Return to

April 2007

Commentary (red) on the article (black) about the Imperial College research.


A bird flu outbreak could spread from one farm to dozens of others within weeks unless a mass cull of infected and healthy poultry was carried out, new research has warned. This is not 'research' it is modelling. It will probably have been validated, if at all, using the same data used to generate the model, but it is impossible to tell without the details.  


An expert said that in areas where poultry farms are particularly close together, like Norfolk in the UK, it would be difficult to impose measures in time to stop one farm infecting at least one other, and that two or three others will be affected within a week. This is nonsense as it is such a generality - did they not look at what happened at the Bernard Matthews (BM) enterprise. Disease comes from Hungary, infects turkeys, turkeys killed, disease stopped in its tracks - despite wild birds and rodents etc potentially scavenging/spreading. H5N1 detected beyond the shed where infection started only because there was no great regard for biosecurity between shed as all turkeys were to be slaughtered. In the BM outbreak the reproduction number was zero (as they point out) - in a fairly densely populated area - which would tend to invalidate at least some of their assumptions. Either the infection is less transmissible than they are modelling, or the 'measures' imposed were indeed adequate. Either way, this one example does not correspond to their model.


If widespread culling of infected birds and even "pre-emptive" culling of uninfected birds in surrounding farms is not carried out, the contagious disease would keep spreading at an increasing rate. Should refer to 'infectious birds' rather than 'infected birds'. The post hoc analyses of the foot and mouth disease (FMD) data have shown that the importance of pre-emptive culling was overestimated. Does no one at Imperial ever think of vaccination and if not, why not? Vaccination is rarely perfect and it is not necessary to prevent infection totally, just to reduce disease and prevent between-farm transmission. Vaccination is the way we prevent human diseases 'spreading at an increasing rate'. Even without vaccination or other 'measures', unless the population is replenished (unlikely to be allowed in an infected area) the rate of transmission will peak and then slow as susceptibles become removed by disease or immunity.


And even with measures like culling imposed to stop spreading, two recent cases in Italy and Netherlands showed up to ten new farms infected within just two weeks of the initial outbreak. This extrapolation is invalid unless the populations are similar - and they have already stated that this is not the case.


In a new study researchers from Imperial College and the London School of Hygiene and Tropical Medicine analysed recent cases of bird flu strains H7N3, H7N7 and H7N1 in Italy, the Netherlands and Canada to come up with a model to estimate a farm-to-farm reproductive number - a measure of how many other farms the virus is transmitted to from one affected farm. It is obvious that current concerns relate to H5N1 and as H5N1 is probably less transmissible, it would not generate the 'helpful' numbers. The fact is that H5N1 did NOT spread like this in the recent outbreak, so the 'researchers' have obviously had to use other viruses in order to justify the proposed cull.


Ideally in an outbreak the number would be below one, meaning no spread had taken place. But scientists found that the number was on average between 1.1 and 2.4, meaning that in most cases an outbreak at one farm led to the infection of a herd at at least one other farm, and in areas where poultry farms were packed close together, one farm infected on average more than two others. The reference to 'herds' is interesting and not usually applied to collections of poultry  'flock' would be more usual. Have the authors taken a chunk of text straight out of an FMD paper? It would be interesting to know the density of poultry farms modelled and compare that to the situation in the UK. Time frames (would there have been time to vaccinate?), control measures used, environmental conditions and potential mechanisms of between-farm transmission all need to be considered, and all will vary greatly between virus strains - and countries... All very interesting, but hardly relevant to H5N1 and the UK - and certainly not something that should be used to drive policy.


To eradicate the disease and stop any further spread the reproductive number needs to be brought below one. The statement is correct, but R is heterogeneous - the average needs to be brought below one, but that will not necessarily stop all spread if a local epidemic has R above 1 - and an average R above 1 may still allow most spread to stop if local R is below one. It is very simplistic modelling, and easy to understand - but bears little relation to the heterogeneity present in the real world and to practical disease control measures.


However, researchers found that standard measures to stop spreading - enhanced bio-security, movement restrictions and culling of infected birds - did not succeed in making the number drop below that threshold. This is a model - how was it validated? How was the accuracy of this statement verified? What were the input data? Why did they not consider vaccination? Was their population homogeneous, heterogeneous, clustered, spatially realistic?


