Anemia as a risk factor for tuberculosis: a systematic review and meta-analysis

Background Tuberculosis is a major public health problem caused by Mycobacterium tuberculosis, occurring predominantly in population with low socioeconomic status. It is the second most common cause of death from infectious diseases. Tuberculosis becomes a double burden among anemic patients. Anemia increases an individual’s susceptibility to infectious diseases including tuberculosis by reducing the immunity level. Therefore, the purpose of this study was to determine whether anemia is a risk factor for tuberculosis. Method Relevant published articles were searched in electronic databases like PubMed, Google Scholar, EMBASE, and Cochrane Library using the following MeSH terms: risk factor, predictors, tuberculosis, TB, Anaemia, Anemia, hemoglobin, Hgb, and Hb. Articles written in the English, observational studies conducted on the incidence/prevalence of tuberculosis among anemic patients, or papers examined anemia as risk factors for tuberculosis were included. From those studies meeting eligibility criteria, the first author’s name, publication year, study area, sample size and age of participants, study design, and effect measure of anemia for tuberculosis were extracted. The data were entered using Microsoft Excel and exported to Stata version 11 for analysis. The random-effects model was applied to estimate the pooled OR and HR, and 95% CI. The sources of heterogeneity were tested by Cochrane I-squared statistics. The publication bias was assessed using Egger’s test statistics. Results A total of 17 articles with a 215,294 study participants were included in the analysis. The odd of tuberculosis among anemic patients was 3.56 (95% CI 2.53–5.01) times higher than non-anemic patients. The cohort studies showed that the HR of tuberculosis was 2.01 (95% CI 1.70–2.37) times higher among anemic patients than non-anemic patients. The hazard of tuberculosis also increased with anemia severity (HR 1.37 (95% CI 0.92–2.05), 2.08 (95% CI 1.14–3.79), and 2.66 (95% CI 1.71–4.13) for mild, moderate, and severe anemia, respectively). Conclusion According to the current systematic review and meta-analysis, we can conclude that anemia was a risk factor for tuberculosis. Therefore, anemia screening, early diagnose, and treatment should be provoked in the community to reduce the burden of tuberculosis.

to iron deficiency. Acute and chronic infections, including malaria, cancer, and HIV are also the cause of anemia [1,3].
Anemia is classified into three categories based on the severity; mild, moderate, and severe anemia. Mild anemia is defined as Hb concentration of 10.0-10.9 g/dl for pregnant women and 6-59-month children, 11.0-11.4 g/dl for 5-11-year-old children, 11.0-11.9 g/dl for non-pregnant women, 12-14-year-old children, and 11.0-12.9 g/dl for adult men. On the other hand, moderate anemia defined as the Hb value of 7.0-9.9 g/dl for pregnant women and 6-59-month children and 8.0-10.9 g/dl for 5-11-year-old children; 12-14-year-old children, non-pregnant women, and adult men while severe anemia is defined as the Hb value less than 7.0 g/dl for pregnant women and 6-59 months of children; and less than 8.0 g/dl for 5-11-yearold children, 12-14-year-old children, non-pregnant women, and adult men [2].
Anemia reduces health-related quality of life, increases morbidity and mortality in patients with chronic disease. It also predisposes an individual to some infectious diseases including tuberculosis (TB) [4]. It is a common hematological finding among TB patients with the prevalence of 44-89.1% [5][6][7][8][9]. On the other hand, the proportion of TB among anemic patients is higher than non-anemic patients (the highest burden in severe anemic patients) [10,11]. In anemic patients, cell-mediated immune response and bactericidal capacity of leucocytes are significantly suppressed [12,13].
Tuberculosis is an airborne chronic infectious disease caused by Mycobacterium tuberculosis and predominantly occurred in low socio-economical segments of population. It is the second most common cause of death among infectious diseases. A total of 8.7 million new active TB cases and 1.4 million TB-related deaths were estimated globally [14][15][16].
The assessment of potentially modifiable risk factors is a vital for the development of TB control policies [17]. Accordingly, some TB risk factors have been known for decades, including systemic diseases such as diabetes mellitus and chronic kidney disease as well as tobacco smoking, alcohol use, body mass index, silicosis, human immunodeficiency virus (HIV) infection, splenectomy, and gastrectomy [17,18]. Under-nutrition, refugee, homeless, and direct contact with active TB are also risk factors for TB [18]. People exposed to these factors are called the risk groups for TB in which the prevalence or incidence of TB is significantly higher than in the general population. The World Health Organization (WHO) recommended and established guidelines for these risk groups to be prioritized for screening of active TB than the general population [19].
However, even the diagnosis of anemia with Hb measurement is a low cost, and more widely available in clinical settings to know the anemic status of the individual [20], there is no any established guideline or policies to consider anemic patients as the risk group for TB and to be prioritized them for screening for active TB. But, studies have investigated the link between anemia and TB prevalence (the relationship between anemia and the risk of contracting TB). Accordingly, some studies showed that anemia is risk for TB [21][22][23]. In contrast, others showed that anemia is not the risk of development of TB [20,24]. These contradict findings and the absence of systemic review and meta-analysis conducted about the risk of contracting TB among anemic patients; motivate the authors to conduct this systemic review and meta-analysis. Thus, the main objective of the current systematic review and meta-analysis was to determine the pooled risk effect of anemia on the development TB.

