Figure 1

Figure 1. . Selection procedure for reports included in this analysis. A list of all identified 1,203 potentially relevant reports and the core dataset is available as supplemental material (

A systematic multilingual literature search was performed according to international guidelines with certain modifications (http://www.cochrane-handbook.org/). Potentially relevant reports on putative zoonotic TB caused by M. bovis or M. caprae were identified by a search of 32 bibliographic databases by using a highly sensitive search syntax. All publications/reports documented in the various databases and published until March 2010 were considered (Table 1; Technical Appendix 1, 2). Reference Manager v11.0.1 bibliographic software was used to store and remove duplicated reports, leaving 12,176 records (Figure 1). Titles and abstracts of these reports were screened to remove studies unlikely to contain pertinent information. Altogether, 1,203 potentially relevant reports were identified (Technical Appendix 1, 2) of which 447 (37%) were available online and assessed for eligibility (Figure 1).

Figure 2

Figure 2. . Proportion of zoonotic tuberculosis (TB) among all TB cases (A) and estimated overall TB incidence (B) stratified by World Health Organization (WHO) region. Overall TB incidence rates estimated for 2010...

Eligible records (written in English, French, German, Spanish, or Portuguese) reported data for at least 50 persons tested on the frequency (prevalence, incidence) or death rate of putative zoonotic TB and contained data from no earlier than 1990. In connection with other ongoing studies, records reporting information on disease sequelae and transmission routes were also included. For 100 randomly selected reports of the 1,203 potentially relevant reports, availability and eligibility was assessed independently by 3 operators. Differently appraised records were reassessed, and the screening procedure was harmonized accordingly. The remaining records were randomly assigned to 1 of the 3 individual operators for further assessment; 140 records were considered eligible and subjected to data extraction. The data were stratified by multiple variables, if possible (e.g., country or province, HIV status). If any of the reports included referred to relevant external data or eligible reports that were not identified during the earlier steps of our literature search, the respective data were also included in this analysis. For 15 randomly selected reports of the 140 eligible records and additional data cited by or referring to these records, data extraction was performed independently by the 3 operators. Differently extracted data was reassessed and the extraction procedure was harmonized accordingly. The remaining records were randomly assigned to one of the 3 operators for data extraction. Of 140 eligible records, 91 reported or referred to data on disease frequency or mortality rates of putative zoonotic TB (Technical Appendix 1 Figure 2). If accessible, additional population estimates and official TB notifications were included in the database. This core dataset was used to analyze, by geographic region and country, the global occurrence of zoonotic TB in humans and its contribution to the overall TB prevalence in affected settings. A subset of data stratified by HIV status was used to assess a potential association between TB caused by M. bovis and HIV co-infection. For each of these analyses, duplicated data, present in >1 report was removed from the core dataset (Figure 1). Statistical analyses were performed in IC Stata 10.0 (StataCorp LP, College Station, TX, USA).

Naturally, the occurrence of zoonotic TB is greatly dependent on the presence of TB in cattle. Information on the global distribution and prevalence of bovine TB is scarce, but available data suggest that TB in cattle is prevalent in virtually all major livestock-producing countries of the developing world and Africa, specifically (2,3,7). Disease control in cattle is largely absent in these regions. Consequently, the majority of the human population is at risk for exposure to bovine TB; and globally, the occurrence of zoonotic TB likely mirrors TB prevalence in cattle.

Our systematic literature search on the occurrence of zoonotic TB provided no data for the WHO region of Southeast Asia, including major cattle producing middle- and low-income countries (e.g., India, Bangladesh, Pakistan, Myanmar, Indonesia) (Technical Appendix 1) (33). Moreover, except for Europe, data were acquired for few countries of the regions represented in this analysis (Table 1) with some, in terms of livestock production, particularly relevant countries missing (e.g., Canada, Kenya, Russia, South Africa, Sudan, and Turkey) (33). Nationwide surveillance data were almost exclusively available for high-income countries that have programs in place for bovine TB control and regular milk pasteurization (Table 1). Although ample data were obtained for many low-risk, high-income countries, the lack of nationwide surveys in potential high-risk settings precluded a credible estimation of the global occurrence of zoonotic TB.

Recorded incidence rates for zoonotic TB in Europe, the United States, Australia, and New Zealand were consistently below 1/100,000 population/year (Technical Appendix 2). Incidence rates were unavailable for other countries. However, a crude estimate could be obtained by multiplying the observed country-specific median proportion of zoonotic TB by the respective overall TB incidence rates. This suggested incidence rates of zoonotic TB of ≈1/100,000 population/year or lower for all countries outside Africa included in this survey, except for the Republic of Djibouti that reported ≈4 zoonotic TB cases/100,000 population/year.

