This manuscript presents BM Ab, BM PCR and YS Ab (Check Test) data from twenty six working farms that were recruited to a pilot BVDV eradication programme. Although there is general agreement in the current literature that BM Ab levels and youngstock Check Tests are appropriate ways to assess BVDV status at the herd level [20, 25–27, 35, 36], there remains some confusion about their reliability and the impact of historic infection and vaccination. We discuss these issues in the context of our study herds below.
The recruited herds were predominantly located in Somerset, UK and the large proportion of dairy herds in the study population reflects the region from which herds were sampled. The average study herd size appears large when compared to the average herd size of 120 animals quoted by Defra for the Taunton area at the time of recruitment . However, the Defra figure may be falsely low as this will include small holdings and in the authors’ experience, the herd sizes represented here are typical of the range of farms in this region.
Within the analyses conducted, the wide 95 % confidence intervals quoted throughout highlight that there is a degree of uncertainty surrounding the results and thus the figures should be used with caution. This is largely due to the small sample size of 17 herds which underwent statistical analysis. Despite these potential limitations, the improvements demonstrated in sensitivity and specificity and also in the ROC and probability curves by removing the delay between screening a herd and acting on those results are substantial and worth discussion. Of the 26 herds that were recruited, only 17 were suitable for statistical analysis since BM and YS samples were collected in addition to individual animal samples to determine the BVDV status of all animals on this subset of farms. Two farms that had conducted a WHT were deemed unsuitable for inclusion in the statistical analysis as ten animals were not available for the initial YS screens and are included in Figs. 1 and 4. The seven remaining farms did not undergo WHTs and were therefore excluded from the statistical analysis, but are presented for completeness in Figs. 1 and 4. Whilst we believe that the regular surveillance would have detected infection on these farms over, and beyond, the course of the study period analysed here, it remained an assumption that they were BVDV free and so it would have been inappropriate to influence the ROC, probability, sensitivity and specificity analyses on this basis. This point is especially pertinent when we consider that Farm 15 returned a mid positive BM Ab titre at the WHT and also negative YS results in both the initial Check Test and the simulated Check Test from the WHT yet a PI was present throughout the period. This was primarily due to herd dynamics and is discussed in relation to this herd’s results below.
The Use of BM Ab Tests for determining Herd BVDV Status
BM Ab tests provide an initial, rapid and cost effective assessment of the BVDV status in dairy cattle at the herd level. It is largely accepted that high BM Ab levels correlate with a high probability of the presence of a PI. However, we found several instances where herd BVDV status would have been incorrectly classified if based on an interpretation of the BM Ab result alone. The sensitivity and specificity of the test throughout this study was 80.00 and 85.71 % respectively however, the probability curves indicate that it is difficult to accurately define a cut-off point where we can say with reasonable certainty that a herd does or does not contain a PI animal. The removal of the delay between the BM Ab screening and further investigations does improve the confidence that one can have in the interpretation of the test but as a sole screening method, the usefulness of BM Ab remains limited in many circumstances. We know that both historic BVDV infections and herd vaccination can have a significant effect on correct BM Ab interpretation [33, 38] and both of these factors are likely to have played a significant role in the test results and interpretation here.
BVDV antibodies following a natural infection may persist for 3 or more years in an individual animal [7, 34]. When one considers that BM Ab assesses all milking animals including the oldest in the herd, it explains why it can take in excess of 1000 days for BM Ab to show any real decrease following the removal of PI animals . Booth et al. 2013  demonstrated that Farms 13 & 24, which were BVDV-free farms that neither vaccinated nor experienced active BVDV infection during the study period experienced a slow yet consistent decrease in the BM Ab titre; in this type of herd BM Ab can be a sensitive surveillance test for monitoring for disease incursion into the milking group since an increase in Ab titre can be considered of significance. Vaccination can also confound interpretation of BM Ab results further. With the exception of Farms 13, 24 & 39, all BVDV-free dairy herds in this study vaccinated during the period data was collected (Table 1) and all began the study with mid to high BM Ab titres (Fig. 1). In the longitudinal analysis of BM Ab trends by Booth et al. 2013 , BVDV-free vaccinating herds did not demonstrate a clear decrease in BM Ab. The fact that the majority of the study herds were utilising BVDV vaccines, in combination with the likelihood that many of them may have experienced historic infection and thus contain older seropositive animals may explain the higher level of misclassification of false positive farms observed with the BM Ab tests compared to YS Check Tests. In practice, careful thought needs to be given to the application and timing of BM Ab testing in herds utilising BVDV vaccines and the use of this test may be deemed inappropriate in some systems. This is discussed in further detail by Booth and et al. 2013 .
