The computer modelling team from Imperial College, headed by Professor Roy Anderson, has led government policy throughout the foot and mouth disease (FMD) epidemic. It was their analysis that proposed the greatly increased slaughter of the contiguous cull, and warned that anything less would allow the epidemic to spiral right out of control, with the eventual loss of half the total UK livestock population. On 11th May, a paper detailing this work was published in Science magazine.

Five months later, on 4th October, a second paper was published in Nature magazine, updating their work and claiming to justify their earlier predictions. Media reports have widely publicised these claims -- for example, Channel 4 news reported "If what was then the Ministry of Agriculture had acted immediately on their advice to slaughter infected farms in 24 hours and neighbouring or contiguous farms in 48 hours, a million of the 4 million slaughtered animals would have been saved, and 400 of the 2,000 outbreaks would have been averted." But no independent scientific review of their work has taken place, and no one in the media is questioning whether their claims have any substance.

As a small-scale specialist sheep keeper, I have studied every aspect of the foot and mouth epidemic, and successfully resisted the contiguous culling of my own flock in April. I do not believe that there is any scientific basis for the extended culling policies, and find the Imperial College papers entirely unconvincing and scientifically flawed. Moreover, an independent analysis can reasonably draw the opposite conclusions to those that have been given such public prominence.

The Imperial College team are biomathematicians who apply standard mathematical techniques to complex biological phenomena, in this case epidemiology - the spread of disease. Modern computing power enables the rapid calculation and display of many possible outcomes, described as "modelling". This is a useful technique but as with all computing, it relies entirely upon the quality of the input data - put rubbish in, get rubbish out. So to appreciate the true value of the Imperial team's work, a close look at their input data is required.

Back in May, the first paper declared "Contact tracing for all FMD-affected farms has produced unique data on the spatial scale of disease transmission (provided by MAFF), clearly demonstrating that farms closest to index cases of FMD are at greatest risk of infection." This, of course, promoted the concept of contiguous culling to remove those farms at highest risk before they, in turn, became infectious to others. But in October, the second paper contradicts this by stating "The newly estimated spatial kernel differed significantly from that previously derived from the infectious contacts identified by MAFF, with considerably more long-distance transmission events being predicted. This implies significant biases in the MAFF contact-tracing process, with closer contacts being more easily identified." Worse, it goes on, "The median distance of the newly estimated kernel is about 4 km, suggesting that most transmission probably occurred through the movement of animals, personnel or vehicles, rather than through animal contact or windborne spread."

It is hardly necessary to look any further, since this admission, that the original data were completely wrong, has entirely destroyed the rationale for contiguous culling. Contiguous premises are now found to be no more at risk than any others within the locality of an infected premises; indeed the average transmission distance of 4 km argues powerfully against slaughter of contiguous premises since this cannot influence disease spread over such long distances, except by the very negative effect of swallowing up valuable resources of time and manpower for no benefit.

By looking further into these reports, more such contradictions are easily found. The May paper states "We did not differentiate between host species but instead used a time-varying infection-to-report distribution averaged over species." However, by October the Imperial team are saying "Farm susceptibility and infectiousness both varied significantly, with smaller farms being substantially less infectious and less susceptible than larger ones. Stratifying farms by the most prevalent species revealed a trend for cattle farms to be more susceptible, and pig farms to be least susceptible."

This, of course, is precisely what veterinary scientists had been saying all along, but it was completely ignored in the earlier modelling work. In fact, Professor Alex Donaldson of Pirbright laboratory has shown that there is wide variation between the livestock species in their capacity to catch and to transmit the disease; while the current strain of FMD virus is extremely unlikely to spread by windborne means from a source farm to its contiguous neighbours in most practical situations, whatever the livestock species involved. So here again, the conclusion is clear - that the original input data to the computer modelling programme were completely wrong.

A key parameter, central to the Imperial team's calculations, is that of farm infectivity, which they denote as (r). This is a ratio "representing the infectiousness of a farm after the disease has been reported, relative to that just before reporting". As with any viral disease, livestock exposed to FMD go through an incubation period lasting a few days before they themselves become infectious to other animals. Symptoms do not develop for a further day or two, so that by the time clinical signs are seen in the first animal on a farm, it is likely that several others are also incubating the disease. By the time these several individuals develop symptoms, more animals are entering the incubation and infectious phases; and so on, in a progressively accelerating pattern. During the heightened awareness of a large-scale epidemic, early recognition of FMD can be expected because farmers are looking closely for the clinical signs on a daily basis. So the farm infectivity ratio, r, will be much greater than one as the disease spreads at an increasing rate through the herd or flock following recognition of the first signs.

