Global prevalence of Mycobacterium bovis infections among human tuberculosis cases: Systematic review and meta-analysis

Abstract

Tuberculosis(TB) is a chronic communicable bacterial disease caused by Mycobacterium tuberculosis complex (MTBC) species. M. tuberculosis is the main causative agent of human TB, and cattle are the primary host of Mycobacterium bovis; due to close interaction between cattle and humans, M. bovis poses a zoonotic risk. This review summarizes and estimates the prevalence of M. bovis infection among human cases. Studies reporting TB prevalence data that were published in English during 10 years from 20 April 2009 to 17 April 2019 were identified through search of PubMed and other sources. Quality of studies and risk of bias were assessed using standard tools for prevalence study reports. Characteristics of included studies and their main findings were summarized in tables and discussed with narrative syntheses. Meta-analysis was performed on 19 included studies, with a total of 7,185 MTBC isolates identified; 702 (9.7%) of them were characterized as of subspecies M. bovis, but there was a large prevalence difference between the studies, ranging from 0.4% to 76.7%. The genotyping-based studies reported significantly lower prevalence of zoonotic TB than did the studies based on older techniques. The overall pooled prevalence of M. bovis aggregated from all included studies was 12.1% of the total MTBC isolates, while the corresponding pooled figure from the 14 genotyping-based studies was only 1.4%. Generally, human M. bovis cases reported from different countries of the world suggest that the impact of zoonotic TB is still important in all regions. However, it was difficult to understand the true picture of the disease prevalence because of methodological differences. Future investigations on zoonotic TB should carefully consider these differences when evaluating prevalence results.

1 |. INTRODUCTION

Limited disease management and uncontrolled movement of infected cattle are factors contributing to the ongoing transmission of Mycobacterium bovis and that probably increases risk of zoonotic TB (Ameni et al., 2013; Torres-Gonzalez et al., 2013). However, human TB due to M. bovis is not common in countries where bovine TB in cattle is controlled through national eradication strategies such as test-and-slaughter of infected animals (Michel et al., 2010; Radunz, 2006). It was projected that the current epidemiological condition of bovine TB in most developing countries could be very similar to that of Europe in the 1930s, during the pre-pasteurization era (Ayele et al., 2004). Particularly, some population groups who have close contact with infected animals could be at higher risk than the general population (Alemayehu et al., 2008; Gumi et al., 2012; de la Rua-Domenech, 2006). Most of those studies were conducted in resource-limited countries, however, relied on comparative observational research findings using tuberculin skin tests in both humans and animals, and these procedures could not determine the specific causative agent.

Not only in resource-limited countries, until the end of the 20th century, investigations of zoonotic TB in many of the developed countries were mainly based on phenotypic characteristics of the species such as colony morphology and use of biochemical tests (Sreevatsan et al., 1996). However, over the last three decades, many nucleic acid-based methods have been developed to differentiate strains within the Mycobacterium tuberculosis complex (MTBC), such as restriction fragment length polymorphism (RFLP) analysis, DNA sequencing of target genes and polymorphic DNA analysis (Cousins et al., 1991; Lee et al., 2000; Ramos et al., 2014; Watterson et al., 1998), as well as several PCR-based methods, such as spoligotyping and mycobacterial interspersed repetitive unit-variable number tandem repeats (MIRU-VNTR), which have been extensively used in parallel with commercially available test systems (Bidovec-Stojkovic et al., 2011; Christianson et al., 2010; Mathuria et al., 2008; Shi et al., 2018).

Applications of these advanced molecular diagnostic procedures have been improving the quality of scientific evidence and addressing research gaps in the area of zoonotic TB (Ramos et al., 2014). Indeed, genotypic variations of MTBC strains, as well as lack of in-depth knowledge on host-pathogen co-evolutionary relationships, could hide the fundamental information and evidence needed to design effective preventive interventions (Legesse et al., 2011).

Although diagnosis and treatment of every person with TB is one of the key activities of the End TB Strategy, it has recently been recognized that people at risk of zoonotic TB have been neglected. To shift the existing paradigm, there is a need to have updated and credible scientific evidence. Accordingly, the recently published document ‘Roadmap for zoonotic tuberculosis’ announced that conducting systematic surveys and reporting high-quality data on the incidence of zoonotic TB is one of the priority areas (Dean et al., 2018). Therefore, this review and meta-analysis study aimed to summarize recent evidence and used to estimate the global prevalence of human M. bovis cases reported from epidemiological studies published in the last ten years.

