Mycobacterium bovis in rural Tanzania: Risk factors for infection in human and cattle populations

ARTICLE IN PRESS Tuberculosis (2007) 87, 30–43

Tuberculosis http://intl.elsevierhealth.com/journals/tube

Mycobacterium bovis in rural Tanzania: Risk factors for infection in human and cattle populations Sarah Cleavelanda,, Darren J. Shawa,b, Sayoki G. Mfinangac, Gabriel Shirimad, Rudovick R. Kazwalae, Ernest Eblatee, Michael Sharpf,1 a

Wildlife and Emerging Disease Section, Royal (Dick) School of Veterinary Studies, University of Edinburgh, Easter Bush, Roslin, Midlothian EH25 9RG, Scotland b Veterinary Clinical Studies, Royal (Dick) School of Veterinary Studies, University of Edinburgh, Easter Bush, Roslin, Midlothian EH25 9RG, Scotland c National Institute for Medical Research, Muhimbili Research Station, Dar es Salaam, Tanzania d Department of Veterinary Clinical Studies, University of Glasgow, 464 Bearsden Road, Glasgow G61 1QH, Scotland e Department of Veterinary Medicine and Public Health, Sokoine University of Agriculture, Morogoro, Tanzania f Moredun Research Institute, Pentlands Science Park, Bush Loan, Penicuik, Midlothian EH2 0PZ, Scotland Received 22 July 2005; received in revised form 13 January 2006; accepted 1 March 2006

KEYWORDS Mycobacterium; Tuberculosis; Risk factors; Tanzania; Cattle

Summary Although bovine tuberculosis is widespread throughout Africa, very little is known about risk factors for Mycobacterium bovis infection in either human or cattle populations. A human case–control study was conducted in northern Tanzania, comparing risk factors and prevalence of cattle interdermal test positives of cases (cervical adenitis cases from which M. bovis was isolated) with age- and sexmatched controls (selected at random from potential hospital attendees within the community). A cattle cross-sectional study was also set-up involving 27 villages selected at random in four districts, with 10,549 cattle and 622 herds tested, and questionnaire surveys conducted in 239 households. M. bovis was confirmed in seven of 65 (10.8%) human cervical adenitis cases, of which only one came from a household owning infected cattle. M. bovis in human patients was associated with families in which a confirmed diagnosis of tuberculosis had previously been made (po0:001) and with households far (4100 m) from neighbours (p ¼ 0:003). In cattle, overall prevalence of intradermal test positives was low at 0.9% (0.70–1.06%), but

Corresponding author. Tel.: +44 131 650 6404; fax: +44 131 651 3903.

E-mail addresses: [email protected] (S. Cleaveland), [email protected] (D.J. Shaw), [email protected] (S.G. Mfinanga), [email protected] (G. Shirima), [email protected] (R.R. Kazwala), [email protected] (E. Eblate), [email protected] (M. Sharp). 1 Present address: Veterinary Laboratories Agency, Lasswade Laboratory, Pentlands Science Park, Bush Loan, Penicuik, Midlothian EH2 0PZ, Scotland. 1472-9792/$ - see front matter & 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.tube.2006.03.001

ARTICLE IN PRESS Mycobacterium bovis in rural Tanzania

31

widespread, with 11.8% (8.44–13.17%) herds containing at least one reactor. Prevalence of intradermal test positives increased significantly with cattle age (po0:001). Herds with the following risk factors had a significantly greater prevalence of intradermal test positives: 450 cattle in the herd (p ¼ 0:024); herds housed inside at night (p ¼ 0:021) and herds in contact with wildlife (p ¼ 0:041). Furthermore, villages that experienced annual flooding had a higher prevalence of infection (p ¼ 0:043). & 2006 Elsevier Ltd. All rights reserved.

Introduction Mycobacterium bovis is a member of the group of Mycobacterium species classified as the Mycobacterium tuberculosis complex and the cause of bovine tuberculosis. Worldwide, M. tuberculosis is considered to be the most common cause of human tuberculosis, causing an estimated 1.8 million deaths in 2002.1 However, there is often little awareness about the potential for zoonotic Mycobacterium species, such as M. bovis, to cause human tuberculosis and our understanding of the contribution of M. bovis to the human epidemic worldwide is very limited. Tanzania is one of few developing countries with quantitative data on the prevalence of M. bovis in humans. In the southern highlands region of Tanzania, M. bovis was isolated from 1/23 (4%) cases of pulmonary TB2 and from 6/21 (28.6%) cases of cervical adenitis,3 and, more recently, M. bovis was isolated from 7/65 (10.8%) of culture-positive cases of cervical adenitis in the Arusha Region.4 Since extrapulmonary TB cases comprise 15–20% of the 54,000 new cases of TB recorded each year in Tanzania,5 M. bovis may not be a negligible component of the human tuberculosis epidemic. The Arusha Region is an area of particular concern in Tanzania, with more than 30% of TB cases recorded as extrapulmonary TB5 and having large numbers of livestock-keepers.6 These findings have led to suggestions that risk factors associated with livestock-keeping play a major role in the epidemiology of M. bovis in humans.7 In many parts of the developing world, frequent opportunities exist for zoonotic transmission of M. bovis, with widespread consumption of unboiled milk and extremely close human–livestock contact in many rural communities.7–9 However, the relative importance of different sources of infection and routes of transmission in Africa is still largely unknown. A preliminary analysis of some data from patients diagnosed with cervical adenitis in the Arusha Region indicated that consumption of raw milk and poor ventilation in houses were both linked with M. bovis adenitis in humans.4 This current study

builds on this preliminary analysis, presenting results from a formal case–control study of risk factors for human M. bovis infection, including more comprehensive analysis of factors associated with livestock ownership and management, M. bovis prevalence in cattle, food consumption practices, housing and socio-economic factors. Earlier studies in Tanzania have reported a wide variation in prevalence of M. bovis infection in cattle in different parts of the country, ranging from 13.1% intradermal skin positives in randomly selected herds in the southern Highlands,10 1.7% in the eastern zone,11 0.9% in dairy farms in Dar es Salaam, 0.6% in indigenous cattle in the coastal zone12 to 0.2% in intensively managed herds in the Lake Victoria zone.10 However, little is known about the factors governing the prevalence of bovine tuberculosis in Tanzania, or the principal risk factors for infection in cattle. Other studies conducted elsewhere have identified several factors, including cattle introductions in Ireland13 and Italy,14 large herd size in Ireland,13 Zambia,15 Eritrea,16 and Michigan,17 and proximity/access to wildlife (e.g. badgers in Ireland13; white-tailed deer in Michigan17). In Tanzania, data relating to these risk factors are equivocal. For example, higher prevalences have been reported in larger herds in the southern Highlands10 but not in the eastern zone.11 Finally, although Tanzania is renowned for the abundance and diversity of its wildlife, no information is yet available on the role of wildlife as potential sources of infection for cattle and humans. Aerosol transmission is generally considered the main route of cattle-to-cattle transmission, largely based on the finding that 90% of tuberculous cattle have lesions in the thoracic cavity and head.18 But many questions remain about the relative importance of different sources of infection for cattle and the routes and mechanisms of within- and between-herd transmission. The objective of this project was therefore to conduct an integrated study of M. bovis epidemiology in humans and cattle in the Arusha Region of Tanzania in order to determine the prevalence of