Yet this very obvious conclusion eludes the Imperial college team. Back in May, they state, "Because data do not exist with which to estimate the infectiousness of a farm as a function of time since infection, prudence dictates that in addition to more rapid culling of infected farms, it is necessary to consider other interventions."

They continue, "If r is greater than one, ring culling still accelerates the decline of the epidemic but at the cost of a larger cull than rapid index case slaughter alone". Clearly, a contiguous cull cannot be justified if the net result is a higher slaughter total overall. So, they simply assumed that r wasn't greater than one. In the notes of small print at the end of the paper, they state, "We assumed constant infectiousness from 3 days after infection until slaughter, for an average of eight infectious days."

Using this most extraordinary assumption, they calculate figures that appear to support a contiguous culling policy. But for values of r greater than one, which in reality it clearly is, they admit "Our analysis shows that ... slaughtering on all farms within 24 hours of case reporting can significantly slow the epidemic ... and results in rapid control if we assume that infectivity increases throughout the time from infection to slaughter."

Yet five months later, the October paper re-states "Infectiousness is assumed to be constant from the day after a farm is infected until the day the animals are culled". There is no attempt made to model for alternative, realistically higher values of the ratio, r - the original erroneous assumption is simply repeated. Again, the inescapable conclusion is that the wrong input data have been used.

Those of us who actually experienced the holocaust of mass slaughter have seen for ourselves the inconsistencies between official control policy and reality. Farms with cattle were seen to be infected far more often than sheep holdings, yet sheep were blamed throughout for the spread of disease (a claim now being disproved by widespread blood testing). Most contiguous farms that were slaughtered tested negative for disease; and of those that refused the cull, none that I know of went on to develop disease.

My own experience was not untypical. Two neighbouring dairy farms reported FMD within a few hours of each other; both recognised clinical signs in one cow. These two herds were slaughtered within 24 hours before any other animals developed signs. Large numbers of sheep and cattle were then slaughtered on contiguous farms - except for my own small flock, which had been protected with commonsense biosecurity arrangements. After a long struggle against slaughter, the flock was eventually cleared as disease-free by blood testing, demonstrating that the animals slaughtered all around us on other contiguous farms were also uninfected.

How had the disease reached our parish? The nearest infected premises lay 6 km away downwind, and no other farms in between were affected, so windborne spread was illogical. The farms concerned had taken biosecurity precautions, but being dairy farms could not avoid regular visits from milk tanker and feed lorries. Time and again, dairy farms became infected at some distance from previous outbreaks, yet the obvious link with these vehicles was consistently denied by the authorities, until the later outbreaks in Yorkshire and Northumberland eventually led to the introduction of "blue box" high-security measures; these included the direct supervision of milk tanker and feed lorry movements and disinfection procedures, with the result that these regional outbreaks quickly subsided.

Even the Imperial College team, so remote from the reality of the situation on the ground, observed in their October report "Our analyses have demonstrated that ... changes in culling policies explain less than 50% of observed past variation in transmission rates, indicating that effective movement restrictions and rigorously maintained biosecurity are equally vital."

But the final condemnation of the computer modellers' claims is simply this - that their proposed policies have never actually been implemented. Their targets of slaughter within 24 hours on infected premises, and within 48 hours on contiguous premises, have never come close to being met on a national basis.

As the October report admits, "In no week of this epidemic have more than half of contiguous premises been culled within the specified target of 48 hours from report of a new case." In other words, the gruesome experiment has been so incomplete that no conclusions may reasonably be drawn from it.

In summary, I submit that the computer modelling analyses of the Imperial College team, under Professor Roy Anderson, are fundamentally flawed through the use of wrong input data and wrong assumptions. The original field data from MAFF were wrong; assumptions about airborne spread were wrong; crucial differences between livestock species were ignored; the significance of long-distance spread via movements of animals, personnel and vehicles was underestimated; and assumptions of the change in farm infectivity over time were wrong.

There can be only one, inevitable, result of using such inaccurate input data -- the output data obtained are wrong, no matter how clever and convoluted the statistical analysis applied.

This is appallingly bad science, the unquestioning acceptance of which has led to the unjustified slaughter of millions of healthy animals, and has actually hindered disease control by diverting resources away from the more important task of rapid slaughter on infected farms. Upon close examination, therefore, Roy Anderson turns out to be a Professor, not of zoology, but of mythology.

alan.beat@talk21.com