2 |. MATERIALS AND METHODS

2.1 |. Protocol development and review approach

This review and meta-analysis study has been conducted in accordance with published protocol (ID = CRD42017076409) (Hawult Taye et al., 2017). The overall review approach was done based on condition-context-population (CoCoPop) review method. Each section of the review was done and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guideline (Moher et al., 2009).

2.2 |. Eligibility criteria

Inclusion and selection criteria were defined based on the CoCoPop components of the review method. The prevalence of M. bovis that was identified from human TB cases was deemed the relevant disease condition. Initially, the protocol was restricted to include studies conducted in developing countries only. However, as there were few molecular studies, we reformulated the review question and searching strategies to identify published articles reported from any country in the world. Therefore, all cross-sectional studies conducted globally through either a prospective or retrospective data collection method that reported the relevant disease condition were eligible to meet the context of the review. The source population of the relevant disease condition, the third component of the review method, was set as all clinical isolates of the MTBC species identified through either genotyping or phenotyping techniques.

Therefore, studies that targeted identification of MTBC species from human TB cases with specific aim of reporting the prevalence of M. bovis were deemed the relevant inclusion criteria. Hence, regardless of their socio-demographic and behavioural risk factors, all human population groups were considered. Species identification techniques based on either biochemical tests (phenotyping) and/or molecular methods (genotyping) were the main preconditions. The latter included molecular techniques such as mycobacterial interspersed repetitive unit-variable number of tandem repeats (MIRU-VNTR), spoligotyping and/or any PCR-based amplification methods using specific primers.

2.3 |. Sources and Search strategy

To increase the sensitivity of the searching strategy, all observational studies published in English from 20 April 2009 up to 17 April 2019 were identified through searches of PubMed, Google scholar, CINAHL and EMBASE databases. To increase sensitivity of each search, all Medical Subject Heading (MeSH) terms and the respective alternative words and/or phrases that could explain the condition, population and the method were listed and constructed within a gird table. Then, terms within the same concepts were connected with Boolean ‘OR’ and combined with other components of searching terms using Boolean ‘AND’. Through that, the final searching term (Table S1) was built at PubMed database, which is the primary searching engine. All identified articles from the different databases were imported to the Endnote reference manager software (sofrev no. SR15).

2.4 |. Screening and study selection

After removal of duplicates, three levels of screening based on Title, Abstract and Full-text review were performed. Two independent reviewers (HT and SB) performed full-text review and excluded articles based on specific requirements, which are listed in the PRISMA flow chart (Figure 1). Although molecular species identification technique was considered as a major precondition, only methods that can identify and differentiate known MTBC members were included. Particularly, studies that did not target identification of M. bovis from human samples or population-based comparative studies that estimated prevalence of zoonotic TB based on intradermal tuberculin skin test (TST) were excluded. Thus, all included studies should have at least one specific objective aiming on reporting the epidemiology of MTBC species. Whereas molecular studies that used to identify and differentiate M. bovis isolates for other purposes such as investigating drug resistance or laboratory-based biomedical research activities were excluded.

FIGURE 1.

PRISMA flow diagram shows the searching strategy and screening of eligible studies at different levels of the review process

2.5 |. Data extraction

Two of the authors (HT and SB) independently performed data extraction using piloted and structured data extraction form. Before collecting the relevant information, the full text of selected studies was further reviewed; and then six (31.57%) of the total 19 eligible studies were randomly selected and used for piloting the data collection tool. As most of the studies have no categorical data related to gender and age of the study participants, the two variables were removed from extraction tool. Finally, the name of authors, year of publication, country where primary studies were conducted, site of tuberculosis infection, study populations and sources of human sample, molecular techniques applied to identify the etiologic agent, total numbers of identified MTBC and rate of M. bovis isolates, and authors’ conclusion or notes related to the review question were collected and summarized in Table 1a,b. Consistencies of extracted information obtained from the two authors were compared. There was no discrepancy, except minor differences related to notes listed in the last column of Table 1a,1b, which presented. That was managed through a collaborative decision and agreed on the most relevant messages that further explained in result and discussion section.

TABLE 1.