ARTICLE IN PRESS 32

S. Cleaveland et al.

M. bovis infection in cattle populations and to identify and quantify risk factors for infection in human and cattle populations. This has required involvement of the medical, veterinary and wildlife sectors within Tanzania. Risk factors for human infection were identified through a case–control study. The prevalence of bovine TB infection in cattle was determined through cross-sectional skintest surveys in a random sample of cattle herds. Cattle risk factors were determined from crosssectional surveys at the individual-, herd- and village-level. Additional information on the relative importance of different routes of transmission was obtained from lesion distribution data in cattle abattoir samples.

Mbulu) within the Region, which was formerly called Arusha Region and which is now called Manyara Region, Tanzania, between the latitudes of 3.381S and 4.521S and longitudes of 35.01E and 35.91E (Fig. 1). Each district contained one principal urban centre (district centre) with the remaining settlements comprising agropastoral village communities. The form of land-use throughout all four districts was predominantly subsistence farming, involving mixed crop–livestock systems, with a small proportion of land dedicated to commercial arable and coffee farming. The study area contained two wildlife-protected areas (Tarangire and Manyara ecosystems), with the Ngorongoro-Serengeti ecosystem bordering the study area to the north (Fig. 1).

Materials and methods

Human case–control study

Study area

Risk factors for human M. bovis infection were examined by carrying out a case–control study. Ethical clearance was obtained from the Medical Research Co-ordinating Committee in Tanzania.

For all components of the study, work was carried out in four districts (Babati, Hanang, Karatu and

Figure 1 Map showing location of villages sampled in the cross-sectional cattle study and the various national parks (shaded areas). Crosses refer to missionary hospitals; triangles district headquarters, and the size of the circles represents the prevalence of cattle intradermal test positives in the village (from 0 to 42%).

ARTICLE IN PRESS Mycobacterium bovis in rural Tanzania

Definition and selection of cases Cases were defined on the basis of culture and isolation of M. bovis from lymph node biopsies of patients diagnosed clinically with adenitis according to national19 and international guidelines.20 Human cases were enrolled in this study through the National Tuberculosis and Leprosy Control Program (NTLP), with verbal consent obtained from each patient enrolled following explanation of the study purpose and counselling on the laboratory tests for the specimens collected. Data on adenitis cases were collected from seven referral hospitals in four districts of the Arusha Region (Fig. 1), described in detail elsewhere4. Data and samples were collected by the District Tuberculosis and Leprosy Co-ordinators (DTLCs) or clinical officers responsible for tuberculosis in the referral hospitals. Electric and kerosene deep freezers were supplied for storage of specimens at approximately 20 1C. On presentation at the hospital, patients were interviewed by the DTLCs and clinical officers, and clinical forms completed. Lymph node biopsy specimens were taken from all patients fulfilling the diagnostic guidelines for extra-pulmonary TB21 before starting anti-TB chemotherapy. Specimens were placed in universal containers, stored in deep-freezers (as above) for periods of up to 2 months and transported to the diagnostic laboratory on ice packs in cool boxes. Frozen biopsy specimens were processed for culture and identification tests at the Central Tuberculosis Reference Laboratory, Dar es Salaam. All processing of samples was carried out using aseptic techniques in a safety cabinet. Processing, culture and identification tests were carried out at the Central Tuberculosis Reference Laboratory, Dar es Salaam, using aseptic techniques in a safety cabinet. Details of methodology have been described elsewhere4,22 using established protocols.23–25 Briefly, following decontamination and neutralization, suspensions were centrifuged and aliquots of the remaining sediment inoculated on blood agar and onto Lowenstein–Jensen media with pyruvate and Lowenstein–Jensen media with glycerol. Cultures were incubated at 37 1C overnight. If there was evidence of contamination, the decontamination procedure was repeated and processing time increased to 1 h. Cultures were incubated at 37 1C for 6–12 weeks while observing for signs of growth on a weekly bases. Established protocols were used to identify Mycobacteria species.23–25 The observation of isolates with eugenic growth on both Lowenstein–Jensen media were tentatively identified as M.

33 tuberculosis while those with eugenic growth in the pyruvate-containing media were regarded as suggestive of M. bovis. The chemical identification tests23,24 were further carried out for identification of Mycobacteria species according to standard methodology.23–25 Patients receiving a positive diagnosis of human tuberculosis received DOTS therapy according to NTLP guidelines with supervision of treatment by the DTLC. All patients were provided with pre-test counselling prior to being asked for permission to carry out HIV testing, and post-test counselling, according to guidelines on ethics for health research in Tanzania.26 All sera submitted for HIV serology were labelled with identification numbers only and tested anonymously using single Behring ELISA tests (Dade Behring Marburg GmbH, Emil-von-Behring Marburg/ Germany) and positive samples retested using Wellcozyme HIV Recombinant (Murex Biotech/UK). Post-test counselling was provided for all patients enrolled however at the time of the study antiretroviral therapy was not available through government medical services.

