Do we need a badger cull to control bovine TB?

What is bovine TB?

Bovine tuberculosis (TB) is a chronic bacterial disease of cattle, found worldwide. The causative organism (Mycobacterium bovis) can also infect humans, particularly through drinking milk from infected cows. In most countries pasteurisation of milk, coupled with close inspection of cattle carcases at slaughterhouses, has eliminated transmission to humans.

Most European and some Latin American countries claim to have successfully controlled or eradicated bovine TB in the cattle population through a 'test and cull' strategy.

Cull:  To remove rejected members or parts from (a herd, for example). Definition by the Free Online Dictionary

Bovine TB is diagnosed with either the tuberculin skin test, or by the gamma interferon blood test. Neither test can be relied upon to detect all infections, and both give some false positives.

Movement restrictions are imposed on herds which have a reactor, and reactors and animals in close contact with the reactor are isolated from the rest of the herd and removed to slaughter. Affected herds are re-tested periodically and restriction is only lifted after the herd has one or two clear tests. Diagnosis is confirmed (or otherwise) post mortem by laboratory techniques.

Jersey cattle at the milking yard on an East Sussex farm (Photo: InfluentialPoints)

In a few countries 'test and cull' has failed to 'eradicate' the disease in cattle. This has been ascribed to the the involvement of a wildlife reservoir: white tailed deer in the USA, possums in New Zealand and badgers in Ireland and Great Britain.

These countries have therefore included wildlife culls in an attempt to eliminate the disease, but with varying degrees of success.

Eradicate:  Permanent reduction to zero of the worldwide incidence of infection caused by a specific agent as a result of deliberate efforts. Definition by Dowdle (1999) Note: this term is frequently misused to apply only within a specified geographic area - which is correctly termed disease elimination.

Possibly the least successful attempts have been made in Great Britain, where badger culling as become a highly contentious issue which arouses passions on all sides.

• Most farmers believe badgers are the major source of bovine TB outbreaks in cattle, and want a badger cull, especially in areas where disease incidence is high.
• Many wildlife experts oppose such a cull arguing that it may make the problem worse and is anyway uneconomic.
• Opinion polls also suggest most of the public are strongly opposed to a cull.

What is the history of Britain's attempt to 'eradicate' the disease?

Efforts have been made to eliminate the disease in UK since 1935 when voluntary tuberculin testing and culling of cattle was introduced. A compulsory testing programme began in the 1950s.

• Areas were declared to be attested after all animals with a positive tuberculin skin test reaction (so-called reactors) had been removed for slaughter, and two successive tests of each animal had shown that all herds in the area were TB free.

By 1960, the whole of the UK had been declared attested. The disease had not been eliminated, but it had been effectively controlled - yearly herd incidence had been reduced to about 2%. A continuing cattle testing and cull programme further reduced this to well below 1% in most of Great Britain, but in south-west England it appeared to level off at about 1.5%.

Control:  The reduction of disease incidence, prevalence, morbidity or mortality to a locally acceptable level as a result of deliberate efforts; continued intervention measures are required to maintain the reduction. Definition by Dowdle (1999)

The UK Ministry of Agriculture concluded in 1973 that this was because wild badgers were providing a reservoir of infection. Since the policy in Britain was (and still is) to 'eradicate' rather than control animal diseases, the government embarked on a badger cull policy that was to extend on-and-off to the present time. The initial removal method was to permit affected farmers to kill (all) badgers on their own farms by shooting (the method to which the UK government now proposes we return).

Concern about the welfare implications of these methods led to the government taking over control operations. Once infection in cattle had been attributed to badgers, populations were sampled up to one kilometre from the farm boundary to identify infection status. Setts (dens) of infected social groups, and other social groups in contact with them, were then gassed with hydrogen cyanide.

Gassing operations began in August 1975. These measures appeared to be successful in further reducing the cattle TB infection rate, with prevalence in 1979 in the south-west dropping to 0.5%, and the rest of the country to 0.1%. However, from 1980 the number of herd breakdowns started to increase again in the south-west, whilst temporarily remaining fairly stable at around 0.1% in the rest of the country.

Figure reproduced here for critical appraisal is from Krebs (1997)

Between 1982 and 1985, a clean ring strategy was introduced to replace the gassing strategy. Under this strategy social groups were identified by bait marking - an improved method to identify groups since one group can use several different setts. Those groups, found on laboratory examination to be infected, were culled, extending out to successive social groups until a clean ring of uninfected social groups was found. During this period, the proportion of affected herds slowly increased in the south-west (where most of the culling was going on), but remained steady elsewhere.

In 1986 the situation was reviewed by the Dunnet Committee and the so-called interim strategy was introduced. This was intended to offer a means of controlling badgers on infected farms, pending the development of the live test, when only infected badgers would be killed (in other words true culling rather than removal of all badgers). All badgers on the breakdown farm were removed by cage-trapping and shooting, but no removal was carried out on neighbouring farms. This new strategy failed to stop the increase and the proportion of herd breakdowns increased five fold in the south-west, to reach about 2.5% by the mid 1990s. In the rest of the country a similar increase started and by 1996 had reached 0.6%.