Whilst vaccination and historic infection would explain those farms that returned high positive BM Ab results in the absence of PIs, it is more difficult to explain those herds (Farms 15 and 19) that returned only mid-positive BM Ab results in Fig. 4 when PI animals were present within the herd. For these farms, the reasons are likely due to the location of the PI animal(s) within their herds:
For Farm 15, the lack of a high positive BM Ab result is harder to explain since the PI was of milking age. She was however a poor animal and often separated from the milking herd in a hospital pen due to illness.
Farm 19 contained a PI animal of approximately 1 year of age and yet returned a mid positive BM Ab result. Upon initial interpretation, it may simply be possible that the PI did not have contact with the milking herd, however, analysis of the BM PCR results bring this interpretation into question since the herd was positive on BM PCR testing. It is difficult to provide an explanation for this combination of results especially given that at the initial screen for Farm 19 (conducted 5 months prior to the WHT), the herd returned a high positive BM Ab result (Fig. 1).
Farm 2 highlights the need for the practitioner to consider the epidemiological units that make up the herd structure in detail both before testing and before coming to a decision on the status of the herd based on the results of the tests. At both the initial herd screen and at the point of the WHT, the milking herd on Farm 2 had not been recently exposed to BVDV, hence the mid positive result. If a single assessment had been performed on a bulk milk sample, and decision reached that the herd was uninfected, then infection in the heifers would have gone unnoticed until subsequent BM Ab titres had increased. Our Check Test results indicated that there was likely exposure in the heifer group at the calf rearer and that this was almost certainly due to the biosecurity breakdown involving mixing with animals from another farm. The return of three heifers carrying PI foetuses to the main farm in combination with the farmers reluctance to cull PI animals once identified and the failure to engage the owner of the other animals at the heifer rearer in the scheme eventually resulted in infection of the main herd and an increase in the BM Ab titre to a high positive (data not shown). This farm is discussed in more detail by Booth and Brownlie 2012 .
The use of young stock (YS) check tests for determining herd BVDV status
The real value of YS antibody tests is that young animals will be seronegative if unexposed and maternally derived Ab (MDA) have waned; if they do have antibodies demonstrating an active immune response (in contrast to MDA), they will have been infected only within their short lifetime (e.g. 9 months). That would indicate a recent, or present, infection on the farm. Assessing animals at, or after, 9 months of age is a valuable time to initiate a YS screen as MDA will have has waned and it is typically pre-vaccination.
YS below six months of age should not be selected for Check Testing since MDA are highly likely to be present. Chamorro et al. 2014  demonstrated that in extreme cases, BVDV MDA can take 11 months to wane and that MDA are commonly present up to 6 months of age; hence have the potential to confound interpretation of the test. In the authors’ experience, testing animals at least 9 months of age, especially in beef suckler herds where colostral transfer of MDA is more reliable and higher quality than in dairy systems, provides a safer margin for ensuring that MDA is not present.
