The Trouble with the Mathematical modelling in March 2001
From: Use and abuse of mathematical models: an illustration from the 2001 foot and mouth disease epidemic in the United Kingdom
R.P. Kitching , M.V. Thrusfield & N.M. Taylor
Rev. sci. tech. Off. int. Epiz., 2006, 25
"...The epidemic and its control resulted in the death of approximately ten million animals, public disgust with the magnitude of the slaughter, and political resolve to adopt alternative options, notably including vaccination, to control any future epidemics.
The UK experience provides a salutary warning of how models can be abused in the interests of scientific opportunism...models were used as strong support for the implementation of the contiguous cull.
The authors of this paper argue that the models were not fit for the purpose of predicting the course of the epidemic and the effects of control measures.
The models also remain unvalidated... .... scientific experts must be accountable, not only to government ministers but also to other experts. To date, this has not occurred in the context of the 2001 epidemic. ... ."
The mathematical model that drove the contiguous cull was written - not by available experts who were familiar with the Pan Asia 'O' strain - but by epidemiologists and bio-mathematicians who were not
The team whose model was used by the government, from Imperial College, had not even been asked to provide one. The close association between Royal Society Fellows: Imperial College's Roy Anderson, the Chief Scientific Adviser David King and others who had the ear of the Prime Minister has been thought to have a lot to do with this. So Roy Anderson joined the Science Group. He and his team elbowed their way in, frankly.·
The Pirbright members of the Science Group - Dr Paul Kitching and Dr Alex Donaldson - really knew a great deal about the virus and had studied its spread but were not listened to as they should have been within the Science Group.
Yet Kitching and Donaldson were internationally recognised experts. Several weeks later, a worried Dr Kitching said in an interview on Channel 4 News 21 April 2001:
"... The modellers produced some very seductive graphs... The problem has been that virtually none of the models have been able to predict what has actually happened, and I feel this is because there hasn't been the data input available, and there hasn't been the expert advice sought to feed into these models. There hasn't been, for instance, (a) distinction between the different species affected with foot and mouth disease and how they would influence the model; the fact that the virus has been hidden in sheep for some time and that many farms have been infected for 6 or 7 weeks before they've even been identified. These type of things haven't been addressed by the modellers and clearly this has had an influence on their output. The alarming thing is how it seems to have influenced policy to such an extent...(Read full interview)Dr Kitching left the UK soon afterwards to take up his position as director of the National Centre for Foreign Animal Disease in Canada.
The Imperial College model's premise was based on a faulty calculation.
The basic reproduction number, R0 determines whether or not an infectious disease will spread through a population. (Disease is in decline once it gets to R0 =1 or less.) Imperial said that R0 was well above 1 and only the killing of all the animals 3 to 4 kilometres around each infected farm would stop the disease from spreading to the whole country. Perhaps because sheep don't get very ill with FMD and the disease had evidently been rippling through flocks for some weeks without clinical signs, when sheep were tested, 'old' infections came to light and were added to the reported R0 figure as if they were new local spread. In fact the disease had already peaked by the time the untried and unvalidated - and unnecessary - contiguous cull was introduced.
Much of the data that the Imperial College model used for its predictions and policy were flawed.
Data had been hurriedly gathered and converted from many different sources to try to create a database of all UK farms. Examples are: the June 2000 agricultural census (not consistently collected particularly since in Feb 2001 numbers of sheep were constantly changing to maximize premiums), local MAFF office databases known as "Vetnet" but overall geographical mapping data loaded from current sources did not necessarily match the Vetnet data collected from a number of sources over a period of years, and a database created for Swine fever outbreaks. Data was often seriously out of date with regard to crucial factors such as map coordinates, addresses, postcodes, local authorities and counties. Roy Anderson actually commented to the Parliamentary agriculture committee that several of the farms were, according to MAFF's figures, "situated in the North Sea". Holding data had not been removed for farms that had stopped farming many years before. Holding numbers were not unique. There were no computer systems available in many local MAFF offices at the start, and inexperienced personnel were drafted in to copy all this data in from the Vetnet system. Inadequate procedures were followed for checking the data for accuracy. It was a mess.
