Ministry of Health, Community Development, Gender, Elderly and Children
Dar es Salaam | Tanzania, United Republic of
Primary Affiliation: Ministry of Health, Community Development, Gender, Elderly and Children - Dar es Salaam , Tanzania, United Republic of
Int J Infect Dis 2019 Feb 3;79:142-151. Epub 2018 Dec 3.
Faculty of Veterinary Medicine, Norwegian University of Life Sciences, Oslo, Norway. Electronic address:
Download full-text PDF
R Soc Open Sci. 2018 Sep; 5(9): 180479.
The Royal society Open Science
Bacillus anthracis is an aerobic, gram-positive and spore-forming bacterium belonging to the family Bacillaceae . The release of this bacterium (the causative agent for anthrax) from a dead infected host into the environment induces spore formation , enhancing the agent's ability to survive in the soil for a long time . Despite being well controlled in developed countries, anthrax continues to have a devastating global effect on the poor and marginalized populations that depend on small-scale livestock farming in rural areas . Anthrax is continuously ranked as a significant poverty-related neglected zoonotic disease, defined by the World Health Organization as a disease that ‘perpetuate poverty by affecting not only people's health but also their livelihoods’ [4,5]. Flooding, drought and biological vectors (birds, insects or scavengers) or areas of temporary stagnant water may exacerbate anthrax outbreaks . The release of B. anthracis from an infected host into an aerobic environment with insufficient nutrients to sustain bacterial replication induces sporulation . The B. anthracis spores are resistant to extreme conditions such as pH , heat, cold, desiccation and chemical agents, and may, in specific environments, survive up to 200 years . Owing to the extended persistence of B. anthracis spores in the environment, regular epidemics may occur after a long time, such as a recent outbreak in Sweden after 27 years .
Anthrax affects all mammals, but wild and domesticated herbivorous dominate the numbers, as they are often infected through ingestion or inhalation of spores while grazing . The susceptibility to infection differs depending on the host species , with cattle and sheep being the most vulnerable species followed by goats, dogs and horses . Humans are considered to have a moderate susceptibility, while pigs and carnivores are more resistant . Upon ingestion, spores enter macrophages of a susceptible host and are transported to lymph nodes where they germinate into vegetative form  and migrate into the bloodstream and release toxins which cause systemic effects .
Humans typically get infected with B. anthracis through oral, cutaneous and respiratory routes , and the infection could occur during direct contact when butchering, eating raw or undercooked meat, or handling products from infected animals . Cutaneous anthrax is the most frequently diagnosed form of the disease in humans and occurs within 2–6 days after direct contact with anthrax spores . It presents as a papular to a vesicular ulcer which forms a depressed black eschar which is accompanied by oedema .
The first anthrax outbreak in Tanzania was documented among the wildlife species in the national parks during 1962–1998, causing the death of 1200 impalas, and posed a great risk to humans and susceptible livestock . Later on, sporadic human cases have been reported in different parts of the country. In 1985, a total of 239 human anthrax cases were reported in the Rukwa valley in southwest of Tanzania , and in 1988, a total of 11 human cases of cutaneous anthrax were admitted and treated at Mvumi Hospital in the Dodoma region of central Tanzania after patients came into contact with the infected animal carcasses .
In 1985, hundreds of different species of wildlife carcasses were laboratory-confirmed to have died from anthrax in the Selous game reserve , and in 1988, a big anthrax outbreak in wildlife was reported in the Tarangire national park in which 142 impalas, three zebras, four wildebeests and one giraffe were counted dead . Since then, different species of wildlife and livestock and humans have frequently been affected by B. anthracis, with varying disease patterns between years in terms of the size of outbreaks and species affected .
Anthrax is a notifiable zoonotic disease in Tanzania, and it is a disease of public and animal health importance . Despite the seriousness of anthrax outbreaks in animals, there is a poor surveillance system in the animal sector leading to under-reporting of reportable diseases , including anthrax.
Moreover, episodes of anthrax outbreaks are increasingly becoming a threat to humans, livestock and wildlife in Northern Tanzania, specifically in the Arusha and Kilimanjaro regions. For instance, in November 2016, anthrax outbreaks were reported in Monduli district, Arusha region in Northern Tanzania in which 131 carcasses of wild animals were disposed of and 39 carcasses of domestic animals were reported to be consumed . In the Serengeti ecosystem of Northern Tanzania, serological reactions have been reported in herbivorous species often hunted for bushmeat that comes from wildlife which is smuggled in for human consumption . Spillover infections in wildlife can sustain the disease and become a source of spill-back infection to humans and livestock .