Whilst the majority of the herds in this study did vaccinate, they only initiated the primary course of vaccine prior to first service. The result was that we were able to assess the YS cohorts in these herds without the influence of vaccine. There is the possibility on Farms 8 and 25, that some of the older YS selected from the WHT in order to simulate a Check Test on that day could have been vaccinated as they were at the older end of the 7–13 month range, but in all the other herds, this was unlikely to have influenced the results. In this study, we achieved both a high sensitivity and specificity when assessing the ability of the Check Test to determine whether PI animals were present in the seventeen study herds analysed when the Check Test was simulated from the WHT results. Furthermore, the probability curves displayed an increasing ability to predict the presence of a PI animal within the study herds at the suggested cut off of 2.5/10 animals Ab positive once the delay between screening and the WHT had been reduced. Whilst we have sampled randomly within the animals tested at the WHT in order to generate the ‘without delay’ Check Test, it is important to mention that in the field, careful thought should be given to the contact of the cohort(s) tested with the rest of the herd and this is discussed in the context of Farm 15 below. Previous epidemiological studies have shown that when two or fewer YS of the ten tested are Ab positive, there are unlikely to be PIs present on a farm [27–29] and the optimal cut off from this manuscript of 2.5/10 YS Ab positive supports this finding, even in larger UK herds. The authors would urge readers to use some caution when applying this cut-off value in those herds that return 1/10 or 2/10 positive animals since this may be the start of seroconversion in a group, detection of a declining Ab response to vaccine or MDA. The most effective way to distinguish this is to re-test the same animals at least 28 days later to determine whether seroconversion is underway.
Farm 15 was consistently classified as negative when interpreting YS Check Tests when, in fact, it contained one adult PI animal. The explanation is that the PI animal was bought in as a heifer and, after taking considerable time to get in calf, finally produced a dead calf that was quickly removed. All calves born on Farm 15 were routinely raised on a different unit to adult stock and thus the adult PI animal never had contact with young animals and did not produce a live PI calf that was mixed with the YS. This again highlights the importance of considering the mixing of groups within a farm unit. If the cohorts selected for Check Tests have had reasonable exposure to the rest of the herd, the results may be extrapolated to provide an indication of herd status. However, cohorts of animals reared as isolated groups (or even in some cases as separate herds until first calving) are not suitable for selection for Check Tests to determine whole herd status but only group status. Furthermore, Farm 15 highlights the importance of testing purchased stock however, this animal was purchased before Farm 15 was recruited and was therefore only detected once the study commenced.
In a herd with a high positive BM Ab titre due to historic infection and/or vaccination, sampling cohorts of YS may be the only way to accurately obtain an up to date assessment of herd status and the predicted probability and ROC curves in Figs. 3d and 5 respectively support that this should provide a relatively accurate analysis. In beef herds, monitoring antibody levels in YS cohorts is the only method of assessing the herd hence it is re-assuring that this test performs well. In traditional cow:calf suckler units, Check Tests are potentially much more reliable than in dairy units since there is a greater degree of mixing of animals thus separate epidemiological groups are less likely to occur.
The use of bulk milk PCR for determining herd BVDV status
BM PCR provided a highly sensitive and cost-effective test for adult PI animals contributing to the milk tank. Unfortunately, our PCR data is incomplete as the test was not widely available at the time of recruitment of many farms.
Over the course of the study, our limited BM PCR results did identify 5 farms with a positive test; on two of these farms (Farms 1 & 38), PIs were found within the milking herd. On Farm 40, no PIs of milking age were identified and only two older animals had left the milking herd in the time between screening and the WHT. Whilst it is possible that either/both of these culled animals had been PI, another potential hypothesis for the positive BM PCR could be that a number of pregnant animals were carrying PIs and thus the level of virus circulating in the milking herd was sufficient to create detectable levels of viral RNA in the bulk milk. On Farm 19 a positive BM PCR was also returned at the whole herd test, but no milking PI animals were identified, nor were further PIs beyond the initial one identified in the youngstock so it would seem unlikely that this could be due to pregnant animals carrying PIs in this instance and may simply be due to contact between the milking animals and YS group containing the PI on this unit. This concept would be supported if the matching milk sample had returned a high positive Ab titre, but as described above, it is difficult to reconcile why this was not the case and a mid positive titre was returned. In contrast, Farm 29 returned a positive BM PCR result yet no PIs were identified at the WHT. On this farm it is highly likely that the delay in conducting a WHT (12 months) resulted in the death or removal of any PIs prior to identification. The fact that 5/10 YS on Farm 29 tested Ab positive at the initial screen and that the herd had a high positive BM Ab titre support the conclusion that PIs had been present on the farm.