The obvious question to ask is why did a method (cattle test and culling only) that had worked so well up to 1975 (and in many other European countries) become so inneffective post-1980, even when used in conjunction with badger culling.

What can we learn from these descriptive data?

One should be able to learn something about how to control bovine TB from examining these time trend data - but there are two very major constraints on this.

• Firstly just because B follows A does not necessarily mean B is caused by A. (Wakefield made the disastrous error of concluding that the measles-mumps-rubella (MMR) vaccine causes autism in children because the MMR vaccine is given at around 12-15 months of age and autism in a child tends to become evident at about 18-19 months.) Regarding badgers and bovine TB, we could argue that the proportion of cattle infected dropped sharply in 1975 because killing of badgers was allowed. But we could similarly argue that the steady increase in proportion of TB infected cattle from 1980 onwards was a result of culling badgers.

• The second constraint is that this sort of routinely collected data is often of very poor quality. If we are to understand what is going on, we need measures of the population sizes and proportion infected for both cattle and badgers.

  • The cattle data are relatively good, but even here the proportion infected with bovine TB only applies to herds, not to individual cattle. In addition the test is certainly not 100% accurate and we may be missing low grade infections.

  • There is much less information available on badger population sizes. Cresswell et al. (1990) surveyed 2455 1-km squares throughout Britain for badger setts and signs of badger activity, and estimated (making some very dubious assumptions) that the overall population of badgers in the UK was about 250,000. Wilson et al. (1997) did a repeat survey mostly of the same squares which suggested that the density of badger setts had increased significantly in two areas in south-west England and the west midlands.

  • On badger infection rate, a small study in 1979 on just two farms that had recently experienced disease outbreaks, revealed very different prevalences in badgers of 11.1% and and 31.6% (Barrow & Gallager, 1981 ). However, sample sizes were very small. Samples obtained during cull operations were similarly small and unrepresentative. Once the government was involved in gassing, samples became much larger, but remained unrepresentative. Samples from road accidents were regarded as the least biased, but sick badgers may well be more likely to get killed than healthy badgers.

    

    A dead badger (Meles meles) by the side of the road (Photo: InfluentialPoints)

    The only reliable data on badger infection rate were from very restricted areas. For example Delahay et al. (2000) found that prevalence from 1982 to 1996 varied between 10 - 17% in one site in south-west England. There was some evidence of an increase in prevalence in the late 1980s, but temporal trends in disease were not synchronized amongst neighbouring groups. By the time the Krebs trial was initiated (1988-2002), average prevalence in badgers in the ten trial areas was 11.3%, although it varied between areas from 1.6% to 37.2%. (Bourne, 2007a).

In other words we know that cattle infection rate reached its lowest level in 1979 - but why it then increased again despite all the control measures is unclear. Badger numbers may have increased over the period, but there are insufficient data to draw any conclusions about trends in badger infection rates.

The turn to (and away from) science

As the number of disease outbreaks in cattle continued to climb, something clearly had to be done - but what? In 1996 the then Conservative government set up an independent scientific review under the chairmanship of Professor John Krebs to review and make recommendations on government policy on badgers and bovine tuberculosis. Krebs (1997) concluded that it was not possible to state quantitatively what contribution badgers made to cattle infection, because the relevant data had not been collected. The main recommendation was to set up a randomized trial to directly compare the effects of three 'treatments' on the number of disease outbreaks. Those treatments were:

1. proactive badger removal irrespective of whether there were cattle infections or not,
2. reactive culling where culls were only carried out when tuberculosis was found in the cattle (the current policy at that time), and
3.   no badger culling.

It was recommended that in both reactive and proactive removal areas there should be "total removal of complete badger social groups" from the specified areas (that is extermination). All three treatments included regular cattle testing, and culling of infected cattle. An Independent Scientific Group (ISG) led by Professor Bourne was set up to oversee the trial.

Despite serious problems in its execution (see below), the trial was eventually completed in 2005, albeit without total removal of badgers from culling areas. Donnelly et al. (2007) concluded from the trial that badger culling was only likely to be beneficial if conducted systematically over large areas, and sustained over several years. Reactive culling had overall detrimental effects because of the perturbation effect on badgers where they dispersed and spread the infection outside the original affected area. The ISG in their final report to government (Bourne et al., 2007a ) concluded that badger culling can make no meaningful contribution to bovine TB control in Britain, and that some policies under consideration (those similar to reactive culling) were likely to make matters worse rather than better.

The Chief Scientific Adviser (King, 2007) did not accept these conclusions, in particular the negative effects of reactive culling. He instead stated that removal of badgers should take place alongside the continued application of controls on cattle.

The ISG responded (Bourne et al., 2007b ) that King's remit from government did not include economic [and] practical issues, which were absolutely critical in determining whether culling would reduce or increase the incidence of bovine TB. In addition King's report contained fundamental flaws in interpretation of the data. Hence the ISG maintained its previous position.