Instead, "pre-emptive" culling of healthy birds on nearby farms was the only method that brought the level below one and completely eradicated the spread. This seems to be exactly the same as the flawed FMD model - which came to the same (incorrect) conclusions.


Dr Tini Garske, who led the study, said: "After the first infection, a whole flock of birds is likely to be killed within a day or two. Killed or died rapidly of disease? What does this mean?


"And in areas which have lots of poultry being farmed close together, unless measures were imposed, including culling of uninfected birds, there could be dozens of new farms infected within two weeks. What 'measures'? Have they looked at transmission routes?  How do they think the virus is transmitted between these closely located farms? This is pure guesswork and unworthy of a serious scientist - the only certainty is that such statements will be a cause of huge distress to livestock owners. It is not based on any data pertaining to H5N1.


"In Italy there were three new cases after one outbreak within a week and in the Netherlands two new cases within a week. Then, even with measures imposed, there were another 9 or 10 infected farms within two weeks. Cannot extrapolate these results from H7N3, H7N7, H7N1 to H5N1.


"In a very dense poultry farming area pre-emptive culling and de-population of nearby at-risk areas succeeded in containing the outbreak, where other less drastic measures had failed. Can they explain to us how it is possible to depopulate an area of wild birds, cats, and other miscellaneous carriers?


"It's a difficult situation that healthy birds may have to be culled, both for animal cruelty campaigners and also for farmers and the economic implications. This is hardly a coherent argument as healthy birds do not necessarily have to be culled  a proper modelling exercise would examine all the options.


"But I think that if people see an epidemic running out of control then they would be prepared for a large scale cull, even of healthy birds, to stop the devastation to increasing numbers of birds. "In the long-run it would save more birds than it kills." This is an odd and wholly unconvincing argument and is strangely reminiscent of the phraseology used in the somewhat discredited FMD models. Is it an attempt at justifying a pre-emptive killing approach yet again  post hoc, ergo propter hoc?


Dr Garske said though there were not many areas with a high density of poultry farms in the UK, there were some and added this system could prove vital in containing an outbreak. Exactly so, there are not many areas with a high density of poultry farms in the UK. Is this a way of obtaining funding through fear?


She said: "There's not quite the same level of danger in Britain because poultry farms are not so densely packed together, for example with the Bernard Matthews outbreak the reproductive number would have been 0, because no other cases were reported. Well spotted - and of course the virus was of a different strain.


"But there are some more densely packed poultry areas like Norfolk and Devon and this system is very easy to carry out and doesn't need much data. And In the event of an outbreak the reproductive number can be carried out very quickly - it can take from an hour to a day - and DEFRA can then carry out the appropriate measures to stop the spread. 'Doesn't need much data' is hardly high quality research and does not generate much useful information might be added. The reproductive number was calculated  incorrectly - almost daily during FMD, so it would seem unwise to advocate the same approach for avian influenza. If the disease is spreading, you will act to stop it. The size of the reproductive number should not be used to determine control measures.


"This is a very important first step in planning procedures to control a spread." First step? Hardly relevant to planning, especially if they are modelling the wrong virus in the wrong country.


And she added that the method could be used in high risk areas where avian flu is most likely to develop into a strain transmitted easily between humans, like South-East Asia and Africa. This really is awfully naive - I think a trip to China (especially its borders) border or Africa (particularly Nigeria) is what is needed.


She said: "In those areas data about outbreaks is scarce, but this method of assessing the urgency needed to deal with outbreaks doesn't need much data. This is quite appalling - who has allowed such ill judged and ignorant statements to be made?


"In South-East Asia and Africa people have birds at home and keep them to sustain their lives and worryingly there is a lot of the H5N1 strain going around in those areas." It would be helpful to know just how much Dr Garske actually knows about avian influenza - and if she is as ignorant as it would appear she should talk to some real experts, carry out some field work in different parts of the world, examine different husbandry systems and carry out a systematic review of avian influenza before she thinks of 'communicating' anything further - this is dangerous and ill-informed nonsense.






































Archive July and August 2005

Archive Jan 2005 - July 2005

Archive Oct 2004 - Dec 2004

Archive August 2004 -October

OTHER WARMWELL ARCHIVES(opens in new window)