It is important to note that BM PCR can only test those animals contributing to the bulk tank on the day the sample was taken. Our results show that some of the herds that did contain PIs did not return a BM PCR positive test result. PIs later identified on these farms had not contributed to the bulk tank. For this reason, provided the upper limit of milking contributors to a sample, as recommended by the testing laboratory, is not exceeded, a negative BM PCR is a reliable indication that any animals contributing to the BM tank, are not PI. A negative result is not, however, a reliable indication that there is not a PI(s) animal elsewhere in the herd.
The effect of delays between herd screens and whole herd testing
Apart from the 19 month delay between screening and WHTs on Farm 39, most herds were investigated within 8 months of their initial screen. Farm 39 consistently appeared BVDV-free in the regular surveillance conducted and so there was no urgency to perform a WHT in this herd so perhaps it skews the recorded delays unnecessarily. The mean interval is not ideal but in part reflects the reality in practice where results are awaited, reported to the farmer and then a period of discussion and organisation, often avoiding harvesting or silaging, has to occur in order to arrange whole herd testing. If the delay between screening the herd and taking action based on those results is prolonged, the results of the initial screens may not accurately reflect the BVDV status of the herd in question. The data presented here reflects this and in all cases shows an increase in the diagnostic accuracy and relevance of the screens once these delays are removed.
Farms 7, 8, and 29 had delays between initial screens and WHT of 9, 3 and 12 months respectively. Based upon the original YS criteria of >1/10 YS Ab positive indicating an infected herd all were classified as infected. Despite this, no PIs were identified on these three farms during the course of the study (Fig. 1). Farm 8 had a relatively low number of YS seropositive at the initial screen (2/10) which declined further to 1/10 YS Ab positive at the WHT three months later and so is unlikely to have been an infected herd. Whilst no PIs were found, it is highly likely that both farms 29 & 7 were infected due to the number of seropositive YS detected at the initial Check Test; 6/10 and 5/10 YS Ab positive for each farm respectively (Fig. 1). An infected status is further supported for Farm 29 since it returned a positive BM PCR result at the initial screen (Fig. 1). Due to the delays in conducting WHTs on these two farms, it is possible that any PIs had unknowingly been removed from these units prior to undertaking WHTs since both farms show less evidence of BVDV exposure in Fig. 4. Figure 4 demonstrates that, based on Check Tests, herd screens performed closer to the WHT would have indicated that herds 7 & 29 did not contain PI animals at the point of the WHT since their results were 1/10 and 2/10 YS Ab positive respectively (Fig. 4) – this is below the cut off of 2.5/10 animals indicated in the ROC curve in Fig. 5. The result for Farm 29 is further supported by the fact that it was also BM PCR negative at this stage. These results indicate that significant alterations in herd infection status may have occurred on Farms 7 and 29 in the time between the initial screen and WHT.
On Farms 15 & 27 the delay between the initial screens and whole herd tests was 8 & 9 months respectively. Both farms 15 & 27 bought PI animals shortly before the screens presented in Fig. 1 were performed (data not shown). It is evident that for Farm 27, the initial Check Test was likely performed before enough time had elapsed for seroconversion to have occurred within the group screened since continued surveillance on Farm 27 would have detected the presence of the PI animal, since the YS seroprevalence increased from 0/10 YS Ab positive in Fig. 1 to 10/10 YS Ab positive (Fig. 4). This further highlights the need to develop an ongoing surveillance plan. Farm 15 would however remain undetected through YS surveillance and the reasons for this have been discussed above.