Annual conference of the UK Health Protection Agency

Warwick University, 12 September 2005

Annual conference of the UK Health Protection Agency

Professor Jacqueline McGlade
Executive Director, European Environment Agency

Speech at the annual conference of the UK Health Protection Agency on 12 September 2005 at Warwick University

The main EU bodies that are responsible for public health are DG Sanco, the JRC and the European office of the WHO. The EEA has no explicit mandate for health but given the links between environments and health, and the need to provide our information within the context of sustainable development, the Agency has been contributing to the environmental Action Plans of the EU and WHO by

• providing information that helps to "frame" the approach to these complex issues, particularly the underlying cause /effect framework which is a key goal of the EU Action Plan; and by
• helping to increase awareness of the developing environmental health sciences through the "diffusion of policy relevant research results" --a specific mandate of the EEA that is particularly relevant to this issue.

I want to share with you some of our contributions to :

• the emerging approaches to risk assessment and risk management;
• the identification and use of appropriate levels of proof for different hazards, which is essential for the successful application of the precautionary principle;
• the management of risk ,uncertainty and ignorance, each of which require different approaches;
• the implications of multi-causality for both risk assessment and risk management.

I would like to cover these points by using firstly case studies from our report , "Late Lessons from Early Warnings: the precautionary principle 1896-2000", and secondly EEA background papers developed for the EU Environmental Action Plan process.

The Late lessons report looked at the histories of 14 case studies --7 being mainly health related, and 7 being mainly environment related. In the report we tried to see if we could learn anything from these histories that could help us to better manage environmental and heath issues over the next 100 years rather more successfully than we did over the last 100.

The case studies were structured around 4 questions. The first tried to identify "when was the earliest, scientifically credible early warning"?
Perhaps the most illustrative example comes from the report on antibiotics used as growth promoters in farm animals which came from the UK's Medical Research Council in 1969:

This recommendation was initially heeded but then was largely forgotten as the pharmaceutical companies successfully argued that there was an insufficiency of scientific evidence to justify such tight controls- a position they still maintained in 1999 when the EU eventually banned the most hazardous growth promoters, 30 years after this first early warning. Pfizer took the EU Commission to the European Court arguing their point on the sufficiency of scientific evidence, but lost it when the ECJ established that the Commission's use of the precautionary principle, which is a duty under the EU Treaty, was appropriate.

The contribution of trace amounts of antibiotics in food chain to the current serious rise in antibiotic resistance in people is likely to be quite small when compared to the over-use of medicinal antibiotics, but nevertheless it is seen as a credible, if unquantifiable, risk that should be avoided.

Other lessons from this early warning are relevant to today's risk assessments: that action to curb exposures can be justified even where there are large gaps in knowledge about both causality and about the mechanisms of action; and that waiting for further research results may not only take decades of expensive effort but is likely to uncover new sources of uncertainties and ignorance in the process of filling some gaps in knowledge, thus providing new grounds for inaction. As Infant observed in the benzene case study: analysing an issue to death can often result in the death of the people or eco-systems under study.

Another lesson from the antibiotics case study, as well as from the asbestos ,CFCs, PCBs, DES and TBT case studies, is that "surprises", particularly from states of ignorance ( which can be distinguished from uncertainties about known parameters) are to be expected, and therefore to be better managed. Ignorance is not knowing what we don't know. The TBT case is a good illustration of the surprises: that follows its widespread use TBT accumulated in the micro layer at the surface of the sea; it bio-accumulated in several species; parts per trillion were sufficient to trigger sex changes in sea snails via endocrine disruption; and timing of the dose was more critical than the dose itself.

There therefore needs to be greater recognition of the contingent nature of knowledge, (Slide 2) of ignorance as well as uncertainty. Rather more humility and less hubris amongst scientists would contrast with the "misplaced certainties" that characterised many of the case studies in the Late Lessons report. Managing "surprises" is obviously difficult. However, the greater use of substance properties, like persistence, spatial range and bio-accumulation, as well as greater use of scenarios, would have helped us better manage the CFCs TBT, PCBs and other hazards, and could also help us to avert future surprises. Earlier recognition of the impacts of surprises could come from more long term (ie for decades ) monitoring of , for example amphibians and other "surprise sensitive" parameters. Despite, they are substantial funding, institutional and cultural barriers to such monitoring, overcoming them will inevitably be beneficial, as the use of their results in the CFCs, DES and TBT case studies illustrates.

The second structuring question for the "Late Lessons" case studies was: "what were the actions or inactions that followed the early warnings? Given that the first credible warnings on medical X-rays, asbestos, and benzene came in 1896, 1897 and 1898 respectively you will not be surprised to hear that there was much inaction in most of the case studies for many decades inaction that was not justified given the knowledge available at the time.

Authors of the case studies (who were usually people who had played a key role in the unfolding of their case studies such as Joe Farman, who discovered the "hole" in the ozone layer) were asked to judge actions or inactions based on "the spirit of the times" and not with the benefit of hindsight. In some cases this was not difficult. Listen to the plaintiff observation of the chief Medical Inspector of Factories:
"Looking back in the light of present knowledge it is impossible not to feel that opportunities for discovery and prevention of asbestos disease were badly missed." (Thomas Legge.) What is perhaps surprising is that he made that observation in 1934: asbestos was banned in the EU in 1999.

