2 Krewski D. Reanalysis of the Harvard Six Cities Study and the American cancer society study of particulate air pollution and mortality. Cambridge, MA: Health Effects Institute, 2000.

3 Bellander T, Svartengren M, Berglind N, Staxler L, Järup L. The Stockholm Study on Health Effects of Air Pollution and their Economic Consequences (SHAPE), Part II: Particulate matter, nitrogen dioxide and health effects. Exposure-response relations and health consequences in Stockholm County. Stockholm: Department of Environmental Health, 1999.

4 Sommer H, Chanel O, Vergnaud JCh, Herry M, Sedlak N Seethaler R. Monetary valuation of road traffic-related air pollution: health costs due to road traffic-related air pollution--an impact assessment project of Austria, France and Switzerland: third WHO Ministerial Conference of Environmentand Health. London: WHO, 1999.

In the past, establishment of a link between pollution and the prevalence and incidence of respiratory disease has been difficult, despite the association being intuitively appealing.² We believe that the explanation for this difficulty is that the current methods of assessing bulk pollution exposure by use of a 10 µg/m³ increase in particulate matter, an undeveloped and non-selective measure of particulate loading, might be inadequate. The impact of petrol vehicle emissions on aerosol loading and subsequent public health is probably underestimated in many models of the urban atmosphere. Modern petrol engines are believed to have very low or zero primary particulate emission, but their emissions are responsible for the production of photochemically produced secondary organic aerosols (SOA), which contribute to aerosol loading and, possibly, in turn to respiratory disease.

The composition and density of traffic in most cities creates an atmosphere that is dominated by these petrochemical-related SOA precursors,³ which are not easily measured and are currently not included in the emission-driven models of pollution exposure used in most medical and epidemiological research. What must be appreciated is that the nature and chemical composition of PM₁₀ in our cities has undergone many changes over the past 30 years. Such change leads to the apparent paradox that, although total PM₁₀ concentrations in many cities are remaining static or falling, largely because of diesel-engine emission legislation, the prevalence of respiratory disease, especially asthma, is increasing.⁴ Reliable methods of modelling pollution that incorporate the evolution of urban PM₁₀ composition must be included in any future work looking at the relation between public health and air pollution.

Künzli and colleagues claim that several pollutants are correlated with PM₁₀, hence epidemiological studies are precluded from strictly attributing observed effects to a single pollutant. We acknowledge the difficulty in assessment of effects of exposure to complex mixtures, but claim that all simplifiation strategies must be based on biological plausibility to be taken as valid. PM₁₀ represent only a fraction from a given sampling method. Without a description of particle distribution, PM₁₀ is too unspecific for generalised exposure assessments. In several studies, biological response on air pollution from diesel combustion is significantly different from that seen with mineral particles. Thus, health hazards will vary with the amount of particles inducing biological responses and not the total load in a PM₁₀ fraction. Differences in particle distribution between regions in the report might significantly affect the conclusions. Local events can lead to large variations in particle distribution.^1,2

The investigators also claim that PM₁₀ can be used as an indicator of fossil-fuel combustion. Particles derived from fuel combustion are, however, mostly included in a PM_2·5 fraction. Thus, PM_2·5 would be a more appropriate variable if Künzli and colleagues preferred to focus on combustion particles. As they mention, a stronger association between increased mortality and PMs has been seen for the combustion-related PM_2·5 than for PM₁₀.

The lack of biological plausibility for a cause-effect relation between PMs and increased mortality has also been discussed by other researchers. In 1998, J Gamble³ concluded that there is no substantive basis for a cause-effect relation between long-term ambient PM_2·5 and increased mortality. Studies in animals and also comparison with individual exposure to inhaled cigarette smoke indicate that the risks of outdoor PM_2·5 fractions are overestimated by more than 100-fold in the model used by Künzli and colleagues. J Schwartz and colleagues⁴ showed that mineral particles between 2·5 µm and 10·0 µm were not associated with mortality. The contribution from particles in outdoor air to the total daily load of inhalable particles might be of minor importance.⁵ However, we cannot exclude that criticallly ill or significantly predisposed people might in some situations have their lives shortened slightly.

The conclusions of Künzli and colleagues are based on rough assumptions, probably irrelevant for several regions of Europe. In France, where half of new cars run on diesel, particles from combustion of diesel in a PM₁₀ fraction will be more prominent than in corresponding samples in Scandinavia, where mineral particles dominate in periods. The search for subtle links between environmental factors and diseases has been controversial. In our view, some scientists and WHO seem to have been stretching what can be done with epidemiology.⁵ The long-term effects of exposure to products from combustion of fossil fuel remains elusive and will have to be investigated in studies with more specific design.

Authors' reply

Tom Bellander discusses the issue of lifetime lost among pollution related death. Although not relevant for the attributed number of deaths it is of importance in the economic assessment and the discussion about who is at risk is important. Direct estimates of the time lost or the age distribution of the victims are not available; thus, dependent on the assumptions, indirect estimates come to different conclusions of time lost, ranging from several months to a few years.⁵ Our assumption that all victims die due to cardiopulmonary causes, thus having the age distribution of all cardiopulmonary death, is plausible.

1 Kaiser J. Panel backs EPA and 'Six Cities' Study. Science 2000; 289: 711.