Characteristics of included studies (a) that used genotyping techniques and (b) studies which primarily focused on conventional dentification techniques to report the prevalence Mycobacterium bovis identified from the total detected MTBC species

(a)
Published study	Country	Site of infection	Sample source	Review-related study objectives
Bayraktar et al., 2011	Turkey	PTB & EPTB	Archived samples	To determine the distribution of MTBC species
Belay et al., 2014	Ethiopia	PTB	Direct from patients	Investigating molecular Epidemiology of MTBC species
Ereqat et al., 2012	Palestinian territory	PTB	Archived samples	To identify and molecularly characterize the MTBC strains
Etchechoury et al., 2010	Argentina	PTB	Archived samples	To identify M.bovis and established epidemiological link
Firdessa etal., 2013	Ethiopia	PTB & EPTB	Direct from patients	To define the role of M.bovis in human TB
Gumi et al., 2012	Ethiopia	PTB & EPTB	Direct from patients	To assess the presence of M. bovis
Jabbar et al., 2015	Pakistan	PTB	Direct from patients	To detect the prevalence of Zoonotic TB
Jenkins et al., 2011	Nigeria	PTB	Direct from patients	To elucidate the impact of zoonotic BTB
Khattak et al., 2016	Pakistan	PTB	Direct from patients	To determine the occurrence of active M. bovis cases
Lopez-Rocha et al., 2013	Mexico	PTB	Archived samples	To analyse the epidemiologic distribution, of the MTBC strains
Nuru et al., 2015	Ethiopia	PTB & EPTB	Direct from patients	To identify the MTBC isolates
Nuru et al., 2017	Ethiopia	EPTB	Direct from patients	To investigate the transmission of TB between farmers and their cattle
Traore et al., 2012	Mali	PTB	Direct from patients	To determine the factors associated with M. bovis disease
Yeboah-Manu et al., 2016	Ghana	PTB	Direct from patients	To analysed spatial distribution of MTBC lineages
(b)
Published study	Country	Site of infection	Sample source	Review-related study objectives
Bobadilla-del Valle etal., 2015	Mexico	PTB & EPTB	Archived samples	To describe the trends of M. bovis isolation
Ghariani et al., 2015	Tunisia	EPTB	Direct from patients	To show M. bovis as a major cause of TB lymphadenitis
Portillo-Gomez & Sosa-lglesias, 2011	Mexico	PTB & EPTB	Archived samples	To identify M. bovis in humans and to establish epidemiological importance of ZTB in humans.
Siala et al., 2017	Tunisia	EPTB	Direct from patients	To detect and differentiate MTBC members
Torres-Gonzalez etal., 2016	Mexico	PTB & EPTB	Archived samples	To determine factors associated with M. bovis disease
Culture media supplemented by (Š)	Identification techniques:	Total Isolates	Rate	Notes extracted from the full text, authors’ main conclusion or suggestions
Löwenstein-Jensen MGIT Š ND	RFLP gyrB PCR-RFLP GenoType MTBC	188	4.26%	Patients identified as having M. bovis had either history of working in the stockbreeding industry or from families that had cattle herds
Löwenstein-Jensen ND	Spoligotyping	103	1.94%	It was part of a major project focused on molecular and clinical epidemiology in a pastoral community
Löwenstein-Jensen Š ND	Spoligotyping MIRU-VNTR	31	6.45%	Focused on further characterization of M. tuberculosis sub species; small sample size
Löwenstein-Jensen MGIT Š ND	PCR-Restriction Analysis Multiplex PCR	400	2.25%	Cases of human M. bovis infection suggests person-to-person transmission
Löwenstein-Jensen Š Pyr or Gly	Spoligotyping LSP typinga	950	0.42%	The national contribution of M. bovis is minimal while the reported case shared underline risk factors (Zoonosis)
Löwenstein-Jensen Š Pyr or Gly	Spoligotyping LSP typinga	173	1.73%	Focused on area with highest livestock population density in the country
Löwenstein-Jensen Stonebrink Š ND	PCR based on JB21/22 primers	100	4.00%	PCR amplification differentiate two of the four M. bovis isolated that were identified using nitrate reduction (conventional) test
Löwenstein-Jensen Š Pyr or Gly	Spoligotyping MLVA	24	8.33%	Targeted high-risk individuals who had close contact with livestock
No culturing Direct PCR	PCR based on JB21/22 primers	103	4.85%	Participants were abattoir workers, butchers, livestock farmers, veterinary professionals.
VersaTREK Myco System Löwenstein-Jensen Š ND	Spoligotyping	237	2.11%	Patient recruitment was based on passive epidemiologic surveillance
Löwenstein-Jensen Š Pyr or Gly	Spoligotyping	168	1.19%	Hospital-based patient recruitment
Löwenstein-Jensen Š Pyr or Gly	Spoligotyping	31	6.45%	Targeted individuals having cattle contact but did not confirm direct link of transmission.
7H11 agar MGIT Š ND	Spoligotyping	126	0.79%	Cases and associated risk factors were retrospectively identified from recorded database
Löwenstein-Jensen Š ND	Spoligotyping (G) LSP typinga (G)	2,551	0.59%	Patients were recruited at health facilities and there was no specific selection criteria
Löwenstein-Jensen MGIT Š ND	Biochemical tests (P) Pyrazinamidase deamidation (P) Accuprobe (G)	1,165	26.27%	Isolates taken from the same patient were considered as separate episodes if six months apart from each other.
Löwenstein Jensen MGIT Š ND	Biochemical tests (P) GeneXpert (G) GenoType MTBC (G)	79	75.95%	Influenced by primary objective that was intended to show performance of GeneXpert that does not differentiate M. bovis from other MTBC species
Löwenstein-Jensen MGIT Š Pyr or Gly	Biochemical tests (P) PCR of oxyR gene (G)	124	28.23%	Influenced by primary objective - tool performance and identification reports were based on biochemical findings.
Löwenstein-Jensen MGIT Š ND	Biochemical tests (P) TaqMan real-time PCR (G) GenoType MTBC (G)	99	76.77%	Influenced by primary objective that was intended to show performance of the tool while species identification was based on biochemical tests
Löwenstein-Jensen MGIT Š ND	Biochemical tests and Pyrazinamidase deamidation (P) Spoligotyping and GenoType MTBC (G)	533	30.21%	Patient selection was relying on designed format with specific variables of interest. Identification of species was primarily based on biochemical tests.