The Imperial model made assumptions that were inappropriate.
It seems that the model was based on that created by Neil Ferguson and Christl Donnelly from the calculations they had used previously to calculate and model the spread and transmission of human sexually transmitted diseases - together with knowledge gleaned from their work on BSE.
No differentiation was made in the Anderson model between different susceptible species - even though information about the different infectivity of sheep, cattle and pigs to this particular strain of FMD was readily available.
They assumed that FMD infectivity on a farm is constant from the 3rd day after infection until day 11. It is not. Infectivity changes over time and becomes more significant as delays in culling infected animals (as well as all the contiguous stock) build up.
There was intense pressure from political sources to show the disease in decline by the time the General Election arrived.
There was an Election coming and no one in the government wanted the FMD crisis to last too long. The Imperial model seemed to show that the outbreak could - by aggressive killing - be brought to a peak in time to show it was in decline by the time of the Election.
Because the data was wrong, the science was wrong and the assumptions about spread were wrong, the contiguous cull itself was wrong, and indeed unnecessary..
..since by early March, but unknown to the modellers, the disease had already peaked and was already in decline . This letter to the Veterinary Record in 2006 looks at:
"...the extensive post hoc evidence that compulsory contiguous culling was as unnecessary as it was ineffective (Honhold and others 2004, Taylor and others 2004, Thrusfield and others 2005), ... the significant flaws inherent in the models' construction (Donaldson and others 2001, Taylor 2003, Kitching and others 2006)...
The facts speak for themselves. The 2001 FMD epidemic in the UK is remembered for the massive scale of slaughter, involving 10,400 farms and at least 6.5 million livestock, although fewer than 1500 of 2030 'infected premises' (IPs) were confirmed as being infected on laboratory results. Leaving aside the 3 km cull, a separate though significant issue in itself, almost 70 per cent of farms classified as dangerous contacts (DCs) were culled on the prediction from mathematical models that they were at risk of becoming infected. While it is perfectly reasonable to assume that they were at risk, there was every reason to believe that effective surveillance and a properly applied 'stamping out' policy would substantially reduce and control that risk (Donaldson and others 2001)...." (Read in full pdf for references)
"Remember that they all start as merely the mathematical expression of the model builder's presumptions and assumptions."
Martin Hugh-Jones "Why should I believe you when you have a computer pallor and no mud on your shoes? The truth is in the field, not in the computer. When models are checked and rechecked against reality they can be fine-tuned and may eventually become useful…Remember that they all start as merely the mathematical expression of the model builder's presumptions and assumptions. .... I have built my share of mathematical models. Some fall in love with their models but you should never marry them. And the truth is in the field." - Mod.MHJ] (ProMed moderator) http://www.warmwell.com/july17promed.html
Ten years on, is the UK as unprepared now for FMD as it ever was?
It is time people stopped thinking of FMD as a "deadly" plague. It is not. Literally millions of uninfected animals were killed in 2001 by a control policy based on flawed science - and because FMD is an economic disease for Member States of the European Union. FMD vaccination was banned in the EU 20 years ago because there is a special trading advantage to countries that can show they have no endemic FMD. Lack of vaccination is taken to mean the country has no endemic FMD. Therefore, to be "FMD free without vaccination" confers a huge trading advantage. If there is an outbreak, countries that have this advantage are inclined to kill every animal that might spread disease so that they can get back as quickly as possible to claiming that the disease is not endemic. The mindset continues that vaccination in a country means that FMD could be endemic - and must put that country at a disadvantage for trade. Vaccination was thought to be able to mask disease by creating "carriers" that looked healthy but could still pass on the virus. This has never been shown to happen in the field. This wrongheadedness means that vaccination has been ignored by such countries as the UK when its use could protect national herds and flocks at the first threat of disease It is very important that the Expert Group driving policy should contain experts who are adequately qualified in key areas of modern disease control - it is vital, for example, to have a vaccination expert with field experience of implementing vaccination disease control policy so that those who those who have doubts can be properly informed and reassured. Early on-site detection with portable kits such as Smiths Detection BioSeeq http://www.smithsdetection.com/1025_3326.php can limit the