Therefore, recurrent outbreaks of anthrax in Northern Tanzania are probably due to the extensive interactions of human, livestock and wildlife in the interface areas. Sporadic, non-fatal cutaneous anthrax lesions are common in individuals who handle infected meat or come in direct contact with infected animal materials . Although it is well known that cutaneous anthrax is caused by skin contact with contaminated surfaces , during these outbreaks it was not clear which surfaces were the most important vehicle for transmitting B. anthracis to humans in specific geographical and cultural settings.
Other studies have reported that there is limited knowledge on the community's awareness of the role contributed by the interaction of animals and humans in the transmission of zoonotic diseases [29,30]. Our retrospective study of health facilities and animal diagnostic centres from 2006 to 2016 revealed a list of hotspot districts for anthrax outbreaks in Northern Tanzania, and that most reported human cases pertained to cutaneous anthrax infection . Moreover, the Arusha region had a reported incidence of 7.9 human anthrax cases per 100 000 population followed by the Kilimanjaro region with 6.6 per 100 000 population .
During anthrax outbreaks, the multisectoral teams comprising experts from the ministries responsible for human, livestock and wildlife health were dispatched to the affected regions. In these affected areas, a team collaborated with the regional and district's multisectoral teams to contain the outbreaks by mounting preventive and control measures including intensified surveillance, community awareness, improved diagnostic capacity and livestock vaccination against anthrax in the affected areas . In Tanzania, the coordination of response to disease outbreaks is under the One Health coordination desk within the Prime Minister's Office [32,33].
The current study was conducted to identify demographic and behavioural factors associated with cutaneous human anthrax outbreaks in the anthrax hotspot areas of Northern Tanzania. The study was conducted to better understand the causal relations and improve on potential intervention strategies in the region.
The areas for this study were the hotspot districts for anthrax in the Arusha and Kilimanjaro regions of Northern Tanzania. The eligible districts for the Arusha region were Ngorongoro, Monduli and Meru, while in the Kilimanjaro region the study was conducted in Siha, Hai, Rombo and Moshi rural districts. The health facilities involved in each district included Wasso DDH, Endulen Mission Hospital, Pinyinyi, Piyaya, Arash and Magaiduru Dispensaries in Ngorongoro District. Selela, Oltukai, Mto wa Mbu, Mungere, Mbaash and Mswakini Dispensaries were the studied health facilities in Monduli District, while Majengo Dispensary was in Arumeru District. Other health facilities in the study were Hai District Hospital, Sanya Station, KIA and Mtakuja Dispensaries in Hai District; Himo, Rauya RC Dispensaries in Moshi rural district, while Kibong'oto Hospital and Manyata Dispensary were in Siha District. Lastly, Huruma DDH, Nanjara and Karume HC were included in Rombo District. Figure 1 shows the wards where the health facilities and villages involved in the study are located. All the studied districts have a majority of residents practising both subsistence farming and animal husbandry. The study districts are also in the interface areas surrounded by different wildlife conservation areas in the northern circuit of Tanzania. There are soft/porous borders between wildlife conservation areas and human settlements, due to an increased interaction between wildlife and livestock during grazing and at water points  in Northern Tanzania. Humans are also posing a risk of zoonotic disease transmission through farming intensification in close proximity to conservation areas, leading to clearance of bushes (change of landscape) and hence destruction of the wildlife ecosystem, causing an increased rate of contact between disease pathogens and humans, livestock and wildlife [35,36]. The data collection for this study was done from 6 October to 5 December 2016.
This study was of a non-matched case–control design, with cases being retrieved from the local health facilities. For each case, a control was selected from a nearby randomly selected household within the same locality as the eligible case.
The minimum sample size was calculated using the Epitools AusVet sample size calculator (http://epitools.ausvet.com.au/content.php?page=SampleSize) with the assumption that the frequency of exposure in controls was 20%, and the odds ratio was to be detected at 3.0, with 80% power and a 95% confidence interval. With these assumptions, the minimum sample size of 61 cases and 61 controls was calculated. A total of 59 cases and 59 corresponding controls were subsequently recruited.
A case was defined as any person residing in the selected hotspot districts of Northern Tanzania who had ever developed skin lesions by itching of the affected area followed by papular lesions and thereafter a vesicular stage over 2–6 days, eventually developing into depressed black eschar sometimes accompanied by mild or severe oedema . A case was eligible for inclusion in the study if records were found in the medical register at a randomly selected health facility in the hotspot districts of the Arusha and Kilimanjaro regions during the preceding two weeks. The patient should have met the case definition for cutaneous anthrax (as defined above) and his/her name found registered in medical records and had resided in the hotspot districts of the Arusha and Kilimanjaro regions for not less than six months before the time of recruitment. A control was defined as any person who resided in a neighbourhood with an eligible case and had not contracted cutaneous anthrax during the preceding six months. This study excluded anthrax suspected cases with a history of coming from other places apart from the Arusha and Kilimanjaro regions in a period of one week before the onset of signs and symptoms of anthrax. Children under 18 years old were included in the study, but their parents/guardians were interviewed as a proxy on their behalf.
