- Top of page
Determining the factors that influence the transmission and persistence of infectious diseases in free-living wildlife is a central question in population ecology, with important implications for disease management. The incidence of infectious disease is defined as the number of new infections out of a total of all susceptibles per unit time. Although this is expected to vary with population size, other factors may also limit host interactions and contact rates (Loehle 1995; Lloyd-Smith et al. 2005). For example, the spatial or social structure of a population may influence the rate of disease spread (Keeling 1999) and disease persistence (Keeling 2000; Hagenaars, Donnelly & Ferguson 2004). Nevertheless, few empirical studies have described the incidence of infectious diseases in socially structured wildlife populations. Such information is, however, crucial to understanding how population structure impacts on disease transmission (Anderson & May 1991) and the implications for management.
Bovine tuberculosis (TB) is a chronic infectious disease caused by the bacterium Mycobacterium bovis and affects a broad range of mammalian hosts (De Lisle, Mackintosh & Bengis 2001). Its re-emergence and the repeated failure to eradicate TB from livestock in several countries have been related to reservoirs of infection in wildlife populations. In Britain, data suggest that bovine TB in both European badgers Meles meles (Lin. 1758) and cattle is a low incidence infectious disease with an apparently relatively low, but variable, transmission rate (Smith et al. 2001; Cox et al. 2005). In the UK, the badger is widely considered to represent a significant wildlife reservoir for the transmission of M. bovis to cattle (reviewed in Krebs 1997). Since 1973, a variety of badger culling strategies have been used by the UK government as the principal tool to reduce risks of transmission to cattle (Krebs 1997). Despite this, however, the number of cattle compulsorily slaughtered owing to TB has continued to rise (Gilbert et al. 2005) and evidence from the UK Randomized Badger Culling Trial (RBCT) showed that localized reactive culling failed to control TB incidence in cattle (Donnelly et al. 2003). The RBCT also showed that, although proactively culling badgers over larger areas did reduce TB incidence in cattle in the culled areas, it increased in adjoining areas (Donnelly et al. 2006). Complementary field studies of badger territoriality suggested that badger culling disrupted the otherwise stable territorial system and could potentially increase rates of contact among badgers and between badgers and cattle (Woodroffe et al. 2005), a phenomenon that could have contributed to the results of the trial.
Further investigation of the factors that influence TB incidence in badger populations is clearly important in informing the development of sustainable approaches to managing risks to cattle. In particular, given the potential for demographic perturbation inherent in management intervention, understanding the role of intergroup movements in the spread of infection is of vital importance. The present paper is based on results of demographic and epidemiological data from the long-term study of the WP badger population and examines relationships between TB incidence and aspects of population structure and movement, at the level of the individual and social group, respectively.
- Top of page
During the period 1989–2004 the number of badger social groups identified in the core of the study area each year by bait-marking remained relatively constant (mean = 25, range = 23–27, Fig. 1a). However, despite the identification by bait-marking of a group range, no resident badgers were captured in one group in 1993, 2002 and 2003. As the number of social groups remained relatively constant, variations in population size during the study (the overall number of badgers varied from 177 in 2004–300 in 2004, n = 15 years, Fig. 1b) were largely driven by fluctuating group size (according to MNA estimation, annual mean group size = 9·61 ± 0·62 individuals, range = 7·00–11·82, n = 15). MNA estimates indicated that the total population size for the core area remained relatively constant from 1990 to 1997, then rose to a peak in 1999 (Fig. 1b), decreasing thereafter. During the study a total of 8981 individual capture events took place, involving 1859 different individuals. Individual badgers were captured on average twice each year (range of annual means 2·12 ± 0·05). In addition, data were included for 467 animals found dead in the study area.
Figure 1. Trends in population size and M. bovis infection in the Woodchester Park badger population (1990–2004) at social group (a,c) and overall population level (b). The term positive group refers to any group where TB excretion was detected in at least in one individual.