Molecular characterization of isolates In extracting DNA from mycobacterial cells, a fresh culture was suspended in 25 ml of distilled water, boiled for 15 min, and then centrifuged for 30 s at 13,000g. A volume of 5 ml of the supernatant was then used for the PCR procedures. The primers used in the PCR were P0982 (50 GTGAGGGCATCGAGGTGGC 30 ) and P0983 (30 AAACAGTGGCTGCGGATGCG 50 ) from the IS986 insertion sequence that amplified a 245 bp fragment and primers S6055 (50 CGGCAACGCGCCGTCGGTGG 30 ) and S6056 (30 GGGCCGCCACGGCACCCCCC 50 ) derived from the mtp40 genomic fragment, which generated a 396 bp amplification product. The PCR amplification was performed in a final volume of 50 ml containing 1  reaction buffer, Promega (50 mM KCL, 10 mM Tris–HCL; pH 9.0 at 25 1C and 0.1 Triton X-100), 2.0 mM magnesium chloride, 0.1 mM each deoxynucleoside triphosphates, 0.5 U of Taq DNA polymerase (Promega), 0.05 mM of each of the primers for IS986 and 0.1 mM of each of the primers for mtp40. A 20 ml of target DNA was first denatured by incubation in boiling water for 10 min, then immediately quenched on ice before adding 5 ml of it in the reaction mixture. A total of 30 PCR cycles were performed. Each cycle was comprised of a denaturation step set at 94 1C for 30 s, annealing at 62 1C for 30 s and extension at 72 1C for 60 s. In the strain of M. tuberculosis two PCR products at 245 and 396 bp

ARTICLE IN PRESS 34 were expected whereas a single product at 245 bp was expected for M. bovis and other strains of M. tuberculosis complex. All the reactions ware carried out in an automated thermal cycler (Crocodile II Appligene, Illkirch). After amplification, a 10 ml of the PCR product was analysed by 7.5% polyacrylamidegel electrophoresis. Electrophoresis was carried at a constant voltage of 200 for 45 min using a vertical min protean IITM apparatus (Biorad USA) filled with 1  TBE buffer. The gel was then stained by the silver nitrate method described by Herring et al.27 and the gel preserved at 4 1C in a sealed nylon bags to avoid drying.

Definition and selection of controls Controls were selected from the Arusha Region population at random (through sequential random selection of villages, ten-cell units, and households) and included in the study only after verification that the household came from a population of ‘hospital-attendees’ (i.e. people who would present to the clinic if suffering from tuberculosis). This approach was considered appropriate given the widespread network of NTLP clinics throughout Tanzania that effectively cover the whole country. The original intention to identify controls from hospital patients presenting with traumatic and other non-infectious conditions was not continued when it became apparent that these people represented only households with a relatively high socio-economic status, usually living close to the medical centre, and did not provide an appropriate control population for the cases. Controls were matched to cases by age class and sex, which have previously been suggested as risk factors for M. bovis in Tanzania.2 Between five and 11 controls per case were selected. The large ratio of controls per case arose from the fact that controls were randomly selected for questionnaire surveys and cattle-testing during the process of culture and identification of Mycobacteria species and assumed that a greater proportion of adenitis cases would yield culture positive results than in fact occurred (see results). The design of this study differed significantly from the earlier cross-sectional study using the same M. bovis positive patients,4 in that community controls were used as the comparison group rather than culture-negative adenitis patients. This provided several major advantages: first, a greater confidence that controls were genuinely tuberculosis-negative, whereas false negatives may have been included in the culture-negative adenitis

S. Cleaveland et al. controls (as a result of low culture isolation rates); second, the design involved traceback of both cases and controls to households, allowing detailed examination of the relationship between human disease and infection in household cattle.

Risk factors for human infection Each of the cases and controls was traced back to the household and a questionnaire conducted to collect information on a range of variables relating to individual and household characteristics, household practices and patterns of consumption of meat and milk products. Milk consumption was assessed according to the frequency of consumption, the source of milk (cattle/goat), and preparation before consumption (boiled, soured or raw). Housing was assessed according to construction (mud/ brick), number of rooms, number of windows in each room, and the number of people sleeping in the same hut/room as livestock at night. BCG status of both cases and controls was recorded and any hospital-diagnosed cases of tuberculosis in other family members documented. The infection status of cattle owned by the household was determined by tuberculin-testing of cattle using the single comparative intradermal test (SCIT), as described below.

Cattle prevalence study Cattle herds were selected by multi-stage random sampling. A complete list of villages was obtained for each district at the district headquarters and 27 villages selected at random out of 242 villages in the four districts. At each village headquarters, a list was drawn up of administrative sub-divisions (vitongoji) and one sub-division was selected at random to represent the village. Herd-testing followed two procedures. In some villages, livestock-keepers from the kitongoji were requested to bring cattle to a central point (e.g. a crush) for testing. In more dispersed villages, cattle were tested by house-to-house visits conducted as early in the day as possible. The single comparative intradermal tuberculin test (SCIT) was used to test all cattle.28 Briefly, 2000 IU each of avian and bovine protein derivatives (PPD), supplied by the Veterinary Laboratories Agency, Weybridge, were injected intradermally into the neck at least 12 cm apart after recording skin thickness at both sites. Second skin-thickness measurements were made approximately 72 h after injection. From previous studies by Kazwala,2 cattle were considered to be reactors when there was a bovine bias of more than

ARTICLE IN PRESS Mycobacterium bovis in rural Tanzania 3 mm. At the time of testing, individual animal data were recorded on age (determined by the number of permanent incisor pairs), sex and breed of the animal and the name of the owner.

Herd-level risk factors—household questionnaire survey Home visits to the selected households were made on the day after tuberculin testing. On arriving at each home, the study was discussed and a verbal consent obtained. In most cases (86%), the interviewee was either the head of the household or his wife. Questionnaire data were collected by one of three members of the team (GS, GM, EE) and a translator used, where necessary, to facilitate communication of tribal languages, mainly KiIraqw, between interviewer and interviewee. The KiIraqw translator was a full-time member of the study team, fully appraised and trained in the interview process. Each questionnaire included validation questions, which allowed questionnaires to be checked for internal consistency. The questionnaire was first field-tested in a village that was not included in the random selection, but within the study districts. Questionnaires were conducted using a mixture of closed and open questions to collect herd-level data for over 100 variables relating to household characteristics and cattle management practices. At the household level, the variables included: family size, herd size, land-use, cases of confirmed or suspected tuberculosis in the family, recent introduction of new cattle into the herd, contact with livestock from other herds, distance travelled for grazing/water, proximity of wildlife, confinement in house at night, and ventilation quality.

Village-level risk factors Data were obtained from village leaders on the occurrence of flooding in the village. GPS coordinates were collected at a central point in each village to determine the distance of each village to national park boundaries. National park boundaries were obtained from the Conservation Information Monitoring Centre (Tanzania Wildlife Research Institute), Arusha, Tanzania.

