Is specificity of the skin test over-estimated?

Summary

Data was supplied by DEFRA in June 2017 which clearly shows that the risk of an animal being disclosed with TB reduces as herd size increases. This page looks at the implications of this on how the specificity of the skin test may be over-estimated when the skin test is performed in high risk areas. If this is happening, the proportion of cattle reactors, which are TB-negative, is higher in high risk areas than in the low risk areas where test specificity was calculated from the results of herd tests.

Impact of herd size on reproductive rate

In December 2016 the Veterinary Head of TB policy in APHA, Dr Malla Hovi, spoke in front of the Welsh committee on Climate Change, Environment and Rural Affairs. At this time, Dr Hovi had been in this role for 5 years. The following⁸ shows what she said regarding the impact of herd size on the reproductive rate at which bovine TB spreads from animal to animal.

[364] Dr Hovi: There has been a huge amount of risk analysis carried out about TB and various different aspects, and it must be said that the herd size always comes out on the top. So, it usually wipes out all the other potential risk factors and there are constant arguments about whether being a beef herd or a dairy herd is a risk factor, but when you start carrying out the analysis and you put in the herd size, it tends to trump all the other risk factors. So, large herds are more at risk of TB. It's a combination of factors as to why that is: (a) we find it more difficult-with the imperfect tests that we have for TB in cattle, we find it more difficult to clear herds of TB when they are larger. So, you have just mathematically a larger chance of having a false negative animal in that herd. We know from herd-level modelling-and this has been carried out in other countries as well, not just here-that in large herds the reproductive rate of TB, of the disease, in large herds is much higher. It's somewhere between 4 per cent and 5 per cent, when in small herds it's very, very low; it's around 1 per cent. There's obviously more contact between more animals in a larger herd. There is more mixing. Larger herds also probably have more contact surface with the environment, whether it's other cattle herds or badgers, in the case of TB, and they tend to have more fragmented-they tend to be on fragmented holdings. So, cattle move around between higher risk areas and lower risk areas, often. So, those are probably the most common explanations as to why large herds are more at risk, but it's a fact that nobody will dispute.

According to the above extract the reproductive rate in large herds is somewhere between 4 per cent and 5 per cent and around 1 per cent in small herds. Another words the reproductive rate increases a number of times as herd size increases.

Impact of herd size on disclosure risk

The following shows a Freedom of Information request submitted to APHA regarding the risk of an animal being diclosed with TB when herds of different sizes are tested.

Please send to me the annual number of cattle reactors per 1000 skin tests on a bovine in England when performing routine whole herd tests on the following herd sizes in each of the last 2 years.

1-10
11-50
51-100
101-200
201-300
>300

As such, please send to me 6 numbers for 2014 and 6 numbers for 2015.

In response APHA returned the following reply⁹.

The graph below illustrates the figures calculated in the above reply.

Clearly the risk of an animal being disclosed with TB reduces a number of times as herd size increases.

How can this happen if the reproductive rate INCREASES a number of times?

The most likely explanation is that when a veterinary performs a skin test in these routine tests and interprets the result by assessing lump size, the veterinary applies discretion so as not to penalise a herd if it contains a large number of animals. If a veterinary did not apply this discretion large herds would continually be under restriction due to the numerical odds of finding a reactor simply on account of the test's specificity (i.e. tendency to give false positives).

If veterinaries are applying their discretion when testing large herds, do veterinaries apply their discretion when testing in areas where TB is not expected? If so, this has implications for the specificity of the skin test because this specificity was calibrated by looking at disclosure in routine herd test results in areas such as Scotland where TB prevalence was extremely low and not expected.

What evidence is there to support the argument that veterinaries apply their discretion when reading lump size?

Firstly DEFRA were extremely reluctant to supply data in response to the above Freedom of Information request. On first receiving the request, they argued that the information was not held¹. Then when challenged on this, they changed their argument to the cost limit being exceeded². After refusing to supply the information the case went before the GRC Tribunal on appeal. During the course of this appeal, DEFRA announced that it intended to file a witness statement in support of its case³. However that statement never appeared. Instead DEFRA relented, through their solicitor asked for the appeal to be withdrawn, and supplied the requested data. They argued that, due to an improved ability to respond to requests, they were able to respond to the request within the cost limit⁴. This whole process took 12 months. Although the following explanation is speculative and may not apply, perhaps DEFRA were reluctant to supply the requested information because it would reveal how veterinaries apply their discretion when interpreting lump size. If veterinaries applied discretion when testing in the low risk areas used to calibrate the specificity of the skin test, the actual specificity of the skin test when applied in high risk areas where veterinaries may not apply such discretion may be substantially lower. This line of reasoning is somewhat speculative. The following argument is less speculative.

The Randomised Badger Culling Trial showed that when badgers are culled in high risk areas, the number of cattle herd breakdowns substantially reduces. However this only applies when the herd breakdown is confirmed. A confirmed breakdown is where infection is confirmed at post mortem inspection, either by lesion or culture, in at least one animal in that breakdown herd. The impact of badger removal on confirmed breakdowns is shown in the graphs below.

For breakdowns which are not confirmed, badger removal was found to make no attributable difference⁵. See the graphs below.

If specificity of the skin test is at least 99.9% as claimed⁶, then badger removal should have had a clear impact on unconfirmed breakdowns as well as confirmed breakdowns. The fact that it did not raises the question of whether or not reactors in unconfirmed breakdown herds were actually infected. The following chart⁷ showed that about half of all herd incidents between 1994 and 2009 were unconfirmed.

Percentage of unconfirmed incidences

Conclusion

The specificity of the skin test may be substantially over-estimated in high risk areas. If so, this implies that a larger number of TB-negative cattle are being slaughtered in these high risk areas than given by the test's specificity. This is due to how the test results may be interpreted in these areas compared to how results may have been interpreted in the low risk areas where the test's specificity was determined.

References

If you would like to send to me your comments, please contact me

Last Modified 07 Feb 2023 14:00