Chapter 4: Risk factors for foot-and-mouth disease at the wildlife-livestock interface
4.3 Materials and methods
4.4.12 Case control model validation
Comparison between model predictions and the data
When predictions were made from study data using the final case-control model, 27 cases and 42 controls were predicted. Predictions were the same as observations 78.3% of the time, with moderate agreement between predictions and observations (Kappa = 0.57) (Figure 4.15). There were three herds predicted to be cases, although they were really controls. Two of these had greater numbers of cattle than would be expected in a control (270 and 110 head respectively) and the third had acquired new animals over the risk period. Twelve herds were predicted to be controls, although they were really cases. None of these herds had acquired new animals over the risk period and herd sizes were smaller than what would be expected in case herds (Median = 25 cattle, IQR = 15 – 33).
Figure 4.15: Probability of infection predicted by the case-control model compared to case or control status of each data point.
Case – control model with virus isolation confirmed cases only
There were nine case herds from five of the villages with confirmed FMD by virus isolation. These nine cases were selected out along with controls from corresponding villages and the modelling process was repeated. Number of cattle in the herd remained a significant explanatory variable (LRT: χ2= 12.9, p= 0.0003). However, as only one out of the nine cases had acquired new animals in the past month, this did not explain the likelihood of being a case in this group (χ2 = 0.3, p=0.87).
Power analysis for case-control study
The power of the case-control study was estimated as follows. A simulated dataset with an exposure level of 50% for buffalo sightings was generated. An odds ratio of 3 for being a case in association with weekly buffalo sightings was simulated. This simulation was
0.00 0.25 0.50 0.75 1.00
No Outbreak Outbreak
Result from data
Model inferred probability of outbreak
simulated weekly buffalo observations were reduced to 20 out of 70 households, the power was reduced to 45%.
Figure 4.16: Power analysis for the case-control study.
The plot shows the power of a case-control study with a sample size between 10 and 300 to detect an effect that increased the odds of being a case by 3 and had an exposure level of 50%. The broken blue line highlights the power simulated for a
study with a sample size of 70.
4.5 Discussion
Both the cross-sectional and the case-control study generated a consistent conclusion:
livestock management characteristics are the most important drivers of FMD infection in northern Tanzania. In this study, there was no evidence of risk factors associated with wildlife contact. The study also demonstrated higher levels of infection in (a) pastoral and agro-pastoral livestock, (b) cattle and, (c) older animals. Herds with larger numbers of cattle and those that acquired new livestock also had increased odds of having an FMD outbreak.
Potential contact with buffalo or other FMD susceptible wildlife did not explain FMD seroprevalence patterns or FMD outbreaks in livestock in this study. This suggests that FMD drivers in our study area may be different from drivers of outbreaks in southern
controlled in livestock, but endemic in buffalo populations. Multiple reports from these countries implicate buffalo as sources of infection of FMDV for livestock (Caron et al., 2013; Miguel et al., 2013; Thomson et al., 2003; Vosloo et al., 2002a, b, 2010). There are also reports of various antelope species acting as intermediary transmitters of FMDV between buffalo and livestock (Hargreaves et al., 2004; Jori et al., 2009; Vosloo et al., 2006). Reported FMDV prevalence is as high in East African buffalo (67 – 93% NSP antibody seroprevalence (Ayebazibwe et al., 2010a, 2012; Bronsvoort et al., 2008), Chapter 6) as it is in southern African buffalo populations (80 – 100% SAT antibody seroprevalence (Caron et al., 2013)). However the crucial difference between the two ecosystems is its prevalence in livestock. In contrast to southern Africa (Brückner et al., 2002), FMD is prevalent in domestic livestock in East Africa (48 – 76.3% NSP antibody seroprevalence in cattle, Ayebazibwe et al., 2010c; Kibore et al., 2013; Mkama et al., 2014; Namatovu et al., 2013a).
There are many reasons why livestock in an FMDV endemic population pose a more significant source of FMDV for other livestock compared to potential wildlife sources.
