Lies, damned lies and twisted statistics - fake science set to kill 100,000 English badgers

|
Another 100,000 English badgers could be shot because of fake science and faker statistics. Photo: Tom Langton. Note that no badgers died or suffered to produce this photograph!
Another 100,000 English badgers could be shot because of fake science and faker statistics. Photo: Tom Langton. Note that no badgers died or suffered to produce this photograph!
The government / NFU badger culling policy is based on a single study, the Randomised Badger Culling Trials (RBCT), which found that area-wide badger killing reduced TB 'breakdowns' in cattle herds. But a robust reanalysis of the RBCT data reveals that culling is entirely ineffective, writes Tom Langton. The only scientifically valid conclusion is that culling badgers has no effect on TB in cattle. Defra and Natural England must think again!
What happens when you add all the unconfirmed test results back into the RCBT model as being correctly identified as having bTB is that there is no statistically significant effect of proactive culling of badgers on new herd bTB breakdown.

Some people will insist that the English badger cull is completely political.

Yet in some senses, everything in society that is large-scale, expensive, complex and at the heart of a national industry is bound to be political to some extent.

Priorities, such as who gets the money and decisions on what success or failure may look like are political decisions.

It is easy to 'blame' hidden agendas and dark forces, and they may to some extent exist. But quite often the answers are simpler, and rooted in the frailties and fallibility of humans - the human factor. And science with all its inner secrets and complexities is not immune at all to that.

There is no shortage of science in bovine tuberculosis (bTB) and badger studies. Whether, it is the molecular biology of bTB strains, the intricacies of bTB testing, badger vaccination, cattle vaccination and the' DIVA' test, the hands-on pathology of disease investigation. Or perhaps population size estimation and trend analysis or the modelling of hypothetical disease pathways.

The list seems endless, and is a mind-boggling arena for the non-scientist. Understanding the inter-relationships of these disciplines requires simultaneous insight into such areas of expertise and uncertainty.

So, before we blame politicians too quickly, there is a need to look through the microscope at the science that is involved. After all, get two experts on any issue in front of you, any politician will say, and they are bound to disagree.

No meaningful contribution?

From 2013 onwards, the rallying cry of the 'anti-badger cull' movement was that badger culling could offer "no meaningful contribution to cattle bTB control in Britain." This was based upon the final (ISG 2007) report on the Randomised Badger Culling Trials (RBCT) that took place between 1998 and 2005 in England.

And that was also my understanding at the time, based on summaries of the published findings. The main conclusion from culling 70%+ of badgers over a six week period was that any reduced bTB transmission from badgers to cattle is offset by a 'perturbation effect', whereby increased movements of surviving badgers, causes an increased transmission of bTB to cattle, in particular around a culling zone.

But by 2015, with Natural England approving badger culling for the first time in Dorset, it was clear that the animal conservation and welfare charities and aligned RBCT scientists, had completely failed to convince government that the culling policy was wrong, and not based on scientific advice. Checking the scientific detail became the last remaining option to try to make sense of the situation.

What happens when you add all the unconfirmed test results back into the RCBT model as being correctly identified as having bTB is that there is no statistically significant effect of proactive culling of badgers on new herd bTB breakdown.

As an applied ecologist who has co-managed a large wildlife disease investigation with the Institute of Zoology at ZSL, London, I had some relevant background for reviewing the RBCT.

To start with, the science relating to badgers moving around more extensively during and after culling looked reasonably straight forward field study and well documented. The change in number of foxes, expanding into empty badger setts though predator release effects; literally the empty niches left by depleted badger populations, seemed well recorded.

So where did they go wrong?

Moving onto the guts of the main 2007 Independent Scientific Group (ISG) report and the published statistical papers of 2005-2007 and beyond to 2013 was next.

This was trickier. It took four months of evening and weekend reading to get fully into the near 300 page summary of the £50 million research project, and a further period with help from senior statisticians to get better grip on what had been done. Along the way, checking with biologists studying mammals, diseases, or natural processes, there were few who had studied it closely, as opposed to just parts of it and most were just generally aware of the various conclusions.

The RBCT distinguished two types of badger culling - 'reactive' and 'proactive' - each planned in ten areas of around 100 square km in size. Reactive culling is where badgers are killed only on land within a few km of a new bTB cattle herd breakdown and not widely over a large area, as in proactive culling.

But from 1998, reactive badger culling experiments had a faltering start, further hampered by the Foot and Mouth crisis in 2001, restricting access to farms. The result was a depleted dataset due to these unforeseen circumstances. The ISG report nevertheless had come up with its hypothesis that badgers were giving bTB to cows rapidly, by catching and passing it on via a 'perturbation effect'.