Eventually a common position was cobbled together (House of Commons Environment, Food and Rural Affairs Committee, 2007) based on the argument that Sir David King's group of experts did not include the practicalities or costs of culling in its considerations. The government of the time then accepted that culling would not be helpful (House of Commons statement by Hillary Benn (2008)) and provided increased funding to a research programme to develop a vaccination approach, including a series of vaccine trials.

The coalition government which took power in 2009 also proposed a 'science-led' policy, but its first move was to scrap all but one of the planned Badger Vaccine Deployment trials (DEFRA, 2010a). The following year the recommendations made by the ISG following the Krebs trial were abandoned (House of Commons statement by Caroline Spelman, 2011) and it was decided to reintroduce badger culls, this time by free-shooting badgers as they emerge from their setts in the evening. Cost issues were settled (supposedly) by farmers and landowners having to pay for the cull themselves.

Without going into the whole sorry story in any more detail, we step back a little to look at the evidence. First we ask how close is the association between infection in cattle and badgers. Then we ask how transmission each way occurs. Lastly we ask whether reducing the number of badgers does actually reduce disease in cattle.

If at this stage, you loose interest and say 'well its obvious isn't it' - consider the disturbing case of hormone replacement therapy (HRT): By 2001 millions of women in Europe and America were taking HRT, based on results from over 30 observational studies that suggested a 44% reduction in coronary heart disease. Yet within a few years a large-scale randomized trial demonstrated that, far from reducing the risk of heart disease, HRT slightly increased the risk (Petitti, 2004 ). When looking at descriptive and observational studies Petiti concluded: Do not turn a blind eye to contradiction. Do not be seduced by mechanism. Suspend belief. Maintain scepticism.

How close is the association between infection in cattle and badgers?

Observing A is spatially associated with B does not, of itself, prove A causes B, any more than observing that A preceeds B.

Nevertheless, if badgers are the primary source of bovine TB in cattle (or vice versa) their infection should be strongly and positively associated.

Since the fallacy is commonplace, a concrete example may help: Observing the number of churches in UK cities is associated or correlated with the number of bars does not imply one causes the other.

1.  Past evidence was reviewed by Krebs (1997).
   • Bovine TB infections in both badgers and cattle are highly clustered, and these regional clusters (sometimes termed hotspots) are geographically associated in the two species.
   • Another line of evidence is that the prevalence of bovine TB infection among badgers culled following bovine TB outbreaks in cattle is higher than among those killed in road traffic accidents.
   • Results from Northern Ireland also suggested there was a positive association between the the number of active main setts and the risk of a bovine TB breakdown.
   From these and other data, Krebs concluded that there was strong evidence for an association between bovine TB in badgers and cattle, but noted that there were many problems with how the data had been gathered, and recommended that more data be collected to properly assess the risk.

2.  More recent evidence has been mostly obtained via the randomized trial (described below). Woodroffe et al. (2005) looked at nearest neighbour distances between infected and uninfected badgers and cattle. This demonstrated that smaller scale patterns of infection in the two species were spatially correlated, and also that there were close linkages in the distribution of M. bovis strain types in the two species. Jenkins et al. (2007) investigated the impact of badger culling on the spatial distribution of bovine TB infection in badger and cattle populations. Bovine TB infection was significantly clustered within badger populations, but clustering was reduced when culls were repeated across wide areas. There was significant spatial association between bovine TB infections in badgers and cattle herds across successive culls, but this became non-significant when the initial observation was excluded. These patterns are consistent with the idea that badgers are less territorial and range more widely in culled areas, allowing disease transmission to occur over greater distances. Continued clustering of bovine TB infection in cattle, even where badgers were repeatedly culled over wide areas, was thought to reflect cattle-to-cattle transmission.

3.  Probably the best information available on this is from Hone & Donnelly (2008) and Donnelly & Hone (2010). They reported that they had analysed data from the 10 sites randomly selected to be proactive culling sites in the UK Randomized Badger Culling Trial. The authors stressed that the data were observational (not experimental) with the badger data only from the initial cull and the cattle TB incidence data for the year prior to the cull. They developed a priori two-host mathematical models of relationships between bovine TB infection in cattle and badger populations, and then evaluated the predictions of such models relative to the data.

   Their model prediction with the most support (based on Akaike's Information criterion) was a positive linear relationship through the origin between the density of infectious cattle and the density index of infectious badgers (shown below). Several of their models - namely those with density-dependent transmission within and between species, and/or environmental transmission - predicted such a relationship. The figure below shows how closely the data fitted that relationship. The coefficient of determination was quite high at R² = 0.869. This could be taken to suggest that a high proportion of the variation was thus explained, but since it was assumed that the line passed through the origin, R² no longer has its usual simple meaning ( Kvålseth, 1985)

   

   Modified figure reproduced here for critical appraisal is from Hone & Donnelly (2008).