Download figure to PowerPoint
population and tb excretion dynamics
Although the number of badger social groups in the study area remained relatively constant, the number of incident and prevalent (TB was detected in at least one individual) groups per year generally increased over the course of the study (Fig. 1a). Both mean group incidence and prevalence steadily increased over time (Fig. 1c). Annual TB incidence and prevalence at the population level were negatively correlated with population size (RS = −0·69, P ≤ 0·01: rs = −0·55, P < 0·05, respectively, n = 15 years) and mean group size (RS =−0·81, P < 0·001; rs = −0·71, P < 0·01, respectively, n = 15 years) in the same year. When compared with population size and group size in the preceding year, relationships with TB prevalence remained negative but were weaker (RS = −0·50, P = 0·06; rs = −0·64, P = 0·01, respectively, n = 15 years) and no significant associations were found with incidence.
The optimal model is shown in Table 1. The model was fitted with the social group (n = 27) and the year (1990–2004, n = 15) both considered as random factors (Z = 1·51, P = 0·06; Z = 2·27, P = 0·01, respectively). Extra-dispersion scale of the model was 0·685. The prevailing prevalence of TB in a group was significantly associated with the risk of an individual becoming an incident case (Fig. 2a). The risk that a susceptible individual would become an incident case was also higher in groups that were declining in size (Fig. 2b). There was also a statistically significant interactive effect of group TB prevalence and group size trend on incidence, such that the influence of the former was less marked in groups that had undergone extreme changes in size (either positive or negative) than in more stable groups (Fig. 2c). Movement behaviour of the individual was also significantly related to the probability of incidence, such that movement within the core area incurred a greater risk than either not moving or moving into the core area from outside (Fig. 3a). In addition, the group movement index was positively related to the individual risk of becoming an incident case in the following year (Fig. 3b) suggesting that individuals in groups that experienced higher levels of movement were more at risk of becoming incident cases. There were also sex-related differences in the probability of becoming an incident case, such that in female-biased groups, males were at a disproportionately higher risk than females (Fig. 4a). A significant interaction between group sex ratio and the movement index indicated that in groups with equivalent or higher numbers of males the positive association between the group movement index and the individual probability of incidence increased (Fig. 4b).
Table 1. Results of a GLMM to partition variation in the individual risk of a badger becoming a TB incident case in year t. Parameter estimates for the level of fixed factors were calculated considering a reference value of zero for female level in the variable ‘sex’ and for movement from outside in the variable ‘individual movement’. P-values appear in bold where less than a priori significance limit of 0·05.
|Sex||1, 2082||−0·85|| 2·20||0·14|
|Age||1, 2067||−0·17|| 0·13||0·71|
|Group size (year t)||1, 514||−0·04|| 2·11||0·15|
|Trend in group size (year t − 1 year t)||1, 1870||−0·01|| 9·30||< 0·01|
|Sex ratio (year t)||1, 1865||−0·001|| 2·25||0·13|
|Group TB prevalence (year t)||1, 1699||0·03||24·53||< 0·0001|
|Individual movement (year t − 1 year t)||2, 2058||0·67,* 1·54†|| 6·47||< 0·01|
|Group movement index (year t − 1)||1, 1547||2·17|| 9·61||< 0·01|
|Number of captures (log10) (year t)||1, 1985||1·35||10·55||< 0·01|
|Sex × sex ratio||1, 2083||0·21|| 5·12||0·02|
|Sex ratio × group movement||1, 2077||−0·04||12·00||< 0·001|
|Trend in group size × group TB prevalence||1, 1622||0·0002|| 4·08||0·04|
Figure 2. Results of a GLMM relating individual M. bovis excretion incidence in badgers to (a) the mean prevalence of excretors in the social group, (b) the mean trend in social group size and (c) the mean group prevalence of M. bovis excretors in relation to the trend in group size (the three classes are based on 33% and 66% percentiles of group size trend). Mean values are shown with 95% confidence interval for SE.
Download figure to PowerPoint
Figure 3. Results of the individual-level GLMM. Incidence (±95% confidence interval for SE) according to individual movement category (a) and group movement index according to individual incident status (b). It should be noted that the partial effects of the other explanatory variables upon mean movement indices are not accounted for.
Download figure to PowerPoint
Figure 4. Results of an individual-level GLMM, showing interactions involving the social group sex ratio. (a) Displays incidence (±95% confidence interval for SE) relative to sex and social group sex ratio categories (percentiles 33% and 66%). (b) Shows the social group movement index (±95% confidence interval for SE) relative to individual incident cases and the group sex ratio.