Design and protocol registration
This systematic review and meta-analysis was designed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Protocols (PRISMA-P 2015 Guidelines) [25]. The protocol has been registered in the PROSPERO with the registration number of CRD42019161729.

Inclusion and exclusion criteria
Articles wrote only in the English language; observational studies like cohort studies, case-control studies, and cross-sectional studies; conducted on the incidence/ prevalence of TB among anemic patients (any type of anemia) or papers assessed anemia as risk factors for TB, human studies, and articles published until December 17, 2019; and available on the searched databases were included. We excluded studies that did not report crude or adjusted HR, crude or adjusted RR, crude or adjusted OR, or unable to calculate these effect measures from 2 by 2 tables.

Searching strategy and research screening
The articles for this systematic review and meta-analysis were retrieved through reproducible and comprehensive electronic searching of major reputable databases (PubMed, EMBASE, and Cochrane Library) and Web search (Google Scholar and Science direct) using the following MeSH terms: tuberculosis, TB, Anaemia, Anemia, Hemoglobin, and Hgb, Hb, risk factor, and predictors. We also did a manual searching of reference lists of already identified relevant articles to retrieve more eligible studies by using the combination of these keywords. The two authors (YG, MM) performed the search independently based on the following key terms: [1] population (anemia, anaemia, hemoglobin, Hgb Hb), [2] outcome (tuberculosis, TB), [3] study design (prevalence, incidence, cross-sectional, observational, cohort study), and [4] location (worldwide). The terms were used both separately and in combination with the help of the Boolean operator like "AND", "OR". We looked for these terms in the abstract, title, or keywords. Finally, the search results were imported to an endnote to find duplicates. Titles and abstracts were examined by 3 independent reviewers (YG, ZG, and MM). Besides, articles referenced by those identifiers were reviewed for relevance.

Study selection and quality appraisal
The quality appraisal of full-length original research articles was assessed in detail by two independent reviewers (ZG and MM) by using the Joanna Brigg Institute (JBI) criteria [26]. The JBI checklist items related to design, setting, participants, and confounders, bias, statistical analysis, outcome measures, results, and generalizability of the study were checked. The independent reviewers sat together and settled the differences in rating through consensus, before the final decision to include or exclude the articles. The scoring system is as follows: 0 (not done), 1 (done), UC (unclear), and NA (not applicable). The score range for this tool is between 0 (lowest quality) to 8 (highest quality) for crosssectional, 0 (lowest quality) to 10 (highest quality) for case-control studies, and 0 (lowest quality) to 11 (highest quality) cohort studies. Articles with average score of 50% and above were included into this study.

Data extraction
For each study, meeting eligibility criteria, the first author's name, publication year, study area/country, sample size and age of participants as adult and children, study design, and results like crude or adjust HR or crude or adjust OR were extracted by YG using Microsoft Excel. The anemia severity status was also extracted according to the individual study report ( Table 1). The logarithm of HR (log HR) and standard error (Se log HR), or logarithm of OR (log OR), and standard error (Se log OR) were calculated from their corresponding effect measure.

