Contact me about my project

Impacts of badger culling in England: New paper published

Industry-led badger culling began in England in 2013 initially in two pilot areas in Gloucestershire and Somerset. A third cull area was licensed in Dorset in 2015, and then since then the number of cull areas has gradually increased.As of the end of 2018 there were 32 cull areas active in England, and this has now increased to a total of 43 (for a list of cull areas see here). Badger culling is hugely divisive and there is a lot of interest in what effect the culls are having on TB in cattle. Previous APHA reports have described the TB statistics for the cull areas (these are interesting but cannot prove if culling is having an impact – see here for the latest published data), and one previous paper has been published investigating the effects of the first two years of culling in the Gloucester and Somerset cull area (Brunton et al. 2017, summarised here).

A new study has just been published in the Journal Scientific Reports by Downs et al. which builds on this earlier work and gives us out best estimate yet of the impacts of the current badger culls. The study uses four years’ worth of data (2013 -2017) for the Gloucester and Somerset cull areas and two years’ worth of data for the Dorset cull area.

Assessing effects from four years of industry-led badger culling in England on the incidence of bovine tuberculosis in cattle, 2013–2017

I am one of the co-authors of this latest study, and together with the other authors I have created a two page summary of the paper which can be downloaded below and on the Tbhub.

It is always a challenge to summarise complex studies in just a few words, and much more detail can be found in the full paper, which I would recommend reading. I have also outlined some additional points relating to different aspects of the study below:

Why use complex multivariable analyses?

You may be thinking…why do the analyses need to be so complicated? Surely you can just look at the TB stats for the cull areas and show if TB is declining or not? Firstly, TB statistics go up and down all of the time (particularly at the scale of a cull area) so to understand effects of culling or any intervention you need to compare trends in the cull areas to comparison areas without culling. The comparison areas used in the current study are selected to be similar to the cull areas, but these areas will never be identical. This is potentially problematic, as TB in cattle is associated with a whole range of risk factors, many of which relate to cattle husbandry and have nothing to do with wildlife or culling (Skuce, Allen & McDowell 2012). This means that there could be differences between cull areas and comparison areas that are unrelated to culling, this could lead to differences in the level of TB in cattle. The multivariable analyses conducted aim test for an effect of culling, while controlling for these other differences. This means that when we describe the results of these analyses we do not simply say “TB declined by X amount” instead we talk about the effect of culling relative to the comparison areas (see below).

Incidence rate ratios (IRR)

The effect of culling and other factors on TB risk is described in the study using incidence rate ratios (IRR). An incidence rate is the rate that a given event (in this case TB breakdowns) happens over time. The statistics in the paper report the incidence rate RATIO (IRR) as this shows the rate of TB breakdowns in the cull areas RELATIVE to the comparison areas where no culling has taken place. For example an IRR of 0.3 would mean that the rate of breakdowns was 30% of that in comparison areas (ie 70% lower), 0.8 would mean it was 80% (or 20% lower) and so on. An IRR of more than one means an increase… 1.50 would mean a 50% increase (ie the rate was 150% of that the comparison areas) and 1.20 a 20% increase.

In the tables in the paper the IRR is reported for the intervention effect (culling) and also for several other variables (Table 4). For variables with a continuous scale (eg 1-100 rather than distinct categories) the IRR is the change in incidence rate associated with a 1 unit increase in that variable. For example, the IRR for the variable ‘percentage of herds that were dairy’ was 1.01. This means for every unit increase in that variable there was a 1% increase in the rate of TB breakdowns.

OTF-W, OTF-S or all TB incidents?

Bovine TB is reported in a number of different ways and this means that there are a number of possible statistics that can be analysed:

  • OTF-W (Official TB Free Withdrawn): these are TB incidents where there was at least one test reactor or inconclusive reactor with post-mortem evidence of M. bovis infection (judged by the presence of visible lesions typical of TB and/or identification of M. bovis in culture), or at least one slaughterhouse case that yielded M. bovis on culture. In some studies these are referred to as ‘confirmed’ breakdowns.
  • OTF-S (Official TB Free Suspended): these are TB incidents where at least one test reactor or two IRs have been identified but evidence of infection could not be confirmed through post-mortem examination or laboratory culture of tissue samples. In some studies these are referred to as ‘unconfirmed’ breakdowns.
  • All TB incidents: – This includes both OTF-W and OTF-S incidents (ie all breakdowns regardless of if they were confirmed or not).