Abattoir prevalence survey Following consultations with regional veterinary officers, protocols were established for the collection of suspected material by district meat inspectors at Mbulu, Babati, Karatu and Hanang, who

35 were trained and supervised by project staff. Data were collected on the presence, size and distribution of visible lesions in each carcass. Samples from tissues containing visible lesions were collected and stored at district headquarters for up to 2 months at about –20 1C and transferred to Sokoine University of Agriculture in cool boxes on ice packs. Samples experienced at least two freeze–thaw cycles prior to culture.

Sample processing and culture Cattle samples were processed in four steps involving decontamination, neutralization, centrifugation, and inoculation according to Scottish Mycobacterium Reference Laboratory (SMRL) protocols.28 Sediments were inoculated onto Lowenstein–Jensen with pyruvate and Lowenstein–Jensen with glycerol. Cultures were incubated at 37 1C for a maximum of 12 weeks. For identification purposes, positive cultures were sub-cultured into Lowenstein–Jensen media with glycerol/pyruvate. Positive cultures were subjected to tests for species identification, which included oxygen preference test, tween hydrolysis test, niacin test, pigmentation and temperature tests.23–25

Data analysis Individual-, herd- and village-level analyses Cattle skin-test data were analysed at the individual-, herd- and village-level. Data were entered into Excel spreadsheets and analysed at the individual-, herd- and village-level using S-Plus (Insightful Corp, Seattle, USA&, 1999). For the individual- and herd-level analyses mixed-effect models with binomial errors were used29 where herd was entered as the random factor and the other variables of interest entered as fixed factors. At the herd and village-level, data were analysed using the number of positives and negatives as a response outcome (binomial variable), whereas at the individual level, data consisted of a binary response variable (positive/negative).

Case–control study Due to the very low number of human cases confirmed by culture, mixed-effect models could not be used due to lack of convergence. Data were analysed as a matched case–control study,30 involving conditional logistic regression using generalized linear models with binomial errors. While the ratio of controls to cases higher than 5:1 is not

ARTICLE IN PRESS 36

S. Cleaveland et al.

likely to result in substantial improvements in power, these data were available and were therefore included in the analysis. The individual herd and case–control analyses were carried out in a two-stage process. Preliminary analyses involved univariate analyses of all variables of interest. In the second stage, multivariate analyses were run, where significant risk factors in the initial screening were evaluated using multiple logistic regression. In all analyses, po0:05 was taken to indicate significance and the appropriate degrees of freedom are quoted with the test statistic. Odds ratios (OR) and 95% confidence intervals (CIs) for variables that were significant in the univariate analyses are shown in Table 2, whereas ORs for variables that were significant in the multivariate analysis are given in the Table 3.

Results Human case–control study The prevalence of M. bovis among cervical adenitis patients in the Arusha Region has previously been reported,4 with seven human M. bovis cases isolated from culture of 457 lymph node biopsies. Mycobacteria species were isolated from only 65 (14.2%) of these specimens, despite the fact that a large proportion (79/100) of the samples undergoing detailed histological examination had evidence of granulomatous changes (histological results reported elsewhere by Mfinanga22). Of the 65 culture-positive samples, seven (10.8%) were M. bovis, 27 (41.5%) M. tuberculosis and 31 (47.7%) non-tuberculous Mycobacteria.4 Forty-nine human isolates, including five of the seven M. bovis isolates, were analysed using PCR methodology as part of a separate study comparing the performance of multiplex PCR on clinical specimens from cattle and human and confirmed cultures. All 5 culture-positive M. bovis isolates were identified as M. bovis on PCR producing the 245 bp band

Table 1

indicative of M. bovis in the mtp40 and IS986 multiplex PCR. HIV status was determined for only a small proportion of cases and controls, with 2/4 (50%) of M. bovis cases HIV positive in comparison with 0/12 controls. The current study focused on collection of traceback data, including results of cattle skintesting and household questionnaire surveys, which were obtained from each of the seven human M. bovis cases identified. Of these, four (57.1%) came from cattle-owning households, but only one of these herds contained reactor-positive animals. Despite the small sample size, the univariate analyses identified several significant risk factors (Table 2). When these significant factors were entered into the multivariate analyses several significant risk factors remained (Table 3), with disproportionately more M. bovis cases in (a) families in which a hospital-confirmed diagnosis of TB had been made in other family members (2/7 cases vs. 0/66 controls) and (b) households living relatively distant (4100 m) from their neighbours (7/7 cases vs. 37/64 controls). Individuals who had been married were significantly less likely to be M. bovis-positive cases (1/7 cases vs. 39/66 controls).

Cattle prevalence study Of 10,549 cattle tested in the cross-sectional study, 91 were positive giving an overall prevalence of intradermal test positives of 0.9% (95%CI 0.70–1.06). Of 622 herds tested, 74 contained at least one positive reactor, giving a herd prevalence of 11.8% (8.44–13.17). There was no significant difference in overall prevalence of intradermal test positives between districts (F 3;22 ¼ 0:39, p ¼ 0:764; Table 1). The prevalence of intradermal test positives increased significantly with cattle age group (as measured by permanent incisor pairs) (F 4;9897 ¼ 7:61, po0:001, OR ¼ 1.27 (1.16–1.39), Fig. 2a), but there was no significant difference between male and female cattle (0.91% (0.63–1.25) vs. 0.84% (0.62–1.09) (F 1;9920 ¼ 0:16, p ¼ 0:681)).

Prevalence and 95% confidence intervals of bovine reactors in districts within Arusha Region.

District

Villages

Pos

Neg

Total

Prevalence (%795% CI)

Babati Hanang Karatu Mbulu Total

6 7 7 6 26

37 23 12 19 91

3620 3285 1645 1908 10,458

3657 3308 1657 1927 10,549

1.01 0.70 0.72 0.99 0.86

Villages—number of villages sampled in each district; Pos—individual cattle with a positive skin test.

(0.71–1.39) (0.44–1.04) (0.37–1.26) (0.59–1.54) (0.69–1.06)

ARTICLE IN PRESS Mycobacterium bovis in rural Tanzania

37

Village-level risk factors Villages that had experienced annual flooding had a significantly greater prevalence of intradermal test positives with 0.86% (0.73–1.31) vs. 0.68% (0.46–0.88) positive (w21 ¼ 4:12, p ¼ 0:043, OR ¼ 1.53 (0.99– 2.37)). This difference was more marked when inconclusive positives were also included with 2.15% (2.01–2.88) vs. 1.56% (1.37–2.03) positive (w2 ¼ 7:63, p ¼ 0:006; OR ¼ 1.45 (1.10–1.91), Fig. 2b). No statistically significant spatial pattern in village prevalence was detected, either in terms of easting and northing (w2 o0:001, p40:970), or in terms of distance to a national park (w2 ¼ 0:48, p ¼ 0:488) (Fig. 1).

