Dining with death

Hugh Pennington, who led the inquiry into the fatal E.coli outbreak, calls for new food safety regulations, and explains how difficult it is to first trace the source of killer bugs and then get the scientific findings accepted by the policy-makers.

Long experience has led many of us to believe in a hidden hand which starts public health catastrophes on a Friday and propels them to develop their full horror over the weekend. At just after noon on Friday November 22 last year three confirmed cases of E.coli 0157 were reported to the Lanarkshire Department of Public Health in central Scotland. By the evening the number of likely cases had risen to 15. The outbreak grew rapidly and went on to become one of the world's largest, with 20 deaths, almost 500 cases and 151 patients admitted to hospital.

Even the most cursory study of this outbreak shows that food safety - or the lack of it - is a subject which is multidisciplinary in the widest sense. This particular outbreak threw up issues relating to interactions between subjects as varied as medicine, science, sociology, politics, philosophy and the law. It provoked the Scottish Office to establish an expert group of specialists in food safety, microbiology and public health to inquire into the catastrophe and seek lessons from it. During the course of its deliberations the group, which I chaired, tackled basic questions about science and politics, such as "is there any truth in the view that the appointment of an advisory committee is the last refuge of administrative incompetence?", "how often are such groups appointed as an easy way for the executive to avoid doing things that it has decided not to do?" and "are the members of advisory groups, including those drawn from the civil service - always sufficiently objective?" Because the outbreak occurred in Scotland its circumstances will be examined, later this year, by a fatal accident inquiry, a legal process conducted by a judge. This will raise equally interesting questions about the interactions between science and the law. Scotland has a well developed set of legislative and administrative structures for the protection of the public health, yet it has a notoriously poor record not only in food poisoning but in cancer and heart disease. Why is this? Is it a matter of the basic principles being right but errors being made in their implementation, or are its fundamental approaches to the prevention of disease seriously flawed?

A less public byproduct of the outbreak was the establishment of a Cabinet committee to handle the issue. Why did the Government do this? The Cabinet Office document Question of Procedure for Ministers states that "cabinet and ministerial committee business consists of (i) questions which significantly engage the collective responsibility of the Government because they raise major issues of policy or because they are of critical importance to the public, (ii) questions on which there is an unresolved argument between departments''. Researchers in politics will probably have to wait the customary 30 years to find out which of these caused the Cabinet to act. Nevertheless, those of us close to the affair might deem it hubristic for food specialists to imagine that an outbreak of food poisoning was significant enough to fall into the first category. Were there unresolved arguments between, on the one hand, the Scottish Office and, on the other, the departments of agriculture and health?

All these queries are important. Many of them address really big debates; not just in public policy, but in biology, such as evolution; in political science (the constitution), and in the philosophy of science, such as when scientific knowledge become reliable.

Because I am a microbiologist I firmly believe that science has shown that microbes cause disease. I have no time for the wilder excesses of any constructivist sociologist of science who thinks otherwise - even if I accept that social factors may be pre-eminent in determining the incidence, spread, and victims of infectious diseases. The central Scotland outbreak was caused by a particular strain of E.coli 0157, characterised by the production of a toxin and a particular kind of chromosomal DNA structure identical to that possessed by a strain that in 1994 caused the world's largest milk borne outbreak. This also happened in central Scotland - in West Lothian - hence its laboratory designation, the West Lothian subclone.

Because E.coli reproduces by binary fission, without the need for sex, all daughter cells are identical twins, and its populations, by definition, are clonal. But because individual bacteria can occasionally exchange pieces of DNA by a sort of coitus interruptus, and because they suffer from random mutations, the original E.coli clone has long since split into many different sub-clones, one of which is the disease-producing 0157 - which itself is now beginning to divide into further subclones.

Most E.coli strains can coexist happily with us in our bowels. They have the greatest difficulty in killing people. Not so E.coli 0157. It regularly causes fatal infections in the very young and the very old, because its toxin attacks small blood vessels. Even more frequently it causes serious disease in individuals in these age groups by damaging their kidneys, often permanently.

This complication is called the haemolytic uraemia syndrome (HUS).

There is nothing remarkable about this. Many other bacteria cause disease through the activity of toxins, although few are as nasty. What makes E.coli 0157 particularly interesting scientifically is its novelty - it is a brand new organism - and its almost complete failure to cause disease in its natural home, the intestines of cattle and sheep. These observations raise unanswered questions about microbial evolution.