Note: Š Supplemented with pyruvate (Pyr) or glycerol (Gly); (ND) = not described; identification techniques: (G) = genotyping, (P) = phenotyping.

^aLSP typing; large sequence polymorphism typing.

2.6 |. Quality assessment

Due to lack of a standard assessment tool that is developed for molecular prevalence studies, the quality of included studies was assessed using the Joanna Briggs Institute (JBI) Critical Appraisal Checklist used for observational studies reporting prevalence data (Porritt et al., 2014). Although the tool does not take into account the main features of molecular studies where the methodological quality is primarily maintained through standard laboratory procedures such as culturing and characterization of species, compared to other alternative quality assessment tools, it was found to be appropriate and sufficient enough to evaluate the overall quality of included studies. Given that appropriate way of sampling procedures, having clear description of settings and target population, appropriateness and adequacy of subject recruitment, reliability and validity of methods used for the identification of the outcome interest (MTBC species) were emphasized. Most importantly, the numerator and denominator should be clearly reported, and percentages of identified M. bovis cases should be calculated. Furthermore, the revised protocol developed for review of diagnostic studies had also been considered to evaluate the reliability and validity of molecular identification techniques (Mclnnes et al., 2018).

Like that of screening and data extraction procedures, quality of individual study and the overall risk of bias was assessed by two independent reviews (HT and SB). Each individual study was evaluated based on the above nine JBI measurement criteria and score was given as Yes = 1, No = 0 or Unclear = U. Further clarification for unclear scores and additional information was sought to solve discrepancies between the two reviewers. Then, the average score (percentage) of each individual study and the overall risk-of-bias score was calculated.

2.7 |. Statistical analysis

Quantitative meta-analysis using STATA-14 (metaprop command) was performed. Event rate (proportion of M. bovis) was estimated based on random distribution assumption of zoonotic TB prevalence. Individual study estimate (ES), standard error (seES) and the lower and upper confidence intervals were calculated and reported with respect to the relative weight given for each study. Here, the standard assumption is that prevalence follows a binomial distribution and expressed as a proportion. That can be used to obtain the individual study weights. Given that, as presented in the following mathematical formula, the pooled prevalence was calculated according to the inverse variance method (Barendregt et al., 2013). Heterogeneity of study findings was assessed by I² test and managed through subgroup analysis based on study characteristics. In addition, publication bias and precision of study estimates were examined visually from the funnel plots and statistically tested using Egger’s test.

3 |. RESULTS

3.1 |. Search results

A total of 3845 articles (3015 from PubMed and 830 from other sources (CIHAL = 273, EMBASE = 404 and Google Scholar = 153)) were selected and exported to EndNote reference manager of which 3686 articles remained after duplicates were removed. After two levels of screening based on Titles and Abstracts, 56 articles were retained, of which 37 were excluded after full-text review based on our search criteria. A final 19 articles were included for qualitative systematic review and quantitative meta-analysis. Using the review tables (Table 1a,b) and the first meta-analysis modelling (Figure 2). we compared the methodological differences and the M. bovis prevalence reported by these studies, which varied in the use of genotyping and conventional identification techniques. Then, further subgroup analysis was done from 14 of the genotyping-based studies to estimate individual study findings and subtotal ES (prevalence of M. bovis) based on specific characteristics. Searching strategy, screening of articles and the overall review process are shown in diagram (Figure 1).