Download figure to PowerPoint
Badgers that became excretor incident cases were more likely to have been ELISA positive in the previous year than those that did not, and these differences remained regardless of whether they had moved within the core area (14·83 ± 5·51% and 45·00 ± 22·37% for negative and incident individuals, respectively, χ2 = 6·3, d.f. = 1, P = 0·01) or not (16·91 ± 1 ·7% and 48·00 ± 10·38% for negative and incident individuals, respectively, χ2 = 255·1, d.f. = 1, P < 0·001). The GLMM to determine whether serological status (ELISA test result) and physical condition in the previous year were related to the likelihood of a badger moving between social groups identified no significant effect after controlling for sex and group size. Individuals moving within the core area did, however, come from groups with higher previous TB prevalence (8·01 ± 1·78%) than did ‘non-movers’ (6·19 ± 0·45%, d.f. = 1, Fanova = 4·77, P = 0·03).
social group-level analyses
Results of GLMMs to partition the variance associated with the risk of a badger social group becoming an incident case are shown in Table 2: model (a) compares incident vs. negative groups (positive non-incident groups were excluded from this analysis, n = 242 group by year combinations), and model (b) compares incident groups vs. non-incident groups (negative or positive but not incident, n = 355 group by year combinations). The models were fitted with social group (n = 27, Z = 1·59, P = 0·06; Z = 0·98, P = 0·16, for model a and b, respectively) and year (1990–2004, n = 15, Z = 1·33, P = 0·09; Z = 1·20, P = 0·12, for model a and b, respectively) as random factors. The extra-dispersion scales were 0·66 for the model comparing incident vs. negative groups and 0·73 for the model comparing incident groups vs. non-incident groups. When comparing incident vs. negative groups (see Table 2a, n = 242 groups by year), the group size trend was negatively related to the probability of group-level incidence (logit parameter estimate =−0·03 ± 0·006). Overall group size, however, was not related to the probability of being an incident group. The proportion of females in the group was found to be significantly negatively correlated with the probability of group-level incidence (logit parameter estimate =−0·02 ± 0·007). Movement index was not found to be significantly associated with incidence, whereas the number of animals moving from another core group was found to be positively associated (logit parameter estimate = 0·58 ± 0·18, mean annual number of animals moving from the core into negative and incident groups were 0·38 ± 0·11 and 0·75 ± 0·32, respectively). No significant association was found between the rate of annual movement and group-level incidence. The model comparing incident groups with non-incident groups (the latter category including negative or prevalent, but not incident cases, see Table 2b, n = 355) yielded similar results to the first analysis.
Table 2. Results of GLMMs to partition the variance associated with the risk of a badger social group becoming an incident case. Model (a) compares incident vs. negative groups (positive non-incident groups were excluded from this analysis, n = 242 group by year combinations), and model (b) compares incident groups vs. non-incident groups (negative or positive but not incident, n = 355 group by year combinations).
| Group size||1, 118||0·08|| 2·88||0·10|
| Trend in group size (year t − 1 year t)||1, 232||−0·03||18·09||< 0·0001|
| Sex ratio||1, 233||−0·02|| 8·21||< 0·01|
| Group movement index (year t − 1)||1, 216||−0·61|| 1·03||0·31|
| Core movements (year t − 1 year t)||1, 223||0·58||10·11||< 0·01|
| Movements from outside (year t − 1 year t)||1, 228||0·26|| 1·006||0·32|
| Group size||1, 119||−0·001|| 0·00||0·98|
| Trend group (year t − 1 year t)||1, 330||−0·02||18·33||< 0·0001|
| Sex ratio||1, 332||−0·02|| 8·35||< 0·01|
| Group movement index (year t − 1)||1, 225||−0·33|| 0·41||0·52|
| Core movements (year t − 1 year t)||1, 324||0·47||10·81||< 0·01|
| Movements from outside (year t − 1 year t)||1, 334||0·35|| 2·25||0·13|
The analysis considering the number of incident cases is shown in Table 3. The model was fitted with the social group (n = 27) and the year (1990–2004, n = 15), both considered as random factors (Z = 2·33, P = 0·01; Z = 1·36, P = 0·09, respectively). The extra-dispersion scale was 0·81. Group-level TB prevalence was found to be positively associated (logit parameter estimate = 0·04 ± 0·009). As in the individual and other group-level analyses, the prevailing trend in group size was negatively related to the proportion of incident cases in a given group and year (logit parameter estimate = −0·01 ± 0·003) while the movement index of the group for the previous year (logit parameter estimate = 1·12 ± 0·51) was significantly and positively related. A significant interaction between group sex ratio and movement index indicated that the relationship between movement and the proportion of incident cases in a group was stronger when the proportion of resident males was higher (parameter estimate = −0·02 ± 0·009).
