- Top of page
Social behaviour has profound effects on the dynamics and evolution of host–pathogen interactions (Alexander 1974; Rand, Keeling & Wilson 1995). Simple epidemiological models predict that the high contact rates which occur within large social groups will elevate the prevalence of directly transmitted infections (Anderson & May 1979), and this prediction is broadly supported by empirical data from a range of social host species (Coté & Poulin 1995). However, transmission is also influenced by other behaviours such as the extent of ranging (Brown et al. 1994), territoriality (Ezenwa 2004; Nunn & Dokey 2006) and dispersal (Brown & Brown 2004). Since these behaviours are often correlated with sociality, relationships between group size and disease dynamics are potentially complex. Modelling studies highlight the potential importance of social behaviour in the ecology and evolution of host–pathogen interactions (Bonds et al. 2005; Cross et al. 2005), but few empirical studies have been conducted.
European badgers (Meles meles) are social mammals which can become infected with Mycobacterium bovis (the causative agent of bovine tuberculosis, TB). Mathematical models of TB within socially structured badger populations predict that infection should persist only above a threshold group size of six (White & Harris 1995) or eight (Smith et al. 1995) members. Although field studies have thus far detected no such effect of group size (Delahay et al. 2000; Vicente et al. 2007), there is abundant evidence that badger social structure plays a critical role in TB dynamics. In undisturbed populations in TB-affected areas, badger movements are largely confined to small group territories, with infrequent dispersal between groups (Woodroffe, Macdonald & da Silva, 1995; Rogers et al. 1998). Patterns of M. bovis infection reflect this social organization: infection occurs in stable clusters of one or a few social groups (Delahay et al. 2000; Woodroffe et al. 2005b), with new infections associated with dispersal between groups (Rogers et al. 1998; Vicente et al. 2007). Culling of badgers (conducted to try to control the disease) disrupts this stable social organization, leading to expansion of badger home ranges (Woodroffe et al. 2006a), increased dispersal (Pope et al. 2007), elevated M. bovis prevalence (Woodroffe et al. 2006b; Woodroffe et al. in press) and disruption of infection clusters (Jenkins et al. 2007b).
Both general epidemiological models (Anderson & May 1979), and those specific to M. bovis in badger societies (Smith et al. 1995; White & Harris 1995), predict that high contact rates within large social groups should lead to high prevalence of infection. However, multiple other correlates of badger social group size could also influence the relationship between group size and M. bovis prevalence; these are summarized in Table 1.
Table 1. Factors predicted to generate a relationship between badger social group size and the prevalence of Mycobacterium bovis infection. A prediction was considered upheld if analyses showed a statistically significant effect in the appropriate direction
|Factor||Reason for potential effect||Predicted relationship with group size||Testable supporting predictions||Supporting prediction upheld?|
|Consequences of group size for M. bovis transmission|
|Group size||Members of large groups experience high intra-group contact rate and hence high risk of infectiona||Positive||Prevalence should be higher in larger groups||No|
|Dispersal||Dispersing badgers are at higher risk of infectionb, and dispersal rates are higher at low population densitiesc,d||Negative||Small groups should contain more immigrants||No|
|Prevalence should be higher in groups containing more immigrants||No|
|Contact with neighbouring groups||Larger groups might invest more in territorial defencee and so experience lower contact rates with neighbours||Negative||Extra-group paternity might be lower in large groups, manifesting in higher mean relatedness||No|
|Prevalence might be lower in groups with higher mean relatedness||No|
|Access to food||Badgers in larger groups might have restricted access to food resourcesf,g and hence higher susceptibility to infection||Positive||Body weight should be lower in larger groups||Marginal|
|Prevalence should be higher in animals with low body weight||No|
|Factors which might affect both group size and M. bovis prevalence|
|Past culling||Past culling could have lowered local badger density in areas affected by TB||Negative||Groups exposed to past culling should be smaller||Yes|
|Groups exposed to past culling should show higher prevalence||No|
|Habitat type||Access to pasture may allow groups to grow largerh, but also entails potentially infectious contact with cattlei||Positive||Groups with greater access to pasture should be larger||No|
|Groups with greater access to pasture should show higher prevalence||No|
|TB-related mortality||Infected groups might suffer higher mortalityj and so become smaller||Negative||Fewer old animals should be found in smaller groups||Yes|
|Factors potentially correlated with group size|
|Home range size||Small groups, in low density populations, occupy large home rangesk and are thus more likely to encounter infection||Negative||Smaller groups should occupy larger home ranges||Yes|
|Prevalence should be higher in groups with large home ranges||No|
Badgers’ daily ranging behaviour could also influence their probability of M. bovis infection. Ranging widely increases the probability of encountering infection in other badgers, in other host species, or in the environment, leading to a possible association between large home range size and high M. bovis prevalence. Badgers live in large home ranges at low population densities, where social groups are also small (Woodroffe & Macdonald 1993); thus an association could be generated between small group size and high M. bovis prevalence.