It is generally accepted that the organism appeared for the first time in the 1970s - with the first cases of disease documented in the United States in 1982. Did it evolve there? What selective advantage does its virulent genes confer? Why does it keep them if it does not use them in its natural host? That these genes confer an advantage seems self-evident because, since its first appearance, the organism has spread to all continents. This has occurred very unevenly, raising another set of questions. Why are E.coli 0157 human infections particularly common in Scotland, but not in Ireland, and in Canada but not in Scandinavia? Why is HUS a major health problem in Argentina but not in South Africa? Are answers to these questions to be found in national differences in carriage rates in cattle, or what happens in slaughterhouses, or are the differences due to chance? One cannot say. So the mantra "more research must be done" has been uttered. It is happening. Nevertheless, public and political pressure requires that action be taken now.

How did the expert group cope with all this? Although our remit prevented us from apportioning blame at the personal level, it did not require a profound analysis to conclude that the occurrence of a massive outbreak, the frequent recording of other smaller ones, and the continued existence of bacterial food poisoning indicates that the food safety system is failing, either because of faults in its design or because of faults in its operation. The only comfort was that, in spite of many unanswered questions about the biology of E.coli 0157, existing knowledge about its virulence (as measured by deaths) and transmissibility (eating fewer than ten bacteria can have lethal consequences) was sufficient to show that it constitutes about the severest test of the food safety system one could imagine. Even so, it is clear that a century and a half of sanitary reform, 100 years of bacteriology, and half a century of antibiotics have not closed the highways that bacteria use to transport themselves from the intestines and faeces of animals to the mouths of people.

The expert group conducted its inquiry pragmatically. We did not have the luxury of time to conduct an in-depth review of general principles. Nevertheless, such principles were there - tacitly - in our thinking. They often derived from areas remote from food safety, such as the principles guiding policy-making based on inadequate scientific information and those governing the reduction of risk by regulation. These are controversial subjects. Contemporary British examples include arguments between doctors and politicians about the nature of the Gulf war syndrome, and the debate about the impact of a deregulatory philosophy on safety in the off-shore oil industry.

It is not difficult to show that most scientific theories are underdetermined by evidence. It follows that policy-making in just about any medical or related area is going to have a scientific basis characterised by incompleteness and error. Yet most professionals aim to base their practice on sound principles. Experience from other fields gives them clues about how to avoid major errors when basing decision-making on incomplete scientific evidence. One way to compensate for uncertainty is to incorporate substantial degrees of redundancy and overdesign into one's plans. These have long been a feature of the design of buildings and bridges, no doubt because of the fatal consequences of failure.

Overdesign and redundancy are an engineer's response to uncertainty, technical fixes that are not always appropriate in policy-making. Philosophers and lawyers can point to the precautionary principle as another way of protecting human life in the face of hazards marked by scientific uncertainty. It is neatly summarised in principle 15 of the Rio Declaration made at the 1992 United Nations Conference on the Environment.

"In order to protect the environment, the precautionary approach shall be widely applied by States according to their capabilities. Where there are threats of serious or irreversible damage, lack of full scientific certainty shall not be used as a reason for postponing cost-effective measures to prevent environmental damage.'' The ringing tones of the declaration and its underpinning by international agreement should not, however, be taken as evidence that it is operational. It is not. One reason is that it presents a major challenge to legal principles, such as the presumption that no harm has occurred until a party can demonstrate harm and causation. This is an issue of immense practical importance in food safety. Every environmental health officer has had to face the situation of finding a hazard but being unable to take formal action because the evidence falls short of the high level of certainty required by the courts. Officers in Scotland have been particularly influenced by the outcome of an application in 1995 for a condemnation order on 44 batches of a blue cheese. This expensive process (evidence from 26 witnesses was heard over 19 days and 253 productions were lodged) culminated in the refusal of the order by the Sheriff and the incurring of heavy costs by the local council. Argument centred on the scientific evidence relating to the ability of certain strains of listeria to cause disease in humans; many expert witnesses were heard, and a high degree of scientific certainty was demanded. The precautionary principle was clearly not in operation.

Another powerful force that prevents the routine application of overdesign, redundancy and the precautionary principle as responses to scientific uncertainty is their cost. A major issue we faced was the trade-off between safety costs and commercial considerations. Industry often finds the scientific community to be helpful to it here because scientists are more interested in avoiding false positive than false negative errors in situations of uncertainty, placing a greater burden of proof on the person who postulates some, rather than no, effect. At least one body we consulted - the Government advisory committee on the microbiological safety of food - took this attitude. Its scientific and industrialist members made representations to us that because we could not produce a minute by minute detailed account of the spread of E.coli 0157 from the gut of a cow to the residents of Wishaw, the epicentre of the 1996 outbreak, our recommendations somehow lacked scientific weight. In fact, this outbreak will almost certainly turn out to have a degree of epidemiological and microbiological data that will make it one of the best documented ever for E.coli 0157.

