This is the largest seroepidemiological study of BHV-1 carried out in cattle herds in the UK, and the first for more than 15 years. The seroprevalence of BHV-1 has increased dramatically since the 1960s in both cattle and herds. In the 1960s the seroprevalence in cattle was 2.1% , whilst in the current study, 42.5% of cattle always tested positive. Similarly, the herd prevalence was 20% in 1974 , and 83.2% in the current study. The differences in herd seroprevalence between the current and previous studies could be explained in part if the herds in this study were biased towards a particular location, age or cattle purpose since these were factors that influenced herd seroprevalence (Table 4). However, the consistency of seropositivity across the farms in this study over the 3 year time period suggests that infection has reached a temporal equilibrium, or possibly that seropositivity is increasing slowly.
A particularly interesting result is the higher seroprevalence in herds that were restocked after FMD. Carrique-Mas et al. (2008)  reported that bovine tuberculosis could be introduced into a farm at restocking when cattle were bought from herds with high incidence of infection. It is highly likely that purchase will have brought BHV-1 into restocked herds. The reason for the apparent increase in seroprevalence of BHV-1 is unknown but it might be that movement and mixing in new herds is stressful for cattle, resulting in recrudescence of virus in infected cattle that then infect susceptible cattle in the newly formed herd. In non-restocked herds, purchased cattle had a lower seroprevalence, which is contradictory to this explanation. It should also be remembered that the vaccination status of individual cattle is unknown so that the difference in seroprevalence might be explained by differences in purchasing vaccinated cattle, especially in dairy herds. There may also be complex interactions between cattle age, purchasing and restocking, since restocking of herds is very different to "routine" purchase. Additionally, the existence of multiple antigenic types (which cannot be distinguished serologically) means that interpretation of seroepidemiology is not straightforward. Whatever the cause, these results suggest that these purchased cattle were associated with a high herd seroprevalence and highlight the point that farmers should consider the antibody BHV-1 status of cattle before introduction, to prevent concomitant introduction or re-introduction of BHV-1 of the same or a different antigenic type, that might influence disease presentation.
The seroprevalence of BHV-1 increased with herd size in the univariate analysis and in the hierarchical model; this is consistent with other studies [16, 26–28]. The association with herd size was greater in herds where seroconversion occurred during the study (Figure 2), suggesting that active transmission is related positively to herd size, perhaps because larger herds have more potential transmission contacts both within the herd and with other herds (e.g. from veterinarians, other farmers and purchased cattle). However, the association between BHV-1 PP values and herd size might be confounded with other factors associated with herd size, e.g. recrudescence of infection through stress , or exposure to more viral types.
The age serological profile is consistent with life-long seropositivity. Antibodies to BHV-1 in calves of ≤ 6 months of age probably correspond to colostrum derived maternal immunity. This wanes by 1 year of age, and starts to rise by 2 years of age because of active infection. This suggests that the optimum age for vaccination of cattle is 10 – 14 months of age, although with some active seroconversion observed in adult cattle and a seroprevalence only exceeding 50% in older animals, vaccination is likely to have a positive impact when administered at any age. All the vaccinated herds had seropositive cattle, while approximately 80% of the unvaccinated herds had seropositive cattle. On average, less than 50% of the cattle in the unvaccinated herds tested positive at each visit, whilst >75% of the cattle from vaccinated herds were positive. These data suggest that either some cattle in vaccinated herds were not vaccinated, that vaccination does not always induce antibody or that antibody titre induced by vaccination wanes over time. It is possible that antibody titre generated by natural infection also wanes over time, but there is no previous work on this hypothesis.
In the current study, when the herd was a dairy herd rather than a suckler herd, the rate of seroconversion in adults was greater. The average age of cattle did not differ between dairy and suckler herds, and so this is a true effect, possibly due to winter housing or other stressors related to dairy cow management.
Housing and winter stressors are also an explanation for the small but significant effect of season on PP values. However, 31% and 47% of visits were 11–13 months apart for routine visits 1 and 2 and routine visits 2 and 3, respectively, so that there was a tendency for the same herds to be sampled in the same season and so herd effect is interpreted as an apparent seasonal effect. Other explanations are that seasonal culling and replacement of cattle might influence herd prevalence of BHV-1 antibodies.
All the variables above became non-significant when the binary variable seroconversion (indicating whether seroconversion was observed in the herd) was introduced into the final model. This indicates that the model risk factors are those associated with virus circulation between cattle on the farm.
The use of different modelling approaches has been discussed elsewhere [28, 30] but not modelling different distributions of the same outcome. We compared these two outcomes because we wanted to investigate change in antibody level that was not necessarily occurring around the cut off. The associated risks were similar to those in Table 4 whether the data were modelled using a continuous or binary outcome, but the presence of grower cattle and purchased or homebred were not statistically significant in the binomial model, although the trends were in the same direction of risk. This suggests that these variables alter the antibody titre either always above or below the cut off and that this change was detectable in the continuous outcome model but not in the discrete model.
In the discrete model, we assumed that the serological measurement of seropositive/seronegative had perfect sensitivity and specificity. It is possible that bias could be introduced into the analysis if the sensitivity or specificity of the test varied with other variables, for example, with cattle age, or time since infection, however this has not been reported for BHV-1 antibody tests. The continuous outcome (PP) calculated from the ELISA will have had some random error (Figure 4), but avoided the need to categorise cattle as positive or negative to BHV-1 antibody, and avoided the possibility of misclassification due imperfect sensitivity and specificity, although would have retained a bias if serological measurement varied with other variables. However, the consistency between the discrete and continuous outcome models suggests that sensitivity and specificity do not alter with any of the explanatory variables in the model.
The results of this study do not necessarily represent all cattle herds in England, because of the locality of the farms sampled and because the analysis was based primarily on the serological values for BHV-1 from adult cattle. Using only unvaccinated herds in the study may have reduced the extrapolation of the results, because these herds may have had different management practices from vaccinated herds, but there were only 7 herds that reportedly used vaccination and so analysis of these herds' managements was not possible. In addition, there might have been an overestimation of the infection derived seroprevalence because purchased cattle might have been sourced from a vaccinated herd and so had antibodies derived from vaccination: we do not know the vaccination status of source herds. The serological test used in the current study did not distinguish between vaccine derived and pathogen derived antibody so this was unavoidable. It is interesting that only 7 herds were vaccinated and that farmers reported little clinical disease from BHV-1, suggesting that this infection is not an obvious production constraint or clinical disorder in most herds.
Finally, in the current analysis we could not distinguish between BHV-1 sub-types, the epidemiology of individual types might be different and this might account for some of the unexplained heterogeneity. BHV-1 subtype 3 is now classified as a new species (bovine herpesvirus-5 or BHV-5) . It is probable that we were detecting antibodies to BHV-5 as well as BHV-1 because these viruses are antigenically related.