14 June 2011

7 minute read

ABSTRACT

Q fever is a worldwide zoonosis caused by the bacterium Coxiella burnetii. The control of this infection in cattle is crucial: infected ruminants can indeed encounter reproductive disorders and represent the most important source of human infection. In the field, vaccination is currently advised in infected herds but the comparative effectiveness of different vaccination protocols has never been explored: the duration of the vaccination programme and the category of animals to be vaccinated have to be determined. Our objective was to compare, by simulation, the effectiveness over 10 years of three different vaccination strategies in a recently infected dairy cattle herd.

A stochastic individual-based epidemic model coupled with a model of herd demography was developed to simulate three temporal outputs (shedder prevalence, environmental bacterial load and number of abortions) and to calculate the extinction rate of the infection. For all strategies, the temporal outputs were predicted to strongly decrease with time at least in the first years of vaccination. However, vaccinating only three years was predicted inadequate to stabilize these dynamic outputs at a low level. Vaccination of both cows and heifers was predicted as being slightly more effective than vaccinating heifers only. Although the simulated extinction rate of the infection was high for both scenarios, the outputs decreased slower when only heifers were vaccinated.

Our findings shed new light on vaccination effectiveness related to Q fever. Moreover, the model can be further modified for simulating and assessing various Q fever control strategies such as environmental and hygienic measures.

DISCUSSION

In this study, we explored the long term effectiveness of three different vaccination strategies in a recently infected dairy cattle herd, through a modelling approach. Mathematical models are nowadays one of the most effective tools to compare control measures for both human and animal infectious diseases. Here, we focused on vaccination since it is widely used in the field and was identified as a long-term control option for C. burnetii infections by the EFSA . Vaccination with a phase I vaccine in cattle was indeed shown to suppress shedding in milk, placenta, uterine fluid, vagina and colostrum. In Rousset et al., the vaccine appeared neither able to prevent infection in exposed kids, nor to clear infection in infected goats, but effectively reduced the level of shedding in a heavily infected herd. Hogerwerf et al. also found that both the prevalence of shedders as well as the bacterial load in uterine fluid, vaginal swabs and milk were reduced in vaccinated dairy goats.

The vaccination scenarios tested in our study were vaccination of (1) the whole herd for 10 years,  (2) vaccination of the whole herd for 3 years, and (3) vaccination of the heifers for only 10 years. Scenario 1 was predicted the most effective control strategy. In fact, all three vaccination strategies reduced the prevalence of shedders, the environmental bacterial load and the number of abortions. However, their effectivenesses were not equivalent. Since the infection was seldom eradicated in the first years of vaccination, an early cessation of vaccination (scenario 2) would be ineffective on the long run. Its short-term effect on infection dynamics depends on the duration of immunity for effectively vaccinated cows. According to Rodolakis et al., in infected herds, more than 80 per cent of the vaccinated cows still had immune markers one year after vaccination.

However, at the same time, less than 60 per cent of the vaccinated heifers were still skin- test positive. In the field, this means that immunity should last between one year (scenario 2A) and life long (scenario 2B). In this context, the increase in the prevalence of shedders, the environmental bacterial load and the number of abortions should not be observable in the first months following the cessation of vaccination. Nevertheless, the infection is spreading again. Thus, before stopping a vaccination programme on a farm, it seems essential to determine the presence or absence of C. burnetii in the herd. Diagnostic tests at a herd level (e.g. PCR in bulk tank milk) can probably be helpful, although they are imperfect.

According to our simulations, when only the heifers are vaccinated yearly (scenario 3), the decrease in the prevalence of shedders, the environmental bacterial load and the number of abortions is slower than when all the animals are vaccinated (scenario 1): it takes between nine and 10 additional months to reach the same level of prevalence of shedders and environmental load, although the two strategies only differ in the initial action of the control programme. The extinction rate is high for both scenarios.

Thus, although scenario 1 seems the best strategy from an epidemiological point of view, the difference between scenarios 1 and 3 is not so marked and a cost-benefit analysis would be useful to better compare the relative interest of these two strategies. It has to be highlighted that the numerical results of our study partially depend on the model structure and parameter values. The model represented the heterogeneity of shedding which is known to affect infection dynamics and hence the intervention efficacies in many diseases. Indeed, sensitivity analysis show that model parameters governing the shedding levels, the characteristics of the bacterium in the environment as well as some of the physiological parameters strongly influenced the C. burnetii dynamics (see Additional File 1: section 2.1.).

Here, parameter values were inferred or calibrated from field data of naturally infected dairy cattle herds. Thus, we took into account the latest knowledge on C. burnetii infections. Although numerical values of the most influential parameters strongly influenced the output of numerical values, they did not change the relative ranking of the vaccination strategies. The only impact was that, for some combinations of parameter values, the differences between simulated effectiveness for the different vaccination scenarios became less marked, especially between scenarios 1 and 3.

The probability of infection for an effectively vaccinated susceptible cow p_v was quantified based on the hazard ratio of the probability of shedding for vaccinated using non pregnant cows provided in Guatteo et al. Since the confidence interval of this parameter was wide, we studied the influence of this parameter value on the model outputs. Although the mean shedder prevalence was highly influenced by the p_v/p ratio, the mean environmental bacterial load (which indirectly represents the infection risk for both animals and humans) decreased by roughly the same rate regardless of the ratio value.

This is likely because the effectively vaccinated animals shed in decreased quantities. Therefore, irrespective of whether the mean prevalence of vaccinated shedders remains high, the prevalence of high shedders was reduced, with a major impact on the environmental load. This result has also been described by Lu et al. [38] who showed that to reduce the Salmonella prevalence in the long term, highly effective vaccines lowering the infectiousness would be a better choice than highly effective vaccines reducing susceptibility. Interestingly, the environmental bacterial load was hardly sensitive to the p_v/p ratio (infection probability for effectively vaccinated cows), whereas the extinction rate was sensitive.

Therefore, if the vaccine is to be used for eradication of C. burnetii from infected farms, both susceptibility and infectiousness of vaccinated animals have to be determined more accurately in order for the model to be used for prediction purposes or decision support. According to Rousset et al. [22], the lowest shedding level in vaginal swabs was shown to be more frequent in vaccinated than non vaccinated goats. However, further studies are needed to determine if a decrease of infectiousness is observed for all vaccinated animals or only for the animals vaccinated when non pregnant and still uninfected, and to quantify this decrease in all the shedding routes.

It should be noted that the extinction rate is highly influenced by the effect of vaccination on the susceptibility, the level of shedding and the mortality rate of the bacterium in the environment (see Additional file 1: section 2.1.2.), which are all not accurately documented variables. This extinction rate should then be interpreted with caution and used to compare different control strategies within the model. However, the behavior of the extinction rate suggests that it takes time to get free from C. burnetii within a herd.

This model was developed for dairy herds and the results presented here can not be straightforwardly generalized to small ruminants. A major difference is flock management. The typical size of a dairy flock is often much higher than the typical size of a dairy herd and kidding is usually synchronized. Therefore, the way to represent demography in the model should be adapted when representing C. burnetii spread in small ruminants. Besides, shedding characteristics and clinical manifestations may be different between species: according to Rodolakis et al., ewes were found to shed mostly in faeces and vaginal mucus while goats seem to shed mostly in milk. Arricau-Bouvery et al. reported that high abortion rates were rare, except in some caprine herds.

However, all considered, our model represents an extensively documented basis for further development. In conclusion, although an additional cost-benefit analysis considering the economic aspects of control programmes is needed to design an optimal control strategy, our modelling approach showed that a long term yearly vaccination would reduce infection risk in vaccinated herds.