INTRODUCTION
Bovine tuberculosis (bTB) caused by Mycobacterium bovis is widespread in many countries, especially in Africa, and has caused significant economic losses (Humblet et al. Reference Humblet, Boschiroli and Saegerman2009). Previous studies have identified various risk factors for bTB transmission at different levels, such as herd size, previous infection status and cattle movements at herd level (Humblet et al. Reference Humblet, Boschiroli and Saegerman2009; Skuce et al. Reference Skuce, Allen and McDowell2012). Increasing evidence suggests that wildlife species play an important role in bTB dynamics as well (Corner, Reference Corner2006; Renwick et al. Reference Renwick, White and Bengis2007; Humblet et al. Reference Humblet, Boschiroli and Saegerman2009; O'Brien et al. Reference O'Brien, Schmitt, Rudolph and Nugent2011; De Garine-Wichatitsky et al. Reference De Garine-Wichatitsky, Caron, Kock, Tschopp, Munyeme, Hofmeyr and Michel2013). This finding invokes further research efforts especially on multi-species transmission in livestock-wildlife interactions, as studies on the role of wildlife species often only involved single-species transmission (Humblet et al. Reference Humblet, Boschiroli and Saegerman2009; Huang et al. Reference Huang, de Boer, van Langevelde, Xu, Ben Jebara, Berlingieri and Prins2013b ).
It has been hypothesized that greater species diversity is able to reduce pathogen transmission, i.e. the dilution effect (Keesing et al. Reference Keesing, Holt and Ostfeld2006, Reference Keesing, Belden, Daszak, Dobson, Harvell, Holt, Hudson, Jolles, Jones, Mitchell, Myers, Bogich and Ostfeld2010; Ostfeld and Keesing, Reference Ostfeld and Keesing2012; Johnson et al. Reference Johnson, Preston, Hoverman and Richgels2013; Joseph et al. Reference Joseph, Mihaljevic, Orlofske and Paull2013). This dilution effect can operate through different mechanisms, such as decreasing encounter rates among competent hosts or regulating host abundance (Keesing et al. Reference Keesing, Holt and Ostfeld2006). Although the dilution effect has been found in many different disease systems, it is still highly disputed whether the dilution effect generally occurs (Ostfeld and Keesing, Reference Ostfeld and Keesing2012; Randolph and Dobson, Reference Randolph and Dobson2012; Huang et al. Reference Huang, de Boer, van Langevelde, Olson, Blackburn and Prins2013a ; Ostfeld, Reference Ostfeld2013). In Africa, many mammal species can be infected by M. bovis, and these species have been classified as maintenance hosts (such as African buffalo, Syncerus caffer), spillover hosts or dead-end hosts (Corner, Reference Corner2006; Renwick et al. Reference Renwick, White and Bengis2007). Since the competence to transmit M. bovis varies among different species, we expect that mammal species diversity can affect bTB transmission. We hypothesized that higher mammal species richness (MSR) may trigger a dilution effect and reduce bTB risk by increasing the abundance of incompetent hosts which are able to interfere with the bTB transmission pathways. On the other hand, there might also be an identity effect, implying that the occurrence of particular species in the community changes the disease risk, either positively or negatively (Hantsch et al. Reference Hantsch, Braun, Scherer-Lorenzen and Bruelheide2013). In this study, therefore, we tested the identity effect of buffalo, considering that the occurrence of maintenance host species might increase bTB risk because of their high competence in transmitting M. bovis. In addition, since studies have not reached an agreement on the role of greater kudu (Tragelaphus strepsiceros) in bTB transmission (Renwick et al. Reference Renwick, White and Bengis2007; Humblet et al. Reference Humblet, Boschiroli and Saegerman2009), we also tested for an identity effect of this species.