Data analysis and interpretation
The extracted data were entered into Microsoft Excel and export to Stata version 11 for analyzing. The data analysis was by two authors (YG and MM). The potential source of heterogeneity across studies was tested by Cochrane I 2 test statistics which shows the amount of heterogeneity between studies. The I 2 provides the percentage of variability due to heterogeneity rather than chance difference or sampling error. I 2 > 50% was considered statistically significant heterogeneity. The random-effects model which assesses the variability within and between studies was applied to estimate the pooled OR and HR and 95% confidence intervals (CIs). The publication bias was assessed using Egger's test statistics with p value < 0.05 considered as the presence of publication bias. A sensitivity test was done to give a quick indication which study is the prime determinant of the pooled effect size and which is the main source of heterogeneity. The test excludes each study one by one in the analysis to show the pooled effect sizes and associated heterogeneity. Subgroup analysis was done by study design for crosssectional and case-control studies.

Description of studies
Our initial search yielded 1272 articles. After 13 duplicates were removed, 1233 articles were excluded because of not relevant title, and 6 articles were excluded after reading the abstract. Finally, 20 articles identified for further assessment. Two studies that did not report the effect measures (HR, RR, OR) or unable to calculate these measures [36,37], and one study conducted on latent TB [38] were also excluded. Finally, 17 studies were used for Meta-analysis ( Fig. 1). Eight of the research articles included for analysis contained information on the association of anemia severity and TB development (5 cohort studies, 2 case-control studies, and 1 cross-sectional study). As a result, these were treated as two or three separate studies considering the anemia severity as a separate independent variable.

Publication bias
The included studies were assessed for publication bias based on their pooled analysis. The subjective   (Tables 2 and 3).

Sensitivity and heterogeneity test
The overall heterogeneity of I-squared statistics showed substantial heterogeneity for studies pooled by both OR and HR (I-squared = 96.1% and 93.4%, respectively) ( Figs. 4 and 7). Therefore, the subgroup analysis was performed for pooled studies showing high heterogeneity. Studies pooled by OR were sub-analyzed by their study design and anemia severity status (Figs. 5 and 6). However, there was still a substantial heterogeneity except for sub-group analysis by anemia severity which showed that heterogeneity was not observed in studies in which anemia severity was not defined (Fig.  6). On the other hand, cohort studies pooled by HR were sub-analyzed by their participants' age and anemia severity status. The result showed that heterogeneity was not observed in studies conducted children and in studies where anemia severity was mild and moderate (Figs. 8 and 9). Sensitivity and heterogeneity test of the included studies was also done to test the effect of each study on the pooled effect size by excluding each study step by step. But the sensitivity results showed that no study was the prime determinants of the pooled effect size (all studies had nearly equal contribution for the pooled estimate). The heterogeneity test also showed that a single study was not the source of heterogeneity (there were no significance reduction of heterogeneity by omitting each study step by step) (Tables 4 and 5).
Pooled risk estimate of anemia for TB: cross-sectional and case-control studies According to the pooled OR of cross-sectional and casecontrol studies, the odds of TB infection among anemic patients is 3.56 times higher than non-anemic patients (95% CI 2.5-5.0). The sub-group analysis of these studies by anemia severity status also revealed that the odds of contracting TB was increased in severely anemic patients. The odd of TB among mild anemic patients is 2.33 times higher than non-anemic patients, while in moderately and severely anemic patients, it was 3.65 and 3.91 times higher than non-anemic patients, respectively (Fig. 6).  Pooled risk estimate of anemia for TB based on cohort studies Ten cohort studies assessed the risk of anemia for TB by using HR effected size and analyzed these studies together to get the pooled effect size of anemia on TB by using HR. Accordingly, their pooled HR showed that the hazared of TB among anemic patients was 2.01 times higher than nonanemic patients (95% CI 1.70-2.37) (Fig. 7). The sub-analysis of the cohort studies by age group revealed that there was no significant change of risk of anemia for TB between anemic children and anemic adults (Fig. 8). However, the sub-analysis of these cohort studies showed that the hazared of TB was increased with anemia severity (Fig. 9). The hazard of TB among mild anemic patients was not statistically significance compared to non-anemic patients (HR 1.4, 95% CI 0.9-1.9). However, the hazard of TB among moderate anemic patients was 2.08 (95% CI 1.14-3.79) times higher than non-anemic patients, whereas in severe anemic patients, it was 2.66 (95% CI 1.71-4.13) times higher compared to non-anemic patients.