FIGURE 2.

Findings of a meta-analysis generated from the full model showing the prevalence of Mycobacterium bovis (ES (95% CI)) stratified by genotyping- and conventional-based mycobacterial identification methods

3.2 |. Characteristics of included studies

Characteristics of included studies, the main component of study variables and findings related to the relevant disease condition, are presented in Tables 1a,b. In accordance with the review strategy, a brief description of those main features has been narrated under the following three sections, as part of the studies context (setting), condition and population (CoCoPop).

3.2.1 |. Study setting

The included studies represented twelve countries from four continents. About half of them (10/19) were from Africa, five from Central and South America, three from Asia, and one study was from Europe. Ethiopia, at the Horn of Africa, was represented by five studies (Belay et al., 2014; Firdessa et al., 2013; Gumi et al., 2012; Nuru et al., 2015,2017) followed by Mexico in the Americas with four studies (Blanco-Guillot et al., 2017; Bobadilla-del Valle et al., 2015; Lopez-Rocha et al., 2013; Portillo-Gomez & Sosa-lglesias, 2011). Tunisia in North Africa (Ghariani et al., 2015; Siala et al., 2017) and Pakistan in Asia (Jabbar et al., 2015; Khan et al., 2016) each contributed with two studies. The remaining five studies were conducted in four different continents.

3.2.2 |. Conditions and species identification methods

All included articles were cross-sectional studies conducted through either prospective or retrospective data collection method. In this review, the techniques used in the 19 included studies for identification of the clinical isolates of the MTBC as well as for differentiation of M. bovis from other MTBC species were classified into two groups. They differed in methodology for subsequent identification using either phenotypic or genotypic typing methods, or both. Five of the studies used phenotyping by biochemical tests as primary identification and differentiation techniques of the mycobacterial isolates (Bobadilla-del Valle et al., 2015; Ghariani et al., 2015; Portillo-Gomez & Sosa-lglesias, 2011; Siala et al., 2017; Torres-Gonzalez et al., 2016). Studies that were categorized into this group also utilized other diagnostic tools, such as GeneXpert (Marlowe et al., 2011), PCR assays (TaqMan real-time and oxyR gene) (Sreevatsan et al., 1996) and pyrazinamidase deamination (Kamerbeek et al., 1997) that do identify MTBC species but are not necessarily specific for M. bovis (Bobadilla-del Valle et al., 2015; Ghariani et al., 2015; Portillo-Gomez & Sosa-lglesias, 2011; Siala et al., 2017; Torres-Gonzalez. et al., 2016). In contrast however, three of these studies used Genotype MTBC (Hain Lifescience GmbH, Germany), a tool that has been shown to also differentiate M. bovis from other MTBC species (Richter et al., 2003). Henceforth, in this review, these mixed types of diagnostic procedures were defined as ‘conventional-based’ identification methods. In the second group, the studies used primarily different PCR-based genotyping techniques (Bayraktar et al., 2011; Belay et al., 2014; Ereqat et al., 2012; Etchechoury et al., 2010; Firdessa et al., 2013; Gumi et al., 2012; Jabbar et al., 2015; Jenkins et al., 2011; Khattak et al., 2016; Lopez Rocha et al., 2013; Nuru et al., 2015, 2017; Traore et al., 2012; Yeboah-Manu et al., 2016) for identification of MTBC species as well as for differentiation of M. bovis. Essentially, spoligotyping was the most common procedure. In addition, large sequence polymorphism (LSP) typing and IS6110-restriction fragment length polymorphism (RFLP) were also performed in some studies. In this review, this group is referred to as ‘genotyping-based’ identification methods. Both groups used selective mycobacterial culture media for cultivation. However, it has been shown that M. bovis cultivation is enhanced when pyruvate or glycerol is used as an additive to the media (Schaefer, 1952). In fact, only five of the genotyping-based studies (Firdessa et al., 2013; Gumi et al., 2012; Jenkins et al., 2011; Nuru et al., 2015,2017) and one study that relied on biochemical test identification techniques (Portillo-Gomez & Sosa-lglesias, 2011) had clearly described supplementation of pyruvate or glycerol to enhance growth of M. bovis on the culture medium (Table 1a).

