Extrapolations and public policy.

Abstract

In the current issue of the journal, Hermann, DiVita, and Lanier calculate the potential risk for the Commonwealth of Virginia and its surrounding region of commissioning 29 proposed fossil fuel plants. They conclude that the risk is sufficient to invoke the “precautionary principle,” which proposes that when information on a risk is incomplete, the wise risk management option is the one that minimizes the risk of unnecessary health effects. Applied to this situation, the precautionary principle might be interpreted as denying a building permit or licensing of the plants and precluding them from becoming operational. The authors lay the groundwork of their argument by calculating the most likely exposure to fine particulate air pollution that would result from the construction of the proposed fossil fuel plants. The exposure assessment is projected using emissions inventory predictions, applying a dispersion model, adding the load to existing pollution levels, and projecting the ambient levels of fine particulate matter that would result. The authors then apply an exposure-response relationship between fine particulate air pollution and mortality representative of many recent studies. In this way, they calculate that there would be 17 excess deaths among almost 38 million residents of the United States. Is this argument convincing? In the absence of information that would suggest otherwise, the authors have demonstrated that under existing environmental conditions, and assuming a comparable population health status, the impact of the plan, under ideal conditions, would be likely to produce a risk of death for 17 people distributed among 12.9% of the population of the United States, or a risk of premature or untimely death in the same population of less than 0.5 in a million. Does this matter? By the time all 29 power plants are constructed, assuming that they are, the mortality profile of the U.S. population will have changed. Deaths from heart disease and stroke will probably continue to fall, as will deaths from all tobacco-related disease. The mortality function is likely to change as well because the general population features fewer individuals with strong risk for cardiovascular disease and more individuals who are immuno-suppressed for a variety of reasons. The current extrapolation may or may not hold but no epidemiological study will ever have the power to resolve 17 events in 38 million. Thus, the predictions of this article are not testable and, if they turn out to be true under current conditions, there is no guarantee that those conditions will be the same in 10 or 20 or 30 years. That is the problem with an extrapolation over time. It holds true only when the conditions that define it are constant, which they never are. Because of competing causes of mortality and the underlying health or morbidity of the population, the contribution of a factor of this relative strength to population health status or mortality is invisible. In another context, an effect this low would be deemed de minimis. Does that mean it is unimportant? Clearly no, if you or someone you care about happens to be one of the 17. However, the reader does not have sufficient information to judge whether ”only” 17 deaths are the best or the worst that can be expected as far as mortality, or somewhere in the middle. With the information given, there is no way of judging, as decision makers must, the alternative risks of the various decision options. For example, decision makers do not know if the reduction in energy availability, which, in a market economy usually finds its expression as an increase in price, will lead to more deaths by discouraging use of air conditioning or by pricing heating oil out of range or making electric automobiles infeasible or increasing the cost of running hospitals. They have no way of knowing how the life-cycle risk of the plants, including their construction hazards, compare with alternatives. If electricity is generated out of the area and supplied to Virginia by way of the Eastern Intertie, would this not simply displace the emissions elsewhere? Might 17 deaths avoided in and around Virginia turn into 17 deaths in the Midwest? There is another key fact that decision makers do not know: whether an alternative energy technology would be feasible, sufficient to meet demand (ideally accounting for voluntary conservation), safer, and available in time. For example, the authors suggest that new-generation nuclear energy would be safer and that wind power and solar energy have negligible mortality. Perhaps so, although such studies tend to neglect occupational risks and mortality needs to be stacked against injury and disease risk during the complete life cycle of the facility. Without a full examination of the