Table 3. Results of a GLMM to partition the variation in TB incidence at the level of the badger social group (n = 355).
|Group size||1, 284||−0·02|| 1·15||0·28|
|Trend in group size (year t − 1 year t)||1, 342||−0·01|| 5·70||< 0·05|
|Group TB prevalence (year t)||1, 340||0·04||11·22||< 0·001|
|Group movement index (year t − 1)||1, 333||1·11|| 4·71||0·03|
|Core movements (year t − 1 year t)||1, 337||0·13|| 3·00||0·08|
|Movements from outside (year t − 1 year t)||1, 341||−0·12|| 0·44||0·51|
|Sex ratio × group movement||1, 336||−0·02|| 7·84||< 0·01|
- Top of page
The influence of the social structure of wildlife populations on the movements of individuals may be complex, and is likely to have significant effects on disease transmission dynamics and persistence (Gudelj & White 2004; Cross et al. 2005). Observational studies of wildlife populations seldom consider such interactions (but see Cross et al. 2004; Brown & Brown 2004) although they are likely to be of profound importance in the management of wildlife diseases. Field studies of badger populations show that culling induced perturbations to the social structure may have substantial demographic consequences (Cheeseman et al. 1993; Tuyttens et al. 2000a,b; Woodroffe et al. 2005) and the results of a recent large-scale field experiment suggest that this phenomenon could be counter-productive to the control of TB in cattle (Donnelly et al. 2003; Donnelly et al. 2006). However, little is known of the epidemiological consequences of the social disruption of badger populations that may drive such counter-productive effects. The current study of a long-term data set from a free-living, high-density badger population focused on identifying the main risk factors associated with the incidence of disease (excretion of M. bovis) at both individual and social group levels. The results strongly suggest that intergroup movements, group structure and stability play an important role in disease dynamics and are consistent with the hypothesis that social disruption may increase incidence.
The single most important factor to explain variations in the risk of an individual badger becoming an incident TB excretor, and in the percentage of incident cases in a social group, was the existing prevalence of excretors in the group (Table 1, Fig. 1a). This is unsurprising given the high levels of sociality among group members, including the shared occupation of setts. The presence of infectious adult female badgers in social groups in the study area has previously been associated with new infections (Cheeseman et al. 1988b), and particularly with the proportion of infected cubs (Delahay et al. 2000b). In addition to direct transmission among group members, it has been suggested that the setts themselves might also act as a potential source of infection (Delahay et al. 2000b; Courtenay et al. 2006).
Despite the importance of existing levels of infection in social groups, the risk of becoming an individual incident case was also influenced by movement, particularly within the core of the study area. Such movement was also associated with the emergence of incident groups and with the proportion of incident cases arising in a group (Fig. 3a,b). These results are consistent with the previous observation that for the period 1978–95 at WP, the incidence of infected badgers increased following years of high rates of between-group movement (Rogers et al. 1998). It seems likely therefore that movement between groups either enhances transmission by increasing the frequency of contacts between previously segregated individuals or is related to the progression of disease to the point of excretion of bacilli. However, the highly aggregated distribution of M. bovis infection in badgers (Cheeseman et al. 1988b; Delahay et al. 2000b) means that not all movements incur the same level of infection risk. In the present study the risk of a group becoming an incident case was related to movements among core social groups but was not associated with movements from outside the core area. This probably reflects the location of a persistent localized cluster of infection in this undisturbed, highly structured core of the study population (see Delahay et al. 2000b).