A previous study tested the dilution effect and found that MSR was negatively associated with the probability of bTB presence (Huang et al. Reference Huang, de Boer, van Langevelde, Xu, Ben Jebara, Berlingieri and Prins2013b ). However, spatial pathogen dynamics has several distinct epidemiological phases including new establishment in an area, persistence, fade-out and recurrence, and the patterns of different phases might be driven by different determinants (Allepuz et al. Reference Allepuz, Casal, Napp, Saez, Alba, Vilar, Domingo, Gonzalez, Duran-Ferrer, Vicente, Alvarez, Munoz and Saez2011; Skuce et al. Reference Skuce, Allen and McDowell2012). Therefore, we tested the dilution effect as well as the identity effect in the persistence and recurrence of bTB and compared the differences in these two distinct phases. In addition to wildlife species, other factors are also responsible for facilitating bTB transmission, such as cattle density, the status of ‘neighbourhoods’ and climate (Humblet et al. Reference Humblet, Boschiroli and Saegerman2009). However, since studies usually investigated the effects of these factors at herd level and relatively fewer studies focus on regional scale (Humblet et al. Reference Humblet, Boschiroli and Saegerman2009), we also tested whether these variables act as risk factors for bTB risk at regional scale. The present study aimed at testing the dilution effect as well as the identity effects of two wildlife species (African buffalo and greater kudu) in the persistence and recurrence of bTB in cattle. We also examined which other factors were associated with bTB persistence and recurrence and tested whether they are associated with different sets of predictors.
MATERIALS AND METHODS
BTB data
Data of bTB in cattle in Africa from 2005 to 2010 were provided by the World Animal Health Information Database (Ben Jebara et al. Reference Ben Jebara, Caceres, Berlingieri and Weber-Vintzel2012) from the World Organisation for Animal Health (OIE), which reported the status of bTB at the administrative level every 6 months during this period. The lowest administrative level of reporting was used as the level of analysis in this study. Only countries with bTB outbreak histories were included in the analyses, since we focused on bTB recurrence and persistence. The compiled dataset has 879 rows covering 11 countries and 190 administration areas (Table S1).
We classified a disease fade-out if bTB was not reported in two consecutive reporting periods (therefore one year) following a presence. This criterion has been used in previous studies to identify bTB persistence and fade-out (Brooks-Pollock and Keeling, Reference Brooks-Pollock and Keeling2009; Karolemeas et al. Reference Karolemeas, McKinley, Clifton-Hadley, Goodchild, Mitchell, Johnston, Conlan, Donnelly and Wood2011). Therefore, a bTB presence in a specific period was considered as bTB recurrence if bTB was not reported in the administrative areas in the previous two reporting periods. BTB persistence was classified if a bTB presence was reported within a year (i.e. within maximally two 6-months reporting periods).
Wildlife data
MSR were calculated from the African Mammal Databank (AMD), an atlas of medium to large mammals (Biotani, Reference Biotani1998). For each mammal species, the AMD includes two polygon coverage files respectively describing the distribution of suitable habitat and the distribution of species occurrence at a 1×1 km resolution (Biotani, Reference Biotani1998). The intersection of these two distribution maps was calculated as the ‘actual distribution’ for each species (Boitani et al. Reference Boitani, Sinibaldi, Corsi, De Biase, Carranza, Ravagli, Reggiani, Rondinini and Trapanese2008). Since small mammals are less often involved in bTB transmission in Africa, only species with an average body mass ≥2 kg from the AMD were used in the analysis. The presence or absence of each mammal species was recorded in each administrative area and were compiled to calculate the species richness of each administrative area. Using species distribution maps from the AMD, the percentage of the area occupied by African buffalo (Buffalo) and greater kudu (Kudu) were calculated for each of the administrative areas to test the identity effects of these two species.
Other predictors
Many studies have identified that cattle density is an important predictor for bTB outbreaks (Gilbert et al. Reference Gilbert, Mitchell, Bourn, Mawdsley, Cliton-Hadley and Wint2005; Mathews et al. Reference Mathews, Lovett, Rushton and Macdonald2006; Humblet et al. Reference Humblet, Gilbert, Govaerts, Fauville-Dufaux, Walravens and Saegerman2010; Vial and Donnelly, Reference Vial and Donnelly2012; Huang et al. Reference Huang, de Boer, van Langevelde, Xu, Ben Jebara, Berlingieri and Prins2013b ). The bTB status of ‘neighbourhoods’ might also play an important role in explaining and predicting bTB risk (White et al. Reference White, Martin, De Jong, O'Keeffe, More and Frankena2013). In addition, previous studies also indicated that climate can substantially influence bTB dynamics through the influence on the survival of M. bovis (Wint et al. Reference Wint, Robinson, Bourn, Durr, Hay, Randolph and Rogers2002; Rhodes et al. Reference Rhodes, De Leij and Dale2007; Humblet et al. Reference Humblet, Boschiroli and Saegerman2009, Reference Humblet, Gilbert, Govaerts, Fauville-Dufaux, Walravens and Saegerman2010). Therefore, we also included these variables to test their effect on bTB persistence and recurrence at a regional scale.