Herd-level risk factors—household questionnaire survey Household questionnaire surveys were conducted for 239 skin-tested herds, comprising a total of 5451 cattle. The prevalence of intradermal test positives in these herds was again low at 0.99%

(0.74–1.29) (54/5451), with 17.2% (12.60–22.54) herds (41/239) containing at least one reactor. For this set of animals, exposure to herd-level risk factors could be determined from questionnaire data and those variables that were significant from the univariate analyses are presented in Table 2. There are three main groups of associations—number of animals owned, contact with domestic animals (cattle and goats) and contact with wildlife. However, the multivariate analyses resulted in only a few of these factors being associated with reactor prevalence (Table 2). The prevalence of reactors was significantly higher in herds with more than 50 cattle (1.34% (0.98–2.20) vs. 0.70% (0.50–1.08)), in those herds that were housed inside during the night (1.10% (0.85–1.66) vs. 0.40% (0.34–1.11)), and where impala were reported in proximity to the household (1.72% (0.92–2.13) vs. 0.69% (0.56–1.22)). In contrast, the prevalence of reactors was significantly less in those herds which were herded with goats (0.8%3 (0.73–1.36) vs. 1.68% (0.63–2.41)) (Table 3).

5 3.0

Percentage of animals truly positive

4

3

2

1

0

1.5

1.0

0.5

1

N

2to3 3to4 4plus

Age class (incisors)

(b)

2.5

2.0

1.5

1.0

0.5

0.0

0.0 <1

(a)

Percentage of animals in a village truly & inclusive positive

Percentage of animals in a village truly positive

2.0

Y Flooding

N

Y Flooding

Figure 2 (a). Changes in the prevalence of intradermal test positives in individual cows as a function of age (incisor pairs). Vertical lines denote 95% confidence intervals. (b) Percentage of animals positive (truly and inconclusively) in villages that do (Y) and do not (N) experience annual flooding. White horizontal lines denote the median, solid boxed the interquartile range, and whiskers the range.

ARTICLE IN PRESS 38

S. Cleaveland et al. Table 2

Herd level and case–control univariate analyses.

Factor Herd level analyses Own more than 50 cows Have a hut (not a house)

p Value 0.016 0.011

Odds ratio

Factor

p Value

1.92 (1.13–3.27) 0.18 (0.05–0.67)

Case control analyses HIV 0.016 District 0.049

0.005

NA Hanang 0.24 (0.01–3.50) Karatu 0.12 (0.01–1.75) Mbulu NA NAy

0.005

NAy

o0.001

NAy

0.009

NAy

0.008

27.19 (1.68–440.57) 12.50 (1.03–151.30)

(odds compared to Babati) o0.001

3.00 (1.61–5.59)

Number of calves in house

0.025

1.03 (1.00–1.06)

Number of cattle

0.006

1.01 (1.00–1.02)

Total animals

0.012

1.01 (1.00–1.02)

430 min walking to water source Contact with wildebeest

0.028

0.67 (0.47–0.96)

0.006

2.83 (1.36–5.93)

Contact with elephants

0.012

2.62 (1.24–5.54)

Contact with leopards Contact with baboons Herd cattle with goats Herd cattle with goats in wet season

0.017 0.028 0.027 0.031

0.49 0.53 0.54 0.54

Animals inside at night

(0.27–0.88) (0.30–0.93) (0.32–0.93) (0.32–0.95)

Distance to village centre (km) More than 2 km from village centre More than 100 m to nearest household Greater than 4000ft in altitude Never married Case in family with symptoms/ signs Case in family confirmed at hospital Keep donkeys

Odds ratio

0.048

o0.001

0.005

NAy

24.24 (1.72–342.72)

Table shows statistical significance (p value), odds ratios and their 95% confidence intervals. NA—odds ratio and confidence intervals not applicable due to either  None of the controls or y All of the cases were associated with a factor.

Abattoir survey Visible granulomatous lesions were detected by meat inspectors in 1504 (19.8%) of the 7589 cattle carcasses examined (Table 4). Of the 1290 carcasses for which the site of lesions was recorded, the majority (791 ¼ 61.3%) had lesions in the gastrointestinal tract, 456 (35.3%) in the thorax, 47 (3.6%) in the lymph nodes of the head, and 66 (5.1%) in other sites, predominantly the prescapular lymph nodes. Mycobacteria species were cultured from 139 (9.2%) of the 1504 samples with visible lesions M. bovis was isolated from 10/139 (7.2%) culturepositive lesions, with the remaining isolates comprising M. terrae (48.2%), M. avium (7.9%), M. cholenei (0.7%), M. gordonae (7.9%), M. fortuitum

(0.7%), M. flavescens (2.9%) and M. smegmatis (2.9%). Of the 10 M. bovis isolates, 5/10 (50%) were isolated from lesions in the gastrointestinal tract.

Discussion This study provides, for the first time, a data set combining quantitative data on prevalence and risk factors of M. bovis infection in both humans and animals. The human adenitis results indicate that M. bovis is a cause of a proportion of human extrapulmonary cases of tuberculosis in rural Tanzania and provides preliminary data on putative risk factors for human infection. The cattle study provides an important contribution to our understanding of determinants of prevalence in cattle,

ARTICLE IN PRESS Mycobacterium bovis in rural Tanzania Table 3

39

Herd level and case–control multivariate analyses

Factor

p Value

Herd level analyses More than 50 cows Herds that were housed inside during the night Contact with impala Cattle herds herded with goats Case–control analyses Families in which a hospital diagnosis of TB had been made in other family members Households living 4100 m from other households Individuals who had never been married

Odds ratio

F 1;146 ¼ 5:21, p ¼ 0:024 F 1;146 ¼ 5:49, p ¼ 0:021

2.00 (1.23–3.26) 3.55 (1.71–7.35)

F 1;146 ¼ 4:26, p ¼ 0:041 F 1;146 ¼ 6:04, p ¼ 0:015

1.78 (1.02–3.08) 0.46 (0.25–0.84)

w21 ¼ 12:11, po0:001

NA

w21 ¼ 8:86, p ¼ 0:003

NAy

w21 ¼ 4:41, p ¼ 0:036

23.90 (0.94–608.20)

Table shows statistical test values, significance (p value), odds ratios and their 95% confidence intervals. NA—odds ratio and confidence intervals not applicable due to either.  None of the controls. y All of the cases were associated with a factor.