Moving between social groups may also be a risky strategy for an individual badger because of the increased likelihood of aggressive encounters from other territory holders (Kruuk 1978). This provides an additional potential route for disease transmission through bite wounding, which has been suggested to lead to a more rapid progression of disease and excretion of bacilli (Gallagher & Nelson 1979; Clifton-Hadley et al. 1993). In addition, the physiological stress of moving between social groups could contribute towards the development of disease in individuals, by, for example, reducing immunocompetence and permitting the reactivation of latent infections (Gallagher & Clifton-Hadley 2000). Particularly, during periods of reproductive activity, individuals may experience heightened physiological stress and their immune systems may be weakened (Griffin & Thomson 1998). Such seasonal effects could influence movement (see Discussion below) and also disease susceptibility, potentially triggering active excretion of previously latent infection (see Gallagher et al. 1998).
The importance of between group movements of individuals raises the question of why animals move. Although a detailed investigation of the factors that may predispose individuals to move is beyond the scope of the present study, we carried out preliminary analyses to determine whether prior exposure to M. bovis (inferred from a positive ELISA result) or body condition were likely to be influential. Although individuals with a positive ELISA result in the previous year were more likely to become incident excretor cases, neither previous exposure or body condition were associated with the likelihood that a badger subsequently moved groups. Nevertheless, susceptible individuals that moved within the core area were likely to have been resident in a group with a higher prevalence of TB excretors than the resident groups of badgers that did not move. This raises the possibility that the presence of TB in a group could promote dispersal to avoid infection, a phenomenon that has been observed in badgers (Butler & Roper 1996; Roper et al. 2001) and birds (Brown & Brown 1992) in response to ectoparasite burdens. However, to our knowledge, there is no evidence in the literature of a similar phenomenon in microparasite infections.
The present study describes a chronic disease persisting within a population that exhibits social structure, whereby infectious contacts may be frequent within groups, and may spread spatially by host movement between groups (see also Cross et al. 2005). The high degree of social structure in badger populations could promote persistence of TB by allowing infection to occur asynchronously in various groups and avoiding deep ‘global’ troughs (Bolker & Grenfell 1996). This is supported by previous work that indicated that temporal trends in TB in the WP population were not synchronized among neighbouring groups (Delahay et al. 2000a). The balance between TB extinction across groups and group recolonization (referred to in the paper as group incidence) allows for disease persistence in WP (Fig. 1a). In the present study the number of incident groups (patches) did not exhibit dramatic oscillations, varying from 1 to 5 per year (Fig. 1a), and later appearing to stabilize at about four groups per year. The number of prevalent groups tended to stabilize at about 12–14 of the 23–27 social groups present each year. Hence, infection remained present in the population in a relatively high number of groups throughout the study period, but apparently disappeared from some, before re-emerging sometimes several years later. While relatively low levels of intergroup movement (Rogers et al. 1998) appear to limit the rate of between-group transmission in this highly structured population, the chronic characteristics of the pathogen ensure that infection persists.
The risks of becoming an incident case at both the individual and group level were also associated with social group dynamics and structure (Fig. 2b). For example, groups that had diminished in size since the previous year were more likely to become incident groups. In groups that underwent substantial changes in size (either increases or decreases), the influence of the pre-existing prevalence of TB on individual risks of becoming an incident case decreased (Fig. 2c). The study population was stable in terms of social organization as indicated by the relatively consistent number of social groups. In spite of this however, pronounced trends in population dynamics and disease incidence and prevalence were observed. These findings differ in some respects from those of previous studies (e.g. Rogers et al. 1997; Delahay et al. 2000b), which reflects the inclusion in the present study of more recent data, and highlights the inherent value of long-term data sets in identifying demographic and epidemiological patterns that may operate over many years. It is particularly notable that after 1999 there was a generally negative trend in the WP badger population size, driven by changes in group size. Simultaneously the percentage of incident cases observed gradually increased (Fig. 1b), with the result that these phenomena became significantly negatively correlated. Reductions in the size of some social groups in WP observed during this study occurred in the absence of any known human disturbance. Groups that were either decreasing or increasing in size received a greater number of movements from the core area than groups that remained stable. In such circumstances the potential influence of intergroup disease transfer on the probability of subsequent incident cases might be greater than that of the pre-existing prevalence of infection in the group. This could explain the significant interaction observed in the present study.
