Using shared needles for subcutaneous inoculation can transmit bluetongue virus mechanically between ruminant hosts

Bluetongue virus (BTV) is an economically important arbovirus of ruminants that is transmitted by Culicoides spp. biting midges. BTV infection of ruminants results in a high viraemia, suggesting that repeated sharing of needles between animals could result in its iatrogenic transmission. Studies defining the risk of iatrogenic transmission of blood-borne pathogens by less invasive routes, such as subcutaneous or intradermal inoculations are rare, even though the sharing of needles is common practice for these inoculation routes in the veterinary sector. Here we demonstrate that BTV can be transmitted by needle sharing during subcutaneous inoculation, despite the absence of visible blood contamination of the needles. The incubation period, measured from sharing of needles, to detection of BTV in the recipient sheep or cattle, was substantially longer than has previously been reported after experimental infection of ruminants by either direct inoculation of virus, or through blood feeding by infected Culicoides. Although such mechanical transmission is most likely rare under field condition, these results are likely to influence future advice given in relation to sharing needles during veterinary vaccination campaigns and will also be of interest for the public health sector considering the risk of pathogen transmission during subcutaneous inoculations with re-used needles.


Modelling approach
To investigate the relationship between C T value and the probability of transmission two models were considered. In the first model, the probability of transmission (p(t)) at time t is related to the C T value (C(t)) as follows, where α and β are parameters. In the second model, the probability of transmission is assumed to be independent of C T value, so that, where α is a parameter (cf. equation (0) with β=0).
The duration of the period between infection and the appearance of a detectable result via qPCR (equivalent to the incubation period) was assumed to follow a gamma distribution with shape parameter k and mean θ (so that the variance of the distribution is θ 2 /k). The times of the last negative qPCR result and the first positive qPCR result were used to define the interval during which the incubation period was completed.

Parameter estimation
Parameters were estimated in a Bayesian framework. The likelihood for the data needs to account for the possibilities that: (i) those animals which generated a positive qPCR result could have been infected by any of the inoculations (and we do not know which); and (ii) those animals which did not generate a positive qPCR result may not have been infected or were infected, but had yet to complete the incubation period before the end of the experiment.
Furthermore, we assume that each challenge acts independently and multiple successful challenges do not result in a shorter incubation period.
For an animal which generated a qPCR result, its contribution to the likelihood is given by, where J is the total number of challenges, t j is the time of the jth challenge, t neg is the time of the last negative qPCR results and t pos is the time of the first positive qPCR result. The first term in the summation (in braces) in equation (0) is the probability that the animal became infected on the jth challenge (and was not infected in any of the previous challenges) and the second term is the probability that the animal would complete its incubation period between the times of the last negative and first positive qPCR result.
For an animal which did not generate a positive qPCR result, its contribution to the likelihood is given by, ) .
where t end is the time at which the experiment ended and the remaining variables and parameters are the same as in equation (0). The first term in equation (0) is the probability that the animal was infected on the jth challenge but had yet to complete the incubation period by the end of the experiment (cf. equation (0)), while the second term is the probability that the animal did not become infected.
Finally, the likelihood for the complete data set is given by, where a identifies the animal (a=1,2,…,A), δ a is a variable indicating whether (δ a =1) or not (δ a =0) animal a generated a positive qPCR result and L POS and L NEG are the individual contributions to the likelihood defined by equations (0) and (0), respectively.
Non-informative priors (diffuse Normal with mean zero and standard deviation 100) were used for the parameters in the probability of transmission (i.e. α and β in equation (0) or (0)).
Informative priors for the incubation period parameters (k and θ) were calculated by fitting a gamma distribution to the incubation periods reported in 1 . All priors were assumed to be independent of one another.
Samples from the joint posterior distribution for each model were generated using an adaptive Metropolis algorithm 2 in which the scaling factor for the proposal distribution was tuned during burn-in to ensure an acceptance rate of between 20% and 40% for efficient sampling of the target distribution 3 . Three chains were run for 500,000 iterations after burnin, with an appropriate burn-in selected based on assessment of chain convergence (both visually and by the Geweke diagnostic method 4 ).
The two models, (0) and (0), were compared for each experiment (i.e. cattle, sheep intradermal and sheep subcutaneous), so that a total of six models were fitted to the experimental data. If the 95% CI for b included zero, we concluded that there was insufficient evidence to infer an effect of donor PCRemia on the probability of transmission and rejected model (1) in favour of model (2).