From the Food and Agriculture Organization (FAO), we collected the data of cattle density in 2005, which had a resolution of 0·05 degree. We calculated the mean cattle density for each administrative area. To investigate the effect of ‘neighbourhood’, the percentage of the border that was shared with neighbouring infected areas in the preceding year (BorPre) was calculated based on the bTB presence/absence data from OIE. Twelve climatic variables (Table 1) representing temperature and precipitation conditions were calculated based on the Climate Research Unit (CRU) datasets (Harris et al. Reference Harris, Jones, Osborn and Lister2013). CRU time-series datasets yield month-by-month variations in climate from 1900 to 2010. These are calculated per (0·5×0·5 degree) grid cell, compiled from an archive of monthly mean temperatures provided by more than 4000 weather stations distributed around the world. In addition, to investigate the effect of wildlife-livestock interactions, we included the interaction terms between cattle density and three wildlife variables, i.e. MSR, areas occupied by buffalo and kudu, as potential predictors. The data for all predictor variables were acquired from existing databases (Table S2). All data pre-processing analyses were conducted in ArcGIS 10.0.
Table 1. Descriptions, abbreviations, units and summaries (mean±s.d.) of the predictors used in the analysis
Statistical analyses
Generalized linear mixed models (GLMM) with a binary response were used to examine the effects of predictors on bTB persistence and recurrence. In addition, country was included in the models as the random factor to control for possible differences between countries, because of the lack or incompleteness of the data of veterinary service and used control measures. Before performing the GLMMs, we log-transformed, log(x+0·5), Buffalo, Kudu and BorPre.
Using GLMMs, two approaches were applied to investigate the associations between predictors and bTB persistence or recurrence. First, we used a stepwise selection approach to construct a final multiple regression model. The area of the unit was retained in the model to correct for the effect of area size. We initially included cattle density in the model, since this predictor was shown to have significant impacts on both bTB persistence and recurrence. Then, other variables were added into the model by a standard-entry stepwise procedure, using a Likelihood Ratio Test (LRT) to test for the difference in the fit of the nested models. For highly correlated independent variables, only the one causing the largest change in the Log-Likelihood (LL) was added to the model to avoid multi-collinearity. We included interaction terms after including all main factors. Main terms were maintained in the model if they were included in a significant interaction term. Second, a ‘removal approach’ based on multiple regression frameworks (Gilbert et al. Reference Gilbert, Xiao, Pfeiffer, Epprecht, Boles, Czarnecki, Chaitaweesub, Kalpravidh, Minh, Otte, Martin and Slingenbergh2008; Martin et al. Reference Martin, Pfeiffer, Zhou, Xiao, Prosser, Guo and Gilbert2011) was used to test for the consistency of results obtained via the stepwise selection approach. In this approach, all variables, except for those causing multi-collinearity, were forced in the model. Then, the coefficient and the change in −2LL were estimated on the removal of each independent variable. We tested for the spatial autocorrelation of the residuals (final models in stepwise selection approach and full models in removal approach) using Moran's I index and found little evidence of spatial autocorrelation for the models (Table S3). The whole statistical process was conducted in R 2.15.1 with appropriate packages.
RESULTS
Descriptive epidemiology
The regional prevalence of bTB was 36% during the study period (Table 2). The percentage of areas experiencing recurrence of bTB was 14%, and the percentage of areas with a bTB fade-out was 22% (persistence percentage was thus 88%).
Table 2. Infection prevalence, percentages of persistence and recurrence of bovine tuberculosis (bTB) in the African administrative areas during 2006–2010
Risk factors for bTB persistence
Both the stepwise selection approach and removal approach showed that cattle density was positively correlated to bTB persistence, while the interaction of cattle density and MSR was negative (Table 3). Only one of the investigated maintenance hosts, buffalo, played a significant role in bTB persistence, as indicated by the positive interaction with cattle density in both approaches. Removal approach also detected a positive main effect of buffalo. In addition, the mean precipitation of the driest month in the preceding year (PreRainMin) showed a negative relationship with bTB persistence in both approaches, and the mean temperature of the warmest month (PreTemMax) was negatively correlated with bTB persistence in the stepwise selection approach.