Discussion
To the best of our information, this is the first systematic review and meta-analysis conducted to determine the pooled risk factor of anemia for TB. Anemia defined by low Hb or red blood cell (RBC) concentration is a major hematological finding in chronic diseases [39,40]. It is also a known risk factor for some chronic diseases [41]. In the current systematic review and meta-analysis, majority of the included studies showed that anemia was a higher predicator of TB [10, 11, 17, 22, 27-29, 30, 31]. However, some studies showed that anemia was not the risk of TB [20,24,33]. The reason for discrepancy might be different in sample size, study design, diagnostic method, and geographical location. But, according to the current systematic review and meta-analysis pooled effect size estimate, anemia was the risk factor of TB. The pooled effect of cross-sectional and case-control studies showed that the odds of TB among anemic patients were 3.56 times higher than non-anemic patients. Indeed, this may not show the cause and effect association due to the limitation of study design (is anemia the cause of TB infection? or is TB disease the cause of anemia?).
However, pooled analysis of cohort studies, which show the cause and effect association, also revealed that anemia was a risk for TB. According to the pooled effect estimate of 10 cohort studies analyzed by HR, the hazard of TB among anemic patients was 2.01 times higher than non-anemic patients. This might be due to that anemic patients might be nutritionally imbalanced and immuno-compromised. Anemia was used as indirect assessment of nutritional and immune status of the individuals. Most of the included studies did not report the type of anemia. However, the WHO reported that 50% of the cases of anemia are due to iron deficiency [42] which The black dot at the center of the gray box represents the point OR estimate of each study, and the line indicates the 95% confidence interval of the estimates. The gray boxes represent the weight of each study contributing to the pooled OR estimate. The first five blue diamonds represent the 95% confidence interval for subgroup analysis pooled OR estimate, and the last blue diamond represents the 95% confidence interval for overall pooled OR estimate. OR odds ratio, CI confidence interval, I 2 shows the heterogeneity of the included studies, p indicates the statistical significance of the heterogeneity might be true for the included studies. Successful transmission of TB is influenced by a variety of conditions, including proximity and duration of contact with an individual with active TB disease, and the immune-competency of the individual infected with tuberculosis [43]. Individuals with a weak immune response (immune compromise individuals) are at risk of TB [44]. Iron was confirmed to be a vital element not only for erythropoiesis, but also for immune system development and play an important role in the integrity of the immune system; and its deficiency can cause impairment of immunity. Ekiz et al. suggests that the important immunogenic mechanisms like humoral, cell-mediated, and nonspecific immunity and the activity of cytokines are influenced by iron deficiency anemia. Especially, the percentage of monocytes with oxidative burst activity and the ratio of monocytes with phagocytic activity were highly reduced in iron deficiency anemia [13]. Macrophage phagocytic activities are important immunological response in controlling of TB infection by forming granulomas which is an aggregate of immune cells and walls of the mycobacterium which limiting further replication and spread of the tubercle bacilli [44,45]. Aly et al. and Das et al. also showed that iron deficiency anemia impaired cell-mediated immune response specifically T cell-mediated immunity [46,47]. A review by Stephen stated that even all studies did not show a consistence result, there were an impairment of polymorph neutrophil function and intracellular bacteriocidal activity of immunological cells in iron deficient individuals [1]. Iron status may also modulate the type of immune response mounted through its influence on the body's cytokine profile. Experimental evidence has shown that iron deficiency changes the balance between Th1 and Th2 cytokines, promoting a dominant Th2 response that has been associated with clinical TB disease [48]. Generally, morbidity from infectious disease is increased in iron-deficient populations, because of the adverse effect of iron deficiency on the immune system [49,50]. Therefore, the anemic patients probably might have impaired or modulated immune system that favors the replication of TB.
The other cause of anemia are the presence of other micronutrient deficiencies, including vitamins A and B12, folate, riboflavin, and copper [42]. These micronutrient deficiencies can also cause immunological impairment [51]. According to Erkurt et al, vitamin B12 has important immunomodulatory effects on cellular immunity, and abnormalities in the immune system in pernicious anemia are restored by vitamin B12 replacement therapy [12].
The other possible explanation might be the direct involvement of RBCs in maintaining of the innate and adaptive immune system. Evidence shows that RBCs are modulators of T cell proliferation. In particular, RBCs are able to enhance T cell expansion and survival by inhibiting activation-induced T cell death, an effect possibly associated with a decrease of oxidative stress within activated T cells. Optimal T cell proliferation and survival were only observed with intact RBCs and when RBCs were in close contact or proximity with activated T cells [52].
In the current systematic and meta-analysis, the burden of TB was increased with anemia severity. According to the pooled effect estimate of casecontrol and cross-sectional studies, the odds of TB among mild anemic patients is 2.33 times higher than non-anemic patients, while in moderately and severely anemic patients, it was 3.65 and 6.91 times higher than non-anemic patients, respectively. The included cohort studies also revealed that the  [53]. This justifies the finding, the higher incidence of TB among anemic patients than non-anemic patients, and why it was increased in moderate and severe anemic patients compared to mild anemic patients. In anemic patients, there are low RBCs (low concentration of Hb), especially in moderate and severe anemic patients. Therefore, anemic patients might have disturbed immune system and low bacteriocidal activity due to low Hb concentration, which enhances the growth of TB. Generally, anemia is a risk factor for TB; this is because anemic patients might be immuno-suppressed and susceptible to TB, and it is known that TB is The black dot at the center of the gray box represents the point HR estimate of each study, and the line indicates the 95% confidence interval of the estimates. The gray box represents the weight of each study contributing to the pooled HR estimate. The blue diamond represents the 95% confidence interval of the pooled HR estimate. HR hazard ratio, CI confidence interval, I 2 shows the heterogeneity of the included studies, p indicates the statistical significance of the heterogeneity, + anemia severity not defined, * mild anemia, @ mild and moderate anemia, # moderate anemia, $ severe anemia common in immuno-compromised population patients [54].