Yet for many, the speed of bTB transmission from badger to cattle involved looks unrealistically rapid for it to be a genuine phenomenon. The work did not seem to have taken into account when the data on cattle TB incidence was taken - which was immediately after the first proactive badger removals.

The sequence of events proposed after badgers killings would be:

  • increased badger mobility and transmission of TB amongst badgers. Newly infected badgers becoming infectious (a process taking months or longer),

  • then infectious badgers making contact with cattle somehow in a mechanism that is unknown,

  • cows establishing new bTB infection in vulnerable individuals, over months or longer, sufficiently to trigger responsiveness to the tuberculin test

  • detection at slaughter / post mortem culture / microscopy; breakdowns might need to wait six months on average and up to a year for the next testing period to be detected, during which there was a 20-50% chance per cow of it being missed and possibly picked up after a further year or longer.

  • on testing, checking for non-visible disease by culture will then take several months.

Notably, this entire sequence of events requires a considerable time to play through.

Safe science? Growing doubts ...

Reviewing the literature on BTB and badgers, there was a group of six academics including Professor Simon More from the Centre for Veterinary Epidemiology and Risk Analysis in Dublin who had studied the ISG and published an immediate critique of several aspects.

In addition, the record showed Sir David King (Chief Scientist at the time) set up his own expert group that effectively challenged the strength of the statistics concerning reactive badger culling increasing cattle breakdowns via a badger / bTB perturbation effect.

More recently, the Chief Veterinary Officer for Wales has drawn similar conclusions as those before her (see video after 34 minutes and BBC report). So now there are three separate expert appraisals suggesting independently that reactive culling and its related badger / bTB perturbation effect are not safe science.

All in all, the suggested increased bTB transmission would have been likely to take years, if indeed it was real at all. The cold fact is that the reactive culling studies were too interrupted to prove increased bTB herd breakdowns resulting from reactive culling.

This was a fatal blow to the ISG conclusions in 2007 and in fact to badgers, because the other main finding of the RBCT in relation to proactive badger culling was that killing badgers reduces bTB herd breakdowns by around 23%.

So by 2010, during the preparations of the 2011 Government Policy on bovine TB eradication, it was only the proactive cull 'benefit' that was foremost in the minds of the cull designers. This too was the message to the politicians and the farming industry at the time.

A Catch 22 that snared the scientists, campaigners and celebrities

The irony of this situation is that in 2012, the anti-cull movement of charities, voluntary bodies and celebrities got behind the 2007 ISG report findings, and the statements by scientists associated more closely with the research, perhaps not realising or recognising its flaws.

They used the 'no meaningful contribution' and 'government ignoring science' as their campaign headline. But in doing so they were in fact saying: "we agree that it is badgers spreading bTB significantly and with the perturbation effects increased bTB spread by badgers to cattle, and that badger culling can work."

And so those scientists involved were trapped in a Catch 22 of either: agreeing their bTB perturbation hypothesis was unproven, and that proactive culling works to reduce bTB; or trying to prop-up the perturbation effect story, as a balance to the proactive cull bTB reduction.

In the end, most of the anti-cull movement campaigned, unaware of this paradox and without really knowing the details of the real scientific uncertainty underneath.

Checking the science of proactive badger culling

Slightly taken aback, it seemed important next to turn to the aspect of the RBCT that had been suggested was more robust; the proactive badger cull 'benefit' of reducing bTB from mass badger culling across a wide area. This involved looking at the methods and analysis of proactive culling.

The first, and rather shocking thing to notice is that the raw data shows that in four of the ten proactive culling zones, bTB actually went up and not down when compared with its control area. There was no 'benefit' nearly half of the time in terms of what the farmer and veterinarian might see 'on the ground'.

How could that be? To find the answer I next had to tackle mathematical modelling and the assumptions and adjustments made to the raw data that had turned this into a 'significant' result. After a lot of hard work by a statistician who had volunteered to retrace the analysis, there was nothing in the analysis that actually looked 'wrong' in terms of the mechanics of what had been done.

However a number of serious problems began gradually to emerge. Using an alternative but equally valid model on the RBCT data indicated a lack of statistical significance from the proactive cull data. This was simply using, for each comparison between cull and control area the years over which proactive culling was actually carried out, rather than the average number of years, as used in the ISG analysis.

This was a very simple adjustment, using the time that each set of herds had actually been exposed to change rather than the average. As such it was an equally valid, if not more valid approach, to that used in the study.

Confounding variables and unjustified exclusions of data

Other issues cropped up, such as the wide range of confounding variables such as changes to testing and cattle movements that were likely to have been uneven within and between study areas and controls. The RBCT had not been a double-blind trial and landowners had known whether badger culling was taking place or not in the cull and control areas.