Beside methodological differences on species isolation and identification, the proportion of human M. bovis cases reported by primary studies have shown significant variation that could also be due to some underlying conditions. For instance, three of the genotyping studies that reported relatively higher numbers of M. bovis cases targeted very few selected high-risk individuals (Ereqat et al., 2012; Jenkins et al., 2011; Nuru et al., 2017). While two of the facility-based genotyping studies that covered study populations of relatively larger size and who shared some underlying risk factors (such as contact with livestock), reported few human M. bovis cases, suggesting minimal contribution of zoonotic TB (Firdessa et al., 2013; Yeboah-Manu et al., 2016). On the other hand, most of the conventional-based studies, which reported remarkably higher prevalence of M. bovis cases, were highly influenced by their primary objective. Those studies were either more concerned on the importance of Zoonotic TB or need to show the performance of the tool while the procedure could not differentiate M. bovis from other MTBC species (Ghariani et al., 2015; Siala et al., 2017).

3.2.3 |. Populations (human cases, specimen and isolates)

As it is described in Table 1a,b, there were two main sources of samples (specimens). The majority of the studies recruited study participants to collect specimen samples directly from clinical patients while seven studies (Bayraktar et al., 2011; Bobadilla del Valle et al., 2015; Ereqat et al., 2012; Etchechoury et al., 2010; Lopez-Rocha et al., 2013; Portillo-Gomez & Sosa-lglesias, 2011; Torres-Gonzalez et al., 2016) used archived clinical samples through a retrospective data collection approach. For studies using archived clinical isolates, demographic profiles were retrieved from databases. Among a total of 7,223 MTBC isolates, 5,568 (77%) were isolated from sputum samples. The remaining 1,655 isolates (23%) originated from patients with different types of extra-pulmonary TB (EPTB).

3.3 |. Risk-of-bias assessment

Given more emphasis to the main components of the primary studies summarized in Table 1a, b, study-level risk-of-bias score was measured using JBI assessment tool. Average score of two independent reviewers identified 37% (7/19) studies with low risk of bias, 63% (12/19) with moderate risk of bias, and the overall quality score reflected that about 70.8% of assessment criteria were found to be appropriate. However, across the probing questions selected for risk-of-bias assessment, using appropriate sampling frame was not clearly reported from almost half of the studies (Table S2). With that in consideration the critical appraisal and quality assessment score, the selected sets of included studies were considered for quantitative analysis. However, because of the observed methodological differences and heterogeneity in first full model (Figure 2), which was strongly significant among studies using the conventional identification method, further subgroup analysis and final pooled estimates were generated from genotyping-based studies only. Indeed the observed differences between conventional and genotyping identification techniques were critically evaluated and discussed with narrative synthesis.

3.4 |. Prevalence of zoonotic TB

From all 19 studies included in this review, a total of 7,185 MTBC species were isolated of which 702 (9.77%) were identified as M. bovis. Excluding 100 cases with mixed infection reported by Torres-Gonzalez. et al. (2016), the proportion of M. bovis isolates identified from EPTB cases, 440/1,522 (28.9%), was much higher as compared to isolates identified from PTB cases, 230/5,563 (4.1%). There was a vast difference in zoonotic TB prevalence among the individual studies that ranged from 0.42% (Ghariani et al., 2015) up to 76.7% on primary analysis (Siala et al., 2017). However, a clear trend in prevalence could be found when the typing method was considered – all five conventional-based studies reported higher than the average prevalence of zoonotic TB while genotyping-based methods detected lower than the average prevalence of M. bovis (Table 1b). Interestingly, all the three studies (Bobadilla-del Valle et al., 2015; Portillo-Gomez & Sosa-lglesias, 2011; Torres-Gonzalez. et al., 2016) from Mexico reported similar findings with a prevalence range of 26%–30%. The other two studies (Ghariani et al., 2015; Siala et al., 2017), which reported the highest proportion of M. bovis (75.9% and 76.7%) among the included studies, were both from Tunisia. This can be compared with a prevalence of 8.3% reported from Nigeria by Jenkins et al. (2011), which was the highest M. bovis prevalence identified among included studies that were using genotyping methods (Jenkins et al., 2011).

3.5 |. Quantitative meta-analysis findings

The overall estimate of human M. bovis infection, which was generated from the first random model with all included studies, showed a pooled prevalence of 12.1% [ES (95% CI) of 0.121 (0.095, 0.148)]. This is potentially an overestimation as a result of the high average prevalence (47.1%) reported by conventional-based identification techniques. As also described in Table 1b, prevalence data reported by conventional-based techniques were highly influenced by the primary objectives of those studies, where the majority of them were intended to show the performance of biochemical tests as compared to the reference tool. And besides the observed prevalence difference, the significant heterogeneity (I² = 98.06%) recorded from this group, suggests uncertainty in first meta-analysis model estimation (Figure 2). Hence, all of those studies were excluded at this level and they were not considered in the remaining quantitative meta-analysis estimates. The first full model is presented here to show how the prevalence of zoonotic TB has been inconsistently reported in studies using different identification techniques.