The paper largely focuses on changes in OTF-W incidence rates. This is partially because these breakdowns are confirmed as being caused by TB, while it is possible that OTF-S incidents may have been caused by something else. The analyses for the RBCT also focused on changes in confirmed (OTF-W) TB incidents, largely because no significant changes were observed when for all TB incidents or unconfirmed incidents (OTF-S) were analysed.

The paper and the supplementary materials of the paper (a range of extra details and tables published alongside the study) report the results of the analyses using OTF-W and all TB incidents. Analyses of changes in all TB incidents found a decline in the Gloucester and Somerset cull areas (although less than when focusing on OFT-W), but also a significant increase in the Dorset cull area.

Why is the data only up to 2017?

This paper summarises the data up to the end of 2017. This is now nearly two years ago, so why is there such a delay? This is partly because it takes a while to collate the enormous amount of TB data and statistics collected each year. For example the TB statistics covering the period up to the end of 2018 were only published (here) in March of 2019, which is after this paper was submitted for publication. Complex analyses like those reported in this paper also take time to complete, so it is not a simple task of updating the figures (particularly as the licensing of new cull areas may overlap with comparison areas used in the analysis). Finally, the paper also has to go through a process of peer review where it is assessed by anonymous reviewers (typically scientist with expert knowledge of the field). This peer review process is crucially important for ensuring that the science is robust, but it can take many months to complete (anywhere from 3/4 months up to a year or more in some cases).

A few final thoughts on the implications of this study

The key conclusions from the study are outlined at the end of the summary document. However, given the diverse views on badger culling it is inevitable that people will look at this study and come to very different conclusions. In my view it is important that any scientific studies such as these are looked at properly and the limitations and results are fully understood. This study does demonstrate that industry lead culling can result in benefits, measured as a drop in the rate of new breakdowns (relative to other areas without culling). This change in TB incidence is a standard way of evaluating disease management, partly because a fall in the rate of new disease incidents is likely to be the first change that can be detected. Whether culling leads to longer term reductions in disease in these areas, such as a reduction in prevalence (the % of herds under restrictions) we cannot definitively say from these data.

One important limitation of this study relates to the scale of the work. In an ideal world if you wanted to look at the effect of culling (or any disease intervention) you would carry out a randomised, replicated trial. This is what was conducted during the RBCT (for a summary of the RBCT see here) as there were 10 randomly selected areas for each culling treatment. The results from the RBCT analyses then give you an estimate of the average effect of culling (along with a range of uncertainty around this) if it were applied to a given random area. This essentially gives you the power to predict future impacts of culling, assuming it is done in a similar way, place and time (the RBCT was government cage trapping, in areas with high levels of TB in cattle, from 1998-2005).

The current badger culls in England are a programme of wildlife management with the aim of disease control, but they are not a scientific trial aimed specifically at investigating the effects of culling (like the RBCT). So while there may be opportunities to evaluate culling outcomes (like this study), it may be that we are limited in what we can conclude from the results. This latest study provides a good estimate of the effect of industry lead badger culling in the three specific areas studied. As mentioned in the summary document, these areas were not randomly selected and they could be very different from the other licensed cull areas in England. For example, the role that badgers play in cattle TB could be much smaller or larger. The effectiveness of the culling could also be very different due to differences in habitat, badger density, farmer behaviour, protestor activity or government support. This does not mean the results of this study are not valid, but it makes it difficult to predict what effects the culls will have in the other areas. For example, culling in other areas could have a large impact (like a Gloucester), a smaller impact (like Somerset), no impact (yet – as in Dorset) or a totally different impact we haven’t seen yet. It also means that these data cannot be used to confidently evaluate the whole culling policy, just those specific areas studied. Hopefully future analyses will help to shed light on the effect of culling in other areas, as well as the longer term impacts in Gloucester, Somerset and Dorset.