   Whilst recognising that their results were based on observational data, a small data set and two clusters of data points (shown orange and green in our graph), Hone & Donnelly argued that this graph demonstrated a close positive relationship between bovine TB in cattle herds and badgers infectious with M. bovis - with infection passing in both directions. They suggested that this meant that bovine TB in cattle herds could be substantially reduced, possibly even eliminated, in the absence of transmission from badgers to cattle, providing there was no change in the situation during the process of reducing transmission.

   Clearly there is a relationship here - but describing it as 'close' is questionable.

   • In the right hand cluster, the initial proactive culling of badgers was delayed until after the foot and mouth disease (FMD) epidemic in 2001. Over this period all testing for bovine TB ceased. These triplets showed markedly higher levels of bovine TB in both cattle and badgers, interpreted by Bourne et al. (2007) as evidence of cattle to badger transmission when cattle testing and culling was interrupted by the FMD epidemic. Hone & Donnelly felt that a more uniform spread of data across the range of density of infectious badgers would have been desirable. This is true, but what is more important is that the same relationship should hold within each of the two clusters.

   • Examination of the data shows that there was no evidence of a positive relationship within each cluster - separate regressions gave non-significant negative slopes. Of course, if there is lot of measurement error, this would not be surprising. But then as the authors themselves point out, the regression model assumes that the independent variables are estimated without error. This is not a serious problem if that error is small, but then we have just had to assume there is a high level of measurement error to make sense of the within cluster relationships...

   We have some further points of concern:

   • The statement that 'sites were randomly selected to be proactive culling sites' is misleading. The sites for proactive badger culling (at least 9 out of 10 of them) were randomly allocated to this treatment from a total of 30 sites - but those 30 sites appear to have been convenience selected. Hence any bias in the selection of the 30 original areas will still affect the 9 sites randomly allocated to proactive culling.

   • We suspect that there is a high risk of spatial autocorrelation in these data which would result in an overestimate of the strength of any relationship.

     Convenience sampling is where study units are chosen because they are most accessible or convenient, and no attempt is made to obtain a representative sample. A good example is journalists doing 'person in the street' interviews to assess public opinion. Here bias is inevitable, and it is impossible to extend any conclusions beyond the actual samples taken.

In conclusion, there seems to be a relationship between the density of infectious cattle and the density index of infectious badgers, probably resulting from transmission between and within both hosts. But that relationship is not very close, even when the best available data are used, and may not be linear. It is especially worrying that there is no relationship within each of those clusters. This could be because the data are still not very good (high levels of measurement error) or it could be because one or more major explanatory variables and/or confounding factors are not being considered.

How does transmission each way occur?

This section is unfortunately very short, because we still have very little understanding of how bovine TB is transmitted from badgers to cattle, or vice-versa. The International Scientific Group (Bourne et al., 2007) concluded, and King (2007) agreed, that the most likely means of transmission are inhalation of infected droplets from the lungs of other infected animals, or oral ingestion of mycobacteria from farm environments. The only direct experimental evidence comes from the work of Little et al. (1982). They demonstrated that bovine TB infected badgers can transmit M. bovis to cattle, but only under unnatural experimental conditions - cattle & badgers were kept together in a concrete-lined yard where the badgers slept in a metal pig sty.

So one is left with asking how frequently badgers may encounter cattle under natural conditions.

• Benham & Broom (1989) concluded that the normal behaviour of badgers would not result in direct transmission of tuberculosis from badgers to cattle via air expired by badgers or via bodily contact. Similarly Benham & Broom (1991) found that nearly all cattle strongly avoided herbage contaminated with badger faeces or urine.

• More recently Böhm et al. (2009) used novel proximity logging devices to show that badgers and cattle came within 4 m of each other on pasture infrequently, but it was not as rare as previously thought. Tolhurst et al. (2008) used remote surveillance, radio-tracking and faecal analysis to show that badgers do sometimes exhibit close, investigative 'nose-to-nose' contact with housed cattle and also excreted/scent-marked on and around feed.

It is potentially much easier to prevent contact between housed cattle and badgers than when they are out at pasture, but until we know the relative risks of contact in housing versus pasture, it is difficult to predict how effective increased biosecurity measures around housing and feed stations will be.

Badger in a farm building (Photo: InfluentialPoints)

Does reducing number of badgers reduce disease in cattle?

Descriptive studies

We tried above to interpret the descriptive data on changes over time in the incidence of tuberculosis in cattle herds in England and Wales, and found it difficult to reach any firm conclusions. We can instead compare trends in disease incidence (% reactors of total cattle population) between different countries which differed in their use of badger culling for TB control.

•  Northern Ireland has not been culling badgers at all.
•  The Irish Republic has culled an increasing number of badgers each year from around 2000 per year in the mid 1990s to around 6000 per year in the late 2000s each year.
• Great Britain culled badgers up to the mid 1990s, but then it was (officially) restricted to the proactive culling areas in the randomized trial.

How much illegal culling has been carried out is difficult to assess, but it is certainly widespread in Great Britain and the Irish Republic.

Figure reproduced here for critical appraisal is from TBInfo.com et al. (2005a).

The trends in each country are very different.