spread of disease by detecting foot and mouth before clinical signs appear. This allows for an early stand-still of movements and the possibility of ring vaccination rather than whole country vaccination, and quickly identifies the strain in order to produce the appropriate vaccine. On-site diagnosis gives results about infection within around 90 minutes and makes it unnecessary to kill uninfected animals even if they are within the infected area. Smiths Detection now make an on-site diagnostic kit very similar to that available in 2001 - the one refused by David King (the Chief Scientific Advisor who had no veterinary or FMD experience) because he failed to understand its use. But there is no mention of it in the Contingency Plan - even though such technology can detect FMD infection days before clinical symptoms appear.UPDATE Sunday 6th February 2011 Emergency or "ring" vaccination, (for which supplies of vaccine for all strains of the virus can be obtained from Merial at Pirbright or other vaccine producers), can be used in a protective ring around any discovered infection - from the outside inwards - to stop the spread of disease. Had it been used early enough in the 2007 outbreak we should not have had the second wave. Given that the decision not to vaccinate in 2007 led to a second wave of spread in Surrey, it is hard to credit that the recent exercise "Silver Birch" in late 2010 decided not to use emergency vaccination AT ALL - for the same discredited claim put forward by the meat industry in 2001 that people wouldn't buy meat of vaccinated animals. Of course they would. All imports of beef from Argentina or Brazil, for example, come from vaccinated cattle and were being consumed by British consumers in 2001. UK consumers continue to eat the products from vaccinated animals all the time without realising it. We heard no such arguments when we were all vaccinating against Bluetongue. Meat and products from FMD vaccinated animals do not have to be labelled in the UK, as both the Consumer Council and the FSA have made clear.
In 2001 there were people who said, "They're only meat animals and they'd have died anyway. Farmers are crying crocodile tears." Such a statement shows no understanding of good livestock farmers. The manner of handling animals and of their death matters. There were far too many cases of rushed and cruel practices when officials and slaughterers didn't know how to handle animals in both 2001 and 2007 and had not got the correct equipment or expertise. What's more, so many valuable genetic lines, hefted sheep, rare breeds, breeding stock and pets who were not destined for meat were unnecessarily lost because of the policy of killing healthy animals in a map-drawn circle around infection. (The policy makers had not understood that airborne spread was poor for FMD serotype O, the strain of the virus in the UK in 2001.) The Contingency Plan is almost impossible to make head or tail of because of its dense bureaucratic language - but what is very clear is that there is no indication that vaccination (even though produced by Merial at Pirbright in Surrey) would be rolled out immediately - or used at all. Emergency vaccination is most effective if it is deployed early on. The plan should be a real set of actions that everyone understands - and ought to be clear and sensible enough for any vet and farmer to agree with. This is not the case. There is no longer even a separate plan for FMD. The plan simply includes it together with other diseases in a very long document. The Northumberland Report produced after the 1967-1968 epidemic is scientific, simple and succinct; in many respects it is everything that is required of a Contingency Plan. The Northumberland Report recommended the use the diagnostic tests - available even in 1968 - to detect incubating disease in contact animals. Present policy following the Animal Health Act of 2002 makes it possible that diagnositics can be used only for epidemiological purposes, not to prevent killing. This is a regressive move and ignores the benefits of effective, humane technology now available in the twenty-first century. For DEFRA to make such a fuss about "biosecurity" gives the government an opportunity to blame farmers if the worst happens, but this is a red herring - or red tape. No sane farmer or smallholder is going to want disease anywhere near their farm and they know it is in their own interests to take no risks at all whereas the lack of a good clear and modern contingency plan is the responsibility of the Government.. There is a high risk of the virus being introduced and, once here, it is the responsibility of everyone, including farmers, to take every possible measure to reduce spread. It is very unfortunate that as soon as FMD was confirmed in 2001 there was no immediate targeting of animal dealers and halting of animal movements. Where does the virus come from? Bushmeat, illegal imports and so on - but also research and development within establishments thought to be proof against virus escape. Farmers themselves can do nothing about this at all. There is a dangerous complacency in the UK about unsafe imports. By its reluctance to have vaccination as its first line of defence, to