Table 4

Summary of slaughter house survey, see text for description of lesion detection.

District

Cattle slaughtered

Animals with lesions

Proportion lesion positive (%)

Babati Hanang Mbulu Karatu Total/Overall

2064 1339 2092 2094 7589

292 420 588 78 1379

11.80 25.40 23.09 5.60 15.60

routes of within- and between-species transmission, and the development of appropriate strategies of disease control.

M. bovis infection in people Despite the limitations of the human case–control study (with only seven cases of M. bovis), hospital confirmation of TB in another member of the family and distance from other households were both significantly associated with M. bovis infection in people. There are at least three possible explanations for the association between M. bovis infection and hospital-confirmed TB cases in other family members: (a) animal–human transmission of M. bovis to other family members, as well as to the case, (b) confounding risk factors for M. bovis and other causes of tuberculosis (e.g. poverty, malnutrition or HIV), or (c) human-to-human transmission of M. bovis. However, without culture results available to identify the Mycobacteria species

(10.41–13.24) (23.07–27.81) (21.29–24.96) (4.64–6.66) (14.77–16.42)

causing ‘hospital confirmed TB’ in people and with insufficient data on HIV status, further evaluation of these alternative hypotheses is not possible in the current study. The increased risk associated with living in a relatively remote household, which remained significant on multivariate analysis, could be explained by relatively poor access to public health education and lack of knowledge of preventive health measures. Mfinanga et al.7,22 have previously demonstrated significant tribal and cultural differences in levels of knowledge about tuberculosis and disease prevention, which may affect disease outcomes. A recent case–control analysis of human brucellosis, a zoonosis that can similarly be transmitted through consumption of contaminated milk and direct contact with livestock, also identified isolated and remote households as a significant risk factor for infection in the Arusha region (Dr. J. Kunda, unpubl. data). Despite the finding of raw milk consumption as a risk factor in the preliminary cross-sectional analy-

ARTICLE IN PRESS 40 sis carried out by Mfinanga et al.,4 no livestock factors emerged as significant risk factors for human M. bovis infection from either the univariate or multivariate analyses in this study. Indeed, three of the seven cases (42.9%) came from households that did not own cattle, and of those in cattleowning households, only one contained a positive reactor. The failure to detect raw milk consumption, cattle ownership and intradermal test positivity as risk factors may be the result of the low power of this study and/or the low resolution provided by questionnaire data on detailed milk consumption and preparation practices. However, further investigation is clearly required to investigate the relationship between cattle and human infection, and potential sources of zoonotic transmission that involve factors other than owned livestock, for example, consumption of milk from communal sources, such as markets. Although these results provide an indication of potential risk factors for M. bovis infection in humans, the study is clearly limited by the small number of cases. The low sensitivity of culture for isolation of Mycobacteria from human and animal samples collected from local hospital and field sites indicates a clear need for improved field protocols and diagnosis methods. This is likely to be critical for increasing the power of any future case–control studies, particularly in rural African settings, where M. bovis and other Mycobacteria species may contribute substantially to cases of human extrapulmonary tuberculosis.

Prevalence of infection in cattle The a priori expectation was to find relatively high levels of infection in cattle from the Arusha Region because of the high prevalence of extrapulmonary TB cases that have previously been documented in the human population.6 Therefore, the observed intradermal test prevalence of only 0.86% was lower than expected. Despite this low individual prevalence, a relatively high proportion of herds (12%) and most villages (88%) contained at least one reactor animal, and we therefore conclude that, although M. bovis occurs at low levels, infection is widespread in cattle throughout the region. In comparison to the prevalence of reactors (0.9%), the prevalence of visible granulomatous lesions in cattle carcasses was relatively high (19.8%). However, M. bovis was isolated from only a small proportion (7.2%) of culture-positive lesions in the abattoir survey. A caveat remains, however, because a low isolation rate of Mycobacteria may have resulted from reduced sensitivity of culture

S. Cleaveland et al. arising from prolonged storage at field sites and the freeze–thaw cycles that occurred during transportation. Furthermore, contamination of lesions and overgrowth of M. bovis with environmental Mycobacteria remains a possibility, which may lead to an underestimate of the true infection rate of M. bovis from abattoir samples. The age-related increase in prevalence is consistent with an endemic pattern of infection in cattle. In an endemic situation (where infection is continuously present), animals are increasingly likely to have become exposed with time and hence show increasing positivity with age. An alternative explanation is that is that, under stress or old age, latent infections may be reactivated and lead to development of a positive skin test.31 Although this phenomenon has been used to explain the predominance of human tuberculosis in older people,32 it is not known whether this is a major feature of M. bovis infection in cattle. As in other studies in many other parts of the world, higher prevalences of intradermal test positives were recorded in larger herds.2,5,22 This might be expected if densities and contact rate (and hence the potential for transmission) increase with increasing numbers of individuals (for example, if the herd is confined within an enclosure or house at night). An alternative explanation relates to extinction rates; in large herds the probability of infection becoming extinct as a result of stochastic factors is likely to be lower than in small herds. Whatever the underlying mechanism, the finding of an increased prevalence in large herds suggests that within-herd transmission routes are likely to be important in M. bovis transmission in Tanzania.

Transmission routes in cattle This study raises several important questions about the relative contribution of different transmission routes of M. bovis in cattle. Most authors indicate that aerosols are the main route of cattle-to-cattle transmission.9,18,33 Consistent with this is our finding of an increased mean prevalence in those herds kept inside at night. The predominance of tuberculous lesions in the respiratory tract of cattle in previous studies (reviewed by O’Reilly and Daborn9) has suggested that it is the respiratory route that is the principal route of transmission. However, in this study, the majority of visible lesions were recorded in the gastrointestinal tract rather than the respiratory and although culture results were available for only a small proportion of these, 50% M. bovis positive lesions were recorded in the gastrointestinal tract.