Strength and limitation of the study
One of the strengths of this review was being the first systematic review and meta-analysis to determine the pooled risk estimate of anemia for TB. Moreover, the review was conducted according to the preferred reporting items for systematic review and meta-analysis (PRISMA-P statement) protocol. However, this review had limitations. The extent of heterogeneity between included studies was high. We were unable to get adjusted risk ratio from some of the included studies. Therefore, we used crude risk measures from these articles. The other limitation of this systematic review and meta-analysis was, only articles published in English were used for our literature search. Most of the included studies were conducted in Africa which might cause geographical bias.

Conclusion
Anemia is a major public health concerns that may predispose an individual to some infectious diseases including TB. According to the current systematic and meta-analysis, anemia was the risk factor of TB and the risk was increased with anemia severity. Diagnosis of anemia with Hb measurement is a low cost and more widely available in clinical settings to know the anemic status of the individual. Anemia screening, The black dot at the center of the gray box represents the point HR estimate of each study, and the line indicates the 95% confidence interval of the estimates. The gray boxes represent the weight of each study contributing to the pooled HR estimate. The first two blue diamonds represent the 95% confidence interval for subgroup analysis pooled HR estimate, and the last blue diamond represents the 95% confidence interval for overall pooled HR estimate. HR hazard ratio, CI confidence interval, I 2 shows the heterogeneity of the included studies, p indicates the statistical significance of the heterogeneity, + anemia severity not defined, * mild anemia, @ mild and moderate anemia, # moderate anemia, $ severe anemia early diagnose and treatment may reduce the magnitude TB at the community level. Therefore, health professionals should treat anemia as early as possible and policy makers should consider anemic patients as risk for TB and established screening guideline.  The black dot at the center of the gray box represents the point HR estimate of each study, and the line indicates the 95% confidence interval of the estimates. The gray boxes represent the weight of each study contributing to the pooled HR estimate. The first five blue diamonds represent the 95% confidence interval for subgroup analysis pooled HR estimate, and the last blue diamond represents the 95% confidence interval for overall pooled HR estimate. HR hazard ratio, CI confidence interval, I-squared shows the heterogeneity of the included studies, p indicates the statistical significance of the heterogeneity