•  In Northern Ireland the percentage of cattle found to be reactors was around 0.5% in 1999, but it then increased sharply in 2002. This probably resulted from the breakdown of the bovine TB testing system following the 2001 FMD outbreak. This resulted in both increased local cases, and import of untested animals to replace animals culled. Once testing was re-established, TB declined again to its previous level up till 2010.
•  In the Irish Republic the percentage of reactors declined slightly from around 0.6% in 1999 to around 0.4% where it has stayed ever since.
• In Great Britain it was low in 1999, but increased sharply in 2002 (as in Northern Ireland) following the FMD outbreak. However, unlike Northern Ireland, the rate did not drop again, but instead continued to increase, especially in the south-west of the country.

Certainly one would be hard pushed to reach a clear conclusion on the advisability or otherwise of badger culling from these descriptive data. All one can say is that after very different histories, each country seems to have ended up with roughly the same bovine TB prevalence (albeit with great within-country variation)!

Observational badger removal projects

We now consider badger removal projects. These were intended as experiments, but there was no random allocation of treatment to area, sometimes no reference (without culling) area, and sometimes no replication. As such they do not meet the basic requirements for a scientific experiment, and are correctly classified as observational studies.

Three such studies were carried out in Great Britain - at Thornbury in Avon, at Steeple Leaze in Dorset, and at Hartland in North Devon. In the Thornbury trial badgers appear to have been completely removed (at least temporarily) and no new infected herds were recorded for over ten years (Clifton-Hadley et al., 1995; Gallagher et al., 2008) A similar result was obtained at Steeple Leaze. However, documentation of the studies seems to have been rather poor and there were no survey-only areas.

Two studies were carried out in the Republic of Ireland. In the east-Offally project proactive badger took place in one area from 1989 to 1995. Máirtín et al. (1998) compared the proportion of new confirmed tuberculous herd restrictions in that area with cattle from an area where only reactive removal was practiced. The incidence of bovine TB in cattle fell markedly in the removal area and less markedly in the neighbouring reactive culling reference area. The study was expanded post 1995 over a somewhat larger area. By 2004, Kelly et al. (2008) observed a significant decreases in herd restrictions of 22% in the entire proactive removal area and 37% in the inner proactive removal area.

Another non-randomized trial, the Four Areas Study, was carried out in Ireland from September 1997 to August 2000 (Griffin et al., 2005a). Matched removal and reference areas (average area of 245 km²) were convenience selected in each of four counties of Ireland: Cork, Donegal, Kilkenny and Monaghan. In the removal areas badger culling was intensive and proactive throughout the study period, with 0.57 badgers /km² removed. In reference areas it was instead reactive, with badgers culled only in response to severe tuberculosis outbreaks in cattle, with 0.07 badgers /km² removed. The response variable was restriction of cattle herds where tuberculous lesions were detected in one or more animals.

Figure reproduced here for critical appraisal is modified from Griffin et al. (2005a).

In the final year of the study, the odds of a confirmed herd restriction in the removal compared to the reference areas were 0.25 in Cork, 0.04 in Donegal, 0.26 in Kilkenny and 0.43 in Monaghan. At first sight, this trial does appear to provide fairly decisive evidence that relatively intensive culling with minimal reinvasion does reduce the incidence of bovine TB in cattle.

Against that it has to be said that using reactive removal as the control may well have increased the bovine TB rate in the reference areas, so that the effect was overestimated. Griffin et al. (2005a) rejected this arguing that there was no significant increase in levels of tuberculosis in cattle in response to reactive badger removal in the reference areas. Whilst this is true, the graphs in Griffin et al. (2005b) (not shown here) do show a marked peak in 1998-1990 (higher than any previous levels) in three of the four (reactive culling) reference areas.

Griffin et al. also argued that the intensity of badger removal in the reference areas was very low - too low to result in any perturbation of badger populations. But this is only the case when averaged across the whole area - in the affected areas the intensity of removal was very high. The authors concluded by saying that although feasible, widespread badger removal was not a viable strategy for the long-term control of tuberculosis in the Irish cattle population, and therefore there was a need to develop an effective vaccine for badgers.

These observational studies certainly strongly suggested that badgers were involved in the transmission of bovine TB. Where badger removal was near complete, there were no new TB cases for several years. But in other areas the level of tuberculosis reduction did not seem to match the level of badger reduction. And, as we should have learnt from medical trials, observational studies using non-random allocation are almost invariably biased and unreliable.

The randomized badger culling trial (RBCT)

Lastly we come to the Krebs randomized trial. This was the first properly designed trial with randomized allocation of treatment to experimental units (in this case areas). We nowadays expect very high standards of randomization in medical trials, simply because we know that any other approach tends to give the wrong answer. But the Krebs trial was the first (and so far only) instance where the approach has been used to compare methods of controlling bovine TB.

Thirty trial areas, each about 100 km2, were selected as ten matched triplets, all in areas of high cattle TB incidence. Where possible, they followed geographic barriers likely to impede badger movement. Neighboring trial areas were separated by buffer zones at least 3 km wide.