Table 3. Summary statistics (regression coefficient b±s.e.; Adjusted Odds Ratio, AOR and 95% CI, calculated from the first and third quartiles, and P-value) for the predictors correlated with bTB persistence using a stepwise selection procedure and a removal approach (some non-significant variables in the removal approach were not listed)
#Variables are mean temperature of the warmest month (PreTemMax) and mean precipitation in driest month (PreRainMin) in the preceding year, mammal species richness (MSR), percentage of the area occupied by buffalo (Buffalo), the interaction terms CattleD×MSR (CattleMSR) and CattleD×Buffalo (CattleBuff).
* P<0·05; **P<0·01; ***P<0·001.
Risk factors for bTB recurrence
For bTB recurrence, the stepwise selection approach and removal approach generated similar results (Table 4). In contrast to bTB persistence, bTB recurrence was found to be correlated with a different set of predictors (Tables 3 and 4), highlighted by the positive main effect of the area occupied by Buffalo and the absence of climatic effects for bTB recurrence. Similar to bTB persistence, cattle density had a positive effect on bTB recurrence and there was a significant negative interaction between cattle density and MSR.
Table 4. Summary statistics (regression coefficient b±s.e.; Adjusted Odds Ratio, AOR and 95% CI, calculated from the first and third quartiles, and P-value) for the predictors correlated with bTB recurrence using a stepwise selection procedure and a removal approach (some non-significant variables in the removal approach were not listed)
#Variables are mammal species richness (MSR), percentage of the area occupied by buffalo (Buffalo), the interaction term CattleD×MSR (CattleMSR).
*P<0·05; **P<0·01; ***P<0·001.
DISCUSSION
Despite extensive efforts to investigate the influences of wildlife on bTB transmission, the understanding of the impacts of livestock-wildlife interactions and wildlife community structure remains limited (Corner, Reference Corner2006; Renwick et al. Reference Renwick, White and Bengis2007; Humblet et al. Reference Humblet, Boschiroli and Saegerman2009). Here, we demonstrated that wildlife species play a substantial role in both the persistence and recurrence of bTB, and moreover, that these two distinct phases are correlated with different sets of risk factors.
In line with previous studies (Gilbert et al. Reference Gilbert, Mitchell, Bourn, Mawdsley, Cliton-Hadley and Wint2005; Humblet et al. Reference Humblet, Gilbert, Govaerts, Fauville-Dufaux, Walravens and Saegerman2010), our results showed that a higher cattle density facilitates both persistence and recurrence of bTB. This can be explained by the manner of bTB transmission, which is mainly through aerosol transmission (Renwick et al. Reference Renwick, White and Bengis2007; Skuce et al. Reference Skuce, Allen and McDowell2012), a density-dependent transmission (Huang et al. Reference Huang, de Boer, van Langevelde, Xu, Ben Jebara, Berlingieri and Prins2013b ). Increasing host densities can lead to higher encounter rates of susceptible and infectious hosts, thereby promoting persistence and spread of the pathogen (Lloyd-Smith et al. Reference Lloyd-Smith, Cross, Briggs, Daugherty, Getz, Latto, Sanchez, Smith and Swei2005; Brooks-Pollock and Keeling, Reference Brooks-Pollock and Keeling2009).
Recently, the influence of species diversity on disease dynamics has attracted great interest (Keesing et al. Reference Keesing, Holt and Ostfeld2006). A well-known hypothesis is described as the ‘dilution effect’, which states that high species diversity can reduce disease risk (Keesing et al. Reference Keesing, Holt and Ostfeld2006). The negative interaction between cattle density and MSR in our study indicates the occurrence of a dilution effect in this bTB disease system. High MSR can reduce the positive effect of cattle density on both bTB persistence and recurrence. This dilution effect is possibly explained by ‘encounter reduction’, in that the addition of alternative hosts may decrease the risk of pathogen transmission by reducing encounter rates between susceptible and infected hosts (Keesing et al. Reference Keesing, Holt and Ostfeld2006). Many mammal species that can be infected by bTB are spillover or dead-end hosts and do not transmit the pathogen efficiently (Corner, Reference Corner2006; Renwick et al. Reference Renwick, White and Bengis2007). The presence of these non-competent wildlife species might act as barriers to herd movement of cattle and reduce encounter rates among herds, which leads to decreased probabilities of bTB persistence and recurrence in cattle. A few previous studies indicated that herbivores may forage in limited areas in the presence of other species, especially predators (Howery and DeLiberto, Reference Howery and DeLiberto2004). However, the understanding of effects of wildlife on livestock movements is still lacking and needs more studies. In addition, we cannot exclude the possibility that the dilution effect we found might be caused by some other predictors correlating with MSR. For example, higher mammal species might occur in nature reserves where fences are frequently used to prevent contact between wildlife and livestock. These fences are also able to interfere with cattle movements and thus lead to a ‘pseudo’ dilution effect.
