[Why is the "Number of Premature Deaths Due to Environmental Exposures" not Appropriately Quantifiable?]

Epidemiological studies and their applications in regulations of hazardous substances (e. g. by WHO, USA, EU) often quantify effects of environmental exposures on populations ("burden of disease") by calculating "numbers of premature deaths due to exposure". A recent example is the study by Schneider et al., commissioned by the German Federal Environmental Agency (Umweltbundesamt), into the burden of disease caused by exposures to nitrogen dioxide (NO₂) in Germany. The authors assessed the proportion of premature deaths due to exposure by the "Attributable Fraction" (AF). However, true numbers of premature deaths caused by NO₂ could be much higher or smaller. Indeed, Robins and Greenland showed in 1989 that the AF approach is inappropriate. Despite its far-reaching relevance for epidemiology and public health, their seminal work was not adequately taken into consideration, possibly due to its sophisticated level of mathematical argumentation. Our contribution illustrates - with simple examples - unappreciated but important pitfalls. We recommend that the concept of "number of premature deaths" be abandoned and "years of life lost due to exposure" be provided instead, calculated per capita. However, "years of life lost due to exposure" should not be stratified by age or causes of death (diseases). Furthermore, we show that "Disability Adjusted Life Years" (DALY) do not provide a meaningful measure to evaluate the effect of environmental exposures on populations.