The combination of the delay in action and the long-tail latency effects of asbestos disease means that there will be some 250,000-400,000 deaths in Europe over the next 35 years from the three main asbestos diseases (asbestosis, lung cancer and mesothelioma).

Similar inaction, or long delays in actions, were also features of most of the other case studies, including for example medical x-rays, benzene, sulphur dioxide as a cause of acid rain and PCBs.
Whether the delay in action between the first early warning on CFCs and the ozone hole, in 1974, which was based on fairly solid but theoretical science, was unjustifiable, is more debateable. Given the long latencies involved, and the seriousness of the impacts, earlier and significant actions than those initiated by the Montreal Protocol in 1987 could have saved us some of the impacts that we will suffer from over the next 50 years or so. (Slide 3).
Joe Farman's conclusion was that "a plausible case brought only limited action...precautionary management of the environment was not uppermost in the minds of policymakers".

In two of the case studies, beef hormones and TBT in marine paint, early precautionary actions were taken on the basis of preliminary scientific information that provided a credible suspicion of risk, though in the case of beef hormones, that view was not shared by the WTO who found against the EU for its ban on their use. And with TBT the initial precautionary action in France in 1982 to ban such paints for small boats was strongly conditioned by the large economic losses incurred from the collapse of the valuable oyster beds which was attributed to TBT. A lower level of proof was used to justify the action because of the potentially large costs of inaction.

This issue of using different levels of proof that vary with the potential costs of action and inaction lies at the heart of the precautionary principle and its use in the risk assessment and management of hazards. It is not a new idea. Sir Bradford Hill in his classic 1965 paper on "The Environment and Disease: Association or Causation?" recommended three levels of proof that could be used to "convict" a substance or activity of risk:

• "relatively slight evidence" that would be appropriate, for example , "to restrict the use of a drug for early morning sickness" (which did not happen in the DES case study ,with devastating cancer impacts on the daughters of the pregnant women);
• "fair evidence " to justify replacing a probable carcinogen in the workplace with a safer substitute "without too much justice if we are wrong"; and
• "Very strong evidence" needed to justify making the public stop smoking and eating the fats and sugars that they like".

As Bradford Hill pointed out, it is essentially the potential costs of being wrong that determine the appropriate level of proof, which is why of course we use two levels of proof in our courts : "beyond reasonable doubt" for the criminal courts and "balance of probabilities" for civil courts. Here the potential costs and benefits of being wrong in both directions (eg convicting innocent people or letting murderers go free for the criminal courts) are valued differently and therefore justify using different levels of proof.

The IPCC in their reports of 1995 and 2001 spent much time on this issue of appropriate levels of proof. In their later report they provided seven different levels that could be used to characterise the evidence for the various hypotheses involved in the Climate change debate. They had used one of these , "the balance of evidence" level (which is the same as the civil court "balance of probabilities") in their earlier report to justify the conclusion that mankind was disturbing the climate, a conclusion that came 98 years after the first early warning on climate change in 1897.
Transparency in the choice of appropriate levels of proof in the risk assessment and management of current hazards is therefore essential.

This also points to another lesson: reliable science is a necessary but not sufficient factor in decisions on public hazards: the values and economic interests of stakeholders are key ingredients of successful risk assessment and risk management. Slide 4 illustrates that such values and interests are needed, in differential amounts, at every stage of the new paradigm for risk assessment and management. The Royal Commission report on "Setting Environmental Standards" in 1998 did much to promote this shift from the old to the new paradigm in risk assessment and management.
The old linear approach which involved scientists first doing their risk assessments, then passing the results onto the risk managers, who then tell the public what will happen, is being replaced by the circular paradigm depicted in Slide 4, as the lessons of the last 100 years of hazard management are turned into new practices.

If the costs of being wrong in both directions is central to choosing an appropriate level of proof, then the reliable estimation of the costs and benefits of action and inaction becomes even more important. This was the third question that case study authors had to address in the "Late lessons" report. However, this was the weakest section of the report as the authors were not economists. We shall address this more effectively in Volume 2 of Late Lessons, due in early 2007, where there will be a specific chapter on the economic analysis of the costs and benefits of action and inaction.

However, we already know from Volume I that the costs of inaction ( or of ineffective action) on fisheries, asbestos, PCBs, sulphur dioxide, BSE, CFCs etc have been very large, if in some cases unquantifiable, such as the value of an intact ozone layer, or the loss of public trust in scientists and government. Conventional cost/benefit analyses have often been too narrowly drawn to capture some of these items adequately.

Pro and con analyses, which are the terms we used in the Late lessons report to emphasise the broader approach, involve the unquantifiable items as well as the quantifiable, and particularly their distribution across groups, regions and generations. This is an equity point that arises from the differential distribution of benefits and exposures across different groups, a point that this conference is also addressing in more detail and which the EEA will include in its work on pro and con analysis next year.