Nevertheless, it has to be admitted that policy-making in the microbiological safety of food often has to be conducted without the strength of scientific evidence one would like, partly because of the difficulties of obtaining evidence. More often than not when investigators arrive the food that contained the offending microbe cannot be studied because it has been eaten.

Fundamental problems also attend the methods that are used to investigate food poisoning. Heavy reliance is put on epidemiology, which tries to establish associations between factors - such as the consumption of a particular food item - and the occurrence of disease. But a food source was not identified in three quarters of the 1,744 cases of E.coli 0157 reported in Scotland between 1990 and 1996. This was because most of these cases occurred in isolation, not as part of an outbreak, which makes them difficult to investigate by epidemiological methods.

It is plain that epidemiological data will nearly always fall significantly short of providing the sort of evidence that scientists look for when they set out to "confirm" a hypothesis - even if what they mean is not "confirmation" in the strict logical sense, but the provision of strong support for its probability. The Victorian polymath William Whewell showed how a scientist's "confirmation" could be massively strengthened. His "consilience of inductions" is now a forgotten term but the technique is still used all the time. It involves the demonstration of supporting evidence for a hypothesis by two completely different and unrelated methods. In the case of a food poisoning outbreak this is commonly done by showing that cases of infection related epidemiologically, say by having eaten the same suspect food, have also been infected with an identical microbe. The development of techniques to classify bacteria using DNA fingerprinting methods is a very active field for this reason. Lawyers have been among the keenest exploiters of this information but some bacteriologists have been reluctant to use the new methods.

Food safety legislation, even if defective, requires enforcement. The current regulatory system is imperfect, not only because its scientific basis is incomplete and error-ridden, but because it fails more often than it should. Enforcers have to cope with a legal milieu that sometimes places too heavy a burden of proof on them. They also have to face hostility from those they regulate. The anecdotes related to the expert group about abattoir inspectors whose car tyres were deflated after they had ordered the line to be slowed, were probably apocryphal but illustrate the point.

Attempts to control infectious diseases go back to the late Middle Ages. A typical example is the edict of the Aberdeen town Council of 21 April 1497 concerned with the control of sexually transmitted infection - the French disease: "It was statut and ordanit be the Alderman and Counsale for the eschevin (overcoming) of the infirmity cum out of Franche and strang partis, that all licht woman be chargit and ordanit to decist fra thar vices and syne of venerie, and all thair buthis and houssis skalit (demolished)I" Such regulations and their successors have played a key role in directing the evolution of municipal government in the succeeding 500 years.

Nevertheless, and despite its cleanliness, tidiness, and general municipal order, Aberdeen still has a thriving red light district.

Hegel's dictum is often true: "What experience and history teach is this - that people and governments never have learned anything from history, or acted on principles deduced from it.'' Food safety is no exception - the recommendations of the expert group that butchers should be licensed was first made to the Government by the Richmond committee in 1989, many outbreaks ago.

A general principle governing the success and failure of regulatory regimes is that learning from past errors is a powerful way to make progress. Its converse - that not attending to what is known is a source of failure - is well documented. A good example was the Chernobyl disaster. The design fault in the RBMK reactor control rods which caused a transient but massive increase in power when they were suddenly inserted into the reactor to shut it off in an emergency - a positive scram - and which caused the reactor to blow up, was identified in Lithuania three years before, but no compensatory measures were taken and no information disseminated about it.

In contrast, a classical example of a regulatory system which became remarkably effective by learning from disasters was the 19th-century British Railways Inspectorate, established in 1840 by the Regulation of Railways Act. The inspectorate was a branch of the Board of Trade staffed by officers seconded from the Royal Corps of Engineers. It investigated accidents in public and made recommendations - in contrast to food poisoning where even now this only happens sporadically and where learning about rather than learning from incidents still predominates. By the end of the century it could be counted a success because the probability of death in an accident on a train had fallen to 1/100,000,000 per journey. The inspectorate emphasised self-regulation by railway companies. This was not entirely deliberate. Mid-Victorian judges, sympathetic to those who saw the activities of regulators as dangerous to individual liberty, often granted orders of certiorari to quash regulatory decisions, and when parliament responded by including clauses to nullify these in legislation they made mandatory orders instead. Nevertheless it is clear that a mixture of self-regulation and legally supported enforcement of prescriptive measures worked well in an atmosphere of unbridled capitalism and for an industry with enormous political power. The setting of the balance between these elements of regulation is a continuing issue today, and our report has stimulated a debate about it in Scotland, although one not as vigorous, it has to be said, as the effort being put into the construction of claims for increased funding to carry out its recommendations.

Rumour has it that the Government is vigorously pressing ahead with plans to establish a Food Standards Agency. Let us hope so. It has much to do.

Hugh Pennington is professor of microbiology, Aberdeen University.