Badgers’ risk of infection might also be influenced by their level of territorial defence. Territories held by large social groups are vigorously defended through scent marking (Stewart et al. 2001), potentially reducing contact with (and disease transmission from) members of other social groups. Larger social groups might therefore experience lower M. bovis prevalence.
Effects of group size on M. bovis infection could operate through susceptibility as well as through exposure. Badgers in larger social groups have lower body weights (Rogers, Cheeseman & Langton 1997; Macdonald et al. 2002), suggesting that they may be nutritionally stressed and therefore potentially susceptible to infection (Dai, Phalen & McMurray 1998). Such an effect would be expected to cause higher M. bovis prevalence in larger social groups.
In principle, M. bovis infection might influence badger social group size, as well as vice versa. Badgers shedding M. bovis bacilli experience somewhat higher mortality than do those with no evidence of infection (Wilkinson et al. 2000). Such disease-related mortality might potentially suppress the size of infected groups.
The dynamics of M. bovis infection in badgers are further complicated by the existence of an alternative host species. Badgers are able to transmit M. bovis to cattle (Griffin et al. 2005; Donnelly et al. 2006), and cattle-to-badger transmission also appears to be widespread (Woodroffe et al. 2006b). Cattle pasture sustains high densities of badgers’ preferred prey (Kruuk et al. 1979), and the availability of pasture and deciduous woodland have been shown to influence badger group size (da Silva, Woodroffe & Macdonald, 1993). Since foraging on pasture promotes contact with cattle, access to pasture could increase the probability of M. bovis infection in badgers, as well as increasing group size.
The management of TB could also influence the relationship between badger social structure and M. bovis infection. Badger culling was part of British TB control policy for many years (Krebs et al. 1997), lowering local badger densities and hence depressing group size (Tuyttens et al. 2000b; Woodroffe et al. 2008). Since culling was targeted at areas of high TB risk, it could generate an association between small badger group size and high M. bovis prevalence even if no underlying causal relationship existed.
The Randomized Badger Culling Trial (RBCT), a field trial of the effectiveness of badger culling as a control measure for cattle TB in Britain (Bourne et al. 2007), offered a rare opportunity to explore the relationships between host social structure and pathogen prevalence, on a large spatial scale in replicated study areas. We used RBCT data to investigate the relationship between badgers’ social organization and their probability of M. bovis infection. To test the predictions outlined above (summarized in Table 1), we sought evidence of associations between M. bovis prevalence in badgers and (i) the size and age structure of social groups; (ii) evidence of dispersal; (iii) territory size; (iv) body weight; (v) habitat type; and (vi) past culling.
- Top of page
We observed a consistent negative relationship between badger abundance and M. bovis infection, with lower prevalence in large social groups and at high population densities. This contrasts with the predictions of several models of TB dynamics in badgers (Anderson & Trewhella 1985; Smith et al. 1995; White & Harris 1995).
The difference between our empirical findings and model predictions suggests that existing models incorrectly characterize the relationship between badger abundance and M. bovis transmission. Simple models of microparasite infections, assuming either density- or frequency-dependent transmission, predict that infection prevalence should either increase, or remain constant, as host abundance increases (Lloyd-Smith et al. 2005). Although these predictions are upheld for some host–pathogen systems (e.g. Dobson & Meagher 1996; Begon et al. 1999), the negative relationship which we observed suggests that the relationship between badger abundance and M. bovis transmission is fundamentally different from that assumed by existing models. Although host contact rates may be elevated within larger groups of badgers, some other factor appears to influence transmission more strongly, leading to reduced prevalence. Although negative relationships between group size and individual infection risk are often observed where mobile ectoparasites (e.g. biting flies) are shared within a group of hosts (Coté & Poulin 1995), this ‘dilution effect’ is not relevant here, as microparasites cannot move freely between hosts.
We conducted a range of analyses to investigate factors which might explain the relationship between badger abundance and M. bovis prevalence. These analyses should be interpreted with caution, since many are based on indices and indirect measurements. This approach was necessary because our analyses used a ‘snapshot’ of data collected from culled animals; detailed behavioural studies, conducted over several years, would provide more accurate information on factors such as group size, dispersal, and ranging behaviour. Nevertheless, our approach provided data across a wider range of environmental conditions, and with far larger sample sizes, than would have been feasible for a behavioural study. Moreover, post-mortem diagnosis of infection has a higher sensitivity than does the clinical sampling necessitated by longitudinal studies (Clifton-Hadley, Wilesmith & Stuart 1993). Overall, we consider this study complementary to smaller-scale longitudinal studies.