Firstly, cattle shed more infectious FMDV than buffalo (Gainaru et al., 1986). FMDV transmission from buffalo to cattle is very difficult to replicate experimentally (Anderson et al., 1979; Bengis et al., 1986; Condy & Hedger, 1974; Dawe et al., 1994; Gainaru et al., 1986; Vosloo et al., 1996) whereas cattle to cattle transmission is a standard procedure in vaccine testing. Secondly, buffalo are dangerous animals and people avoid them if possible and similarly buffalo avoid people. Consistently with this, the households in the current study reported more frequent reports of sightings of wildlife other than buffalo. A study investigating buffalo-livestock contacts in Zimbabwe also reported that cattle and buffalo utilize shared resources such as watering holes at different times if possible, and contacts between the two species are not common (Miguel et al., 2013). Buffalo movements are predictable, facilitating avoidance by cattle herders (Caron et al., 2011). In contrast to buffalo-livestock contacts, this study shows extensive opportunities for contacts between livestock from different areas through movements, acquisitions and shared resources, especially in agropastoralist and pastoralist settings. Our findings fit with increased
were separated from all susceptible wildlife in northern Tanzania, FMDV would persistently circulate in livestock.
The lack of evidence for buffalo to livestock FMDV transmission in this study is consistent with other studies in East African settings. A study in Kenya, albeit with low sample numbers, found no evidence that buffalo and livestock shared SAT serotype FMDV variants (Wekesa et al., 2015). Wildlife contact was not perceived to be an important risk factor for FMD outbreaks by veterinary services in Uganda, and more FMD outbreaks were reported in districts with high cattle movement compared to districts adjacent to national parks (Ayebazibwe et al., 2010b). A study in Cameroon in West Africa found an association between sightings of forest buffalo and reports of FMDV, but noted that this was confounded by people who travelled further afield with their cattle being more likely to see buffalo (Bronsvoort et al., 2004a). Similarly, in the current study, distance walked to reach grazing and water was positively associated with buffalo sightings. However, buffalo sightings did not improve the explanatory ability of the model.
The consistency of the findings in both the cross-sectional and case-control studies adds weight to the conclusion that contact with buffalo does not play a major role in FMDV infection in livestock in northern Tanzania. The statistical power to detect an effect was interrogated in both studies and the cross-sectional study was shown to be sufficiently large to detect an effect. In isolation, the case-control study had relatively less power, but its consistent conclusions potentiate those of the overall study.
Livestock practice was an important explanatory variable for FMDV exposure. Similarly, a study in Ethiopia reported that pastoralists were more likely than settled farmers to own FMD seropositive livestock (Megersa et al., 2009). Larger herds, and increased movements and potential for contacts of agropastoral and pastoral livestock compared to those of smallholders may explain this finding. Consistent with this, the case-control study in the agropastoral district highlighted herd size and livestock acquisitions as risk factors for FMD outbreaks. Each individual animal in the herd has the potential to be exposed to FMDV from an outside source and then to infect its herd mates with close contact.
Therefore larger herds mean more opportunities for livestock to introduce disease into the herd, and more animals within the herd for further transmission. Herd size was also a risk
important also in the spread of other diseases, for example bovine tuberculosis and brucellosis (Cleaveland et al., 2007; Makita et al., 2011).
Recent acquisitions of livestock were also identified as risk factor for FMD outbreaks, which is consistent with another study in Cameroon (Bronsvoort et al., 2004a).
Furthermore, district veterinary officers in Uganda perceived animal movements and the introduction of sick animals to increase the risk of FMD outbreaks (Ayebazibwe et al., 2010b). Such acquisitions may result in the possible introduction of new FMDV variants to naïve animals.
Our finding that cattle are more likely to be seropositive than sheep or goats is consistent with reports from Uganda (Ayebazibwe et al., 2010a; Namatovu et al., 2013), and makes sense in the context of the experimental literature. Cattle are recognised to be more susceptible to FMD and show longer periods of virus persistence compared to sheep and goats (reviewed by Alexandersen et al., 2002 and Arzt et al., 2011a, b). Furthermore, a recent study suggested that, in a mixed population, sheep played a more limited role in the transmission of FMDV than cattle (Bravo de Rueda et al., 2014). However, highly variable FMD patterns of infection and clinical signs have been reported in small-ruminants, and it is likely that the role of this species varies with different breeds and virus variants (Anderson et al., 1976; Barnett & Cox, 1999; Kitching & Hughes, 2002). Whilst small ruminants in our study had lower FMDV seroprevalence compared to cattle, 48.5%
seropositivity in these species still represents a very high burden of infection, which could result in reduced welfare and milk production and mortality of kids and lambs (Chapter 3).