Buffalo was found to have a positive identity effect on both bTB persistence and recurrence in cattle. The positive interaction between buffalo and cattle density in bTB persistence indicates that the positive effect of cattle density on bTB persistence was amplified under high buffalo occurrence. Theoretical studies suggested that the probability of pathogen persistence is indeed positively related to host density (Lloyd-Smith et al. Reference Lloyd-Smith, Cross, Briggs, Daugherty, Getz, Latto, Sanchez, Smith and Swei2005). As a maintenance host, buffalo can increase the effective encounter rates among cattle herds by frequently contacting different cattle herds, especially because they have similar habitat and dietary requirements, and thereby increase the effect of cattle density on pathogen persistence. For bTB recurrence, the positive association between buffalo and bTB recurrence might be caused by buffalo that spread the pathogen from infected regions to pathogen-free regions. Also, as a reservoir species, buffalo could preserve the pathogen and transmit it to cattle in the regions without cattle infection, triggering bTB recurrence in cattle. Previous studies suggested that maintenance hosts are more likely to remain in the community with decreased biodiversity (Johnson et al. Reference Johnson, Rohr, Hoverman, Kellermanns, Bowerman and Lunde2012; Huang et al. Reference Huang, de Boer, van Langevelde, Olson, Blackburn and Prins2013a ), which indicated that the positive effect of a maintenance host might also be underlain by the dilution effect. In this study, however, we did not find a significant negative correlation between MSR and buffalo occurrence, suggesting the independence of the identity effect from the dilution effect. For greater kudu, neither bTB persistence nor recurrence was correlated with the presence of greater kudu, which might indicate that greater kudu is not a maintenance host for M. bovis.
Climate plays a role in the epidemiology of bTB through affecting the survival of M. bovis (Wint et al. Reference Wint, Robinson, Bourn, Durr, Hay, Randolph and Rogers2002; Rhodes et al. Reference Rhodes, De Leij and Dale2007). Extreme temperatures and precipitation (i.e. mean temperature of the warmest month and mean precipitation in driest month in the preceding year) were found to be correlated with bTB persistence. This is probably because hot and dry weather precludes long-time survival of M. bovis (Wint et al. Reference Wint, Robinson, Bourn, Durr, Hay, Randolph and Rogers2002; Rhodes et al. Reference Rhodes, De Leij and Dale2007; Humblet et al. Reference Humblet, Boschiroli and Saegerman2009, Reference Humblet, Gilbert, Govaerts, Fauville-Dufaux, Walravens and Saegerman2010). However, we did not find any significant relationship between climatic variables and bTB recurrence, which indicates that climate might exert little effect on pathogen re-establishment in cattle at this scale of analysis.
In summary, our study showed that wildlife substantially influences regional patterns of bTB persistence and recurrence. Mammal species richness was negatively correlated with bTB persistence and recurrence, suggesting the existence of a dilution effect, while buffalo as the maintenance host species had a positive identity effect. Our study also underlines the relevance of different ecological/climatic factors in determining bTB risk, and suggests that bTB persistence and recurrence are correlated with different sets of risk factors. These results are relevant for more effective control strategies and better targeted surveillance measures in bTB.
SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit http://dx.doi.org/S0031182013002357
ACKNOWLEDGEMENTS
We appreciate the help from Francesco Berlingieri, Paula Cáceres and Lina Awada for providing assistance with preparing the data at OIE.
FINANCIAL SUPPORT
This research is sponsored by the Chinese Scholarship Council (No. 2010619023).