All trial areas were surveyed for badger activity and then (most) were randomly allocated to treatments. Allocation was done such that each treatment - proactive culling, reactive culling, or survey only (no culling) - was repeated ten times, once within each triplet. Consent was sought from landholders before areas were surveyed for badger activity and culling treatments allocated. The proportion of inaccessible land within proactive treatment areas varied from 15% to 50% (30% overall).

Following treatment allocation, initial badger culls were conducted on all land in the proactive areas for which consent was given. Culling treatment area boundaries were defined beyond trial area boundaries where necessary to ensure that all badgers likely to use land inside the trial areas were targeted. Badgers were captured in cage traps placed primarily at setts. No culling took place in February - April each year to avoid killing mothers with dependent cubs. No attempt was made to remove all badgers in an area, but instead trapping operations were conducted over a fixed number of nights (initially 11 and then 8).

Initial badger culls for each proactive trial area were completed between December 1998 and December 2002, and 'follow-up' culls were repeated approximately annually with longer delays in 2001 because of the FMD epidemic. As soon as the initial proactive cull was complete, data were collected on bovine TB incidence in cattle in and around trial areas, using established veterinary surveillance. Primary analyses were based on the incidence of confirmed breakdowns.

The first results published from the trial were not as expected. An interim analysis carried out before the planned completion of the trial (Donnelly et al., 2003) suggested that localised reactive badger had not reduced bovine TB incidence in cattle, but may well have increased it. The reactive treatment was associated with a 27% increase in the incidence of cattle herd breakdowns when compared with no culling areas. The 95% confidence interval to this was a 2.4% decrease to 65% increase. Hence, whilst the increase was not quite significant at P = 0.05, it seemed very unlikely that this treatment would reduce TB incidence. As a result, the government discontinued the reactive culling treatment, and the trial continued with just two treatments - proactive culling and survey only.

Woodroffe (2008) reported that the proactive cull had been effective in reducing the number of badgers. There was a 73% reduction in the density of badger latrines, a 69% reduction in the density of active burrows and a 73% reduction in the density of road-killed badgers. Localized 'reactive' culling caused reductions of 10 - 32% in the 'signs' of badger populations.

Donnelly et al. (2007) reported on the impact of the culling on the incidence of bovine TB in cattle. During the trial period, bovine TB incidence in cattle was significantly lower by an average 23.2% inside culled areas (red points below), but non-significantly higher by an average 24.5% on land within 2 km of the culled area (green points below), relative to matched unculled areas.

Figure reproduced under a Creative Commons licence from Jenkins et al. (2010).

Subsequent analysis by Vial & Donnelly (2011) using a case-control study design supports the conclusion that localized reactive culling increased the risk of bovine tuberculosis in nearby cattle herds.

Inside the culling area the beneficial effect of culling tended to increase on successive annual culls giving a 35% decrease in bovine TB by the end of the culling period. After culling ended, the positive effects inside proactive culled areas initially became more pronounced to give a 50% decrease in bovine TB. The positive effects declined thereafter, although the sudden increase at months 31-36 resulted from use of incomplete data (see below); the full data indicate a more gradual reversion to the previous TB level.

The detrimental effect in adjoining areas was ascribed to the perturbation effect (Carter et al., 2007). Culling of badgers results in immigration into culled areas, disruption of territoriality, increased ranging and mixing between social groups. The increased contact rates between social groups are thought to exacerbate bovine TB transmission. The detrimental effect tended to diminish on successive annual culls and were no longer apparent by the end of the trial.

A cost-benefit analysis suggested that benefits in terms of reduction of bovine TB in the culling zones did not offset the costs of culling and of the increase in bovine TB in adjoining areas.

One might think that being the 'best' trial so far would lead to general acceptance of the results - but far from it!

• Ecologists and wildlife conservationists have generally accepted the results.
• But the response from most of the farming community has been negative, as it has from much of the veterinary establishment.
• Even the implementing body DEFRA noted that: "The large number of biases inherent in any field trial makes interpretation of the results generated from them difficult." This may seem an odd comment to make about the only randomized trial that has been carried out on this issue. But it is important to acknowledge that there were many problems both in the design and implementation of this trial.

Our own criticisms would focus on the following:

1. Too few replications
   The number of replications was very small, at 10 for each of two (originally three) treatments. This level of replication was set purely by power considerations, which are concerned with demonstrating statistical significance for a given effect size - not with obtaining a result which can be extrapolated to a larger area (unless the original experimental units are randomly selected). If this had been a medical trial, it would be regarded as an (excessively) small matched-cluster randomised trial.

   More trials would have to be run in different areas to obtain generalizability, followed by a meta-analysis to estimate the overall effect size. But the chances of being able to run any more randomized trials on this topic seem vanishingly small. The enormous cost of the trial (estimated at £50 million) seems likely to doom this approach - at least run by a goverment department - to the history books. As it is these are the results of just one trial carried out at 30 convenience-selected sites in south-west England at one point in time - extrapolation is therefore highly speculative.