ARTICLE IN PRESS Mycobacterium bovis in rural Tanzania Alimentary routes of transmission may therefore also be important in rural Tanzania. Stochastic modelling of the transmission dynamics of M. bovis infection may yield information about the relative importance of these different transmission routes. The high frequency of gastrointestinal lesions may result in faecal excretion and widespread environmental contamination, creating an additional source of infection for cattle. As most herds share common grazing and watering points, environmental contamination from a single infected herd has the potential to infect several herds. An external (environmental) source of infection coupled with low transmission rates, is consistent with the observed pattern of relatively large numbers of positive herds, but low prevalences within herds. Many factors affect the survival of M. bovis in the environment, including exposure to sunlight, soil pH, temperature, moisture and the natural microflora.33,34 Although cattle-to-cattle transmission through naturally contaminated pasture has not been demonstrated, pasture irrigated with M. bovis-infected water has induced infection (Schneller, 1959 cited by Morris et al.33). Of particular interest in this respect is the finding that seasonal flooding in the environs of a village is a significant factor for M. bovis infection in cattle. Areas prone to flooding are known to favour the environmental persistence of other Mycobacteria species35 and high moisture levels assist the survival of M. bovis.36 We therefore suggest that faecal contamination of the environment coupled with prolonged environmental persistence in flooded areas may combine to provide an important source of M. bovis infection in cattle in this area. It is interesting to note that some of the major ‘hot-spots’ of bovine tuberculosis elsewhere in Tanzania and Africa are areas that are prone to flooding, such as the Usangu Plains and Ruvu valley in Tanzania2 and Kafue floodplains in Zambia.37 Similarly, in the Rwenzori National Park (now Queen Elizabeth National Park), Uganda, the highest prevalence of infection in both buffalo and warthog was found in marshland areas adjacent to Lake George.38,39

Wildlife Very little is still known about the role of wildlife in the epidemiology of M. bovis in Tanzania. M. bovis has been isolated from wildebeest (Connochaetes taurinus) and topi (Damaliscus lunatus) in the Serengeti National Park and seropositivity has been detected in Serengeti lions (Panthera leo) and Tarangire buffalo (Syncerus caffer).40 With a high

41 prevalence of infection detected in some populations (e.g. 10% of randomly culled wildebeest in the Serengeti were M. bovis-positive)40 combined with wide-ranging movement patterns beyond park boundaries, wildlife clearly have the potential to act as a source of infection for cattle. Wildlife associations in this study were equivocal. From the univariate analysis, there is an indication that presence of wildebeest may be important as a risk factor. However, wildebeest contact was reported for only 10 herds, which may explain the loss of significance of this factor in the multivariate model. The presence of impala in the vicinity of a household emerged as a significant risk factor for infection in cattle, which suggests that wildlife-tocattle transmission may occur. However, no association was detected between village-level prevalence of interdermal test positives and proximity of a village to the borders of wildlife protected areas (all of which are unfenced), as might be expected if wildlife acted as a major source of infection.

Conclusions The growing concerns about tuberculosis in Africa and the knowledge that M. bovis does contribute to the current human epidemic emphasize the importance of integrating veterinary, medical and wildlife sectors in the investigation and control of this disease. We believe that an important result of the inter-sectoral collaboration achieved in this study has been an increased awareness of zoonotic tuberculosis in Tanzania. This study has provided preliminary data on risk factors for M. bovis infection in human and cattle populations in Tanzania, and raises a multitude of additional questions, such as the relationship between human, livestock and wildlife infection, and the relative importance of socio-economic factors, livestockrelated factors and HIV in human disease. The results from this study also have practical implications for the development of strategies for the control and prevention of bovine tuberculosis in Africa. Test-and-slaughter is often considered inappropriate for African settings, because of expected high costs of repeat visits to herds and low compliance of this policy. However, if the within-herd prevalence of reactors is low, compliance may be higher than anticipated. Indeed, following the cattle testing during the crosssectional study, most reactors were voluntarily culled by owners following advice given by the veterinary-medical team. The management of bovine tuberculosis transmitted from wildlife re-

ARTICLE IN PRESS 42 mains an intractable problem in many countries. If wildlife-to-cattle transmission is an important source of infection in rural Africa, any solution to disease control is likely to have to await development of cattle vaccines.34,41 Despite the common finding of herd size as a risk factor in many countries, management of herd size has never been proposed as a potential approach to disease control. However, in this setting, where withinherd transmission is likely to occur during confinement in houses/enclosures at night, reducing of the number of animals kept together may be a sensible recommendation.

Acknowledgements We would like to thank the Department for International Development for funding this work, the Tanzanian Commission for Science and Technology, Sokoine University of Agriculture and National Institute for Medical Research for permission to carry out this study in Tanzania. We would also like to thank the livestock officers from the Ministry of Water and Livestock Development for the invaluable assistance in the field; NTLP staff in the referral hospitals for their enormous efforts and commitment and Paulo Charles for his effort and enthusiasm in the field. We would also like to thank two anonymous referees for their comments on an earlier version of the manuscript. Ethical approval: Ethical clearance for the human case–control study was obtained from the Medical Research Co-ordinating Committee in Tanzania.

References 1. World Health Organisation. WHO factsheeet 104. Information and Resource Centre HIV, TB and Malaria. http:// www.who.int/mediacentre/factsheets/fs104/en, Geneva, Switzerland: World Health Organisation; 2002. 2. Kazwala RR. Molecular epidemiology of bovine tuberculosis in Tanzania. University of Edinburgh; 1997. 3. Kazwala RR, Daborn CJ, Sharp JM, Kambarage DM, Jiwa SFH, Mbembati NA. Isolation of Mycobacterium bovis from human cases of cervical adenitis in Tanzania: a cause for concern? Int J Tuberc Lung Dis 2001;5:87–91. 4. Mfinanga SG, Morkve O, Kazwala RR, Cleaveland S, Kunda J, Sharp MJ, et al. Mycobacterial adenitis: role of Mycobacterium bovis, non-tuberculosis Mycobacteria, HIV infection and risk factors in Arusha, Tanzania. E Afr Med J 2004;81:171–8. 5. NTLP. Annual report, Dar es Salaam, Tanzania, Department of Preventive Medicine, Ministry of Health, 2000. 6. Catley AP. A report on the prevalence and zoonotic implications of bovine tuberculosis in Tanzania. Centre for Tropical Veterinary Medicine, University of Edinburgh, 1992. p. 19.

