealed that DM patients had a two-to-four-fold increased risk of TB ( Al-Rifai et al., 2017 ). Our study also estimated the pooled prevalence of TB based on DM type. The findings revealed an 8.56% pooled TB prevalence among individuals who had type 1 DM, compared with 2.80% for individuals with type 2 DM. The pooled OR revealed that those individuals with type 1 DM had 2.70 times the odds for developing TB compared with type 2 DM patients. Likewise, a higher prevalence of TB among children and adolescents with type 1 DM was reported from South Africa ( Webb et al., 2009 ), while a 10.0% prevalence of culture-positive PTB among type 1 DM patients was reported from India ( Nair et al., 2016 ). The higher risk of TB in patients with type 1 DM might be due to the longer duration and difficulties in controlling hyperglycemia in this group, in addition  Our study also assessed the prevalence of DM among TB patients. According to data collected from 3283 TB patients, 407 developed DM, giving a pooled prevalence estimate of 12.77%. This was lower than a global pooled estimate of 15.3% reported by Noubiap et al. ( Noubiap et al., 2019 ). Likewise, in a study conducted by Workneh et al., the global pooled median prevalence of DM among TB patients was estimated as 16%. This might have been due to the high prevalence of DM among the general population in the countries included in the above studies. This was supported by a study conducted in South Asia, where the pooled prevalence of DM among TB patients was estimated as 21% ( Gautam et al., 2021 ). However, in a study conducted by Alebel et al., the pooled estimate of DM prevalence among TB patients in sub-Saharan Africa was estimated as 9.0%, which is lower than that found by our study ( Alebel et al., 2019 ).
Our study estimated the pooled prevalence of prediabetes among TB patients to be 24.19%. Although there have been no previously reported pooled estimates of the prevalence of prediabetes among TB patients, it has been commonly reported in individual studies. A comparable finding of 24.5% has been reported in India ( Viswanathan et al., 2012 ). Higher prevalences have been reported in Vietnam (29%; Hoa et al., 2018 ) and Kenya (37.5%; Owiti et al., 2017 ), with a lower prevalence reported in Eritrea (10%; Araia et al., 2021 ). Generally, our study and previous studies conducted in different countries have revealed that TB patients are at high risk of developing DM, which necessities early DM detection among TB patients.
Of the 12 variables assessed using pooled OR estimates, two -older age and family DM history -were found to have a statistically significant association with DM in TB patients. Older TB patients had 2.25 times the odds for developing DM compared with their counterparts. A link between DM and older age was also reported in an earlier systematic review ( Workneh et al., 2017 ). Likewise, a global meta-analysis study revealed that when the age of TB patients increased, the prevalence of DM also increased ( Noubiap et al., 2019 ), while a nationwide cohort study in Portugual revealed that the odds for DM among TB patients increased by 4.7% per year of age ( da Costa et al., 2016 ). The elderly are at risk of developing DM -mainly type 2 -due to decreased insulin secretion and impaired pancreatic islet functioning. A lack of physical activity and modern lifestyles, especially with regard to food choice, are the main factors resulting in obesity and consequent development of DM ( Kirkman et al., 2012 ).
This study also revealed that TB patients with a family history of DM had 3.65 times the odds for developing DM co-occurrence compared with TB patients with no family DM history. This finding has been supported by a global systematic review ( Workneh et al., 2017 ), while an individual study conducted in Tanzania revealed that TB patients who had a family history of DM had up to 17.5 times the odds for developing DM compared with their counterparts ( Mabula et al., 2021 ). Other studies have also shown a family history of DM to be a major risk factor in developing DM in the general population ( Zuo et al., 2014 ;Ismail et al., 2021 ). Important limitations should be considered when interpreting the results of this study. First, the study was based on primary studies conducted only in the English language, which might have introduced bias. In addition, the majority of the studies were cross-sectional; this might have a limited the assessment of risk factors for co-occurrence. Furthermore, the small number of available primary studies might also have introduced bias. Finally, all the studies were hospital-based, which might not have reflected the nature of co-occurrences in the general population.

Conclusion
Co-occurrence of TB and DM is a major public health problem in Ethiopia. The prevalence of TB among DM patients (4140/100 000 population) estimated in our study was far higher than the national TB prevalence (140/100 000 population) among the general population. Individuals with type 1 DM had a higher TB risk (8560/100 000 population) compared with those individuals with type 2 DM (2800/100 000 population). The estimated DM prevalence among TB patients (12.77%) was far beyond the national DM prevalence among adults (3.2%). This suggests the need for an integrative approach to decreasing the dual burden. Elderly TB patients with a family history of DM had a higher risk of developing DM; this needs to be considered during anti-TB treatment follow-up. Thus, active screening of DM patients for TB, and vice versa, is recommended.

Author contributions
AA conceptualized, designed, and drafted the manuscript. AA, GD, and ZWB performed article searching, data extraction, and quality assessment. AA and ZWB conducted data analysis and wrote the manuscript. BG reviewed the final manuscript. All authors read, reviewed, and approved the final manuscript.