Therefore, statistical estimates (ES; 95% CI) generated from the following three meta-analysis models represented findings of the genotyping-based studies only. In the first subgroup analysis, which is grouped based on the regions where primary studies were conducted, a relatively higher prevalence was observed in Asia and Europe with subtotal ES of 0.046 (0.019–0.073) and 0.043 (0.022–0.082) respectively. While subtotal estimates of seven studies conducted in Africa showed a lower contribution of M. bovis with an ES of 0.006 (0.003–0.008). Nevertheless, the highest prevalence was also recorded from studies conducted in Africa (Jenkins et al., 2011; Nuru et al., 2017) (Figure 3). The second subgroup analysis determined the proportion of zoonotic TB with respect to site of infection (Figure 4). In this regard, the average prevalence of M. bovis infection among PTB cases was lower than that from EPTB; with a subtotal ES of 0.013 (0.007–0.020) and 0.041 (0.004–0.078) respectively.

FIGURE 3.

Meta-analysis findings of the random model that shows individual study estimates (ES (95% CI)) of genotyping-based studies and subtotal prevalence of Mycobacterium bovis across regions

FIGURE 4.

Meta-analysis findings of the random model that shows individual study estimates (ES (95% CI)) of genotyping-based studies and subtotal prevalence of Mycobacterium bovis with respect to site

3.6 |. Publication bias

Publication bias was assessed through diagrammatic presentation and statistically tested using Egger’s test. The funnel plot (Figure 5) was constructed from study estimates with pseudo 95% confidence limit against standard error of the estimates. According to this plot, the reviewed studies seem symmetrically distributed on the right and left sides of the vertical line representing the pooled estimate, suggesting minimal publication bias. However, conventional funnel plots used to assess for potential publication bias are assumed to be inaccurate for meta-analyses of proportion studies with low proportion outcomes (Hunter et al., 2014). Given that Egger’s test indicated as there is a significant small study effect at p-value of p >|t|= 0.000.

FIGURE 5.

Funnel plots of standard error and precision used to assess any publication bias

4 |. DISCUSSION

In this review, the 19 included studies showed a large prevalence difference in reported zoonotic TB that ranged from 0.4% (Ghariani et al., 2015) to 76.7% (Siala et al., 2017). All studies that used primarily conventional-based identification methods reported significantly higher prevalence of zoonotic TB than studies that used primarily genotyping-based methods. We also observed a high variation in reported prevalence among studies conducted in the same country where the methodology differed. For instance, in Mexico, Lopez-Rocha et al., 2013 (Lopez-Rocha et al., 2013) found a low prevalence of M. bovis (2.1%) using genotyping-based methods while the three conventional-based studies from Mexico showed prevalence that reached up to 30% M. bovis among the total number of TB cases (Torres-Gonzalez. et al., 2016). Similarly, the highest (76.7%) and lowest (0.42%) prevalence were reported from two African countries, Tunisia (Siala et al., 2017) and Ethiopia, respectively (Firdessa et al., 2013); however, again the study from Tunisia was performed using conventional-based methods while the one from Ethiopia utilized genotyping for strain identification. Likewise, the influence of identification techniques was prominent in previous studies done in Ethiopia, Nigeria and Tanzania, where the median proportion of M. bovis cases was significantly higher than for the findings of other recent molecular-based studies (Cadmus et al., 2006; Cleaveland et al., 2007; Kidane et al., 2002). For example, the specificity of a PCR-based method using primers JB21/22 for identification of M. bovis (see Table 1b) was shown to be low by Sales et al (Sales etal., 2014).

In spite of the above methodological differences, the pooled prevalence (1.4%) of M. bovis aggregated from genotyping-based studies was comparable with estimates previously reported in the global systematic review by Muller et al (Müller et al., 2013). This is also similar to that of the previous findings reported more than two decades ago in Europe and USA. A retrospective analysis of TB cases in the Netherlands during 1993–2007 (Majoor et al., 2011) and United States, from 1995 to 2005, showed an average prevalence of 1.4% (Hlavsa et al., 2008). Muller and colleagues projected that the pooled contribution of zoonotic disease in the African region was estimated to be 2.8% of clinically confirmed cases (Müller et al., 2013). In the current review, the average prevalence of M. bovis cases estimated from genotyping-based studies conducted in Africa was lower than in the other regions covered. Nevertheless, the highest prevalence of M. bovis was documented from studies conducted in two African countries, Ethiopia and Nigeria (Jenkins et al., 2011; Nuru et al., 2017), while two of the largest studies with very low individual estimates of zoonotic TB did have an influence on the pooled (subtotal) African prevalence (Firdessa et al., 2013; Yeboah-Manu et al., 2016).

