Middlemiss and Henderson say sorry for getting it wrong
As previously blogged on 18th March here, the respected journal Veterinary Record published a new scientific appraisal of the effect of badger culling on bovine tuberculosis (bTB) in the High Risk Area of England using government data collected from farmers and vets for over a decade. This extensively peer reviewed paper is available open access online, in full here. The paper concluded that badger culling has not been associated with reductions in bovine TB (bTB) incidence or prevalence among cattle herds.
Alongside a one-page summary of the paper in the Vet Record print edition, the Chief Veterinary Officer (CVO) Christine Middlemiss and Chief Scientific Adviser (CSA) Gideon Henderson published an un peer-reviewed letter rebutting the paper’s main findings. They produced their ‘alternative analysis’ in the form of a graph, and claimed it showed that badger culling was ‘working’ in reducing bTB in cattle. The graph indicated very rapid declines in bTB in culled areas following the commencement of culling, with little change in unculled areas. The CVO Christine Middlemiss also posted a blog on the Defra website using the same graph.
The graph could not be reconciled with publicly available data. There followed repeated requests for Defra to supply the data and methodology, but these were not met. Then last week, six weeks after publication, Middlemiss and Henderson sent an email to the authors of the original paper stating:
“Following your recent correspondence about how incidence in unculled area was calculated we have re-examined our analyses and discovered an error we wish to bring to your attention. The incidence in the area unculled throughout the period was calculated incorrectly. The incidence in cull areas is unchanged. We attach a corrected graph, with the corresponding data and workings as previously requested. We apologise for this error..”
A new graph was provided (see below). After further requests and delay, we have data from Defra to allow us to reproduce their corrected graph but not to check its origination. Defra’s original published graph shows bTB herd incidence higher in unculled areas in four of the five years, while in the new one it sits at the same levels as in culled areas.
As previously, Defra are still disregarding huge areas of unculled land in their blue-bar ‘never culled’ areas, which is problematic. Notably, however, the error bars between ‘culled’ and ‘never culled’ overlap more extensively, so the difference between the two is unclear. It seems that Defra’s corrected calculations corroborate the findings in the Langton et al., and that there has indeed been no significant impact from badger culling on bTB incidence among cattle herds.
Defra’s graphs from their 19th March letter and 5th May email:
Defra’s “never culled” areas are likely to include significant land areas where bTB is less of an issue, with landowners having a lower incentive to coordinate a cull, whilst the “waiting to be culled” portion of the unculled area will have significant areas where bTB is a major problem. Defra is engineering a highly selective use of the available data. It adds up to a misleading picture that is bringing Defra into disrepute. Without access to their full data source, it is not possible to fully understand their rationale.
Further, when you don’t limit the data as Middlemiss and Henderson did, and add “all culled areas” bars (green), and include 2013/14 and 2014/15 (see below), it shows the true extent of decline of bTB incidence in unculled areas that mirrors culled areas.
The 5th May ‘apology’ email from Middlesmiss and Henderson maintains that “this does not change the overall argument in the letter”, yet over six weeks on, they have failed to address a response by the authors to this criticism (published in Vet Record on 2nd April). This response shows that their main argument on ‘incorrect grouping’ of data does not undermine the peer-reviewed statistical analysis.
Specifically, Middlemiss and Henderson claimed that using data from the first two years of culling ‘masks’ any overall effect from badger culling, making it ‘impossible to see’. But Defra’s counter argument rests upon a steep decline in herd incidence over those first two years! Defra’s argument falls and the answer is that taking all the data, herd breakdowns reduce in culled and unculled areas at similar rates, due to cattle measures both before and after badger culling is rolled out.
So, the senior Defra scientists have no answer, and continue to use delaying tactics, while still providing only limited access to the available data that might enable independent researchers to assess their new graph. This is shocking and does not serve the public interest. Cattle-based measures implemented from 2010, and particularly the introduction of the annual tuberculin skin (SICCT) test have been responsible for the slowing, levelling, peaking and decrease in bovine TB in cattle in the High Risk Area (HRA) of England during the study period, before badger culling was rolled out in 2016.
Last week, the authors of the badger culling paper, Tom Langton, Mark Jones and Iain McGill wrote to George Eustice about the continuing fiasco and asking for badger culling to be suspended and for additional clarification and dialogue.
This is what all stakeholders and the public deserve. Clear, open government responding to the facts in an honest and professional way. No more delay, secrecy, and avoidance of the real issues. It is time things changed.
On Friday 20th May, the paper’s authors response to Defra’s apology and clarification was published in Vet Record. You can read this here:
Farming Today featured the debate around Defra’s data miscalculation on 20th May; you can listen here from 7:18 minutes in.
You can donate to the legal fund for the Badger cull ecological impacts case below.Donate