S. Cleaveland et al. 7. Mfinanga SG, Morkve O, Kazwala RR, Cleaveland S, Sharp JM, Shirima GM, et al. The role of livestock-keeping in tuberculosis trends in Arusha, Tanzania. Int J Tuberc Lung Dis 2003;7:695–704. 8. Cosivi O, Meslin FX, Daborn CJ, Grange JM. Epidemiology of Mycobacterium bovis infection in animals and humans, with particular reference to Africa. Rev Sci Tech 1995;14:733–46. 9. O’Reilly LM, Daborn CJ. The epidemiology of Mycobacterium bovis infections in animals and man—a review. Tuber Lung Dis 1995;76:1–46. 10. Kazwala RR, Kambarage DM, Daborn CJ, Nyange J, Jiwa SFH, Sharp JM. Risk factors associated with the occurrence of bovine tuberculosis in cattle in the Southern Highlands of Tanzania. Vet Res Commun 2001;25:609–14. 11. Shirima GM, Kazwala RR, Kambarage DM. Prevalence of bovine tuberculosis in cattle in different farming systems in the eastern zone of Tanzania. Prev Vet Med 2003;57:167–72. 12. Weinhaupl I, Schopf KC, Khaschabi D, Kapaga AM, Msami HM. Investigations on the prevalence of bovine tuberculosis and brucellosis in dairy cattle in Dar es Salaam region and in zebu cattle in Lugoba area, Tanzania. Trop Anim Health Prod 2000;32:147–54. 13. Griffin JM, Martin SW, Thorburn MA, Eves JA, Hammond RF. A case–control study on the association of selected risk factors with the occurrence of bovine tuberculosis in the Republic of Ireland. Prev Vet Med 1996;27:75–87. 14. Marangon S, Martini M, Dalla Pozza M, Neto JF. A case–control study on bovine tuberculosis in the Veneto Region (Italy). Prev Vet Med 1998;34:87–95. 15. Cook AJC, Tuchili LM, Buve A, Foster SD, Godfrey-Faussett P, Pandey GS, et al. Human and bovine tuberculosis in the Monze District of Zambia—a cross-sectional study. Br Vet J 1996;152:37–46. 16. Omer MK, Skjerve E, Woldehiwet Z, Holstad G. A crosssectional study of bovine tuberculosis in dairy farms in Asmara, Eritrea. Trop Anim Health Prod 2001;33:295–303. 17. Kaneene JB, Bruning-Fann CS, Granger LM, Miller RA, PorterSpalding BA. Environmental and farm management factors associated with tuberculosis on cattle farms in northeastern Michigan. J Am Vet Med Assoc 2002;221:837–42. 18. Menzies FD, Neill SD. Cattle-to-cattle transmission of bovine tuberculosis. Vet J 2000;160:92–106. 19. Ministry of Health Tanzania. Manual of the National Tuberculosis and Leprosy Control Program in Tanzania, 1995. p. 8. 20. World Health Organisation. Treatment of tuberculosis: guidelines for national programmes. Geneva, Switzerland: World Health Organisation WHO/TB/97.220; 1997. 21. World Health Organisation. Global tuberculosis control. WHO Report. Geneva: World Health Organisation; 2001. p. 8–34. WHO/CDS/TB/2001.287. 22. Mfinanga SG. Mycobacterial adenitis in Arusha, Tanzania: epidemiology and diagnostic challenges. Norway: University of Bergen; 2003. 23. Collins CH, Grange JM. A review: the bovine tubercle bacillus. J Appl Bacteriol 1983;55:13–29. 24. Grange JM, Collins CH. Bovine tubercle bacilli and disease in animal and man. Epidemiol Infect 1987;99:221–34. 25. Office International des Epizooties (OIE). Bovine tuberculosis. Manual of diagnostic tests and vaccines for terrestrial animals. Paris, France: OIE; 2004. 26. National Institute for Medical Research. Guidelines on ethics for health, research in Tanzania. Dar es Salaam, Tanzania: National Institute for Medical Research; 2001. 27. Herring AJ, Inglis NF, Ojeh CK, Snodgrass DR, Menzies JD. Rapid diagnosis of rotavirus infection by direct detection of

ARTICLE IN PRESS Mycobacterium bovis in rural Tanzania

28. 29. 30. 31. 32.

33.

34.

viral nucleic-acid in silver-stained polyacrylamide gels. J Clin Microbiol 1982;16:473–7. Watt B, Rayner A, Harris G. Modern methods in mycobacteriology. Rev Med Microbiol 1993;4:97–105. Pinheiro JC, Bates DM. Mixed-effects models in S and S-plus. New York: Springer; 2000. Woodward M. Epidemiology: study design and data analysis. Boca Raton, FL: Chapman & Hall/CRC Press; 1999. Pollock JM, Neill SD. Mycobacterium bovis infection and tuberculosis in cattle. Vet J 2002;163:115–27. Manabe YC, Bishai WR. Latent Mycobacterium tuberculosis—persistence, patience, and winning by waiting. Nat Med 2000;6:1327–9. Morris RS, Pfeiffer DU, Jackson R. The epidemiology of Mycobacterium bovis infections. Vet Microbiol 1994;40:153–77. Krebs JR. Bovine tuberculosis in cattle and badgers: a report by the independent scientific review group. London: Ministry of Agriculture, Fisheries and Food; 1997.

43 35. Ashford DA, Whitney E, Raghuanathan P, Cosivi O. Epidemiology of selected mycobacteria that infect humans and other animals. Rev Sci Tech 2001;20:325–37. 36. Wray C. Survival and spread of pathogenic bacteria of veterinary importance within the environment. Vet Bull 1975;45:543–50. 37. Stafford KJ. A review of diseases of parasites of the Kafue lechwe (Kobus leche kafuensis). J Wildl Dis 1991;27:661–7. 38. Woodford MH. Tuberculosis in wildlife in the Ruwenzori National Park, Uganda (Part I). Trop Anim Health Prod 1982;14:81–8. 39. Woodford MH. Tuberculosis in wildlife in the Ruwenzori National Park, Uganda (Part II). Trop Anim Health Prod 1982;14:255–60. 40. Cleaveland S, Mlengeya T, Kazwala RR, Michel A, Kaare MT, Jones SL, et al. Tuberculosis in Tanzanian wildlife. J Wildl Dis 2005;41:446–53. 41. ISG vaccine scoping Sub-Committee. Development of vaccines for bovine tuberculosis. DEFRA: London; 2003.