make no mention of portable rapid diagnosis and to make no proper preparations for the use of either, DEFRA makes an outbreak of FMD in the UK a very dangerous risk. The Government has never carried out a proper risk assessment analysis, looking carefully and with independent, informed input at a variety of possible policies. That is real risk taking. Even if vaccination were agreed by a more enlightened Ministry, it would take vital days or even weeks to get it going. That correctly administered FMD vaccination works is not in doubt by any informed scientist anywhere. In 2001, Uruguay had a very similar outbreak to the UK's but their farmers vaccinated their own cattle. The virus was soon stopped in its tracks - whereas in the UK it dragged on for almost the whole year with disastrous consequences for farmers, smallholders and owners of pet animals of susceptible species. At least 10 million animals died (evidence on warmwell.com) in 2001 and the trauma suffered and social consequences have never been forgotten (See Lancaster University Research study). The waste in terms of money and resources was quite out of proportion to any supposed political or economic gain. In the present global financial situation, such a waste should be thought unthinkable. Research showed that the epidemic in 2001 had peaked before the 3km contiguous cull policy began. Our mass cull policy was as unnecessary as it was cruel. It was also illegal in 2001 to kill animals that had not been exposed to the virus. The Animal Health Act of 2002 sought to legitimise retrospectively the mass killing that had been done and make legal what could be done in future. The thought that it could all happen again is sickening. There are now DIVA vaccines and tests proven to differentiate vaccinated from infected animals - and vaccinated animals are not a risk as carriers even when, as is rare, some vaccinates might be sub-clinically infected. We are STILL waiting for validation for these DIVA vaccines so that post outbreak testing can be carried out swiftly. The political assumption is that the "FMD free without vaccination" trading status is too valuable to lose because if one uses vaccination it takes longer for that trading status to be returned. There is no scientific reason for this trading disparity. If it is assumed that 6 months is the time for a country that has used vaccination to resume trade, it is just as valid that it should be 6 months for a country that has not (where the disease could be lurking anywhere and testing should be just as stringent, if not more stringent, before trading is resumed.) When mistaken policy is decided on by people taking charge over things that they do not understand, that policy can disastrously affect the lives of others. The political centre always wants to keep control - but this results in the sidelining of the real FMD experts and also of people on the ground with local knowledge and veterinary and farming expertise. It was a tragedy that informed vets and scientists were not listened to - and are still trying to make their voices heard. The tragedy of FMD is being played out still. In 2010 the Japanese farmers of Miyazaki were given the same reasons and suffered the same distress when thousands of animals were killed - simply because Japan clung to its economic trading status. In South Korea now the disease has raged out of control and vaccination is being used - but too late for the thousands of farmers who have had their animals snatched away or for those animals already infected. In Bulgaria the authorities are blaming Turkey, rebuilding border fences - and refusing to vaccinate. In all cases this reluctance is because of the trading status of "FMD free without vaccination". The trading status gives an advantage to big business. We and our animals are the ones who pay for this and the price is too high. Rinderpest has been defeated globally by the correct use of science. There is no reason why FMD should not be eradicated globally too
BBC's Countryfile failed to mention that the half million sheep there were killed by a flawed mathematical model - not a "terrible disease".Great Orton:
" It will be remembered that 451,000 sheep (four hundred and fifty one thousand) were slaughtered at Great Orton. 5786 animals from 115 farms were tested between the 7th and 23rd of April 2001. This was not gentle euthanasia but a messy scramble to kill terrified sheep and lambs as quickly as possible. One farm had a positive test (not active virus but antibody positive) on 9 sheep. One farm had 2 sheep with what Andrew Hayward (DVM Cumbria) termed "mild" positive results. Three farms were inconclusive. All the rest, 110 farms, returned negative results. There were no clinical signs on any of the sheep slaughtered at Gt Orton."Countryfile called the Nature Reserve at Great Orton
"a monument to the victims of this terrible disease."No BBC. The sheep, many pregnant or with lambs at foot, were not victims of disease. They were victims of a despicable, ignorant policy, based on bad science, a hopelessly flawed mathematical model - and politicians anxious about the Election.
(Please contact the website for more information on any of the statements made above.)