The last question to authors of the case study chapters asked for lessons learned. I have already covered many of these "Twelve Late lessons" but they all essentially call for the broadening of the framing, assumptions, values and knowledge base used in risk assessments and management.
Volume 2 will use another 16 case studies, including lead in petrol and climate change, to address particularly the generation, financing communication, use, and sometimes misuse, of the relevant science. It will also have a chapter on false positives to see what lessons can be drawn from over --reacting to weak and sometimes mistaken signals.

The second area of EEA work, for the EU Action Plan on Environment and Health, largely concentrated on the cause-effect framework. This has drawn attention to the emerging focus on multi-causality and its implications for both the evaluation of scientific evidence on hazards and for their prevention.

Much of the multi-causality debate has taken place within the research communities that are addressing the gene/environment interface. The conclusion of Johnjoe McFadden, Professor of Molecular Genetics at Surrey University, supports the reality of multi-causality caused by "a network perturbation generated by small, almost imperceptible changes in lots of genes" rather than a few genes playing a dominant role in such complex diseases as heart disease, autism, cancer, and neuro-developmental disturbances.

We can illustrate this with asthma and with the inhibition of IQ development were there are likely to be several different causal chains involving different mixtures of several co-causal and dependent factors that operate differently in different groups of people. Some of the implications of the multi-causality approach are:

• Confounders can be co-causal factors
• Consistency of scientific results into the same disease is not to be expected, as Professor Needleman pointed out in the Lead in petrol debate
• "Small" environmental co-causal factors can have substantial beneficial impacts on diseases if removed from inter-dependent causal chains, or if, like traffic fumes, they can trigger negative impacts in diseases like asthma that are caused by many other factors.

The precision with which genes can be identified compared to the imprecision of exposure assessment means that research results from gene/environment interface studies are likely to be biased against finding that environmental exposures play a significant role in disease causation, as Professor Vineis has recently pointed out.

Multi-causality and complexity may seem to be obvious but in reality they are often neglected, as the SETAC (Society of Environmental Toxicology & Chemistry) report on the "Interconnections between Human Health and Ecological Integrity" pointed out:
"Historically, the cumulative and interactive effects of direct and indirect stressors and other multiple exposures have been neglected, in part because of the sheer complexity of these realities. (SETAC, 2002).

Research however must address these complexities if it is to be realistic. Neither genes nor their environments but their interactions are usually causal. It follows that believing that the causes of common disease like cancer, diabetes, heart disease, etc are the result of independent, rather than dependent actions of multiple agents is holding back progress in both understanding and prevention.

It also follows that the "burden of disease" approach, which currently finds the environment responsible for only some 2-5% of European mortality and morbidity is likely to be flawed. If multi-causality is reality then interdependent causal chains means that all co-causal factors are 100% necessary (but not sufficient) causes, and as noted removing them can have substantially beneficial effects on common diseases. The EEA plans to do further work on multi-causality and the burden of disease, in partnership with the JRC over the next two years.

Multi-causality also means that some of the well known criteria used to help us move from association to causation, which Bradford Hill bequeathed to us, are not robust in denying causality if the criteria are absent. For example "temporality" ie A must come before B if it is to be causal, can be misused to deny the role of say chemicals in falling sperm counts if the fall began before chlorine based chemistry took off, as has been argued by the WHO in its report on the Science of Endocrine Disruptors (2002). However, the other causes of falling sperm counts could have initiated the fall before chemicals join in with their impact and the resulting sperm count could be rising ,falling, or static depending the interactions between the all of the causal factors. Multi-causality therefore means that criteria like "consistency" and "temporality " are only robust in confirming causality but not in denying it, an asymmetry that Bradford Hill was aware off but which his followers have sometimes ignored.

In this ,the 40th anniversary of his classic paper, the EEA in partnership with Collegium Ramazzini and WHO will hold a workshop in London in December, to re-examine his criteria in light of multi-causality and today's knowledge.

Asymmetry is also a feature of the directions of bias within the environmental health sciences. For example greater frequency of false negatives compared to false positives in the main direction of bias makes for sound science but often for unsound public policy on prevention, and both the public and policymakers need to be more aware of these biases.

In conclusion, future approaches to the identification and prevention of environmental stressors by the HPA could benefit from taking account of both the lessons from the histories of "known" hazards and of the issues emerging from the research communities concerning the gene/environment interface and multi-causality. This should help you to better manage the emerging hazards of, say, pharmaceuticals in the environment, hazards from electronic equipment, endocrine disrupting substances and new types of infectious diseases. Such new approaches should also help you to better manage the inevitable surprises that our interventions in nature will undoubtedly generate.

Thank you.

Acronyms used in the speech:
CFC -- Chlorofluorocarbons
PCB - Polychlorinated biphenyls
DES - Di-2-Ethylhexel Sebecate
TBT - Tributyl Tin
BSE - Bovine Spongiform Encephalopathy
IPCC - The Intergovernmental Panel on Climate Change (IPCC) and scientific consensus
SETAC - Society of Environmental Toxicology & Chemistry
WHO -- World Health Organisation