During the present study, social group size varied substantially over time and between groups but was not significantly correlated with either the risk of becoming an incident group or the percentage of resident incident cases. Modelling studies have suggested that infection will only persist in groups above a certain threshold size (Smith et al. 1995; White & Harris 1995). However, the influence of social structure and movement may confound this simple relationship (Cross et al. 2005; Lloyd-Smith et al. 2005) and impede disease invasion and persistence (Loehle 1995). Results from the present study suggest that transmission and persistence of TB in this badger population may be more constrained by levels of transfer of individuals between groups than by group size. Hence, it may be the process of group size change rather than group size itself that has most influence on levels of infection. In this context it is interesting to observe that Roper, Ostler & Conradt (2003) found that the process of dispersal in badgers often involved an intermediate stage during which an animal might move back and forth between two groups before making a permanent change. Clearly this process would enhance the opportunities for transfer of infection as a consequence of dispersal.
The sex ratio of social groups was also related to the risk of an individual becoming an incident excretor, and its influence varied between the sexes (Fig. 4a). When there were more females present in a group, males were disproportionately more likely to become incident cases than females (Fig. 4b). Consequently, as the adult sex ratio of a group became more female biased, the risk of it becoming an incident case increased. This may relate to the observation that males are more likely to move to social groups with a higher proportion of females in residence (Rogers et al. 1998), and that infection may be more prevalent in males (Gallagher & Nelson 1979; Cheeseman et al. 1988b). Previous work has suggested sexual dimorphism in susceptibility to infection and disease (e.g. Zuk 1990), with males tending to be more susceptible. Alternatively (or additionally), sex differences may relate to the different behaviour of males and females in activities related to breeding. Susceptible male badgers were significantly more likely to move (this was the case for 58·8% of cases moving within the core area and only for 40% of susceptible cases that stayed in the social group). As males generally exhibit higher levels of bite wounding than females (Gallagher & Nelson 1979; Delahay et al. 2006) they may be more susceptible to infection via this route, particularly while making extra-territorial movements. In addition, a higher proportion of females in a group could increase the number of potential breeding opportunities for males, and promote aggressive competitive interactions between males, and also the risk of horizontal female–male transmission. Male badgers are probably more actively involved in territorial defence, which may involve aggressive encounters and enhance opportunities for between-group transmission (Cheeseman et al. 1988a; Rogers et al. 1998). The observation that a male-biased sex ratio was related to a more marked positive relationship between the group movement index and the individual risk of being an incident case in the following year (Fig. 3b) may reflect their potential importance for between-group transmission. On the other hand, females may be more important in the maintenance of infection within groups. Hence the pattern of incidence found in relation to individual sex and group sex composition is broadly consistent with sex-related differences in movement, territorial and reproductive behaviour (e.g. Roper, Shepherdson & Davies 1986).
The long-term study of the WP badger population has produced a unique and valuable data set for the investigation of disease dynamics in wildlife. However, field methods for the study of wild mammals and the diagnosis of M. bovis infection in live animals are imprecise and impose limitations on interpretation of the data. For example, live-trapping records provided information on the movement of badgers between social groups, although this can only provide a relative measure of when and how often badgers moved (Rogers et al. 1998). Similarly, contemporary tests for the diagnosis of M. bovis infection in badgers are relatively insensitive (see Delahay et al. 2000b), and so are likely to underestimate true levels of infection. Nevertheless, these limitations should not affect the main conclusions of the present study, as the repeated sampling of a large number of individuals and the duration of the study should allow broad trends to be identified.
Overall, the findings indicated that within-group dynamics, movement rates and the structure of host contacts were more important in driving disease incidence than density-dependent transmission in isolation (Ball & Neal 2002; Cross et al. 2005). The evidence suggests that movement of individuals between groups may be instrumental in driving disease dynamics at the population level, and adds further support to the contention that the social disruption of badger populations, for example by culling, is likely to promote disease spread. Past badger culling policies have been accompanied by an inexorable rise in the incidence of TB in cattle. Indeed, it has become apparent that the various strategies may actually have been a contributory factor to the increase in disease through the phenomenon of perturbation (Woodroffe et al. 2005). The results presented in this paper lend weight to this argument, demonstrating that stable social structure at least mitigates against new incident cases of disease. Hence, the development of successful strategies for the control of TB in badgers and transmission to cattle will require serious consideration of the likely impact of any interventions on badger social organization.