Africa is assumed to bear the highest consequences of zoonotic TB worldwide because of the frequent and concurrent presence of multiple risk factors (2,3,7). This is supported by the highest reported median proportions of TB caused by M. bovis in connection with the worldwide highest overall TB incidence rates (Figure 2). Given an observed median proportion of zoonotic TB of 2.8% and the continental average overall incidence of TB of 264/100,000 population/year (Figure 2), an incidence rate of 7 zoonotic TB cases/100,000 population/year could be estimated (≈20 times lower than the global overall TB incidence rate) (1). Presumably, this is an overestimate as high-risk livestock producing countries of Africa (e.g., Ethiopia, Madagascar, Nigeria, and Tanzania) were overrepresented in this analysis (33) and because studies reporting the presence of zoonotic TB can be expected to be overrepresented among the surveys included (see limitations of the study below). Together, this suggests a low incidence of zoonotic TB in Africa.

Individual studies from various regions reported high proportions of zoonotic TB for specific population groups and settings (Figures 3–6). For example, in the Hispanic community in the United States, zoonotic TB appeared to be a considerable proportion of all TB cases (Table 3) and was associated with the consumption of unpasteurized cheese from Mexico (8–12). The highest median proportions for TB caused by M. bovis were observed in countries in Africa: Ethiopia, Nigeria, and Tanzania (Figure 3). However, the specific populations affected and risk factors of zoonotic TB in these settings remain largely elusive. The highest proportions of zoonotic TB in Africa were reported in studies investigating cases of extrapulmonary TB (Table 2). For example, 9 of the 11 studies reporting proportions of zoonotic TB >10% included cases of extrapulmonary TB, and 4 of the 15 studies reporting a proportion of <3.3% extrapulmonary TB cases were included (Table 2). This may mirror a widely stated association of zoonotic TB with extrapulmonary disease, perhaps reflecting the consumption of contaminated animal products as one of the main drivers of zoonotic TB (2,3,7). It has been postulated that pastoralist and rural communities would be at greatest risk for zoonotic TB (2,3,7), but the lack of data for these population groups prevents confirmation of this assumption. Collected individual studies reporting high proportions of TB caused by M. bovis suggest pockets of zoonotic transmission of TB for specific population groups and settings.

Outside Africa, large proportions of M. bovis infections among TB case-patients have been found mostly in low-TB incidence settings such as Mexico. In Cyprus, Iceland, and Malta, proportions of TB caused by M. bovis of >10% were observed (Figure 5); however, these countries, nationwide, reported <20 human TB cases in the respective years. Similarly, a study in San Diego County, California, USA, showed an overall decreasing incidence of human TB while the incidence of zoonotic TB and therefore also the relative proportion of zoonotic TB has steadily increased (9,12). This suggests that commonly applied control efforts targeting M. tuberculosis transmission have little effect on the occurrence of zoonotic TB and probably reflects the distinct drivers of M. tuberculosis and zoonotic TB infection (e.g., aerosol transmission vs. foodborne infection) (2,3). Similarly, differences in the epidemiology of zoonotic TB are likely to exist between different regions. This could be mirrored by the association of zoonotic TB with HIV in the United States, but not in other areas included in this analysis (Table 4). Ascertaining the factors contributing to an association between HIV and zoonotic TB in some regions will require more in-depth research, thus eliminating potential confounders such as socioeconomic status, education level, national origin, and other factors.

The current study is affected by several biases. The sensitivity of this systematic literature review was affected by the selection of eligible reports and data extraction by a single operator. Also, only reports available online and written in English, French, German, Spanish, or Portuguese were included. Selected reports are biased toward surveys which identified or aimed to identify TB cases caused by M. bovis, possibly resulting in an overestimation of the proportion of zoonotic TB cases. Data from low-income countries included in this study were rarely comparable and not representative of the respective nationwide populations. Nonetheless, it seems unlikely that our conclusions were fundamentally affected by these biases. Lastly, our results are influenced by the technical constraints of the studies included. Specifically, biochemical methods may be relatively unreliable for the identification of M. bovis or M. caprae strains and routine culture methods for M. tuberculosis are suboptimal to detect strains of M. bovis (2,3). Thus, TB cases caused by M. bovis may be systematically underreported.

Reports published after the completion of this systematic review revealed information from countries not covered by this study. A study from Bangladesh analyzed isolates from 350 TB patients but did not identify any infections by M. bovis (34). In a study from Bamako, Mali, 0.8% of TB cases analyzed were caused by M. bovis (35). In Turkey and the West Bank, Palestine, respectively, 5.3% and 6.5% of clinical TB cases analyzed were caused by M. bovis (36,37); however, zoonotic TB can be considered rare in these areas, given the low overall incidence rates of TB of 28 and 0.7/100,000 population/year (1,38). Together, available data suggest a minor global importance of zoonotic TB. However, pockets of more frequent zoonotic transmission of TB seem to be present in certain population groups. More research is needed to identify the main transmission drivers in these areas.