A semi-structured questionnaire was developed in English to be administered to both the cases and the controls. The questionnaire included questions related to potential biological exposure to B. anthracis as well as information about demographic factors such as age, sex, occupation, ethnic group, level of education, district/place of residence, and potential risks linked to travelling outside the village in the last two weeks before onset of the disease. The questionnaires were pretested, and necessary changes were made based on the identified ambiguities. The questionnaire was subsequently translated into Kiswahili, the national language spoken by almost every resident.
Before visiting the eligible households, a brief interview was conducted with the ward and village executive officials. Locally available public health officers, livestock extension officers and natural resources officers were also interviewed to document their views on the occurrences of the human and animal anthrax cases in their areas within a period of one month before the time of data collection.
In each household of an eligible human anthrax case, interviews were conducted using the questionnaire and in the event of an underage case (less than 18 years), the proxy (parent or guardians) was interviewed in the same household. After the case interview, the questionnaire was administered to the head of households near the cases, which served as the control.
The data obtained were entered into a Microsoft Excel® spreadsheet by allowing comparison for duplicate data entry errors, and data cleaning was done to ensure the quality of the information entered in the dataset. The cleaned dataset was then transferred into STATA (Stata15/SE for Windows, StataCorp, College Station, TX, USA) for statistical analysis.
Essential demographic, biological and other characteristics were described for cases and controls. The relationship between anthrax transmission and potential risk factors or other covariates was initially assessed using univariable logistic regression. As many cases were younger compared to the controls, the analysis was split into four age quantiles. Further recoding of all exposure factors as dichotomous (yes/no) variables was done. Candidate variables with p < 0.25 from the initial logistic models were subsequently assessed for collinearity in a cross-tabulation using a Goodman and Kruskal's gamma test. For highly correlated variables, only one of them was selected for further analyses. Variables were identified as confounders and included in the final model if including or excluding the variable altered the effect estimate for another variable by more than 10%.
The first statistical model was developed using a multivariable logistic regression, with a backward elimination strategy among candidate variables. The models were built based upon the Wald test and the likelihood test (p < 0.05). We finally used a Hosmer–Lemeshow test for the goodness of fit and the area under the curve of the receiver operating characteristics to assess the reliability of the final constructed model.
As many exposure variables were correlated, we were faced with the difficulty in establishing a realistic and stable statistical model. Considering the questionnaire used in this study, we noted that there were groups within our investigated population. Those groups were characterized by different patterns of behaviour, caused by disparate preferences, which could lead to anthrax infection. Still, we could not identify any specific variable which describes such behavioural dichotomy. Hence, we adjusted our statistical analyses using a latent class analysis (LCA) method. All variables linked to the expected exposures to anthrax were used to construct two latent classes using the generalized structural equation modelling (gsem) command in STATA. The binomial family and the logit link function defined the variables. The study subjects were classified with a probability of belonging to an Exposed class and the rest as Not Exposed by using a posterior probability of greater than 0.5 as a cut-off/threshold between the two classes. Based on a directed acyclical graph (DAG) model drawn in the DAGitty software , the final statistical model was established using a structural equation model (SEM). The SEM was also built on the gsem platform with a logit link function between the anthrax cases and the Exposed class. Initially, the primary model was built using the graphical interface in the sembuilder, before modifying the model in the gsem command syntax. Demographic factors such as age, sex, occupation and education as well as the history of travel were used as predictors for Exposed and were not linked directly to anthrax cases. As there was a strong age bias in the dataset due to the high number of young cases, separate SEM models for the first age quantile (less than 20 years) and older study subjects were established.
Cases were recruited from Hai (n = 6, 10.2%), Meru (n = 3, 5.1%), Monduli (n = 20, 33.9%) Moshi DC (n= 3, 5.1%), Ngorongoro (n = 12, 20.3%), Rombo (n = 7, 11.9%) and Siha (n = 8, 13.5%). Figure 1illustrates the relative density of cases recorded in each of the wards from the hotspot districts. The timeline for the cases recorded in the different districts is found in figure 2. Table 1 gives the main categories of the demographic and biological variables recorded. Among the study participants, there were more male (n = 70, 59.3%) than female participants (n = 48, 40.7%). The age range of participants was 1–80 years with a median age of 32 years. Figure 3 shows the distribution of age across cases and education groups. A total of 83 (70.3%) of the study subjects had no formal education. During analysis, it was realized that younger cases (1–20 years) were more recruited, with 26 (44.1%) of the 59 cases, while only four controls (6.8%) were from this group.