However, without proof of variables causing statistical skew within and between triplets, it is hard to prove the relevance without tracking down new data from the RBCT period. There was no time to do that - something for the future, perhaps government would take an interest.

It also became clearer that the RBCT had actually been a study in a period during which bTB was very rapidly increasing, not declining. Many of the study areas were very heavily infected before the study had started. What was being concluded upon was not an actual decline but a slower rate of increase.

However, perhaps the biggest shock of all was that the RBCT analysis had only used 'confirmed' herd breakdown rather than 'all' cattle breakdown data in its final 2007 presentation. This is highly significant because our understanding of disease prevalence has improved since the 2007 ISG report.

The lack of visible lung lesions or laboratory 'culture test positive' made during cattle slaughter and post mortem checks, is no longer viewed as meaning that the animal is free of bTB.

Furthermore, whilst the tuberculin skin test misses many infected cows, it very rarely gives false positives. Those RBCT reactor cows with no visible lesions at post-mortem had, all along, been infected with bTB.

The fatal RBCT / ISG proactive culling oversight

This dilemma comes up in the 2007 ISG report (see pages 93-96). It points towards difficulties with post-mortem culturing of bTB as the reason that the disease would be overlooked in unconfirmed reactors.

In hindsight, and what may be seen now as the disastrous move, the ISG analysis decided just to use 'confirmed' breakdown-only data as opposed to 'all' breakdown (confirmed and unconfirmed). Simon More in Dublin had also spotted this.

What happens when you add all the unconfirmed test results back into the model as being correctly identified as having bTB is that there is no statistically significant effect of proactive culling of badgers on new herd bTB breakdown.

Here was a lethal blow to the ISG proactive cull analysis and conclusions. The ISG should have concluded that the RBCT had failed to find a link between proactive badger culling and a reduction in bTB herd breakdowns - the exact opposite of its finding.

Instead it concluded that badgers do pass bTB to cattle at a significant rate. It also said badger culling was not worth doing because of a balancing effect resulting from perturbation effect causing herd breakdown. That was a story that the public and government of the day embraced - but one that extended way beyond the limits of safe scientific conclusion, and one that the government kicked into the long grass.

An inconvenient truth becomes increasingly obvious

The strength of the RCBT had slowly crumbled to bits over my year of study. Several scientists along the way advised me that the RBCT was 'not strong science'. But busy with their own issues, they had tended to see the 'pro-badger' ISG conclusion that the advice not to cull badgers was possibly 'right for the wrong reason' and so fairly harmless - not realising what a change of government might then do.

Others had entertained doubts but felt no need to comment over the bTB 'hot potato', especially as much of their funding was provided by government.

By August 2016, as more badger killing was announced, the awful truth was becoming ever more obvious. The badger protection movement, with few exceptions had joined with the ISG scientists to uphold 'the ISG science', based upon badgers giving bTB to cattle with significant frequency.

They were supported by several Oxford University related academics, although I noted this was often cautiously on more general terms than the ISG specific findings. Speaking out were some who were behind the scenes of the 1997 Krebs review and its RBCT.

On checking and double checking, and testing the frailties, many closest to the issue did not want to talk about it, which just seemed suspicious. Some wanted it covered up for tactical reasons. The phrase 'reputational damage' was used.

A lack of mutual understanding between veterinarians and modellers?

Looking back to the RBCT design, it does seem odd that John Krebs and Roy Anderson at Oxford University had concluded the need for a trial of the kind undertaken, given the clear uncertainties at the time over the disease and the role of wildlife.

Robert May (Oxford and Imperial) who was Chief Scientific Adviser (1995-2000) at its origination, and who worked closely with Krebs, has acknowledged that (see page 302) the use of mathematical models during the 2001 Foot and Mouth Disease epidemic had created controversy based on a "lack of mutual understanding between veterinarians and modellers."

It looks perhaps as if the RBCT may have been a prelude to such problems, but this time involving zoologists. Krebs was recently quoted at the Royal Society as saying "We must acknowledge as scientists that we don't always get it right. Models make assumptions, labels slip in freezers."

Was this perhaps a message regarding the trials that bear his name? Now is the time to find out - before £100 million that would be better spent helping cattle farmers is used to kill and injure 100,000 or more English badgers, all because weak science, and weaker statistics, failed the farmer, cow and badger.

 


 

Tom Langton is a consulting ecologist to government, business and industry who provides advocacy support to charities and pressure groups seeking justice where environmental damage is being caused to species and habitats.

This article is co-published with the Badger Trust. It is scheduled for publication in the next edition of Badger News.

Other articles by Tom Langton

More from this author