INSTITUT Veolia Environnement

  • Home
  • Report n°4: How Much to Spend for the Protection of Health and Environment

Report n°4: How Much to Spend for the Protection of Health and Environment

Dose-response functions

General considerations
The dose-response function (DRF) relates the quantity of a pollutant that affects a receptor (e.g. population) to the physical impact on this receptor (e.g. incremental number of hospitalizations). In the narrow sense of the term, it should be based on the dose actually absorbed by a receptor. However, the term dose-response function is often used in a wider sense where it is formulated directly in terms of the concentration of a pollutant in the ambient air, accounting implicitly for the absorption of the pollutant from the air into the body. The functions for the classical air pollutants (NOx, SO2, O3, and particulates) are typically of the that kind, and the terms exposure-response function or concentration-response function (CRF) are often used.

The DRF is a central ingredient in the impact pathway analysis and merits special attention. A damage can be quantified only if the corresponding DRF is known. Such functions are available for the impacts on human health, building materials, and crops, caused by a range of pollutants such as primary and secondary (i.e. nitrates, sulfates) particles, ozone, CO, SO2, NOx, Benzene, dioxins, As, Cd, Cr, Ni and Pb. The most comprehensive reference for health impacts is the IRIS database of EPA [http://www.epa.gov/iriswebp/iris/index.html]. For the application in an IPA that information often has to be expressed in somewhat different form, accounting for additional factors such as the incidence rate [ExternE 1998, Spadaro & Rabl 2004]. Unfortunately, for many pollutants and many impacts the DRFs are very uncertain or not even known at all. For most substances and non-cancer impacts the only available information covers thresholds, typically the NOAEL (no observed adverse effect level) or LOAEL (lowest observed adverse effect level). Knowing thresholds is not sufficient for quantifying impacts; it only provides an answer to the question whether or not there is a risk. The principal exceptions are carcinogens and the classical air pollutants, for which explicit DRFs are known (often on the assumption of linearity and no threshold).

By definition a DRF starts at the origin, and in most cases it increases monotonically with dose, as sketched schematically in Fig.3. At very high doses the function may level off in S-shaped fashion due to saturation, but that case is not of interest here. DRFs for health are determined from epidemiological studies or from laboratory studies. Since the latter are mostly limited to animals, the extrapolation to humans introduces large uncertainties.

A major difficulty lies in the fact that one needs relatively high doses in order to obtain observable nonzero responses unless the sample is very large; such doses are usually far in excess of typical ambient concentrations in the EU or North America. Thus there is a serious problem of how to extrapolate from the observed data towards low doses. Fig.3 indicates several possibilities for the case where the point P corresponds to the lowest dose at which a response has been measured. The simplest is the linear model, i.e. a straight line from the origin through the observed data point(s). The available evidence suggests that a dose-response function is unlikely to go above this straight line in the low dose limit. But the straight line model does appear to be appropriate in many cases, in particular for many cancers. In fact, most estimates of cancers due to chemicals or radiation assume this linear behavior.

Another possibility is the "hockey stick": a straight line down to some threshold, and zero effect below that threshold. Thresholds occur when an organism has a natural repair mechanism that can prevent or counteract damage up to a certain limit.

There is even the possibility of a "fertilizer effect" at low doses, as indicated by the dashed line in Fig.3. This can be observed, for example, in the dose-response functions for the impact of NOx and SO2 on crops: a low dose of these pollutants can increase the crop yield, in other words the damage is negative. Generally a fertilizer effect can occur with pollutants that provide trace elements needed by an organism.

In practice most DRFs used by ExternE, in particular all the ones for health, are assumed to be linear (without threshold). Note that for the calculation of incremental damage costs there is no difference between the linear and the hockey stick function (with the same slope), if the background concentration is everywhere above this threshold; only the slope matters. For the particles, NOx, SO2, O3 and CO the background in most industrialized countries is above the level where effects are known to occur. Thus the precise form of the ER function at extremely low doses is irrelevant for these pollutants; if there is a no-effects threshold, it is below the background concentrations of interest.

Health Impacts
In terms of costs, health impacts contribute the largest part of the damage estimates of ExternE. A consensus has been emerging among public health experts that air pollution, even at current ambient levels, aggravates morbidity (especially respiratory and cardiovascular diseases) and leads to premature mortality [e.g. Wilson & Spengler 1996, ERPURS 1997], see Table 1. There is less certainty about specific causes, but most recent studies have identified fine particles as a prime culprit; ozone has also been implicated directly. The most important cost comes from chronic mortality due to particles, calculated on the basis of Pope et al [2002] (this term, chosen by analogy with acute and chronic morbidity impacts, indicates that the total or long term effects of pollution on mortality have been included, by contrast to acute mortality impacts that are observed within a few days of exposure to pollution). Another important contribution comes from chronic bronchitis due to particles [Abbey et al 1995]. In addition there may be significant direct health impacts of SO2, but for direct impacts of NOx the evidence is less convincing.

In ExternE the working hypothesis has been to use the DRFs for particles and for O3 as basis. Effects of NOx and SO2 are assumed to arise indirectly from the particulate nature of nitrate and sulfate aerosols, and they are calculated by applying the particle DRFs to these aerosol concentrations. But the uncertainties are large because there is insufficient evidence for the health impacts of the individual components or characteristics (acidity, solubility, ...) of particulate air pollution. In particular there is a lack of epidemiological studies of nitrate aerosols because until recently this pollutant has not been monitored by air pollution monitoring stations. All DRFs for health impacts have been assumed linear at the population level, in view of the lack of evidence for thresholds at current ambient concentrations. By contrast to the homogeneous populations of cloned animals studied by toxicologists, the absence of a no-effect threshold is plausible for real populations because they always contain individuals with widely differing sensitivities (for example, at any moment about 1% is within the last nine months of life and thus extremely frail).