2. Non-random treatment allocation
   Treatment was only allocated randomly in 9 of the 10 triplets for undefined 'security' reasons. The inconvenient fact that 10% of allocations were done non-randomly, rather than in a (stratified) random manner was completely ignored in all analyses. Results should have been analyzed with and without that triplet so that any bias could be assessed.

3. Implementation problems
   Some landowners refused access for badger culling, and (no doubt) some illegal culling continued in 'non-culling' areas. Treatment and monitoring schedules were severely disrupted by demonstrators and especially by the FMD outbreak. This meant that badger culling could not be carried out at annual intervals, and the routine cattle TB test and cull programme fell apart.

   Others have also commented on the implementation problems.

   • One of the Field Managers (Caruana, 2006) felt that compulsory entry on to farms was essential, and that eight days per year was totally inadequate to trap-out the badger. He concluded that 'the trial has far too many flaws in it to be trusted to produce meaningful answers'.
   • Gallagher et al. (2008) also noted that serious questions remained concerning the efficiency of culling in the RBCT. They stressed that past badger removal projects had been much more successful because complete social groups had been removed.

   Whilst accepting that the trial had 'problems', this does not invalidate the results providing we treat it as a pragmatic trial. In the real world there are likely to be even more implementation problems than was the case here. In medical trials it is usual analyze results by intention to treat - in other words all experimental units randomized to treatment are subjected to analysis, irrespective of errors in treatment assignment, breakdown in blinding and withdrawals. If we treat the badger trial as a pragmatic trial, then the numerous implementation problems (in particular the failure to kill all badgers in the cull areas) are simply included as representative of the sort of the problems there would be if the approach was used in practice.

4. Reactive culling treatment should not have been terminated.
   This decision was highly questionable given that it was made just before the treatment effect reached statical 'significance'. Reactive culling was the method of control that had been used to date, and would be the only method of control that could realistically be used in future. The early termination of this treatment means that some (for example More et al., 2007) can still argue that the data do not provide sufficient evidence for adverse effects of reactive culling. Perhaps the treatment was only terminated for political reasons - although it must be said that the researchers lost credibility by going along with it post hoc.

5. Cost-benefit analysis was simplistic
   The cost-benefit analysis has been widely criticised for initially using the costs of cage-trapping rather than the cheaper methods of 'free shooting' or snaring. However, Jenkins et al. (2010) did consider alternate methods, including licenced culling, but concluded this approach would almost certainly be patchy and unsustained, and hence most likely prompt increases, rather than reductions, in the incidence of bovine TB in cattle. Our own criticism of the cost-benefit analysis is more serious, in that no value was assigned to badgers - an issue we address below.

6. Premature publication of incomplete analyses
   Some of the analyses (including some economic analyses) were published prematurely, with incomplete data for some time periods. Whilst this is understandable, given the pressure from government for 'answers', it was still unwise since it means that graphs in one scientific paper suddenly change in the next paper - leading to criticism and loss of credibility (bovinetb.info, 2011).

Given the fact that we only have one small randomized trial, we have to draw conclusions from all studies, but give the result of the randomized trial more weight. Complete removal of all badgers (and in some areas deer) from an area, coupled with rigourous test and cull of cattle and elimination of reinvading wildlife, would reduce bovine TB to a low level, and may in the long run eliminate the disease in the area concerned. Removing most badgers (say 70%) over small areas in response to outbreaks (reactive culling) is counter-productive and may increase infections. Removing most badgers (say 70%) over a large area will reduce cattle infection rates by 20-50%, but is probably not cost effective, even if the wildlife resource is assumed to have no value.

Disease eradication or control

The official policy in Britain (and the rest of the European Union) is to eradicate bovine TB from cattle. This is laid out in 'The Bovine TB Eradication Programme For England' (DEFRA 2011b). It is true that disease eradication has been achieved for smallpox in humans, and has recently been claimed for rinderpest in cattle. These diseases, however, have single maintenance hosts and hence eradication is a meaninful objective (CFSPH, 2008). But no disease with multiple maintenance hosts has ever been eradicated - and may never be. Moreover global eradication programmes are extremely expensive and can have very adverse side-effects, especially in relation to diverting resources from effective control methods (see Caplan, 2009 on 'Is eradication ethical?').

Even disease elimination - namely reduction to zero in the incidence of infection within a specified geographical area - is impractical when you have several wild maintenance hosts as with bovine TB. TB infected cattle can be removed using the 'test and cull' approach, with affected herds put under movement restriction and re-tested periodically to eliminate cattle that may shed the organism. But this approach cannot be used for wildlife reservoir species, which in Britain means badgers and fallow deer. Because sick badgers are more likely to get culled, large scale pro-active culls (actually a misuse of the term 'cull') may sometimes reduce the disease prevalence in badgers (Corner et al., 2008), but cannot possibly eliminate infections in a wild population.

We have no reliable diagnostic test for an individual badger, and no way to keep 'contacts' under observation. Hence, if one is really trying to eliminate all infections, the only culling option is the total elimination of badger & deer populations.

Observant readers may note the definition of cull has expanded from its earlier meaning of selective removal of infected individuals, to encompass area-wide extermination of badgers. Let us be grateful similar methods are not applied to human disease.

