While seeking to establish current levels of COVID-19 infection in sheep and cattle in Scotland, historic samples from animals collected prior to the pandemic were found to be unusable.  In consequence, it was necessary for BioSS scientists to use a computationally-intensive method to quantify the probability of samples being positive, where these came from the (potentially) mixed population of positive and negative animals sampled after the beginning of the pandemic.  This analysis indicated that there were no positive samples from sheep, and less than 10% of cattle samples were potentially positive.  Subsequent confirmatory lab tests were able to rule out infection with SARS CoV-2, the causal agent of COVID-19, and conclude that there is no evidence of COVID-19 infection in sheep and cattle in Scotland.

It is known that cattle can be infected with COVID-19 experimentally and will produce detectable antibodies.  Moredun Research Institute researchers set out to quantify the level of COVID-19 seroconversion in Scottish ruminants. Samples from both sheep and cattle were available for analysis.  A total of 210 samples from each species were historic samples, collected prior to the start of the COVID-19 pandemic, and were therefore designated as known negative samples.  A further 2640 samples from sheep, and 2200 from cattle, were available with collection dates after the start of the COVID-19 pandemic.  A typical approach would be to test the level of antibody in the negative samples and use this to define a cut-off, above which samples are unlikely to come from a negative animal, with these being designated as positive.  Where known positive samples are also available then this approach can be optimised to control the numbers of false positives and false negatives.

In this study, however, the known negative samples were found to be unusable as they reacted differently in the assay, possibly due to having been stored. In the absence of known negative samples, a cut-off for use of the ELISA in ruminants could not be defined using simple methods; a Bayesian latent class analysis was used instead.  In this approach, animals are assumed to be either sero-converted or not, but with the status of each animal unknown.  We assume that the test results from negative animals come from a statistical distribution with a lower mean than that associated with positive animals.  By jointly fitting the two distributions and the population-level sero-prevalence, it is possible to identify those samples with unusually high antibody levels.  This exercise highlighted a small group of 184 cattle, but no sheep, with unusually high levels of antibody to COVID-19.

Cattle are subject to their own coronaviruses and it was suspected that these could cross-react with the COVID-19 assay; further laboratory tests were used to definitively rule out COVID-19 infection in these samples.  However, it would have been prohibitively expensive in time and reagents to have performed these tests on all the samples.  The two-stage approach used in this project allowed a very large number of samples to be screened in a cost-effective fashion, while overcoming the practical issues arising from the status of the historic samples, allowing the project team to conclude that there is no evidence of COVID-19 infection in ruminants in Scotland.  

This work was done in collaboration with Tom McNeilly, David Griffiths, David Frew and Philip Steele from the Moredun Institute and George Caldow from SRUC and was funded under the Scottish Government's Strategic Research Programme for environment, agriculture and food.