The second subgroup analysis (Figure 4) found that the average prevalence of M. bovis detected in EPTB cases was three times higher (4.1%) than the subtotal estimates found in PTB cases (1.3%). In studies conducted by Nuru et al (Nuru et al., 2015) and Bayraktar et al (Bayraktar et al., 2011), all of the M. bovis cases were of EPTB origin. However, caution must be applied as the observed heterogeneity and small sample sizes in these two studies could mean that these findings might not be representative of their study population. Indeed, there are also studies supporting the occurrences of higher prevalence of zoonotic TB among populations that had close contact with animals, farmers and pastoralists (Ameni et al., 2013; Legesse et al., 2011). Similarly, as it is stated by Bayraktar et al. and Nuru et al., concomitant chronic infection and settings with possible potential zoonosis were noted as possible risk factors. Consecutively, Nuru and his colleagues reported higher risk of zoonotic TB among farmers, which is supported by Jenkins et al that relatively higher proportions of M. bovis were found in humans who had close contact with the primary hosts (cattle) (Jenkins et al., 2011; Nuru et al., 2017). Correspondingly, a PCR-based study done by Bapat et al confirmed that the prevalence of zoonotic TB in three high-risk districts of India was 11.4%, 8.9% and 12.6% (Bapat et al., 2017). Nevertheless, these high proportions of M. bovis isolates were detected from blood samples collected from individuals who were not clinically diagnosed as TB cases. On the contrary, molecular studies identified very low numbers of M. bovis in sputum samples collected from clinical patients living in the same region (Asia) (Devi et al., 2015; Jiang et al., 2015).

This review is limited by the lack of a general agreement over a recognized methodological standard in molecular techniques used to differentiate M. bovis from other MTBC species. Because of its complexity with many specific procedures in these studies, they were broadly classified as conventional- and genotyping-based methods and we found that there was great discrepancy on prevalence data reports between the two methods. Moreover, there was paucity of primary studies, where only articles published in English were used for our literature search. As a result, very few genotyping studies were included in the final meta-analysis model. Although there is significant improvement on heterogeneity of genotyping studies (I² and p-value) from each of the subgroup analytical models, subtotal estimates have been influenced by studies having large sample size (weight) but very low prevalence of identified M. bovis strains. As a result, the pooled estimates cannot be extrapolated to draw conclusions about the general global burden of zoonotic TB.

5 |. CONCLUSION AND RECOMMENDATIONS

This systematic review covered 19 studies published during 10 years from 2009 to 2019. Although the proportion of M. bovis detected in studies that primarily used conventional diagnostic methods suggests significant impact of zoonotic TB in particular regions, most studies based on genotyping techniques propose that the contribution of human M. bovis cases elsewhere seems substantially lower. However, the pooled estimate of the genotyping-based studies is comparable with previous molecular study findings and systematic review reports. Therefore, the prevalence discrepancy in this review indicates that some of the identification methods might not correctly differentiate M. bovis from other subspecies of the M. tuberculosis complex. In general, our systematic review suggests that zoonotic TB is still widespread in the world. Indeed, depending on the existing epidemiological and clinical conditions, some segments of the world population could have greater risk of zoonosis, suggesting that further investigation of cases among peoples at higher risk of exposure should be emphasized, and with the use of better diagnostic tools. The standardization of typing methods, including development of new molecular methods or validation of the existing ones is necessary for proper estimation of the zoonotic TB burden.

Impacts.

• Here, we critically reviewed and aggregated prevalence of zoonotic tuberculosis (TB) reported by studies published in the last 10 years, with the aim to summarize recent findings. We also examined and compared methods used for identification of the causative agent of zoonotic TB, Mycobacterium bovis.

• Although zoonotic TB is still reported from most continents in the world, it remains challenging to understand the real impact of the disease because of differences in identification methods.

• Overall, this review suggests that diagnostic tools specific for detection of Mycobacterium bovis are applied for studies to determine the impact of zoonotic TB.

DATA AVAILABILITY STATEMENT

This is a systematic review and meta-analysis study and the data extracted from included studies are available from Table 1a and 1b (characteristics of included studies).