This of course brings vehement protestations from farmers, veterinarians and politicians that killing all badgers is unthinkable, and that they only want to reduce the number of badgers. If that is the case, then perhaps it would be better all round if the European Union, the British government and the veterinary establishment finally stopped talking about disease eradication and thought rationally about disease control.

In fact, as we have seen, the only reduction that would have much effect upon the bovine TB infection rate in cattle is a massive reduction in badger numbers, a point apparently accepted by the government scientific establishment. Sir David King told a Parliamentary Committee that, in his opinion, a reduction by 70 to 80% would be perfectly acceptable and would be within the terms of the Bern Convention on the Conservation of European Wildlife and Natural Habitats (House of Commons Environment, Food and Rural Affairs Committee, 2007). His statement is frighteningly reminiscent of the drive to eradicate tsetse flies from southern Africa, where for many years game-elimination was viewed as a viable strategy to achieve that end. Between the 1920's and 1960, 1.3 million game animals were killed, including many species (such as rhino) which are now endangered.

The classic quote was given by John Ford:

"[The failure of game shooting operations to eradicate tsetse in the Sabi Valley] does not imply any intrinsic failure of the method, but indicates only that to achieve control much more intensive killing would have been required." Ford, J. 'Control by destruction of the larger fauna' in The African Trypanosomiasis (Ed. C. H. W. Mulligan), Allen and Unwin, London, p. 563.

The mass killing of wildlife that went on in Africa is now regarded as an appalling waste of resources - as would be the widescale elimination of Britains largest remining wild predator. African governments are now well aware their wildlife resources have a value, and allow for this when comparing the cost-effectiveness of disease control strategies. Perhaps the British and Irish governments could do the same.

So do badgers have any value?

Perhaps the most common response to this from advocates of badger culling is that badgers are not endangered in Britain, with a population size of over 300,000. But this implies that a species has no value unless it is close to extinction. The British badger population has high value in international conservation of the species, simply because it is the largest most stable population in Europe not adversely affected by hunting.

Its importance is recognised by the the Bern Convention on Conservation of European Wildlife and Natural Habitats (1979). It is even accepted by the UK government that mammals such as badgers, otters and seals have great cultural significance (DEFRA, 2011a). As Sir Gordon Conway pointed out, in a talk at Imperial College in 1975, we have to assign a value to our natural resources - or we will undoubtedly loose them for ever.

Young badger found poisoned near badger sett (Photo: InfluentialPoints)

Bennett & Willis (2007) carried out a choice experiment survey in England and Wales to assess what value the 'public' placed on badgers. Whilst people gave a relatively low value to modest reductions in the size of badger populations to control bovine TB, they had a relatively high willingness to pay for a policy that did not involve intentionally killing large numbers of badgers.

Certainly if 'free shooting' of badgers leads to major disruption of the countryside by protesters, it seems likely that the economic consequences for the tourism and recreation components of the rural economy will be severe. This mistake was made during the FMD epidemic in 2002, when huge economic losses were sustained as a result of 'shutting down the countryside' (Blake et al., 2003 ). We now expect - and pay - farmers to play an important role in the maintenance of biodiversity, which implies that we value both the wildlife and the farming.

So where does that leave us? If both badgers and cattle have value, it takes us firmly to the vaccination option for both cattle and badgers. Corner et al. (2002) argued that vaccination is useful "wherever animals of high economic, social or conservation value are involved and test and slaughter or culling programs are not applicable". The badger is about as good a candidate as you can get for having high social and conservation value. A licensed injectable vaccine for badgers is already available for use in 2010, and an oral formulation of the vaccine could be available from 2012. Whilst nearly all the planned trials of badger vaccination in Britain were scrapped in 2010 by the incoming government - one of the most retrogressive and anti-science actions of any new government - a major badger vaccination trial is planned in Ireland (Corner et al., 2009).

As for cattle vaccination, research is continuing on a new vaccine, but a strong case can be made to use the current BCG vaccine in cattle now (see also bovinetb and Torgerson & Torgerson (2009)). Derogation from EU regulations would have to be sought, but the price for this would be small compared to the gains, and other countries would soon follow suit. The Small Farms Association also now supports vaccination of cattle with the current BCG vaccine on the basis that endless prevarication will only further depress the British farming industry. Combining vaccination with cage-trapping of badgers - with a cull only of those with infectious lesions of tuberculosis - may help give speedier disease control in disease hotspots.

We are not saying here that vaccination is the silver bullet to eradicate bovine TB - it is not. But, if we want feasible cost-effective control of bovine TB, we need to vaccinate both badgers and cattle now. The proposed reactive (and no doubt very patchy) badger cull makes no sense scientifically, and may well worsen the situation. Given these facts, could it be the UK government has adopted a dysfunctional disease control policy merely to placate wealthy livestock farmers and avoid spending money? If so, it is almost as unfair to farmers as it is to the badgers...

Robert D. Dransfield (Senior Partner, InfluentialPoints.LLP)