It is difficult to construct a scenario in which large group size per se could cause low prevalence, and we therefore hypothesized that some third factor might be causally related to both measures. Our analyses confirmed that immigrant badgers were particularly likely to be infected; this is consistent with the findings of previous studies (Rogers et al. 1998; Pope et al. 2007; Vicente et al. 2007). However, immigration, as estimated using microsatellite markers, was not significantly correlated with group size and appeared not to explain the relationship between group size and infection.
We likewise found no evidence that the level of relatedness among group members was associated with either group size or M. bovis infection. Mean relatedness was investigated partly because it might provide a long-term index of extra-group paternity and, hence, contact with neighbouring groups. However, lacking detailed behavioural data, we could not fully investigate contact rates between members of neighbouring groups. Such contacts would occur when animals cross into neighbouring territories, or encounter neighbours intruding into their own territories; although we estimated the likely extent of territories, we could not measure home range overlap or the frequency of movement beyond territory boundaries. Radio-telemetry studies have observed such movement patterns regularly in low-density populations (Sleeman, 1992; Tuyttens et al. 2000a) but they may occur less frequently at higher population densities (Woodroffe 1992; Garnett, Delahay & Roper 2005). If badgers do encounter their neighbours less frequently at high population densities, this could explain the lower M. bovis prevalence reported here. However, systematic data are not available to test this hypothesis.
Although past culling was associated with small group size, this appears not to have caused the relationship between group size and M. bovis prevalence. On first inspection, this result appears to contrast with our previous finding that repeated culling, conducted in the course of the RBCT, elevated M. bovis prevalence in badgers (Woodroffe et al. 2006a; Woodroffe et al. 2009) by disrupting their territorial structure, expanding their ranging behaviour and encouraging immigration (Woodroffe et al. 2006b; Pope et al. 2007; Woodroffe et al. 2008). However, the effects described in this paper refer to the start of the RBCT, when ecological conditions for badgers were much more stable than they became once RBCT culling was established. Pre-RBCT culling occurred on a very localized scale (average 1 km2 targeted (Krebs et al. 1997), compared with 113 km2 for proactive and 9 km2 for reactive RBCT culling (Bourne et al. 2007)) and removed comparatively small numbers of badgers (average 15 badgers/trial area/year (Donnelly et al. 2006), compared with 314/trial area for initial proactive culls). Moreover, on average 5 years had elapsed between the most recent ‘past’ culls and the proactive culls analysed here, which exceeds the average badger lifespan (Wilkinson et al. 2000) and is sufficient to allow substantial recovery of the badger populations (Cheeseman et al. 1993; Tuyttens et al. 2000b). Under these circumstances, it is perhaps unsurprising that we detected no effect of past culling on M. bovis prevalence.
The hypothesis that small group size might be a consequence, rather than a cause, of high M. bovis prevalence could not be fully tested in this study. Members of small groups were younger, on average, than members of large groups, a pattern which might in principle be caused by disease-related mortality. However, several other mechanisms could generate the same pattern. A very small proportion of badgers culled in the RBCT showed severe pathology (Jenkins et al. 2007a), consistent with the finding of only modest increases in mortality associated with M. bovis infection (Wilkinson et al. 2000); demographic modelling would be needed to determine whether such mortality would be sufficient to suppress group size. Since females in smaller groups experience higher fecundity (Woodroffe & Macdonald 1995; Macdonald et al. 2002) and potentially higher cub survival (Woodroffe & Macdonald 2000) than females in larger groups, increased recruitment might compensate for elevated mortality.
Although high badger abundance was associated with low prevalence of M. bovis infection, conditions of high badger density would not necessarily reduce the risks of TB transmission to cattle. Odd ratios suggest that doubling badger group size (or density) reduced prevalence by 20–25%: this means that the absolute number of infected badgers would still increase with group size (or density), even though the proportion of badgers infected would decline. This suggests that the risk of transmission to cattle should still be lower in areas with naturally low badger density, unless some other aspect of badger behaviour or ecology, such as wider ranging or use of farm buildings, increased contact between badgers and cattle at low badger densities. Unfortunately for managers, there is strong empirical evidence that attempting to achieve low badger density artificially, by culling, prompts a cascade of behavioural responses which increase badger-to-badger transmission (Woodroffe et al. 2006b; Woodroffe et al. 2009) and undermine benefits for cattle (Donnelly et al. 2003; Donnelly et al. 2006).