Summary of Meeting Paper

The 1996 Annual Meeting of the Society for Risk Analysis-Europe

Epidemiological Equivocacy as a Source of Uncertainty in Risk Assessment. T. Sly, School of Environmental Health, Ryerson Polytechnic University, Toronto, Canada

Epidemiological inquiries frequently result in less enlightenment than expected. While details of new risks to health appear to be published almost daily, the impact is muddied by weak effects, inconclusive associations, and apparent ambiguity in the research. Contradictions between studies, in particular, are the source of media debate and even popular satire. Society's perception is one of questioned credibility, regardless of how rigorous or flawless the investigation, and society often fails to detect what it is looking for in published science: a clear, definitive, conclusion that will inform decision making.

At the same time, the recognition and interpretation of uncertainty is highly desirable in today's science, and if the uncertainties were routinely explicit, and public debate about them encouraged, much of the anxiety about risks to health might actually diminish.

Investigators are thus obliged to tread cautiously between the dangers of declaring (unfounded) confidence on the one hand, and inadvertently generating concern from the method of inquiry on the other. This paper identifies five sources of uncertainty and ambiguity in health and medical research.

1. Equivocacy due to the framework for investigation

Awareness of the importance of environmental and social factors in the disease process had been the norm prior to the mid-nineteenth century, but the emerging concept of ill health as a purely biological phenomenon, and the rise of "scientific" - particularly allopathic - medicine, narrowed the focus. During the second half of the twentieth century, however, social, economic, behavioural, and environmental determinants returned to a more central position, but at a cost. While investigations of specific agents in most cases follow a straightforward sequence of steps, with results that exonerate or condemn that agent, explorations into the roles of interdependent factors are complex in design, difficult in analysis, and frequently enigmatic in outcome. As the base for inquiry has broadened, the conclusions have become increasingly equivocal.

2. Equivocacy inherent in the nature of the current topics and of their risks to health

Early investigations in epidemiology sought to identify and contain disease entities largely of an acute and infectious nature. Although infectious diseases still demand the attention of public health workers, the focus of much of today's epidemiology is upon cardiovascular diseases and cancers, and prospective studies into these conditions may need to gather 10-20 years' worth of data. As our understanding of synergism and interaction between biologic, genetic, and psychosocial factors increases, the results of research will inevitably appear more complex.

3. Equivocacy from the conservative methods, models, and tools of modern research

The credibility of the "scientific method" stems from the conservative statistical analyses by which hypotheses are tested. Instead of assembling supportive evidence to validate hypotheses, they are tested in a "null" form: - that there is no effect, no difference, no association, etc. Only when the null hypothesis is rejected at a satisfactory level of significance, can the original hypothesis be entertained. Such a process will allow us to state, for instance, ".. that the exposure and the outcome appear to be associated - but with a probability that this association might still have arisen by chance alone in up to 5% of repeated trials". A result in this form, while accurate, remains inherently inconclusive for legal and other non-scientific arguments.

Although the usefulness of the result can be improved by measuring dose-response Gradients between exposure and outcome, or the strength of the association, the inferences drawn from hypothesis-testing can easily misinform the lay person.

4. Equivocacy from emphasis, de-emphasis, omission, or addition

Subtle adjustment of reported results from proprietary research or research sponsored or commissioned by interested parties has been the subject of extensive examination. It is not beyond the realm of possibility that for political, or perhaps commercial reasons, the commissioners of any study may choose to stress "inconclusive" findings.

The lay press frequently misconstrues results to be more conclusive than they really are: the "breakthrough" phenomenon. But blame must also be attributed to researchers and their institutions who for diverse reasons often issue news releases far too early, and with not enough attention to how the message might be received.

5. Equivocacy from poor design and implementation

Insufficient attention to the design, sampling, data Gathering, analysis, and interpretation, can manifest as a weak, ambiguous, or even completely spurious finding when the result should have been quite definite - or as strong association when in reality there was none. In either instance, the misleading results can be expected to increase the apparent inconsistency between studies. Both circumstances are unacceptable. Several criteria with potential for weakening the validity and accuracy of studies in the chronic, non-infectious realm include:

Final considerations

Even with optimal data and analysis, a conclusion "beyond all reasonable doubt" may not be attained, nor in some instances should one be expected. To postpone decisions affecting the health of people or the environment until such time as incontrovertible evidence becomes available and accepted may be irresponsible and unethical. In such instances, the "burden of proof' should be weighed against the "burden of prudence."

At all stages of the reporting and publication of research, however, the full extent of the uncertainties and limitations must be stated and interpreted Science cannot provide unequivocal assurance of safety, yet society may become more reassured were it to hear that admission more clearly.