As well as innate host factors as reasons for differences in susceptibility to FMDV, species-specific management factors may explain why cattle have higher FMDV seroprevalence than small ruminants. In the dry season, agropastoralists and pastoralists commonly take their cattle far afield to locate sufficient grazing and water, whereas the small ruminants are left at home (Dr. Lugelo, personal communication). Longer distance movements may result in increased opportunities for contacts with infected livestock.
2001; Lankester et al., 2015b). On the hills, cattle from many different areas mix, presenting a suitable environment for FMDV transmission. This study also shows that cattle are frequently swapped between herds for temporary care new acquisitions are more common than for small ruminants.
Our finding of age as a risk factor for FMD-NSP sero-positivity is consistent with another risk factor studies in Ethiopia (Jenbere et al., 2011)(Bayissa et al., 2011; Megersa et al., 2009) . There are several possible explanations for age as a risk factor for seropositivity to FMDV.
a) Firstly, older animals are as likely as younger animals to succumb to FMDV infections and they have been exposed to other risk factors for longer, and therefore are more likely to be seropositive. Herd owners report FMD lesions in adult cattle at least as often as in young cattle (Figure 3.8, Chapter 3). A study in Kenya reported that FMD lesion incidence rates in an outbreak did not decline with age (Lyons et al., 2015). Older cattle may be as likely to succumb to FMDV infection as:
i. Post-infection immunity against one serotype wanes quickly leaving the animal susceptible to further infections by that serotype.
ii. Many different antigenic types of FMDV are circulating, so past infection with one type will not confer immunity against serial infections with different serotypes.
b) Secondly, NSP antibodies decay slowly. Therefore animals will remain seropositive for many years, as has been demonstrated by (Elnekave et al., 2015).
In persistently infected animals, FMDV in the oropharynx may serve as a long-term immune stimulant for NSP antibody generation (Parida et al., 2005).
c) Thirdly, even if animals have less virus replication in FMDV infections subsequent to their first infection, the anamnestic antibody response will boost NSP antibody levels.
These three hypotheses will be explored further through investigations of serotype-specific
The findings of this study suggest that, while FMD is circulating widely, control efforts should focus on control of livestock related risk factors for infection. There is no indication from this study that measures to separate wildlife from livestock will reduce the FMD burden in northern Tanzanian livestock in the early stages of a control programme. The conventional ranch based fencing and biosecurity measures used in Southern Africa, may not be appropriate for the East African system and alternative vaccine based strategies may be a more workable solution in these settings.
Despite the results of this study, occasional transmission events from buffalo to livestock cannot be ruled out. Whilst FMD is prevalent in livestock, any signal of wildlife-to-livestock FMDV transmission in this study is likely to be drowned out by the dominance of livestock related risk factors. An intervention study, as proposed by Haydon et al. (2002) and Viana et al. (2014), where livestock related risk factors are controlled, or a more in depth study into the rural small holder area where there are fewer livestock related risk factors may be the approach necessary to investigate wildlife-livestock transmission.
Investigations into FMDV circulation in northern Tanzanian buffalo populations (as described in Chapter 6), and antigenic and genetic comparisons between the FMDVs infecting wildlife and livestock are also necessary to address patterns of cross-species transmission.
Although several robust explanatory variables for FMDV seroprevalence and outbreaks were identified, there were also data-points that the models failed to explain, meaning that not all of the variability in the system as been accounted for. The likely reason for this is that the study has addressed only some aspects of FMDV epidemiology. To understand the complex web of host, virus and environmental factors present in this system, a wider range of the epidemiological and ecological tools are necessary. Further work is required to address the dynamics of herd immunity against different variants and serotypes of FMDV, the ecology and diversity of the FMDV population in the region, and the interaction between these two elements. Investigation into which antigenic and genetic types of
The outcome variable in this study, NSP seropositivity, comes from a commercial test that was originally designed for differentiating FMDV infection from vaccination for trade purposes rather than for unraveling the epidemiology of the disease in endemic countries (Chung et al., 2002; Sorensen et al., 2005). Some of the challenges of maximizing the information obtained from the NSP ELISA and other diagnostic tests in the context of FMD in endemic countries are addressed in Chapter 5. The case-control study avoids some of the issues of serological test interpretation by prospectively monitoring the study area and observing clinical signs of FMD rather than diagnosing infection retrospectively. The conclusion of both study types described in this chapter are consistent; livestock management practices are the most important risk factor for FMDV and there is no evidence for wildlife contact related risk factors in the study area.